[Ian Bell]

I am designing a mic pre based on three 6AU6 tubes. The first two are
operated as triodes with no NFB with an overall gain setting pot divider
between them and the last is a cathode follower. I have discovered that by
judicious selection of the component values in the CF that the
predominantly 2H distortion of the first two stages can be reduced by a
factor of 10. This effect is not frequency or output level sensitive but
does depend on the setting of the inter stage pot - it works best at
highest overall gain.

I was very surprised to see this effect and wondered if anyone else had come
across it. I seem to remember an article somewhere about a two stage amp
where the second stage adds the same sort of distortion to the inverse of
the waveform and hence reduces overall distortion but I cannot at present
lay my hands on it.

OK, admittedly these are just simulations at present and yes, Patrick, I
have pretty much all the components on hand and I WILL be building it as
soon as possible. It is so intriguing I cannot resist.

Cheers

Ian

[Eeyore]

Ian Bell wrote:

I am designing a mic pre based on three 6AU6 tubes. The first two are
operated as triodes with no NFB with an overall gain setting pot divider
between them and the last is a cathode follower. I have discovered that by
judicious selection of the component values in the CF that the
predominantly 2H distortion of the first two stages can be reduced by a
factor of 10. This effect is not frequency or output level sensitive but
does depend on the setting of the inter stage pot - it works best at
highest overall gain.

I was very surprised to see this effect and wondered if anyone else had come
across it. I seem to remember an article somewhere about a two stage amp
where the second stage adds the same sort of distortion to the inverse of
the waveform and hence reduces overall distortion but I cannot at present
lay my hands on it.

OK, admittedly these are just simulations at present and yes, Patrick, I
have pretty much all the components on hand and I WILL be building it as
soon as possible. It is so intriguing I cannot resist.

If you simply want low distortion I suggest you use semiconductor circuitry.

The main reason valve / tube circuitry is used by those who like it is for the
*added distortion* that gives the sound more 'colour'.


Graham

[Nick Gorham]

Ian Bell wrote:
I am designing a mic pre based on three 6AU6 tubes. The first two are
operated as triodes with no NFB with an overall gain setting pot divider
between them and the last is a cathode follower. I have discovered that by
judicious selection of the component values in the CF that the
predominantly 2H distortion of the first two stages can be reduced by a
factor of 10. This effect is not frequency or output level sensitive but
does depend on the setting of the inter stage pot - it works best at
highest overall gain.

I was very surprised to see this effect and wondered if anyone else had come
across it. I seem to remember an article somewhere about a two stage amp
where the second stage adds the same sort of distortion to the inverse of
the waveform and hence reduces overall distortion but I cannot at present
lay my hands on it.

OK, admittedly these are just simulations at present and yes, Patrick, I
have pretty much all the components on hand and I WILL be building it as
soon as possible. It is so intriguing I cannot resist.

Cheers

Ian

Yes, that sort of thing happens in may amps, expecially SET's with two
single ended stages. However it may not work that well in reality,
especially if you have reactive loads.

--
Nick

[Ian Iveson]

Ian Bell wrote

I am designing a mic pre based on three 6AU6 tubes. The first two are
operated as triodes with no NFB with an overall gain setting pot
divider
between them and the last is a cathode follower. I have discovered
that by
judicious selection of the component values in the CF that the
predominantly 2H distortion of the first two stages can be reduced
by a
factor of 10. This effect is not frequency or output level sensitive
but
does depend on the setting of the inter stage pot - it works best at
highest overall gain.

I was very surprised to see this effect and wondered if anyone else
had come
across it. I seem to remember an article somewhere about a two stage
amp
where the second stage adds the same sort of distortion to the
inverse of
the waveform and hence reduces overall distortion but I cannot at
present
lay my hands on it.

OK, admittedly these are just simulations at present and yes,
Patrick, I
have pretty much all the components on hand and I WILL be building
it as
soon as possible. It is so intriguing I cannot resist.

Here's some blether:

http://www.enjoythemusic.com/magazin...audiopax88.htm

look for "Timbre Lock", a couple of para's ahead of heading "Promise
Delivered?".

What you have, with judicious choice of valves, is the opportunity to
tailor the proportions of even and odd harmonics. You can reduce the
even but not the odd, by the same principle as in a PP stage...except
in the series case you may have dissimilar valves with different input
levels and different loads...lots to play with.

If you are using SPICE, especially with many of the rough-and-ready
models kicking around, then beware that the real spread of distortion
will not be very precise, and will miss out some foibles of valve
operation.

If you *are* using SPICE, then all should surely be clear? You should
be able to see how two similarly asymmetrical but opposite signals sum
to a more symmetrical one. Symmetrical signals have no even harmonics.

Anyway, you may enjoy minimising THD, but an alternative goal would be
to optimise the proportions of harmonics to some "golden ratio". Or by
ear, which may be another way of doing the first, or the second,
depending on your ears, or perhaps not.

BTW, using similar valves may not be the best way to go, because they
won't be in similar situations. And it's cheating. All the fun is in
finding compatible partners.

cheers, Ian

[Ian Iveson]

er...series cancellation isn't so simple because multiplication is
involved. All the same, you should be able to see what's happening if
you make the right pictures.

Ian

[Andre Jute]

Ian Bell wrote:
I am designing a mic pre based on three 6AU6 tubes. The first two are
operated as triodes with no NFB with an overall gain setting pot divider
between them and the last is a cathode follower. I have discovered that by
judicious selection of the component values in the CF that the
predominantly 2H distortion of the first two stages can be reduced by a
factor of 10. This effect is not frequency or output level sensitive but
does depend on the setting of the inter stage pot - it works best at
highest overall gain.

I was very surprised to see this effect and wondered if anyone else had come
across it. I seem to remember an article somewhere about a two stage amp
where the second stage adds the same sort of distortion to the inverse of
the waveform and hence reduces overall distortion but I cannot at present
lay my hands on it.

