[Patrick Turner]

I have examined my HP33A for about 5 hours armed with the HP service
manuals just to understand some of the basics involved with this
instrument, and to try to find out why the 333A distortion
measurements were plain wrong compared to my own design of THD
measurer built about 12 years ago.

But even in voltmeter mode, the use of my 333A was became pointless
because after it had been turned on for 1/2 an hour the voltage
measurement strated to drift about 15% high, and then If I tried to
use it to measure THD while monitoring on my CRO, and inevitably
overloading the circuits when changing range settings the voltmeter
would become unstable and give readings which were up to 3 times what
they should have been. The instrument is fairly immune failures if Vin
is only 10Vrms which I used, 1kHz, for my circuit testing and probing.
Nothing became overheated. But really bad drift occurs. There are many
electrolytic coupling caps. The meter amp appears to be faulty.

Now the meter amp is a damn strange concoction of discrete transistors
with both positive and negative loops of FB and which have different
effects at different F, so that it can operate from 5Hz to several
MHz, with a very flat response **providing** all the vari-caps and
gain pots and the biasing pot are all set just right. I found the bias
pot was somewhat out, but when adjusted right according to the
schematic Vdc voltages the bad meter drift still occurred.

It would have been shortly after the 333A was developed by HP that
opamps were invented and VOILA, all the efforts HP went to became
immediately obsolete. I began to think the HP meter circuit was like a
horse and buggy compared to a good motot car, and the circuit working
seems to rely on all resistors never changing in value, and on many
other things like cap leakage, and perhaps bjts and other bits
becoming dodgy and almost unidentifiable in the years since 1967.

Anyway, methinks blind freddy could have told HP that they had little
chance of making the meter amp reliable and precise over a long time.

The meter amp needs only approximately 1mVrms to develop 100mV of
output signal which will swing the meter full scale. Whatever is
drifting to cause the meter reading to wander about the workshop when
it wants to remains difficult to fathom, but I can see drift could be
very easy with all that open loop gain and with an absurd closed loop
gain of about 100. If the meter amp gain was 10, and there was less
attenuation by switching and resistors before amplifying a signal back
up again for metering, methings reliability would be better.

Before the drift began, I set the unit going in Voltmeter function
with a known 10Vrms input and measured all Vac and Vdc all over the
circuit boards. There was less than 1mV of input to the meter amp, and
the NFB at the emitter of the first bjt also measured the same, so
differential input between base and emitter was a lot less than
1mV,and not easy to measure. So open loop gain at 1khz was very high,
like an opamp, but all devices are set up in SE mode, something one
does not see in many opamps because its so difficult to get it to work
well compared to using differential pairs and a PP output stage.
OPamps are good to use in DIY test gear and with sockets so that if an
opamp dies, just unplug&chuckit, and in goes another for $3. I've
rarely ever replaced any opamps in the many bits of test gear I made.
But the ordinary sockets for opamps are more likely to stay TIGHT than
the kind of wonky HP lead to pin connectors.

I found the meter readings were difficult to interpret, and I have
tried a few times to read the manual but it is bloody obtuse, and just
what is going on and exactly what is being measured remains vague to
me There were plenty of early generation GEEKS who managed to stay
employed at companies like HP but who could not write manuals to be
easily be understood.

Now my own THD meter is FAR SIMPLER, with 1/20 of the parts, and far
less likely to go wrong, yet it measured THD easily to 0.002% from a
sample audio signal at 1kHz as low as 0.5Vrms.

I have a low THD signal which is fed to an amp. The amp output can be
applied to a dummy load aross which is a 1k switched attenuator to
reduce the signal to below 10Vrms if it is a high level signal without
attenuation, which could be the case of a 300W amp feeding 8 ohms.
There is a following LC bridged T notch filter which blocks the 1 kHz
signal to a depth of -90dB, approx, and all other F can go past where
they can be applied to a meter and CRO. So %THD is calculated as 100 x
Vdistortion / Vsample, and there are few attenuators in the signal
path, except a hum filter to remove content below 300Hz to be able to
see and measure harmonics above 1 khz better. After the bridged T LC
filter I have a HPF with an opamp 1.5kHz to 10kHz BW. Gain is 10 or
+20dB and this lowers 1khz down another 20dB while amplifying the
distortion to examine low distortion signals on the CRO's lowest and
most sensitive range. For amps where I know the result will be nearly
all above 0.05%, just the passive bridged T is enough and the
following opamp and HPF is not used.
When I use my own THD meter I know exactly what I am measuring.

When using the 333A for THD, I set the input to Distortion and adjust
the meter range to a low level to display the THD on the CRO from the
outpt terminals of the 333A and to be able to measure the 333A output
voltage. But the meter reading has no corelation to what I measure at
the output with an external meter.

Its only when I compare the use of my own simple reliable gear to this
vastly over complex and old fashioned HP junk that I began to feel
some despair, and concluded that I could be wasting days farnarkling
around with something that's better abandoned than restored, or
seriously modified with replacement opamps. Now I know why I was given
the 333A and the 331 and mabye the previous owner couldn't make out
what he was measuring, and as sure as God made lil' apples, he must
have had fun swearing and cursing the crazy behaviour.

I have not fully decided to junk the 333A. The bridged nulling circuit
with wien bridge with precision transistors and a 4 gang tuning cap is
an engineering masterpeice. But HP have light dependant resistors and
lamps that light them up and a complex kind of circuit that defies
easy understanding. While it works, it is a continuing miracle. And
the working voltages are very low, and there's lots of high impedance
circuitry with wide BW and if one tries to test the unit without
covers on there is terrible noise problems.

But all I need is one meter which tells me the voltage input, and
another to tell me the distortion voltage after the fundemental tone
has been either filtered out with a tunable nulling circuit or
filtered out with a tunable bandstop filter. I always use two meters
and a CRO to measure THD, so I can plot the THD easily against output
voltage because we need to know how THD increases between low DUT
levels and high DUT levels, right up to clipping and maybe beyond.
I've got plenty of meters.

The construction of a narrow bandwidth bandstop filter which is
infinitely tunable across a decade of frequencies means one has to
vary C, or R, or pairs of Cs or Rs, anf the tuning gang is actually
very stable and good compared to any pot. But switched caps and
switched resistors are also good but then one has to tune the input
test F to suit the meter filter F settings. I've found the wien bridge
oscillators with C tuning gangs to be the most long lasting and stable
and with a nice spread of F on the dial. But for very low F, the R
associated with a given C of say 1,200pF max from a parallel 3 gang
radio cap has to be above 20Megohms to get the F below 7Hz for amp
testing.

A variable R + fixed C wien bridge oscillator is a nice srable thing
if the range of R variation is small and you just want to test audio
amps with a single fixed F, say 1kHz, and you have a "broad tune range
pot for +/- 50Hz with a second adjust pot for +/- 5Hz to get the
bandstop filter to react perfectly and in phase with the amp output
signal which is being fed with the slightly adjustable single F
oscillator. This careful tuning gives the deep rejection of 1,000Hz.

But one may use a fixed notch filter based around the RC parallel T
and some FB around two opamps. Usually this has limitations to the
depth of the notch, maybe only -55dB. But if you cascade a couple of
such easy simple things then the depth of the notch can become
enormous, and there is nothing to adjust except the oscillator F.

