Hello Brother Rodents,
I would like to run my CD straight into my amp without the aid of a preamp.
If you were to do the same, how would you configure the attenuator for
volume control? Would you use a plastic film 100K pot in the CD's output or
the same in the amp's input. Perhaps you like a multi switched dual
attenuator with a few positions. We can get in quite a debate, I suppose, on
the type of resistors and how many positions do you really need. Perhaps for
the ambitious, you would try an infra red remote assembly? I'm mostly
concerned with not adulterating the signal, keeping it as pure (transparent)
as possible. Opinions? Thank you.
Cordially,
west

west wrote:

Hello Brother Rodents,
I would like to run my CD straight into my amp without the aid of a preamp.
If you were to do the same, how would you configure the attenuator for
volume control? Would you use a plastic film 100K pot in the CD's output or
the same in the amp's input. Perhaps you like a multi switched dual
attenuator with a few positions. We can get in quite a debate, I suppose, on
the type of resistors and how many positions do you really need. Perhaps for
the ambitious, you would try an infra red remote assembly? I'm mostly
concerned with not adulterating the signal, keeping it as pure (transparent)
as possible. Opinions? Thank you.
Cordially,
west

I high quality dual pot would do a good job at this. If you're only using a CD
player, I would suggest to go with a 10K log pot, rather than a 100K, connecting
the pot accross the output of the CD player, with the wiper going to the input
of your amplifier.

The reason I suggest 10K rather than 100K, is that most CD players have a nice
low output impedance, so driving the 10K will be no problem, and if you ever
choose to use long cables or cables with a high capacitance, you won't have to
live with high frequency roll off like you would with a 100K pot.

I reccomend ALPS potentiometers, the one in my preamp is a big square thing, and
it's detended to feel as though it's a stepped attenuator even though it's just
a pot. It's fairly important that you spend some money on this component,
because having a pot that isn't perfectly balanced channel to channel, accross
its whole range, is really annoying.

west wrote:

Hello Brother Rodents,
I would like to run my CD straight into my amp without the aid of a preamp.
If you were to do the same, how would you configure the attenuator for
volume control? Would you use a plastic film 100K pot in the CD's output or
the same in the amp's input.

I would put the attenuator/pot/whatever in the amp if possible. The
output impedance of the CD player is probably quite low - if you add a
volume control the output impedance get's a lot higher.

If the cable between the CD player and amp is short, you could also put
the attenuator in the CD player. But use a value lower than 100k - 10,
20 or 50k would be much better.

Perhaps you like a multi switched dual
attenuator with a few positions. We can get in quite a debate, I suppose, on
the type of resistors and how many positions do you really need.

You could make an attenuator using a few resistors and relays and get a
32 or 64 step attenuator with no problems. If you're rich you could also
use one of the attenuator transformers with 25 2dB steps.

I tried replacing the volume pot in my amp with a resistor/relay
attenuator some time ago, and it did sound better. The difference wasn't
big, but I think it was there - I could have imagined the difference of
course

Perhaps for the ambitious, you would try an infra red remote assembly?

Of course! Why wouldn't you?

I'm mostly concerned with not adulterating the signal, keeping it as
pure (transparent) as possible. Opinions? Thank you.

Then I think as attenuator with perhaps 5 relays would be a "pure"
solution. Use good relays (they are not expensive), and if you want you
could hardwire the circuit to avoid the PCB tracks.

Best regards,

Mikkel C. Simonsen

Mikkel,
I may be missing a point (which isn't difficult for me to do). In your
attenuator concept, why use relays? Can not you use a dual pole
multi-position, audiophile quality switch? This switch would select
different boutique resistors. Why would you need relays? Please advise.
Thank you.
west
BTW: is there such an animal as a audiophile quality multi position, stereo
switch? I would need about 4 to 6 positions. Also, what resistors should I
use for attenuation, if I went that way?

"Mikkel C. Simonsen" wrote in message
...
west wrote:

Hello Brother Rodents,
I would like to run my CD straight into my amp without the aid of a
preamp.
If you were to do the same, how would you configure the attenuator for
volume control? Would you use a plastic film 100K pot in the CD's output
or
the same in the amp's input.