OK, admittedly these are just simulations at present and yes, Patrick, I
have pretty much all the components on hand and I WILL be building it as
soon as possible. It is so intriguing I cannot resist.

Cheers

Ian

Actually, no, what you're seeing is most unlikely to be harmonic
cancellation, especially if the observed phenomenon is not frequency
dependent.

What you're seeing is instead a flattening of the transfer curve
brought about by higher current, higher plate load, higher bias,
higher plate voltage, or any or all in various combinations dictated
by your choice of components. The flattening of the transfer function,
and raising it by higher current, lifts the tube out of the noisy
regions into linearity; higher current also has a beneficial effect on
Miller (there's an article on my netsite which shows how you can use
standard slew rate math to discover the minimum amount of current you
should put on the driver to avoid Miller screwing your design), all of
which adds to the linearity. It is furthermore a very euphonious
shift you have brought about, as you will hear when you build the
device, in that the disturbing third and higher harmonics should be
almost eliminated. Roundabout 1998 an idiot called Chernofsky and his
puppetmaster, Creepy LaFevre of Magnequest notoriety, challenged me to
a design contest; they ran away after only the power stages were
published because their noise generator was simply steamrollered by my
ultra-quiet design. John Byrns calculated the distortion proportions
of both amps; search Google for the "Bubbaland 300" and you'll find,
amid the flamewar, a discussion of my HCHV high-everything concept of
silence in tube amps. Everyone who has the sense to design his amps in
a simulator (I wrote mine in Excel), rather than build them badly on
the bench and spend months trying to gimmick them right by trial and
error, and who thinks about what he sees, eventually arrives at HCHV.

About a decade ago I made myself very unpopular with the
ultrafidelista when I laughed out loud at an article by Reid Welch
about harmonic cancellation, and when abused for my lack of solidarity
with another Joenetter explained bluntly that the answer lies in that
study of the transfer curves which too many of the ultrafidelista are
too artistic to commit. (I also pointed out bluntly that I am a real
artist--one who earns a substantial living by creating works of art--
and that I know from experience that art--including amps that sound
like mine--is created by perspiration, not inspiration.) I'm sorry to
disappoint you as well, because frequency free-harmonic cancellation
would be a three-star free lunch for tubies everywhere, but I don't
think we'll see it in my lifetime. As for claims, heard even on RAT,
that so and so injected noise which cancelled greater noise elsewhere,
I just giggle; there are ways of distributing noise more benificially,
and ways of reducing noise, and above all ways of shaping it
euphoniously by reducing higher and odd harmonics more than you can
reduce the less disturbing second harmonic, and of simply dividing the
noise on the principle of divide and conquer (on a split rail, for
instance, the whole amp will be quieter if you partially bypass the
bottom leg so that it is split below the cathode the same distance as
the bias voltage in the top leg (1)), but none of these are
"cancellations" as your headline has it, they are, alas, merely
reductions.

Andre Jute
Visit Jute on Amps at http://members.lycos.co.uk/fiultra/
"wonderfully well written and reasoned information
for the tube audio constructor"
John Broskie TubeCAD & GlassWare
"an unbelievably comprehensive web site
containing vital gems of wisdom"
Stuart Perry Hi-Fi News & Record Review

(1) Now watch the flame war that erupts when someone claims that the
point of splitting the bottom leg of a split rail supply is to
"inject" the power supply noise at the most beneficial point where it
"cancels" the "load noise" (aka "current noise" as in another
concurrent thread) in the plate circuit... It ain't injection that
matters, of course, but balance of the top and bottom halves, and all
the more so when the justification of your expensive split rail supply
is a fully balanced amp that is very quiet already. I make this point
because what might seem like a subtle distinction of semantics is
actually the fast track to a most productive way of thinking about
tube amps, and one that every tubie must get right if he is ever to be
anything more than a repair hack.

[Andre Jute]

Andre Jute wrote:
Ian Bell wrote:
I am designing a mic pre based on three 6AU6 tubes. The first two are
operated as triodes with no NFB with an overall gain setting pot divider
between them and the last is a cathode follower. I have discovered that by
judicious selection of the component values in the CF that the
predominantly 2H distortion of the first two stages can be reduced by a
factor of 10. This effect is not frequency or output level sensitive but
does depend on the setting of the inter stage pot - it works best at
highest overall gain.

I was very surprised to see this effect and wondered if anyone else had come
across it. I seem to remember an article somewhere about a two stage amp
where the second stage adds the same sort of distortion to the inverse of
the waveform and hence reduces overall distortion but I cannot at present
lay my hands on it.

OK, admittedly these are just simulations at present and yes, Patrick, I
have pretty much all the components on hand and I WILL be building it as
soon as possible. It is so intriguing I cannot resist.

Cheers

Ian

Actually, no, what you're seeing is most unlikely to be harmonic
cancellation, especially if the observed phenomenon is not frequency
dependent.

What you're seeing is instead a flattening of the transfer curve
brought about by higher current, higher plate load, higher bias,
higher plate voltage, or any or all in various combinations dictated
by your choice of components. The flattening of the transfer function,
and raising it by higher current, lifts the tube out of the noisy
regions into linearity; higher current also has a beneficial effect on
Miller (there's an article on my netsite which shows how you can use
standard slew rate math to discover the minimum amount of current you
should put on the driver to avoid Miller screwing your design), all of
which adds to the linearity. It is furthermore a very euphonious
shift you have brought about, as you will hear when you build the
device, in that the disturbing third and higher harmonics should be
almost eliminated. Roundabout 1998 an idiot called Chernofsky and his
puppetmaster, Creepy LaFevre of Magnequest notoriety, challenged me to
a design contest; they ran away after only the power stages were
published because their noise generator was simply steamrollered by my
ultra-quiet design. John Byrns calculated the distortion proportions
of both amps; search Google for the "Bubbaland 300" and you'll find,
amid the flamewar, a discussion of my HCHV high-everything concept of
silence in tube amps. Everyone who has the sense to design his amps in
a simulator (I wrote mine in Excel), rather than build them badly on
the bench and spend months trying to gimmick them right by trial and
error, and who thinks about what he sees, eventually arrives at HCHV.