I soon learnt the LC bridged T notch filter was a real marvel because
it was passive, and could easily take the full 70Vrms applied from any
amp without needing any precision attenuators and worries about frying
delicate solid state active parts, and what came out of the bridged T
could be easily measured and viewed on a CRO, metered, and used for
calculations without having any confusions.

I found most wisdom about audio signal testing in the pages of
Wireless World between the years 1950 and 1980 which was a UK magazine
which attracted articles from many of the brightest stars in analog
electronics over many years. I doubt HP contributed anything, and
methinks the British boffins may have smiled at HP's valiant efforts
which always seemed to use many more devices and be about
unaffordable. But the Yanks were first to the moon in 1969, so despite
the complexity of their gear they could do wonders with it while it
worked, and if the dudes spent enough time to learn how to use it.

I think I saw the price of a 333A for 1967 at $1,250.00, and in
today's money its maybe 20 grand, just for a little box gadget. I
guess the wondrous metalwork and stainless steel screws cost a bomb.

To conclude, If I find the time and can't improve the existing 333A
performance I may well change the meter amp to an opamp or two with
enough bandwidth and fit some diode voltage clamping to prevent the
worst effects of internal lingering Vdc changes due to the C coupling
caps all charging up then discharging slowly after an over load.
I just don't know enough to improve on the HP wien bridge nulling
circuit electronics. But perhaps a slower acting fine tune adjust
mechanism for the tuning cap would be in order.
There is a also a buffer stage to drive the meter amp, and it is unity
gain, and that also could be replaced with a far simpler opamp
arangement. The meter range switching circuit is a veritable wonder
though which seems to work perfectly in 10dB steps and which has no
active circuitry. Switch quality seems to be fair even though my 333A
has had a very hard life judging my external appearance and some mild
corrosion. The internal wiring has many wires crimped onto lead end
grippers which then slide onto what look like gold plated pcb pins.
many came loose, some were intermitent and what is needed is to cut
off the lead grippers and solder the wires to the pcb pins and
restrain them from movement by dressing wire positions and using cable
ties to avoid joint failure.

The more I look at old electronics gear, the more I see how to make it
perform a little better, and I do like my gear to be easier to live
with than my present old HP333A.

One multi band variable wien bridge oscillator I use all the time with
just one opamp has now got intermittent pot problems. I have a couple
of box fulls of old radio tuning gangs, maybe several identical 3 gang
types with each gang 20pF to 400pF which can make 60pF to 1,200pF when
paralleled. Two such 3 gangers made to turn together in a wien bridge
can be made to get from 5Hz to 60Hz, and give 5 ranges with the top
one at 50k to 600kHz. With the variable C, one needs a 5 position two
pole rotary switch and 1% fixed resistors of 22Mohms, 2.2M, 220k, 22k
and 2k2.

I used to have a Topward function generator with sine, square and
triangular waves and with AM and FM and a horrible complex circuit on
a tiny board, horrid plasic case, and a noisy little mains tranny. I
blew its output amp up about 3 times by accidental contact with tube
amp rail voltages. The press button range switches began to be
unreliable, the single gang pot for varying Vdc which controlled the
variable F wore out, twice, and I swapped that pot for a 12 pos
switched attenuator crammed in, but even that gave troubles after time
and the F would drift. The THD was bad, and worse than any of my wien
bridge oscillators, including the tube one I use. Eventually after
several fixes, The Topward output lead came into accidental contact
with some quite high tube amp B+. There was a fizzling sound with
puff of smoke, and the high Vdc pushed way too much current deep into
the many ICs in the unit, and all I could do was put the cooked mess
into my rubbish bin.

Patrick Turner.

[Patrick Turner]

Below are my initial thoughts about my HP333A distortion meter which
has faults I can't fix because I don't have time for the analysis, and
if I was successful, finding exactly the right spare part is
difficult, and the fix might not last considering the age and obsolete
complexity of what is there, including very many electrolytic
capacitors, old transistors, and resistors, diodes, connections and
switch contacts. After a sleep I continued this AM with
appraisals.......

I decided to re-test a old tubed wien bridge oscillator which I know
produces 1.1% THD at 8.8Vo.

The 333A gave the same strange volt meter drift about 10 mins after
after turn on. With Auto mode, the nulled distortion difted up and
down more than in manual mode. Then after leaving it running for 10
mins the Vo became a blur of noise as something somewhere went
beserko, and I came to the conclusion this junk ain't worth fixing
unless I happened to be a retired HP service technician who loved
fixing these things.

For the brief time it seemd to work I recorded the Vin, meter range
setting, and distortion Vo to my CRO which I could measure accurately
enough with my Fluke meter.

Then I switched to a 331A distortion meter I also have and with the
same settings and voltages got different Vdn readings.

Then I used my own THD meter to confirm what I already knew about my
tubes oscillator; it had high THD, but was still a useful tool for
response checks between 20Hz and 200kHz.

There was no way I could see how the Vdn from the outputs of either
333A or 331A could be multiplied or divided to convert the readings to
come close to what my own meter was telling me. So I guess BOTH units
could be faulty.

I myself could be faulty for not being able to interpret what I am
reading.

I investigated what signals were flowing along the 333A signal path
when switched to Distortion function and it seemed like a very small
voltage sample of Input voltage was applied to the wien bridge nulling
amp, about 0.027V out of 8.4Vin and only 0.0006V was the possible
output voltage, which is then amplified x 100 to give 60mV when the
range switch was turned to 0.001. If Dn% = 100 x Vo/Vdn, then I would
get 100 x 0.06/8.4 = 0.7%, but I know Dn% = 1.1%

Anyway, the 331A when used in Voltmeter mode works fine, so I'll keep
it for that use, but I see no reason to use it for THD measurements.

The 333A could possibly be converted to a better circuit with opamps
and utilising the existing PSU.

Or make into a nice 5 range wien bridge oscillator with range from
about 6Hz to 600kHz.

My paper records include some remarkable efforts from British
boffins.......

I have Ian Hickman's schematic details for a **TUNABLE** THD meter
reviewed fully in Electronics World in January 1996. This measures
down to 0.001% THD. Its wien bridge network has tuning capacitor
gangs. The principles were based on an earlier design by Linsley Hood
in 1972. About 5 discrete transistors are used with some opamps for
getting dwon from 0.01% to 0.001%.

The earlier Linsley Hood design of THD meter from Wireless World, July
1972 measures to 0.01% and it also uses a wien bridge network with
switched resistors and tuning gangs, and uses 8 discrete transistors
which no critical types.
If the C gangs are 2 x 470pF - 40pF, 40max, you can test between 22Hz
and 26kHz.

Use of say two identical 3 gang caps each with a total of 1,410pF -
100pF, and R values of 22Mohms, 2M2, 220k, 22k, 2k2
one could get ranges of 5.1Hz to 72Hz, 51Hz to 720Hz, 510Hz to 7.2kHz,
5.1kHz to 72kHz, and 51kHz to 720kHz.

The ability of accurate THD measurents above say 20kHz begins to drop
because of the HF open loop gain reductions of transistors or opamps
used.

So it looks possible to equal the HP performance but with far less
expensive and difficult to understand / fix crap under the bonnet.

Patrick Turner.