I would put the attenuator/pot/whatever in the amp if possible. The
output impedance of the CD player is probably quite low - if you add a
volume control the output impedance get's a lot higher.

If the cable between the CD player and amp is short, you could also put
the attenuator in the CD player. But use a value lower than 100k - 10,
20 or 50k would be much better.

Perhaps you like a multi switched dual
attenuator with a few positions. We can get in quite a debate, I
suppose, on
the type of resistors and how many positions do you really need.

You could make an attenuator using a few resistors and relays and get a
32 or 64 step attenuator with no problems. If you're rich you could also
use one of the attenuator transformers with 25 2dB steps.

I tried replacing the volume pot in my amp with a resistor/relay
attenuator some time ago, and it did sound better. The difference wasn't
big, but I think it was there - I could have imagined the difference of
course

Perhaps for the ambitious, you would try an infra red remote assembly?

Of course! Why wouldn't you?

I'm mostly concerned with not adulterating the signal, keeping it as
pure (transparent) as possible. Opinions? Thank you.

Then I think as attenuator with perhaps 5 relays would be a "pure"
solution. Use good relays (they are not expensive), and if you want you
could hardwire the circuit to avoid the PCB tracks.

Best regards,

Mikkel C. Simonsen

"Mikkel C. Simonsen" schreef in bericht
...

snip

You could make an attenuator using a few resistors and relays and get a
32 or 64 step attenuator with no problems. If you're rich you could also
use one of the attenuator transformers with 25 2dB steps.

I tried replacing the volume pot in my amp with a resistor/relay
attenuator some time ago, and it did sound better. The difference wasn't
big, but I think it was there - I could have imagined the difference of
course

Then I think as attenuator with perhaps 5 relays would be a "pure"
solution. Use good relays (they are not expensive), and if you want you
could hardwire the circuit to avoid the PCB tracks.


Hi Mikkel,

I vaguely remember seeing a "stepped relay attenuator" a while ago.
Whether the circuit was of your hand or somebody else's I don't recall.
Checking my bookmarks also to no avail.

Anyway, using 5 relays and a few resistors for a 32 step attenuator "sounds"
pure and simple.
Could you (re)post your circuit, at f.e. ABSE, please?

Regards,
Jan.


Best regards,

Mikkel C. Simonsen

west wrote:

Mikkel,
I may be missing a point (which isn't difficult for me to do). In your
attenuator concept, why use relays? Can not you use a dual pole
multi-position, audiophile quality switch? This switch would select
different boutique resistors. Why would you need relays? Please advise.

You could use a switch, but then the number of settings would be
limited. With 5 relays you could get 32 steps. A rotary switch is also
difficult to remote control

BTW: is there such an animal as a audiophile quality multi position, stereo
switch? I would need about 4 to 6 positions.

Elma? Grayhill?

Also, what resistors should I use for attenuation, if I went that way?

That ones's easy - just as easy as what capacitors

I use Welwyn metal film resistors. But try different ones - they don't
have to be expensive.

Best regards,

Mikkel C. Simonsen

JKoning wrote:

You could make an attenuator using a few resistors and relays and get a
32 or 64 step attenuator with no problems. If you're rich you could also
use one of the attenuator transformers with 25 2dB steps.

Hi Mikkel,

I vaguely remember seeing a "stepped relay attenuator" a while ago.
Whether the circuit was of your hand or somebody else's I don't recall.
Checking my bookmarks also to no avail.

Anyway, using 5 relays and a few resistors for a 32 step attenuator "sounds"
pure and simple.
Could you (re)post your circuit, at f.e. ABSE, please?

I have posted two different schematics. One uses 8 relays and "emulates"
a pot, so it can be used to make a pot with any taper:
http://stiftsbogtrykkeriet.dk/~mcs/8bitatt.gif

The other uses 6 relays and is an attenuator/volume control replacement:
http://stiftsbogtrykkeriet.dk/~mcs/6bitatt.gif

Both types of attenuator can be controlled by an absolute encoder, if
you want to keep it simple (no microcontrollers). You can also use a
standard encoder or up/down buttons and a microcontroller - which makes
remote control easy.