About a decade ago I made myself very unpopular with the
ultrafidelista when I laughed out loud at an article by Reid Welch
about harmonic cancellation, and when abused for my lack of solidarity
with another Joenetter explained bluntly that the answer lies in that
study of the transfer curves which too many of the ultrafidelista are
too artistic to commit. (I also pointed out bluntly that I am a real
artist--one who earns a substantial living by creating works of art--
and that I know from experience that art--including amps that sound
like mine--is created by perspiration, not inspiration.) I'm sorry to
disappoint you as well, because frequency free-harmonic cancellation
would be a three-star free lunch for tubies everywhere, but I don't
think we'll see it in my lifetime. As for claims, heard even on RAT,
that so and so injected noise which cancelled greater noise elsewhere,
I just giggle; there are ways of distributing noise more benificially,
and ways of reducing noise, and above all ways of shaping it
euphoniously by reducing higher and odd harmonics more than you can
reduce the less disturbing second harmonic, and of simply dividing the
noise on the principle of divide and conquer (on a split rail, for
instance, the whole amp will be quieter if you partially bypass the
bottom leg so that it is split below the cathode the same distance as
the bias voltage in the top leg (1)),

I mean of course "the plate voltage in the top leg", that is, that is,
what is left of the B+ after the load resistor drop.

but none of these are
"cancellations" as your headline has it, they are, alas, merely
reductions.

Andre Jute
Visit Jute on Amps at http://members.lycos.co.uk/fiultra/
"wonderfully well written and reasoned information
for the tube audio constructor"
John Broskie TubeCAD & GlassWare
"an unbelievably comprehensive web site
containing vital gems of wisdom"
Stuart Perry Hi-Fi News & Record Review

(1) Now watch the flame war that erupts when someone claims that the
point of splitting the bottom leg of a split rail supply is to
"inject" the power supply noise at the most beneficial point where it
"cancels" the "load noise" (aka "current noise" as in another
concurrent thread) in the plate circuit... It ain't injection that
matters, of course, but balance of the top and bottom halves, and all
the more so when the justification of your expensive split rail supply
is a fully balanced amp that is very quiet already. I make this point
because what might seem like a subtle distinction of semantics is
actually the fast track to a most productive way of thinking about
tube amps, and one that every tubie must get right if he is ever to be
anything more than a repair hack.

[Phil Allison]

"Ian Bell"

I am designing a mic pre based on three 6AU6 tubes. The first two are
operated as triodes with no NFB with an overall gain setting pot divider
between them and the last is a cathode follower.


** Is there to be an input step-up transformer ??

Be nothing but a damn white noise generator otherwise.



....... Phil

[Ian Bell]

Phil Allison wrote:


"Ian Bell"

I am designing a mic pre based on three 6AU6 tubes. The first two are
operated as triodes with no NFB with an overall gain setting pot divider
between them and the last is a cathode follower.


** Is there to be an input step-up transformer ??


There is 10:1

Be nothing but a damn white noise generator otherwise.


Indeed.

Ian

[Phil Allison]

"Ian Bell"

** Is there to be an input step-up transformer ??


There is 10:1


** Then why the hell have *2* triode gain stages with well over 1000 times
gain ??

Makes your total gain OVER 10,000 TIMES !!

Madness.



..... Phil

[Patrick Turner]

Phil Allison wrote:

"Ian Bell"

** Is there to be an input step-up transformer ??


There is 10:1

** Then why the hell have *2* triode gain stages with well over 1000 times
gain ??

Makes your total gain OVER 10,000 TIMES !!

Madness.

.... Phil

If the mic signal = 0.5mV, then gain of 10,000 gives 5Vrms.
If the pot gain control is to be used at the -20dB central position,
then Vo = 0.5Vrms.

Madness?

Patrick Turner.

[Phil Allison]

"Patrick Turneroid ****wit"
"Ian Bell"

** Is there to be an input step-up transformer ??


There is 10:1

** Then why the hell have *2* triode gain stages with well over 1000
times
gain ??

Makes your total gain OVER 10,000 TIMES !!

Madness.


If the mic signal = 0.5mV, then gain of 10,000 gives 5Vrms.
If the pot gain control is to be used at the -20dB central position,
then Vo = 0.5Vrms.

Madness?


** For serious recording work, mic gain of 1000 times is not usable as it
results in excessive background noise - consisting audible hiss from the mic
and pre -amp PLUS amplified noise from the environment.

Having 1000 times gain at only half knob setting is UTTERLY INSANE !!

The OP has not built his fantasy tube pre-amp yet - when an if he does,
he will soon see it is not a game for mugs like him.



....... Phil

[Ian Bell]

Phil Allison wrote:


"Ian Bell"

** Is there to be an input step-up transformer ??


There is 10:1


** Then why the hell have *2* triode gain stages with well over 1000
times gain ??

Makes your total gain OVER 10,000 TIMES !!

Madness.


Stage gain is about 30dB and transformer gain is 20dB - max gain
20+30+30=80dB - excessive even for a mic pre. Trouble is, for a nominal
+4dBm output, a single 30dB stage and transformer will give an input
sensitivity of just -46dBm which is barely adequate in many situations.
Some more gain would be preferable. Changing the first stage to a pentode
would supply extra gain but would be too noisy. Using local NFB on each
triode to reduce the stage gain to 20dB would work.

Ian

[Phil Allison]

"Ian Bell"

** Then why the hell have *2* triode gain stages with well over 1000
times gain ??

Makes your total gain OVER 10,000 TIMES !!

Madness.


Stage gain is about 30dB and transformer gain is 20dB - max gain
20+30+30=80dB - excessive even for a mic pre. Trouble is, for a nominal
+4dBm output, a single 30dB stage and transformer will give an input
sensitivity of just -46dBm which is barely adequate in many situations.