...Patrick Turner at wrote:
I have examined my HP33A for about 5 hours armed with the HP service
manuals just to understand some of the basics involved with this
instrument, and to try to find out why the 333A distortion
measurements were plain wrong compared to my own design of THD
measurer built about 12 years ago.

But even in voltmeter mode, the use of my 333A was became pointless
because after it had been turned on for 1/2 an hour the voltage
measurement strated to drift about 15% high, and then If I tried to
use it to measure THD while monitoring on my CRO, and inevitably
overloading the circuits when changing range settings the voltmeter
would become unstable and give readings which were up to 3 times what
they should have been. The instrument is fairly immune failures if Vin
is only 10Vrms which I used, 1kHz, for my circuit testing and probing.
Nothing became overheated. But really bad drift occurs. There are many
electrolytic coupling caps. The meter amp appears to be faulty.

Now the meter amp is a damn strange concoction of discrete transistors
with both positive and negative loops of FB and which have different
effects at different F, so that it can operate from 5Hz to several
MHz, with a very flat response **providing** all the vari-caps and
gain pots and the biasing pot are all set just right. I found the bias
pot was somewhat out, but when adjusted right according to the
schematic Vdc voltages the bad meter drift still occurred.

It would have been shortly after the 333A was developed by HP that
opamps were invented and VOILA, all the efforts HP went to became
immediately obsolete. I began to think the HP meter circuit was like a
horse and buggy compared to a good motot car, and the circuit working
seems to rely on all resistors never changing in value, and on many
other things like cap leakage, and perhaps bjts and other bits
becoming dodgy and almost unidentifiable in the years since 1967.

Anyway, methinks blind freddy could have told HP that they had little
chance of making the meter amp reliable and precise over a long time.

The meter amp needs only approximately 1mVrms to develop 100mV of
output signal which will swing the meter full scale. Whatever is
drifting to cause the meter reading to wander about the workshop when
it wants to remains difficult to fathom, but I can see drift could be
very easy with all that open loop gain and with an absurd closed loop
gain of about 100. If the meter amp gain was 10, and there was less
attenuation by switching and resistors before amplifying a signal back
up again *for metering, methings reliability would be better.

Before the drift began, I set the unit going in Voltmeter function
with a known 10Vrms input and measured all Vac and Vdc all over the
circuit boards. There was less than 1mV of input to the meter amp, and
the NFB at the emitter of the first bjt also measured the same, so
differential input between base and emitter was a lot less than
1mV,and not easy to measure. So open loop gain at 1khz was very high,
like an opamp, but all devices are set up in SE mode, something one
does not see in many opamps because its so difficult to get it to work
well compared to using differential pairs and a PP output stage.
OPamps are good to use in DIY test gear and with sockets so that if an
opamp dies, just unplug&chuckit, and in goes another for $3. I've
rarely ever replaced any opamps in the many bits of test gear I made.
But the ordinary sockets for opamps are more likely to stay TIGHT than
the kind of wonky HP lead to pin connectors.

I found the meter readings were difficult to interpret, and I have
tried a few times to read the manual but it is bloody obtuse, and just
what is going on and exactly what is being measured remains vague to
me There were plenty of early generation GEEKS who managed to stay
employed at companies like HP but who could not write manuals to be
easily be understood.

Now my own THD meter is FAR SIMPLER, with 1/20 of the parts, and far
less likely to go wrong, yet it measured THD easily to 0.002% from a
sample audio signal at 1kHz as low as 0.5Vrms.

I have a low THD signal which is fed to an amp. The amp output can be
applied to a dummy load aross which is a 1k switched attenuator to
reduce the signal to below 10Vrms if it is a high level signal without
attenuation, which could be the case of a 300W amp feeding 8 ohms.
There is a following LC bridged T notch filter which blocks the 1 kHz
signal to a depth of -90dB, approx, *and all other F can go past where
they can be applied to a meter and CRO. So %THD is calculated as 100 x
Vdistortion / Vsample, and there are few attenuators in the signal
path, except a hum filter to remove content below 300Hz to be able to
see and measure harmonics above 1 khz better. After the bridged T LC
filter I have a HPF with an opamp 1.5kHz to 10kHz BW. Gain is 10 or
+20dB and this lowers 1khz down another 20dB while amplifying the
distortion to examine low distortion signals on the CRO's lowest and
most sensitive range. *For amps where I know the result will be nearly
all above 0.05%, just the passive bridged T is enough and the
following opamp and HPF is not used.
When I use my own THD meter I know exactly what I am measuring.

When using the 333A for THD, I set the input to Distortion and adjust
the meter range to a low level to display the THD on the CRO from the
outpt terminals of the 333A and to be able to measure the 333A output
voltage. But the meter reading has no corelation to what I measure at
the output with an external meter.

Its only when I compare the use of my own simple reliable gear to this
vastly over complex and old fashioned HP junk that I began to feel
some despair, and concluded that I could be wasting days farnarkling
around with something that's better abandoned than restored, or
seriously modified with replacement opamps. Now I know why I was given
the 333A and the 331 and mabye the previous owner couldn't make out
what he was measuring, and as sure as God made lil' apples, he must
have had fun swearing and cursing the crazy behaviour.

I have not fully decided to junk the 333A. The bridged nulling circuit
with wien bridge with precision transistors and a 4 gang tuning cap is
an engineering masterpeice. But HP have light dependant resistors and
lamps that light them up and a complex kind of circuit that defies
easy understanding. While it works, it is a continuing miracle. And
the working voltages are very low, and there's lots of high impedance
circuitry with wide BW and if one tries to test the unit without
covers on there is terrible noise problems.

But all I need is one meter which tells me the voltage input, and
another to tell me the distortion voltage after the fundemental tone
has been either filtered out with a tunable nulling circuit or
filtered out with *a tunable bandstop filter. I always use two meters
and a CRO to measure THD, so I can plot the THD easily against output
voltage because we need to know how THD increases between low DUT
levels and high DUT levels, right up to clipping and maybe beyond.
I've got plenty of meters.

The construction of a narrow bandwidth bandstop filter which is
infinitely tunable across a decade of frequencies means one has to
vary C, or R, or pairs of Cs or Rs, anf the tuning gang is actually
very stable and good compared to any pot. But switched caps and
switched resistors are also good but then one has to tune the input
test F to suit the meter filter F settings. I've found the wien bridge
oscillators with C tuning gangs to be the most long lasting and stable
and with a nice spread of F on the dial. But for very low F, the R
associated with a given C of say 1,200pF max from a parallel 3 gang
radio cap has to be above 20Megohms to get the F below 7Hz for amp
testing.

A variable R + fixed C wien bridge oscillator is a nice srable thing
if the range of R variation is small and you just want to test audio
amps with a single fixed F, say 1kHz, and you have a "broad tune range
pot for +/- 50Hz with a second adjust pot for +/- 5Hz to get the
bandstop filter to react perfectly and in phase with the amp output
signal which is being fed with the slightly adjustable single F
oscillator. This careful tuning gives the deep rejection of 1,000Hz.

But one may use a fixed notch filter based around the RC parallel T
and some FB around two opamps. Usually this has limitations to the
depth of the notch, maybe only -55dB. But if you cascade a couple of
such easy simple things then the depth of the notch can become
enormous, and there is nothing to adjust except the oscillator F.