I have recently designed yet another "passive preamp" - a lot like this
one: http://stiftsbogtrykkeriet.dk/~mcs/Remote3/index.html but with a
6-relays attenuator instead of the motorpot. But the PCB is on the way
from Bulgaria, so I won't be able to test that circuit before I receive
the PCB.

I could also draw a small volume control only PCB, if anybody's
interested.

Best regards,

Mikkel C. Simonsen

"Mikkel C. Simonsen" schreef in bericht
...
JKoning wrote:

snip

I have posted two different schematics. One uses 8 relays and "emulates"
a pot, so it can be used to make a pot with any taper:
http://stiftsbogtrykkeriet.dk/~mcs/8bitatt.gif

The other uses 6 relays and is an attenuator/volume control replacement:
http://stiftsbogtrykkeriet.dk/~mcs/6bitatt.gif

Thanks Mikkel for the post.
This is exactly what I had in mind but could'nt remember where to find.
For the first mentioned 8 relay pot "emulator" I wonder why you choose
R15/16 & R31/32 to be 12K instead of 12K8?

Regards,
Jan.



Both types of attenuator can be controlled by an absolute encoder, if
you want to keep it simple (no microcontrollers). You can also use a
standard encoder or up/down buttons and a microcontroller - which makes
remote control easy.

I have recently designed yet another "passive preamp" - a lot like this
one: http://stiftsbogtrykkeriet.dk/~mcs/Remote3/index.html but with a
6-relays attenuator instead of the motorpot. But the PCB is on the way
from Bulgaria, so I won't be able to test that circuit before I receive
the PCB.

I could also draw a small volume control only PCB, if anybody's
interested.

Best regards,

Mikkel C. Simonsen

JKoning wrote:

I have posted two different schematics. One uses 8 relays and "emulates"
a pot, so it can be used to make a pot with any taper:
http://stiftsbogtrykkeriet.dk/~mcs/8bitatt.gif

The other uses 6 relays and is an attenuator/volume control replacement:
http://stiftsbogtrykkeriet.dk/~mcs/6bitatt.gif

Thanks Mikkel for the post.
This is exactly what I had in mind but could'nt remember where to find.
For the first mentioned 8 relay pot "emulator" I wonder why you choose
R15/16 & R31/32 to be 12K instead of 12K8?

You just spotted an error - it should be 12k8 of course... I don't
remember what values I actually used in my prototype circuit - I just
found some E12/E24 values I had available.

Best regards,

Mikkel C. Simonsen

"Mikkel C. Simonsen" schreef in bericht
...
JKoning wrote:

The other uses 6 relays and is an attenuator/volume control
replacement:
http://stiftsbogtrykkeriet.dk/~mcs/6bitatt.gif

I really like this one.
Pure and simple set-up.
Yet confusing, forgive my ignorance, as I can't see how you calculate the
various steps of attenuation.
Did you have any second thoughts about channel-separation/cross-talk between
L/R signals by using DPDT relays caused by the parasitic capacitance between
contacts?
Did you possibly do any cross-talk comparision between a 2x6-SPDT version
and the drawn 6x-SPDT version?

Plan to give it a shot using a rotary 5 bit BCD encoder I happen to have
handy.
Nothing build, yet problems arising already.
The encoder does'nt have and end-stop, so..., a 00000 to 11111 transition
in just 1 step... exploded ears...
In case of to many accoustic shocks I'll simply go for the "knob pointer
hits chewing-gum end-stop" solution.
For a test it'll be fine though, we'll see what follows.

Regards,
Jan.


Best regards,

Mikkel C. Simonsen

JKoning wrote:

The other uses 6 relays and is an attenuator/volume control
replacement: http://stiftsbogtrykkeriet.dk/~mcs/6bitatt.gif

I really like this one.
Pure and simple set-up.
Yet confusing, forgive my ignorance, as I can't see how you calculate the
various steps of attenuation.

That's very simple...

Here's the formula for calculating the attenuation of one of the stages:
=ABS(20*(LOG(((1/(D5+((D8^-1)+(E5+(((E8^-1)+(A5^-1))^-1))^-1)^-1))*(((D8^-1)+(E5+(((E8^-1)+(A5^-1))^-1))^-1)^-1)/(E5+((E8^-1)+A5^-1)^-1))*(((E8^-1)+A5^-1)^-1))))

Simple isn't it?