** Your first stage fixed gain of over 300 times is way too high.

Input overload will be a constant problem - particularly with a condenser
mic.

An input pad will make things noisy again.


........ Phil

[Patrick Turner]

Ian Bell wrote:

Phil Allison wrote:


"Ian Bell"

** Is there to be an input step-up transformer ??


There is 10:1


** Then why the hell have *2* triode gain stages with well over 1000
times gain ??

Makes your total gain OVER 10,000 TIMES !!

Madness.


Stage gain is about 30dB and transformer gain is 20dB - max gain
20+30+30=80dB - excessive even for a mic pre. Trouble is, for a nominal
+4dBm output, a single 30dB stage and transformer will give an input
sensitivity of just -46dBm which is barely adequate in many situations.
Some more gain would be preferable. Changing the first stage to a pentode
would supply extra gain but would be too noisy. Using local NFB on each
triode to reduce the stage gain to 20dB would work.

Ian

He has a gain pot in the tube line up.

If for normal use its set at the 12oclock position, its giving -20dB,
so total gain in effect is 60dB, not 80dB,
so a 0.5mV mic signal is raised to 0.5V, which seems plausible.

Patrick Turner.

[Patrick Turner]

Ian Bell wrote:

I am designing a mic pre based on three 6AU6 tubes. The first two are
operated as triodes with no NFB with an overall gain setting pot divider
between them and the last is a cathode follower. I have discovered that by
judicious selection of the component values in the CF that the
predominantly 2H distortion of the first two stages can be reduced by a
factor of 10. This effect is not frequency or output level sensitive but
does depend on the setting of the inter stage pot - it works best at
highest overall gain.

I was very surprised to see this effect and wondered if anyone else had come
across it. I seem to remember an article somewhere about a two stage amp
where the second stage adds the same sort of distortion to the inverse of
the waveform and hence reduces overall distortion but I cannot at present
lay my hands on it.

OK, admittedly these are just simulations at present and yes, Patrick, I
have pretty much all the components on hand and I WILL be building it as
soon as possible. It is so intriguing I cannot resist.

Try a real circuit and perhaps you'll find you won't get the
cancellations
you say you would, could, might, or did get with simulations.

Say you have a first stage gain of 25, then 5mV of mic signal becomes
125mV,
and THD should be 0.02%, 2H. It'll be the same % at the output of the
gain pot
at any setting.
Suppose you have the gain up max, then 125mV x 25 gives 3.125V at the
second stage anode,
and THD produced would be around 0.15%. There will be very little 2H
cancellations
betwen stage 1 and stage 2.

The cathode follower directly coupled with 3V output will add some THD,
but it'll only
be 0.01% or less, and there isn't enough to cancel that made in the gain
stage feeding it, ever.

You should be able to work out what THD you will get on the back of an
envelope.
The spare sides of envelopes comprise some of the hardware of my kitchen
table simulator.
Computer Power for the Simulator is a 1947 Model HO, ( Human Organic ),
which was last tested in 1962 and gave a 135 reading for IQ, although
the guys testing it
said it was slow, and uncooperative to command programs.

But the only time significant 2H cancellation occurs would be when you
turn down the gain pot,
and only one position where cancellations will be almost complete.

Using constant current sources for all your triode loads would reduce
the THD at 3V output
by around 10dB at least, and you'd never have to worry about THD at all.
0.02% at a volt output is usually utterly inaudible.

Not all THD cancels; all the odd numer H sails right on through.
The intermodulation products increase with more stages
but its unavoidable, and as I said if THD 0.02% and you have not had
to rely
on loop NFB, you should get glorious sound.



Patrick Turner.

Cheers

Ian

[Ian Bell]

Patrick Turner wrote:

Try a real circuit and perhaps you'll find you won't get the
cancellations
you say you would, could, might, or did get with simulations.


Agreed. Tube variations for one are not modeled in the simulation.

Say you have a first stage gain of 25, then 5mV of mic signal becomes
125mV,
and THD should be 0.02%, 2H. It'll be the same % at the output of the
gain pot
at any setting.
Suppose you have the gain up max, then 125mV x 25 gives 3.125V at the
second stage anode,
and THD produced would be around 0.15%. There will be very little 2H
cancellations
betwen stage 1 and stage 2.


That is pretty much what I get.

The cathode follower directly coupled with 3V output will add some THD,
but it'll only
be 0.01% or less, and there isn't enough to cancel that made in the gain
stage feeding it, ever.


This is where the cancellation appears to take place in the CF.

You should be able to work out what THD you will get on the back of an
envelope.
The spare sides of envelopes comprise some of the hardware of my kitchen
table simulator.

I use scrap paper from my printer.

Computer Power for the Simulator is a 1947 Model HO, ( Human Organic ),
which was last tested in 1962 and gave a 135 reading for IQ, although
the guys testing it
said it was slow, and uncooperative to command programs.


I find that easy to believe ;-)

But the only time significant 2H cancellation occurs would be when you
turn down the gain pot,
and only one position where cancellations will be almost complete.


Strangely best cancellation appears to be with the gain pot turned full up.

Using constant current sources for all your triode loads would reduce
the THD at 3V output
by around 10dB at least, and you'd never have to worry about THD at all.
0.02% at a volt output is usually utterly inaudible.


I was not particularly trying to avoid it - I was quite looking forward to a
bit of tube sound. I did want to limit it as far as possible to 2H
distortion which is why I avoided local and global NFB. I was expecting
about 0.2% and that is what I get at the 2nd tube anode. I was surprised to
find that by tweaking CF values I could reduce this value at the CF
output - hence the question about distortion cancellation. What prompted me
to tweak the CF values was that my initial CF design produced more
distortion than the prior two stage combined.

Ian

[Patrick Turner]

Ian Bell wrote:

Patrick Turner wrote:

Try a real circuit and perhaps you'll find you won't get the
cancellations
you say you would, could, might, or did get with simulations.


Agreed. Tube variations for one are not modeled in the simulation.