I soon learnt the LC bridged T notch filter was a real marvel because
it was passive, and could easily take the full 70Vrms applied from any
amp without needing any precision attenuators and worries about frying
delicate solid state active parts, and what came out of the bridged T
could be easily measured and viewed on a CRO, metered, and used for
calculations without having any confusions.

I found most wisdom about audio signal testing in the pages of
Wireless World between the years 1950 and 1980 which was a UK magazine
which attracted articles from many of the brightest stars in analog
electronics over many years. I doubt HP contributed anything, and
methinks the British boffins may have smiled at HP's valiant efforts
which always seemed to use many more devices and be about
unaffordable. But the Yanks were first to the moon in 1969, so despite
the complexity of their gear they could do wonders with it while it
worked, and if the dudes spent enough time to learn how to use it.

I think I saw the price of a 333A for 1967 at $1,250.00, and in
today's money its maybe 20 grand, just for a little box gadget. I
guess the wondrous metalwork and stainless steel screws cost a bomb.

To conclude, If I find the time and can't improve the existing 333A
performance I *may well change the meter amp to an opamp or two with
enough bandwidth and fit some diode voltage clamping to prevent the
worst effects of internal lingering Vdc changes due to the C coupling
caps all charging up then discharging slowly after an over load.
I just don't know enough to improve on the HP wien bridge nulling
circuit electronics. But perhaps a slower acting fine tune adjust
mechanism for the tuning cap would be in order.
There is a also a buffer stage to drive the meter amp, and it is unity
gain, and that also could be replaced with a far simpler opamp
arangement. The meter range switching circuit is a veritable wonder
though which seems to work perfectly in 10dB steps and which has no
active circuitry. Switch quality seems to be fair even though my 333A
has had a very hard life judging my external appearance and some mild
corrosion. The internal wiring has many wires crimped onto lead end
grippers which then slide onto what look like gold plated pcb pins.
many came loose, some were intermitent and what is needed is to cut
off the lead grippers and solder the wires to the pcb pins and
restrain them from movement by dressing wire positions and using cable
ties to avoid joint failure.

The more I look at old electronics gear, the more I see how to make it
perform a little better, and I do like my gear to be easier to live
with than my present old HP333A.

One multi band variable wien bridge oscillator I use all the time with
just one opamp has now got intermittent pot problems. I have a couple
of box fulls of old radio tuning gangs, maybe several identical 3 gang
types with each gang 20pF to 400pF which can make 60pF to 1,200pF when
paralleled. Two such 3 gangers made to turn together in a wien bridge
can be made to get from 5Hz to 60Hz, and give 5 ranges with the top
one at 50k to 600kHz. With the variable C, one needs a 5 position two
pole rotary switch and 1% fixed resistors of 22Mohms, 2.2M, 220k, 22k
and 2k2.

I used to have a Topward function generator with sine, square and
triangular waves and with AM and FM and a horrible complex circuit on
a tiny board, horrid plasic case, and a noisy little mains tranny. I
blew its output amp up about 3 times by accidental contact with tube
amp rail voltages. The press button range switches began to be
unreliable, the single gang pot for varying Vdc which controlled the
variable F wore out, twice, and I swapped that pot for a 12 pos
switched attenuator crammed in, but even that gave troubles after time
and the F would drift. The THD was bad, and worse than any of my wien
bridge oscillators, including the tube one I use. *Eventually after
several fixes, The Topward output lead came into accidental contact
with some quite high tube amp B+. There was a *fizzling sound with
puff of smoke, and the high Vdc pushed way too much current deep into
the many ICs in the unit, and all I could do was put the cooked mess
into my rubbish bin.

Patrick Turner.

[Bret L]

As a general rule, HP analog test equipment is the best designed of
its era, excepting only its oscilloscopes which are generally
considered not quite as good as Tektronix. I have not used the 333,
but the similar 334 is still considered a first rate piece of
equipment . I have used them but never worked on them.

If you are not happy with it I think you should pass it on to someone
else.

J L Hood was a competent engineer and I was just reading his "Valve
and Transistor Audio Amplifiers" book yesterday. That said I do not
believe his designs would have satisfied the requirements that HP was
given at the time they were given them. Keep in mind when this box was
designed and the fact that it was years later that IC op amps were of
instrumentation grade.

Some HP boxes were better than others to be sure. HP's most
successful products were its 200xx audio generators, its tube and
solid state RF generators culminating in the 8640B,its frequenc
counters and its digital products which are now obsolete for the most
part such as logic analyzers and such. The VTVMs are also quite good.

[Big Bad Bob]

Patrick Turner wrote:
But even in voltmeter mode, the use of my 333A was became pointless
because after it had been turned on for 1/2 an hour the voltage
measurement strated to drift about 15% high

how are the power supply capacitors and/or voltage regulators doing? My
first impression is a drifting power supply voltage (probably drifting due
to component thermal breakdown). I used to have a projector TV that had
thermal problems in the power supply (I eventually replaced it with a better
TV) and I left the back off and pointed a fan directly at the affected
components, which worked for quite a while until things got unmanageable
(and I bought a new TV).

Now the meter amp is a damn strange concoction of discrete transistors
with both positive and negative loops of FB and which have different
effects at different F, so that it can operate from 5Hz to several
MHz, with a very flat response **providing** all the vari-caps and
gain pots and the biasing pot are all set just right. I found the bias
pot was somewhat out, but when adjusted right according to the
schematic Vdc voltages the bad meter drift still occurred.

if your bias power supply voltage drifts 'in relation to' other voltages
that matter, it's pretty much the same thing as changing its setting. I've
done quite a bit with direct coupled transistor circuits and I know how
dependent on regulated voltages they can be.

If you've got some 'liquid cold' (or one of those "canned air" spray cans
held upside down) you can freeze certain compoments to see if you can alter
the behavior.

[Zilog]

On 2/12/2010 11:27, Big Bad Bob wrote:
Patrick Turner wrote:
But even in voltmeter mode, the use of my 333A was became pointless
because after it had been turned on for 1/2 an hour the voltage
measurement strated to drift about 15% high

how are the power supply capacitors and/or voltage regulators doing? My
first impression is a drifting power supply voltage (probably drifting
due to component thermal breakdown).

This could be, but Patrick is competent enough to have checked that...

I used to have a projector TV that
had thermal problems in the power supply (I eventually replaced it with
a better TV) and I left the back off and pointed a fan directly at the
affected components, which worked for quite a while until things got
unmanageable (and I bought a new TV).

If you found out it was the psu, why not fix it instead of buying a new
one. Unless it wasn't worth it.

Now the meter amp is a damn strange concoction of discrete transistors
with both positive and negative loops of FB and which have different
effects at different F, so that it can operate from 5Hz to several
MHz, with a very flat response **providing** all the vari-caps and
gain pots and the biasing pot are all set just right. I found the bias
pot was somewhat out, but when adjusted right according to the
schematic Vdc voltages the bad meter drift still occurred.

if your bias power supply voltage drifts 'in relation to' other voltages
that matter, it's pretty much the same thing as changing its setting.
I've done quite a bit with direct coupled transistor circuits and I know
how dependent on regulated voltages they can be.

That's true

If you've got some 'liquid cold' (or one of those "canned air" spray
cans held upside down) you can freeze certain compoments to see if you
can alter the behavior.

Spraying "Kalte Spray" can help but in these circuits a drop in
temperature will surely make some components and settings drift.
So what's the cause then: Thermally instable components or the sudden
drop in environment temperature?

I do have some older testgear and some are out of spec or broken.
I wonder if it's worth the effort to fix them.
Some components are specially made for or by the manufacturer and are no
longer available.

Maybe it's better to Ebay them and with the money from it buy a more
recent(and hopefully) better one...

Cheers,

Zilog

[Patrick Turner]

On Dec 2, 6:27*pm, Bret L wrote:
*As a general rule, HP analog test equipment is the best designed of
its era, excepting only its oscilloscopes which are generally
considered not quite as good as Tektronix. *I have not used the 333,
but the similar 334 is still considered a first rate piece of
equipment . I have used them but never worked on them.

The 334 is the same as 333 but has an RF detector so that the AF
signal used to modulate an AM RF or IF wave can be checked for THD.

When brand new, presumably they all worked well, but now they are old,
some 45 years, and I'm sure that if you were alive in 1965, you are
worse now and probably need a couple of things fixed......

*If you are not happy with it I think you should pass it on to someone
else.

Nobody I know wants this stuff now.

*J L Hood was a competent engineer and I was just reading his "Valve
and Transistor Audio Amplifiers" book yesterday. That said I do not
believe his designs would have satisfied the requirements that HP was
given at the time they were given them. Keep in mind when this box was
designed and the fact that it was years later that IC op amps were of
instrumentation grade.

Well, seems to me the Linsley hood design was for the hobbyist with
restricted F range, and standard value resistors used which alone
limits accuracy of voltage measure. But you can get good enough
measure with standard R values for hobbyists.

If you have to read an analog meter then you get visual errors
probably greater than circuit inaccuracies.

HP used a pile of non standard precision R to get accuracy, and
although the properties of the j-fets and bjts they used would have
varied considerably, all their active circuitry has large amounts of
negative feedback to keep everything working linearly so that their
circuit THD stayed below the lowest anticipated distortion in signals.

Quite a few companies made gear to basic recipes found in magazines.
The idea of the wien bridge used as a tunable nulling element for
fundemental F has been around for a long time and may be used with a
few tubes instead of bjts.

* Some HP boxes were better than others to be sure. HP's most
successful products were its 200xx audio generators, its tube and
solid state RF generators culminating in the 8640B,its frequenc
counters and its digital products which are now obsolete for the most
part such as logic analyzers and such. The VTVMs are also quite good.

I have a HP 606A - not bad for a tubed oscillator, and I have a 302A
analyser, all SS which I have yet to try out properly.

Much old stuff is very useful - providing it is not falling to bits
and suffering aging problems which of course are nowhere to be seen
described in the FAQ sections of manuals. I have no time to become an
HP expert. The service manuals for the 333A and my similar simpler
331A don't have working Vac signal voltages listed, and I found it
impossible to re-draw up the circuit as it is used when switched to
measure say a 10Vrms input voltage and show all the working voltages
along the circuit. Such detail in service manuals make it easy to
service the darn things because you do what the manual says and then
measure the voltages at the output and if any part of the circuit
works wrongly, you will spot it.

Patrick Turner.

[Patrick Turner]

On Dec 2, 9:27*pm, Big Bad Bob BigBadBob-at-mrp3-
wrote:
Patrick Turner wrote:
But even in voltmeter mode, the use of my 333A was became pointless
because after it had been turned on for 1/2 an hour the voltage
measurement strated to drift about 15% high

how are the power supply capacitors and/or voltage regulators doing? *My
first impression is a drifting power supply voltage (probably drifting due
to component thermal breakdown). *I used to have a projector TV that had
thermal problems in the power supply (I eventually replaced it with a better
TV) and I left the back off and pointed a fan directly at the affected
components, which worked for quite a while until things got unmanageable
(and I bought a new TV).

After playing with this 333A for a day and switching on/off many
times, the +/- 25V rails never varied, and had low hum.



Now the meter amp is a damn strange concoction of discrete transistors
with both positive and negative loops of FB and which have different
effects at different F, so that it can operate from 5Hz to several
MHz, with a very flat response **providing** all the vari-caps and
gain pots and the biasing pot are all set just right. I found the bias
pot was somewhat out, but when adjusted right according to the
schematic Vdc voltages the bad meter drift still occurred.

if your bias power supply voltage drifts 'in relation to' other voltages
that matter, it's pretty much the same thing as changing its setting. *I've
done quite a bit with direct coupled transistor circuits and I know how
dependent on regulated voltages they can be.

The 333A uses dc feedback loops to stabilise bias as well as copius
positive and negative FB loops to control circuit gains and compensate
for droop in performance at HF. HP knew what they were doing
alright,and when new, the gear all worked well. But like old humans,
such gear becomes almost impossible to fix completely. I have too many
other more important priorities, and I can't afford the time, and its
uneconomic to repair.


If you've got some 'liquid cold' (or one of those "canned air" spray cans
held upside down) you can freeze certain compoments to see if you can alter
the behavior.

Indeed, but meters drifting up 200% high and then back down again and
with intermittencies somewhere which always evades analysis when you
hunt for the cause may not be affected by temporary cooling because
the NFB loops ensure drift in component behaviour is very much
corrected. The bias settings could be altered to allow +/- 33% change
in idle currents of transistors and while things did seem to work a
bit without drift, the bias change made virtually no difference to
working signal voltages which were all very much lower than the
maximum possible without clipping. For example, the meter amp had 4
bjts and uses +/-25V rails. It has only to make 0.1vrms to make the
meter give full swing. The closed loop gain of the meter amp is
nominally 100, but open loop gain is many times more. Only 1mV of
signal at the input of the meter amp is needed to swing the meter
fully. The NFB resistors seemed stable to me, and the gain adjust pot
in the FB loop.

Even if I spent years fixing the 331A and 333A meters, they don't
actually do what I want a distortion meter to do, ie, give a reliable
easy to see measurement of THD down to 0.005% like the simple pile of
**** I designed and built so many years ago in about 1995 because I
could never find any cheap used but working gear then. Most techs was
still using analog stuff and almost none had made the switch to much
more modern methods and gear. When the PC became mainstream and new
gear became cheap and many old techs expired, died or retired, then
there was a lot of their old gear being auctioned or sold. But I never
found anything worth aquiring. One old radio ham gave me a pile of
stuff including a couple of CROs made well before WW2 and these were
truly bloody awful and I chucked them out - real time guzzling
monsters which defined what a boat anchor was.

Patrick Turner.

[Patrick Turner]

On Dec 2, 10:34*pm, Zilog wrote:
On 2/12/2010 11:27, Big Bad Bob wrote:

Patrick Turner wrote:
But even in voltmeter mode, the use of my 333A was became pointless
because after it had been turned on for 1/2 an hour the voltage
measurement strated to drift about 15% high

how are the power supply capacitors and/or voltage regulators doing? My
first impression is a drifting power supply voltage (probably drifting
due to component thermal breakdown).

This could be, but Patrick is competent enough to have checked that...

I used to have a projector TV that

had thermal problems in the power supply (I eventually replaced it with
a better TV) and I left the back off and pointed a fan directly at the
affected components, which worked for quite a while until things got
unmanageable (and I bought a new TV).

If you found out it was the psu, why not fix it instead of buying a new
one. Unless it wasn't worth it.