Say you have a first stage gain of 25, then 5mV of mic signal becomes
125mV,
and THD should be 0.02%, 2H. It'll be the same % at the output of the
gain pot
at any setting.
Suppose you have the gain up max, then 125mV x 25 gives 3.125V at the
second stage anode,
and THD produced would be around 0.15%. There will be very little 2H
cancellations
betwen stage 1 and stage 2.


That is pretty much what I get.

The cathode follower directly coupled with 3V output will add some THD,
but it'll only
be 0.01% or less, and there isn't enough to cancel that made in the gain
stage feeding it, ever.


This is where the cancellation appears to take place in the CF.

I don't think any such cancelation could occur.

If you have a direct coupled CF with the 3rd 6AU6, then the load between
cathode and 0V
for 4mA for Ia will be 30k. In fact with something else connected to the
CF, the
load could end up being 25k, and for 3V output, perhaps Vgk = 3V/gain of
20 = 0.15V.
So the local series voltage NFB with the CF amounts to a gain reduction
of 20 times.
So if the distortion of the triode without the NFB for 3V into 25k = 2%,
then the NFB will reduuce this by a factor of 1/(A+1),
so THD with NFB = 2 / (20 + 1) = 0.095%, or about 0.1%.

THD beloe 3V is approximately proportional to output voltage
so at say 0.2 Vrms for a power needing 2V for clipping,
CF THD = 0.1 x 0.2/3 = 0.0066%, somewhat decent peformance.





You should be able to work out what THD you will get on the back of an
envelope.
The spare sides of envelopes comprise some of the hardware of my kitchen
table simulator.

I use scrap paper from my printer.

Computer Power for the Simulator is a 1947 Model HO, ( Human Organic ),
which was last tested in 1962 and gave a 135 reading for IQ, although
the guys testing it
said it was slow, and uncooperative to command programs.


I find that easy to believe ;-)

But the only time significant 2H cancellation occurs would be when you
turn down the gain pot,
and only one position where cancellations will be almost complete.


Strangely best cancellation appears to be with the gain pot turned full up.

Wait until you do a real amp.....

Using constant current sources for all your triode loads would reduce
the THD at 3V output
by around 10dB at least, and you'd never have to worry about THD at all.
0.02% at a volt output is usually utterly inaudible.

I was not particularly trying to avoid it - I was quite looking forward to a
bit of tube sound. I did want to limit it as far as possible to 2H
distortion which is why I avoided local and global NFB. I was expecting
about 0.2% and that is what I get at the 2nd tube anode. I was surprised to
find that by tweaking CF values I could reduce this value at the CF
output - hence the question about distortion cancellation. What prompted me
to tweak the CF values was that my initial CF design produced more
distortion than the prior two stage combined.

You'd have to use a very low value load to force the CF to produce
enough THD
to cancel that of the preceeding gain stage.

I will bet you a dozen schooners that just arrangeing the circuit to get
maximal
cancellations won't give you the open effortless warm detailed dynamic
clear crisp hifi
that well set up tubes will give.

The main cauise of any THD is the gain tube after the pot.

Now if B+ was +300V, and you had a 47k load, and Ea = +130V, then Ia
would be 3.61 mA. Neglect the few volts across the cathode R with say
470uF bypass.
Ra = approx 10k, µ =36, so gain = 36 x 47 / ( 47 + 10 ) = 30
approximately.

Now if you were to use a CCS, the gain = µ = 36 and the THD would be
about at least 1/3 of the value
for the 47k, and should you wish, B+ could be +250V, Ea at 120V, and Ia
= 5ma,
and THD would fall even further, because with the high RL and higher Ia,
the linearity
would peak at more than 3.6mA.

If you have the anode curves for a 6AU6 in triode, the CCS load appear
as a horizontal line through whatever value of idle current you choose,
and the amount of variation between each Ra curve line for each value of
grid bias
diminishes to a minimum when the load line is a horizontal one.

I can assure you the very best "tube sound" comes from triodes operating
in pure *voltage* mode with NO anode current change that is easily
measured,
which is the case with a CCS using a single pnp transistor where the
actual load
is around 30Mohms, and regarded as a source of current which does not
vary
regardless of the variation in anode voltage.

There will still be some THD even with a CCS.

You can calculate what this might be from measuring the spacings between
the Ra lines.
But usually, the Ra data lines cannot be taken as gospel and are an
approximation
of the actual curves one might get with a given tube sample.
When they may have drawn up the curves 60 years ago, the gear they used
may have had 5%
errors, and usually the actual THD of a triode with CCS is better than
what the curves might suggest.
To establish just what goes on you should have a distortion meter with
input impedance of at least
500k and capable of measuring
0.005% THD in any voltage between 0.5Vrms and 100Vrms.
0.005% of 0.5Vrms = 25uV, and so to see this on a CRO you need to
have a clean test signal with THD = 0.001% max, and be able to remove
the 1kHz signal
from the signal being examined to below the THD level, and then have a
bandpass filter from 1.5kHz to say 10kHz and have an amp to raise the
THD
by say 20 times to 500uV, or 0.5mV to be able to just see it on your
CRO.

I built my own oscillator, filters and amp to be able to measure THD
this well.
But if you asked me to analyse further just what spectral content was in
the 0.005%,
I would be unable to do it. Its safe to suggest that if the CRO shows
the THD as a mainly
simple sine wave of 2kHz, then the 3H might be say 10dB down, or be
0.0016%, rather low,
and not a major worry, IMHO.

Noise and hum will fight against you all the way to trying to get a
decent measurement.
One could have a tunable band pass filter with a Q = 50, and then the
harmonics can be picked out one by one
because the noise across such a tiny BW becomes less than the harmonic.
I built such a filter, which has a dual gang pot for tuning and 3 opamps
and a complex arrangement of
NFB and PFB and is incomprehensible to understand. But it works, and
cost
me only a week to make it work.

Using the right sort of sound card and PC and program also can give you
what I get with primitive
analog test gear I made BC, ( before I had a computer ).