Now the meter amp is a damn strange concoction of discrete transistors
with both positive and negative loops of FB and which have different
effects at different F, so that it can operate from 5Hz to several
MHz, with a very flat response **providing** all the vari-caps and
gain pots and the biasing pot are all set just right. I found the bias
pot was somewhat out, but when adjusted right according to the
schematic Vdc voltages the bad meter drift still occurred.

if your bias power supply voltage drifts 'in relation to' other voltages
that matter, it's pretty much the same thing as changing its setting.
I've done quite a bit with direct coupled transistor circuits and I know
how dependent on regulated voltages they can be.

That's true

Yes but HP knew all about drift in transistor and tube circuits. Their
experts knew just how difficult it always was to make a partially DC
coupled and partially CR coupled circuit full of discretes work with
good stability. 1965 wasn't long after analog computer circuits had
been used, with caluclations and so on reliant on the NFB loop
accuracy.



If you've got some 'liquid cold' (or one of those "canned air" spray
cans held upside down) you can freeze certain compoments to see if you
can alter the behavior.

Spraying "Kalte Spray" can help but in these circuits a drop in
temperature will surely make some components and settings drift.
So what's the cause then: Thermally instable components or the sudden
drop in environment temperature?

I do have some older testgear and some are out of spec or broken.
I wonder if it's worth the effort to fix them.
Some components are specially made for or by the manufacturer and are no
longer available.

Maybe it's better to Ebay them and with the money from it buy a more
recent(and hopefully) better one...

I might rationalise the metering circuit with a couple of opamps to at
least get the meter circuit to work far better than it ever was
designed. This is very easy these days, and I have already done it in
another metering circuit I made which powers a 4" wide meter where the
needle is centred with no signals. The meter amp is logarithmic and I
measure +/- 20dB each side of centre. This makes recording and
plotting the acoustic response af a speaker easy.

The opamps can be made to function as very linear modulation detectors
driving diodes which charge up caps and the DC is fed back to the
opamp input so that the opamp has not got to handle linear wide
bandwidth voltage amplification and then make a dc signal at the
output. If the opamps function as linear detectors, voila, simple and
stable. The Vdc outputs of the opamps follow the peak levels of the
wave forms. Both positive and negative going Vdc voltages from two
opamp outputs which are firmly biased for 0V can be made to provide 2
Vdc signals with the meter slung between the + and -.

Many better ways of doing things have been dreamed up since 1965, so
rather than spend a month of sundays repairing to the old ingenious
standards of 1965 which have serious limitations, its better to remove
some 1965 ingenuity and replace with later invented simple
brilliance.
Some things, like the HP meters do indeed seem like very nice kit and
they are a pleasure to read, although a light inside would be nice.

The 1972 Linsley Hood distortion measuring circuit could be used with
the nicely made HP wien bridge. This would allow a 0.01% THD
measurement, using about 1/4 of the parts HP thought appropriate.

The box quality and mechanical parts HP had made were a lot better
than anything Linsley Hood might have had access to.

In 1965, the US was a real industrial powerhouse with a huge number of
highly skilled tradesmen and women beavering away to make fabulous
gear that cost a kings ransom. Britain was still hurting after WW2.

But Britain didn't lack ideas.

Patrick Turner.


Cheers,

Zilog

[Big Bad Bob]

On 12/02/10 04:54, Patrick Turner so witilly quipped:
Indeed, but meters drifting up 200% high and then back down again and
with intermittencies somewhere which always evades analysis when you
hunt for the cause may not be affected by temporary cooling because
the NFB loops ensure drift in component behaviour is very much
corrected. The bias settings could be altered to allow +/- 33% change
in idle currents of transistors and while things did seem to work a
bit without drift, the bias change made virtually no difference to
working signal voltages which were all very much lower than the
maximum possible without clipping.

direct coupled bipolar circuits are inherently vulnerable to weird
instabilities. If it's a 'discrete built' op-amp circuit, you might end
up with significant changes in forward transconductance over time, which
is likely to cause LOTS of bizarre problems, all of which would be
correctable by restoring the circuit to its design specs. Thermal
breakdown and aging of transistors might be just enough to create these
kinds of problems. Sometimes a little 'liquid cold' might at least
verify which component(s) are likely culprits, but from your description
you might have to replace all of the transistors to get it to work again.

One old radio ham gave me a pile of stuff including a couple
of CROs made well before WW2 and these were truly bloody awful
and I chucked them out - real time guzzling monsters which
defined what a boat anchor was.

in the navy we had a term "float test". It's what you do to said
equipment to see if it's "still good". If it floats, it's ok (but
you'll have to throw it out because of the seawater contamination). If
it sinks, it needed replacing anyway. I only heard of one instance
where 'float testing' actually happened, and it was totally by accident.
And the divers had to go find it (and it ended up being replaced).
Ah, well. That 'reality' thing getting in the way.

For THD measurements I would be inclined to use a 24-bit sound card to
generate a perfect sine wave at the desired frequency, then capture the
audio on the output with the same sound card and do a discrete fourier
transform on the waveform and calculate the THD as a geometric sum of
the harmonic amplitudes divided by the fundamental (I think that's the
right calculation). It's been a while since I've done wave analysis.
DFT is better than FFT since it lets you have sample counts of "not a
power of 2" for a single cycle, so ANY frequency can be used. And we all
have fast computers nowadays so a DFT for 50 harmonics would take less
time than the audio capture.

[Patrick Turner]

On Dec 5, 8:32*pm, Big Bad Bob BigBadBob-at-mrp3-
wrote:
On 12/02/10 04:54, Patrick Turner so witilly quipped:

Indeed, but meters drifting up 200% high and then back down again and
with intermittencies somewhere which always evades analysis when you
hunt for the cause may not be affected by temporary cooling because
the NFB loops ensure drift in component behaviour is very much
corrected. The bias settings could be altered to allow +/- 33% change
in idle currents of transistors and while things did seem to work a
bit without drift, the bias change made virtually no difference to
working signal voltages which were all very much lower than the
maximum possible without clipping.

direct coupled bipolar circuits are inherently vulnerable to weird
instabilities. *If it's a 'discrete built' op-amp circuit, you might end
up with significant changes in forward transconductance over time, which
is likely to cause LOTS of bizarre problems, all of which would be
correctable by restoring the circuit to its design specs. *Thermal
breakdown and aging of transistors might be just enough to create these
kinds of problems. *Sometimes a little 'liquid cold' might at least
verify which component(s) are likely culprits, but from your description
you might have to replace all of the transistors to get it to work again.

During a few days of poor health, I managed to cobble together the
copies of the HP333A and HP331A schematics and to study how they work
in more detail. I could find no explanations why voltages I measured
led me be quite bamboozled when I should not have been, and when both
333A and 331A should have given identical consistent outcomes.

From the manuals I now have a list of voltages and the procedure in
mind required to check the circuit working of this gear.

The 331A looks like being the most redeemable of the two units, and
most easily fixed. The add on functional features of the 333A are
utterly excessive to my needs, and so I will endeavour to make the
333A into a wien bridge oscillator plus a working voltmeter using my
own simpler designs. I did build a voltmeter in about 1993 using an
old but good meter plus $20 worth of cheap SS parts. Its BW is 1MHz,
and has ranges between 10mV and 1,000Vac, with Rin at about 1M.
I made a simple discrete SS voltage amp with NFB to give the BW, but I
didn't include the voltmeter and its diodes in the NFB path. It meant
the meter scale I drew looked a little non linear right above 0.0V for
each range so measuring below 1mV was dodgy. I also didn't have the
intermediate ranges of 0V to 3.2V, which allow better readings of
voltages less than 10% on a 0V to 10V scale. The HP meter is a very
nice peice of kit for a meter.

The wien bridge in the 333A should make a fine oscillator with 5
ranges and 5Hz to 600kHz. The tuning caps have 1,060pF for the 2 gangs
used for the parallel and series wien C elements. The largest value
resistance needed is 29Megohms with 1060pF to get down to the 5Hz, and
just low enough for most of my testing.
I already have a TLO72 opamp in several wien bridge oscilators I have
built. The last one I made gets up to 200kHz with C = 470pF and R
about 1k7. It makes 2.7Vrms output.
If I make another discrete transistor amp with maybe 4 x 2n2222 plus
high Rin j-fet the HF should reach the 600kHz capability of the HP
wien bridge now used as a THD nuller. The existing minimum R value =
2k9, and for 600kHz, C must be 91pF, and methinks the tuning gang used
plus its trimmer caps would have total minimum C 100pF.

One old radio ham gave me a pile of stuff including a couple
of CROs made well before WW2 and these were *truly bloody awful
and I chucked them out - real time guzzling monsters which
defined what a boat anchor was.

in the navy we had a term "float test". *It's what you do to said
equipment to see if it's "still good". *If it floats, it's ok (but
you'll have to throw it out because of the seawater contamination). *If
it sinks, it needed replacing anyway.

And some Germans and Japs helped the replacement rates.

Owning one's own personal yacht is bad enough; one finds it hard to
find good deckhands and ppl to serve the drinks without spilling them,
or drinking it all themselves, but owning a whole damn navy would be
nothing but a ghastly expense.

*I only heard of one instance
where 'float testing' actually happened, and it was totally by accident.
* And the divers had to go find it (and it ended up being replaced).
Ah, well. *That 'reality' thing getting in the way.

For THD measurements I would be inclined to use a 24-bit sound card to
generate a perfect sine wave at the desired frequency, then capture the
audio on the output with the same sound card and do a discrete fourier
transform on the waveform and calculate the THD as a geometric sum of
the harmonic amplitudes divided by the fundamental (I think that's the
right calculation). *It's been a while since I've done wave analysis.
DFT is better than FFT since it lets you have sample counts of "not a
power of 2" for a single cycle, so ANY frequency can be used. And we all
have fast computers nowadays so a DFT for 50 harmonics would take less
time than the audio capture.

I still enjoy the irrationality of not having a PC with fragile
software in my dirty filthy work shed. I agree about the modern
approach, but I am too addicted to fixin good old things which once
fixed, tend to stay fixed.

I have a 302A analyser and I expect it can sniff along bands for stray
F. I hope it works, but won't be surprised if old age has eaten into
it. So far only had to replace the input fuse and holder which got
smashed by previous owners.

Meanwhile the wien bridge oscillator I was using gave up because of
yet another failing pot. So in my junk boxes I found an old good
quality wafer rotary switch, 2 poles, 12 postions, previously used in
a valve tester but luckily not worn or fried, and with firm contacts
and the R can be switched for 12 different F per decade of F range
between 2Hz and 20Hz and so on up to 200kHz.

When I make the 333A into a variable wien bridge oscilator, it should
give 5Hz to 600kHz, and the HF is just high enough to cover AM radio
IF at 455kHz, and to test the reactive behaviour of OPTs at extreme
HF. 200kHz is not quite enough for me. The light dependant resistances
used as part of the HP wien bridge for tracking some signal with
slightly varying F may be useful for the oscillator NFB path which
needs to have varying R somewhere to keep the output level constant at
all F. Usually a light globe is used or a special thermistor mounted
inside a vacuum tube. I've never seen any really suitable thermistors
like the good old ones that were made years ago. If the idea of using
a light dependant resistor and lamp to control the oscillator does not
work I can always use a solid state device as a variable R controlled
by a rectified output Vdc. I have this in my oscillator with TL072.
Such a thing helps to lower THD. While the pot for varying F was good
the signal bounce was low, but it became worse as the pot slowly
became intermttent, and open on one small section of a track. Cleaning
and wrapping cardboard to keep out dust didn't help. Plus the contour
of the log pots I have is a bit odd, maybe real logarithmic contouring
in Taiwanese log pots leaves much to be desired. These cheap pots gave
part of the F band all crowded together. I tried another Taiwan
replacement but this had such a difference between track R at about
midway along the band that the oscillations would not start. A
switched set of F determined by carefully calculated R values will be
OK for most of my testing work.
I have found WB oscillators to generally be very frequency stable and
once warmed up the F variation is negligible while testing some amp.
One Topward oscillator I had which used a pot to vary a DC level to a
digital chip had quite awful distortion, 0.3%, and such high
frequency drift it was hopeless to try to use when testing THD and
needing to null out the fundemental.

Home made oscillator calibration is easy though; I have a modern
digital F meter which I built from a cheap kit, goes up to 50MHz. Just
rotate the pot or the switch until the F numbers come up and write the
marks in ink on the cardboard scale glued to the front panel.
When done, a coat of varnish seals it. Lasts for years and will see me
out, and can be re-done any time.
A 24 pos switch would be luxury........

Patrick Turner.

[JosefMiller]

It would have been shortly after the 333A was developed by HP that
opamps were invented and VOILA, all the efforts HP went to became
immediately obsolete. I began to think the HP meter circuit was like a
horse and buggy compared to a good motot car, and the circuit working
seems to rely on all resistors never changing in value, and on many
other things like cap leakage, and perhaps bjts and other bits
becoming dodgy and almost unidentifiable in the years since 1967.

[Big Bad Bob]

Patrick Turner wrote:
Usually a light globe is used or a special thermistor mounted
inside a vacuum tube. I've never seen any really suitable thermistors
like the good old ones that were made years ago. If the idea of using
a light dependant resistor and lamp to control the oscillator does not
work I can always use a solid state device as a variable R controlled
by a rectified output Vdc. I have this in my oscillator with TL072.

You could create a wein bridge VCO using an LED current-driven by a signal,
lighting up two identical photo-resistors with a few wraps of black
electrical tape around the assembly to keep the light out. That should do
for your wein bridge if the circuit I have in mind is used, i.e. 2 caps, 2
resistors for an RC bandpass filter and oscillator startup via a non-linear
device, often a low voltage incandescent lamp. I've seen this done with 2
triodes and a notch filter also, with positive feedback using the lamp (an
old heathkit sine/square generator specifically). There are also MOSFET
opto-isolators (analog switches) that should effectively give you the same
effect. Some claim proportionally controllable output as I recall, based on
light intensity, which is what the photocell thing would also do.

Such a thing helps to lower THD. While the pot for varying F was good
the signal bounce was low, but it became worse as the pot slowly
became intermttent, and open on one small section of a track.

noisy pots. Bleah. I suggest good quality contact cleaner sprayed into any
openings followed by a rapid succession spinning the knob from one extreme
to the other, several dozen times. Works for guitars and guitar amps.

Regarding cheap Taiwanese parts, can Digikey ship to your area? They tend
to have decent quality stuff, a bit pricey but reliable.