I now don't worry too much about the composition of the THD.
Tube THD and the resulting IMD isn't ever a problem with a preamp if you
are
careful with triodes.
Pentodes are also quite nice used as such, but need shunt NFB to reduce
their high gain.
And they have much more noise if used as input tubes for mics.

The 6AU6 should be a fine triode for an input situation and compete with
1/2 a 6DJ8
for noise and gain.

But you do need a reasonably high input signal for a microphone signal.
With a phono amp, the HF are attenuated at about an approximate slope of
6dB/octave
above 50Hz, and the phono signal has had its HF boosted, so noise can be
very low,
but for a mic the signal is flat, and unboosted, and to get low noise
the
input signal must be high.

Using microphones for recording isn't something I have ever done much
and if I did i'd probably not use a tube at all but use a j-fet.
These are far quieter than any triode.
Most triodes have a hissy noise signal of maybe 2uV if you are lucky.
DC on the heaters is imperative. Shielding may be needed.
Microphony is avoided by careful tube choice.
The 2SK369/2SK147 has about 0.14uV in an average sample.
The step up tranny can still be used.
Once the signal is at over 100mV, tubes are fine.


Patrick Turner.





Ian

[Ian Bell]

Patrick Turner wrote:



Ian Bell wrote:

Patrick Turner wrote:

Try a real circuit and perhaps you'll find you won't get the
cancellations
you say you would, could, might, or did get with simulations.


Agreed. Tube variations for one are not modeled in the simulation.

Say you have a first stage gain of 25, then 5mV of mic signal becomes
125mV,
and THD should be 0.02%, 2H. It'll be the same % at the output of the
gain pot
at any setting.
Suppose you have the gain up max, then 125mV x 25 gives 3.125V at the
second stage anode,
and THD produced would be around 0.15%. There will be very little 2H
cancellations
betwen stage 1 and stage 2.


That is pretty much what I get.

The cathode follower directly coupled with 3V output will add some THD,
but it'll only
be 0.01% or less, and there isn't enough to cancel that made in the
gain stage feeding it, ever.


This is where the cancellation appears to take place in the CF.

I don't think any such cancelation could occur.

If you have a direct coupled CF with the 3rd 6AU6, then the load between
cathode and 0V
for 4mA for Ia will be 30k. In fact with something else connected to the
CF, the
load could end up being 25k, and for 3V output, perhaps Vgk = 3V/gain of
20 = 0.15V.

In the present design the CF is AC coupled to the preceding triode stage.
The CF cathode connects to ground via a 1K and 3.3K in series. The junction
of these two goes to g1 via 330K. The 1K is bypassed with 47uF. The cathode
sits at about 21V and the anode current is about 5mA and g1 is about -5V
wrt the cathode.

So the local series voltage NFB with the CF amounts to a gain reduction
of 20 times.
So if the distortion of the triode without the NFB for 3V into 25k = 2%,
then the NFB will reduuce this by a factor of 1/(A+1),
so THD with NFB = 2 / (20 + 1) = 0.095%, or about 0.1%.

THD beloe 3V is approximately proportional to output voltage
so at say 0.2 Vrms for a power needing 2V for clipping,
CF THD = 0.1 x 0.2/3 = 0.0066%, somewhat decent peformance.


I ordered the last of the components today so I hope to be building and
measuring this later next week.





You should be able to work out what THD you will get on the back of an
envelope.
The spare sides of envelopes comprise some of the hardware of my
kitchen table simulator.

I use scrap paper from my printer.

Computer Power for the Simulator is a 1947 Model HO, ( Human Organic ),
which was last tested in 1962 and gave a 135 reading for IQ, although
the guys testing it
said it was slow, and uncooperative to command programs.


I find that easy to believe ;-)

But the only time significant 2H cancellation occurs would be when you
turn down the gain pot,
and only one position where cancellations will be almost complete.


Strangely best cancellation appears to be with the gain pot turned full
up.

Wait until you do a real amp.....

Using constant current sources for all your triode loads would reduce
the THD at 3V output
by around 10dB at least, and you'd never have to worry about THD at
all. 0.02% at a volt output is usually utterly inaudible.

I was not particularly trying to avoid it - I was quite looking forward
to a bit of tube sound. I did want to limit it as far as possible to 2H
distortion which is why I avoided local and global NFB. I was expecting
about 0.2% and that is what I get at the 2nd tube anode. I was surprised
to find that by tweaking CF values I could reduce this value at the CF
output - hence the question about distortion cancellation. What prompted
me to tweak the CF values was that my initial CF design produced more
distortion than the prior two stage combined.

You'd have to use a very low value load to force the CF to produce
enough THD
to cancel that of the preceeding gain stage.


See above.

I will bet you a dozen schooners that just arrangeing the circuit to get
maximal
cancellations won't give you the open effortless warm detailed dynamic
clear crisp hifi
that well set up tubes will give.


I am sure you are right.


The main cauise of any THD is the gain tube after the pot.

Now if B+ was +300V, and you had a 47k load, and Ea = +130V, then Ia
would be 3.61 mA. Neglect the few volts across the cathode R with say
470uF bypass.
Ra = approx 10k, µ =36, so gain = 36 x 47 / ( 47 + 10 ) = 30
approximately.

Now if you were to use a CCS, the gain = µ = 36 and the THD would be
about at least 1/3 of the value
for the 47k, and should you wish, B+ could be +250V, Ea at 120V, and Ia
= 5ma,
and THD would fall even further, because with the high RL and higher Ia,
the linearity
would peak at more than 3.6mA.


First stage B+ is 170V, Ea=86V and Ia=0.8mA and 2nd stage has B+=250V,
Ea=82V and Ia=1.7mA at present. Anode load is 100K in both cases.

If you have the anode curves for a 6AU6 in triode, the CCS load appear
as a horizontal line through whatever value of idle current you choose,
and the amount of variation between each Ra curve line for each value of
grid bias
diminishes to a minimum when the load line is a horizontal one.