You might be able to improve THD of a wein bridge by tweeking the response
curve with active filter elements.

Anyone consider swapping out the incandescent bulb for a correctly biased
tunnel diode? Just a thought.

[Nordic Breeds WA4VZQ]

There is a very good reason for the lamp in a Wein Bridge oscillator.
The _thermal_ time constant of the lamp must be long compared to the
period of the lowest frequency of oscillation. I do not have the time to
fully discuss this here, but Jim Williams of Linear Technology
Corporation has written quite a bit on low distortion oscillators. I
suggest both Patrick and Bob download Linear Technology's Application
Note 43 on bridge circuits,
http://www.linear.com/pc/downloadDoc..., P1213,D4134.
Jim presents a number of simple Wein bridge oscillator designs that can
produce THD levels from below 0.01% to such a low level (around 3 parts
per million) that most commercial harmonic distortion analyzers have a
noise floor greater than the distortion being measured. Appendices to
this application note include a discussion of Bill Hewlett's oscillator
built for his Master's thesis, and "Understanding Distortion
Measurements" by Bruce Hofer of Audio Precision, Inc. I think you will
find this LTC application note to be very educational yet presented in a
way that is easy to understand.

If this link does not work for you, you may have to register at the LTC
site to download it. I suggest looking through their extensive library
of application notes for many good ideas when you have time.

73, Dr. Barry L. Ornitz WA4VZQ

[Patrick Turner]

On Dec 31 2010, 10:41*pm, Big Bad Bob BigBadBob-at-mrp3-
wrote:
Patrick Turner wrote:
Usually a light globe is used or a special thermistor mounted
inside a vacuum tube. I've never seen any really suitable thermistors
like the good old ones that were made years ago. If the idea of using
a light dependant resistor and lamp to control the oscillator does not
work I can always use a solid state device as a variable R controlled
by a rectified output Vdc. I have this in my oscillator with TL072.

You could create a wein bridge VCO using an LED current-driven by a signal,
lighting up two identical photo-resistors with a few wraps of black
electrical tape around the assembly to keep the light out. *That should do
for your wein bridge if the circuit I have in mind is used, i.e. 2 caps, 2
resistors for an RC bandpass filter and oscillator startup via a non-linear
device, often a low voltage incandescent lamp. *I've seen this done with 2
triodes and a notch filter also, with positive feedback using the lamp (an
old heathkit sine/square generator specifically). *There are also MOSFET
opto-isolators (analog switches) that should effectively give you the same
effect. *Some claim proportionally controllable output as I recall, based on
light intensity, which is what the photocell thing would also do.

For a variable F sig gene and for the NFB regulating R I ended up
using the on resistance of a j-fet biased with dc derived off the op-
amp output with diode detector. For low THD, the adjustments for level
became touchy, and bounce was bad if range was changed or F was
changed. To avoid bounce and to get quick settling, THD may be allowed
to be up to 1% with limiting clamps on the op-amp output.

But in an oscillator I am working on now to get 6V at 0.001% or less
THD, I have a wien bridge using NE5534 plus fixed R and C but with a
double gang radio tuning C with series 250pF padders to get a C
variable range of about +/- 30Hz each side of 1kHz.

The use of a lamp is then fine. I have one which is 60 ohms cold and
200 ohms at 2V, so the NFB network at 6V output is about 600 ohms and
does not overload the opamp. With rails of +/-14.7Vdc, I am getting up
to 7Vrms with THD under 0.005%, with best THD of 0.0041% at 6vrms
output.

The F adjustment with tuning cap does not make signal bounce, and F
may be tuned for best null into distortion measuring notch filter.

To get oscillator THD down to less than 0.001% at 6Vrms, I am using a
couple of cascaded BPFs with R+parallel tuned LC, about 22mH + 1.1uF,
with Rw of the coil at 5 ohms for Q about 10. There is -6dB insertion
loss but two opamps are used each with NFB to give gain of 2 to
overcome insertion loss.


Such a thing helps to lower THD. While the pot for varying F was good
the signal bounce was low, but it became worse as the pot slowly
became intermttent, and open on one small section of a track.

noisy pots. *Bleah. *I suggest good quality contact cleaner sprayed into any
openings followed by a rapid succession spinning the knob from one extreme
to the other, several dozen times. *Works for guitars and guitar amps.

Yes, but not for long.

Regarding cheap Taiwanese parts, can Digikey ship to your area? *They tend
to have decent quality stuff, a bit pricey but reliable.

The operation and stability of variable tuning caps is superlative.


You might be able to improve THD of a wein bridge by tweeking the response
curve with active filter elements.

Maybe variable state types with multiple FB paths and 3 opamps will do
if the THD produced is low.
But I am using opamps with simple R+R series voltage NFB paths and
then passive R+LC filters.
I can get 40dB attenuation of any 2H in the oscilator signal, and more
atten for all other H..

Anyone consider swapping out the incandescent bulb for a correctly biased
tunnel diode? *Just a thought.

I have never seen this.

Patrick Turner.

[Patrick Turner]

On Jan 1, 3:07*pm, "Nordic Breeds WA4VZQ"
wrote:
There is a very good reason for the lamp in a Wein Bridge oscillator.
The _thermal_ time constant of the lamp must be long compared to the
period of the lowest frequency of oscillation. *I do not have the time to
fully discuss this here, but Jim Williams of Linear Technology
Corporation has written quite a bit on low distortion oscillators. *I
suggest both Patrick and Bob download Linear Technology's Application
Note 43 on bridge circuits,http://www.linear.com/pc/downloadDoc...C1,C1154,C1009,....
Jim presents a number of simple Wein bridge oscillator designs that can
produce THD levels from below 0.01% to such a low level (around 3 parts
per million) that most commercial harmonic distortion analyzers have a
noise floor greater than the distortion being measured. *Appendices to
this application note include a discussion of Bill Hewlett's oscillator
built for his Master's thesis, and "Understanding Distortion
Measurements" by Bruce Hofer of Audio Precision, Inc. *I think you will
find this LTC application note to be very educational yet presented in a
way that is easy to understand.

If this link does not work for you, you may have to register at the LTC
site to download it. *I suggest looking through their extensive library
of application notes for many good ideas when you have time.

* * * * 73, *Dr. Barry L. Ornitz * WA4VZQ

Thanks Mr Doctor Barry Ornitz.

I would say the pdf there is the *most informative thing on wien
bridge oscilators" I have ever seen.

BUT, it has ocurred t me that a bridge formed with R + LC on one side
and R + lamp on the other side should also work well and with low THD
because the Q of the LC is far higher than the Q for the RC wien
bridge arrangement. So therefore any harmonic of the oscillator
fundemental fed back via the PFB path is much more attenuated than
using the wien RC array. The NFB passes all the harmonics back
according to the R ratio. So with LC the ß fraction of distortion
signal applied to the op-amp diff inputs is effectively much greater
with LC.

Changing the F with LC is more difficult which is why nobody uses LC
for audio F oscillators unless one likes to have switched taps on a
range of large air cored coils and switched capacitors which is
actually not a bad idea. Q will vary, but its always more than wien
RC. But for between 100kHz and 2MHz the normal radio tuning cap and
lamp regulation in a bridge scheme might work very well, but I ain't
seen any sample of such a thing.

Patrick Turner.