I can assure you the very best "tube sound" comes from triodes operating
in pure *voltage* mode with NO anode current change that is easily
measured,
which is the case with a CCS using a single pnp transistor where the
actual load
is around 30Mohms, and regarded as a source of current which does not
vary
regardless of the variation in anode voltage.

There will still be some THD even with a CCS.


Sounds very interesting and I guess I could still use a 6AU6 for the CSS?

You can calculate what this might be from measuring the spacings between
the Ra lines.
But usually, the Ra data lines cannot be taken as gospel and are an
approximation
of the actual curves one might get with a given tube sample.
When they may have drawn up the curves 60 years ago, the gear they used
may have had 5%
errors, and usually the actual THD of a triode with CCS is better than
what the curves might suggest.
To establish just what goes on you should have a distortion meter with
input impedance of at least
500k and capable of measuring
0.005% THD in any voltage between 0.5Vrms and 100Vrms.
0.005% of 0.5Vrms = 25uV, and so to see this on a CRO you need to
have a clean test signal with THD = 0.001% max, and be able to remove
the 1kHz signal
from the signal being examined to below the THD level, and then have a
bandpass filter from 1.5kHz to say 10kHz and have an amp to raise the
THD
by say 20 times to 500uV, or 0.5mV to be able to just see it on your
CRO.

I built my own oscillator, filters and amp to be able to measure THD
this well.
But if you asked me to analyse further just what spectral content was in
the 0.005%,
I would be unable to do it. Its safe to suggest that if the CRO shows
the THD as a mainly
simple sine wave of 2kHz, then the 3H might be say 10dB down, or be
0.0016%, rather low,
and not a major worry, IMHO.

Noise and hum will fight against you all the way to trying to get a
decent measurement.
One could have a tunable band pass filter with a Q = 50, and then the
harmonics can be picked out one by one
because the noise across such a tiny BW becomes less than the harmonic.
I built such a filter, which has a dual gang pot for tuning and 3 opamps
and a complex arrangement of
NFB and PFB and is incomprehensible to understand. But it works, and
cost
me only a week to make it work.

Using the right sort of sound card and PC and program also can give you
what I get with primitive
analog test gear I made BC, ( before I had a computer ).

I now don't worry too much about the composition of the THD.
Tube THD and the resulting IMD isn't ever a problem with a preamp if you
are
careful with triodes.
Pentodes are also quite nice used as such, but need shunt NFB to reduce
their high gain.
And they have much more noise if used as input tubes for mics.

The 6AU6 should be a fine triode for an input situation and compete with
1/2 a 6DJ8
for noise and gain.


RDH seems to think so too.

But you do need a reasonably high input signal for a microphone signal.
With a phono amp, the HF are attenuated at about an approximate slope of
6dB/octave
above 50Hz, and the phono signal has had its HF boosted, so noise can be
very low,
but for a mic the signal is flat, and unboosted, and to get low noise
the
input signal must be high.


Agreed. Normal practice for tube mic pres seems to be to have a fairly high
turns ratio (10:1 or more) transformer at the input which can provide 20dB
or more of relatively low NF gain.

Using microphones for recording isn't something I have ever done much
and if I did i'd probably not use a tube at all but use a j-fet.
These are far quieter than any triode.
Most triodes have a hissy noise signal of maybe 2uV if you are lucky.
DC on the heaters is imperative. Shielding may be needed.
Microphony is avoided by careful tube choice.

Yes, it is a fascinating design challenge.


Ian

[Patrick Turner]

This is where the cancellation appears to take place in the CF.

I don't think any such cancelation could occur.

If you have a direct coupled CF with the 3rd 6AU6, then the load between
cathode and 0V
for 4mA for Ia will be 30k. In fact with something else connected to the
CF, the
load could end up being 25k, and for 3V output, perhaps Vgk = 3V/gain of
20 = 0.15V.

In the present design the CF is AC coupled to the preceding triode stage.
The CF cathode connects to ground via a 1K and 3.3K in series. The junction
of these two goes to g1 via 330K. The 1K is bypassed with 47uF. The cathode
sits at about 21V and the anode current is about 5mA and g1 is about -5V
wrt the cathode.

The CF load is way to low.

CF loads should never be lower than what you use for an anode load.

The low CF load forces it to distort.
loop gain becomes lower as RL for the CF load is reduced, and the local
CF NFB also becomes lower, so less correction
of the THD occurs.
CF open loop gain = cathode output signal / Vgk signal.

CF closed loop gain = (cathode output signal + Vgk) / Vgk

Patrick Turner.

[Patrick Turner]

flipper wrote:

On Sat, 16 Jun 2007 21:13:40 +0100, Ian Bell
wrote:

Patrick Turner wrote:



Ian Bell wrote:
trim


The 6AU6 should be a fine triode for an input situation and compete with
1/2 a 6DJ8
for noise and gain.


RDH seems to think so too.


I'm curious. With the 6DJ8 having 3 to 4 times the gm what's the
advantage to a trioded 6AU6 for the first stage rather than the 6DJ8?

Cost?

I have accumulated about 40 6AU6 in my junk boxes.
I didn't pay for any of them, and don't know how they got there.

When i turn 150, I will be able to draw on the stocks of tubes I still
will have.

I will be able to pretend I can hear the music, for I will have become
quite deaf.

Patrick Turner.

[Ian Bell]

flipper wrote:

On Sat, 16 Jun 2007 21:13:40 +0100, Ian Bell
wrote:

Patrick Turner wrote:



Ian Bell wrote:
trim


The 6AU6 should be a fine triode for an input situation and compete with
1/2 a 6DJ8
for noise and gain.


RDH seems to think so too.


I'm curious. With the 6DJ8 having 3 to 4 times the gm what's the
advantage to a trioded 6AU6 for the first stage rather than the 6DJ8?

Nothing other than I happen to have some ;-)

Ian

[Ian Iveson]

Ian Bell said:

This is where the cancellation appears to take place in the CF.

Then there is something very badly wrong with your CF or your
simulation.

Why don't you post your circuit diagram, perhaps to ABSE? Easy done.
Maybe you could show pictures of your graphs too. That would be
instructive.

cheers, Ian

[Ian Bell]

Ian Iveson wrote:

Ian Bell said:

This is where the cancellation appears to take place in the CF.

Then there is something very badly wrong with your CF or your
simulation.


Personally I strongly suspect the simulation.

Why don't you post your circuit diagram, perhaps to ABSE? Easy done.
Maybe you could show pictures of your graphs too. That would be
instructive.

cheers, Ian

I assume ABSE is a binaries news group? if so my ISP does not provide access
to binary groups.

Ian

[Patrick Turner]

Ian Bell wrote:

Ian Iveson wrote:

Ian Bell said:

This is where the cancellation appears to take place in the CF.

Then there is something very badly wrong with your CF or your
simulation.


Personally I strongly suspect the simulation.

Why don't you post your circuit diagram, perhaps to ABSE? Easy done.
Maybe you could show pictures of your graphs too. That would be
instructive.

cheers, Ian

I assume ABSE is a binaries news group? if so my ISP does not provide access
to binary groups.

Most binaries usernet groups convey a vast huge amount of pornography to
the
sex staved millions of men who demand it.

All other usernet traffic is a small amount of data by comparison.

data costs money to transfer.

Find yourself an ISP who gives you access to the Internet and includes
access
to all binaries.
I had many rows with ****ing ISPs who promised good access to NGs
included in their
internet access price but who failed to give it.

I had to shop around about 6 times before I could get abse reliably, and
even then
its never instant; image posts take maybe 2 days to be seen after folks
posts them
or m,aybe I miss out, becays ebinaries, or images are only held in
servers for 48 hours,
and the delays in refresh times means that when my ISP source updates
the list of posts, the binaries posts
have expired from upstream.
The amount of traffic also has an effect. less flow = less updates =
expired articles.
You never get problems with alt.binaries.pictures.erotica.blondes
There are hundreds of posts each day.
But abse and some others about radio and phonogrpahs also don't have
much content flow and
don't get through in the system.

Patrick Turner.


Ian

[Iain Churches]

"Ian Bell" wrote in message
...
Ian Iveson wrote:

Ian Bell said:

This is where the cancellation appears to take place in the CF.

Then there is something very badly wrong with your CF or your
simulation.


Personally I strongly suspect the simulation.

Why don't you post your circuit diagram, perhaps to ABSE? Easy done.
Maybe you could show pictures of your graphs too. That would be
instructive.

cheers, Ian

I assume ABSE is a binaries news group? if so my ISP does not provide
access
to binary groups.

Ian. I can host them for you if required.

Iain

[Patrick Turner]

Ian Iveson wrote:

Ian Bell said:

This is where the cancellation appears to take place in the CF.

Then there is something very badly wrong with your CF or your
simulation.

Why don't you post your circuit diagram, perhaps to ABSE? Easy done.
Maybe you could show pictures of your graphs too. That would be
instructive.

cheers, Ian

He already said how he has a total RL of the CF = 4.3k.

Hence the high THD of the CF.

Patrick Turner.

[Ian Iveson]

Patrick Turner

Then there is something very badly wrong with your CF or your
simulation.

Why don't you post your circuit diagram, perhaps to ABSE? Easy
done.
Maybe you could show pictures of your graphs too. That would be
instructive.

cheers, Ian

He already said how he has a total RL of the CF = 4.3k.

Hence the high THD of the CF.


And the lost opportunity for direct coupling.

The problem was unlikely to be spice, although I find measuring
distortion via fourier a bit hit and miss. I should make a distortion
meter I suppose.

I thought you might have been more forthcoming wrt 2H cancellation. If
valves 1 and 2 had the same load and operating point then 2H would
cancel if the attenuation between the two resulted in no overall gain
for the first stage, so the second would produce the same amount, but
inverted. If the valves are at different operating points, then a
minimum 2H would be somewhere else on the volume setting. At high
volume settings, the 2H from the valve 2 would predominate. Assuming
that to be the case, it is interesting to note that the distorting CF,
although it is a non-inverting stage, reduces the total 2H at high
volume settings.

Anyway, Ian might try the CF as it is, with a clean signal, and
measure the ensuing output, for a range of amplitudes and loads.

We should have nipped this CF fetishism in the bud. It'll be the death
of us.

I hadn't really looked at a sharp-cutoff pentode before now. Do they
do things like that but bigger? Why don't I have a model of one? I
fancy PP transformer coupled with distributed load and cross-coupled
screens.

cheers, Ian

[Martin Schöön]

Ian Bell writes:

I am designing a mic pre based on three 6AU6 tubes. The first two are
operated as triodes with no NFB with an overall gain setting pot divider
between them and the last is a cathode follower. I have discovered that by
judicious selection of the component values in the CF that the
predominantly 2H distortion of the first two stages can be reduced by a
factor of 10. This effect is not frequency or output level sensitive but
does depend on the setting of the inter stage pot - it works best at
highest overall gain.

I was very surprised to see this effect and wondered if anyone else had come
across it. I seem to remember an article somewhere about a two stage amp
where the second stage adds the same sort of distortion to the inverse of
the waveform and hence reduces overall distortion but I cannot at present
lay my hands on it.

Pre-distortion linearisation is nothing strange. It is used
routinely in industry. Pre-distortion come is two flavors:
analogue and digital.

The former is about designing stage 1 to have AM/AM and
AM/PM characteristics that take out some of the non-linearity
of stage 2. Simplistically put: let stage 1 have a gain
expansion countering stage 2's gain compression...

Digital pre-distortion is about altering the signal in
the digital signal procession of the equipment to compensate
for the non-linearity of the analog part.

Both techniques are used in the transmitters of radio
base stations of cellular phone systems.

--
Martin Schöön "Problems worthy of attack
show their worth by hitting back."
Piet Hein