Don't ask me how I came up with that formula - I haven't got a clue...
But I have a working spreadsheet, so just tell me the stepsize and input
impedance you want and I'll find out the values.

Did you have any second thoughts about channel-separation/cross-talk between
L/R signals by using DPDT relays caused by the parasitic capacitance between
contacts?
Did you possibly do any cross-talk comparision between a 2x6-SPDT version
and the drawn 6x-SPDT version?

Yes, there may be a problem with crosstalk - I'll know after I test the
circuit. I should receive the PCB in a couple of days, and I'll do some
measurements as soon as the board has been assembled. If it's possible
to use only one relay for both channels it will keep the size down, so
if it works that would be great. Stay tuned...

Plan to give it a shot using a rotary 5 bit BCD encoder I happen to have
handy.

You'll probably still need the ULN2003 (or 2004) - I don't think an
encoder is meant for driving relays.

Nothing build, yet problems arising already.
The encoder does'nt have and end-stop, so..., a 00000 to 11111 transition
in just 1 step... exploded ears...
In case of to many accoustic shocks I'll simply go for the "knob pointer
hits chewing-gum end-stop" solution.

That's always the problem. You can order the encoders with stops, but
you'll probably have to order 1000 units

For a test it'll be fine though, we'll see what follows.

You could just use a standard rotary encoder and a microcontroller - not
that complicated.

Best regards,

Mikkel C. Simonsen


Regards,
Jan.

Best regards,

Mikkel C. Simonsen

Mikkel C. Simonsen wrote:

Did you have any second thoughts about channel-separation/cross-talk between
L/R signals by using DPDT relays caused by the parasitic capacitance between
contacts?
Did you possibly do any cross-talk comparision between a 2x6-SPDT version
and the drawn 6x-SPDT version?

Yes, there may be a problem with crosstalk - I'll know after I test the
circuit. I should receive the PCB in a couple of days, and I'll do some
measurements as soon as the board has been assembled. If it's possible
to use only one relay for both channels it will keep the size down, so
if it works that would be great. Stay tuned...

I finally had time to complete the PCB. It looks like it works, and
-63dB is actually very quiet even with my much too sensitive amp... The
relays I used for the (stereo) attenuator are the smallest I have ever
seen.

But how do you measure crosstalk? Apply a signal to one channel and
leave the other input open? Or shortcircuited? And what signal levels
are normally used? If anybody has a simple recipe I'll make some
measurements...

Plan to give it a shot using a rotary 5 bit BCD encoder I happen to have
handy.

Did you give it a shot?

Best regards,

Mikkel C. Simonsen

But how do you measure crosstalk? Apply a signal to one channel and
leave the other input open? Or shortcircuited?

Connect a resistor to the other input and to ground. The resistance
should be
about the same as the impedance of sources likely used with the equipment.

And what signal levels
are normally used?

Probably signal levels you'd see in normal use.

"Mikkel C. Simonsen" schreef in bericht
...
Mikkel C. Simonsen wrote:

Did you have any second thoughts about channel-separation/cross-talk
between
L/R signals by using DPDT relays caused by the parasitic capacitance
between
contacts?
Did you possibly do any cross-talk comparision between a 2x6-SPDT
version
and the drawn 6x-SPDT version?

Yes, there may be a problem with crosstalk - I'll know after I test the
circuit. I should receive the PCB in a couple of days, and I'll do some
measurements as soon as the board has been assembled. If it's possible
to use only one relay for both channels it will keep the size down, so
if it works that would be great. Stay tuned...

I finally had time to complete the PCB. It looks like it works, and
-63dB is actually very quiet even with my much too sensitive amp... The
relays I used for the (stereo) attenuator are the smallest I have ever
seen.

Cross-talk is frequency dependant. At what f. did you obtain -63dB?
I suggest to determine it at 2 frequencies at least, f.e. 1K and 20K.
Also, possible better cross-talk performance of SPDT relays versus
DPDT relays becomes more perceptable at higher frequencies*.

But how do you measure crosstalk? Apply a signal to one channel and
leave the other input open? Or shortcircuited?

As Robert suggested already, terminate the input, of the
channel to be measured, with a resistor equal to the output
impedance of the preceeding stage.
But also terminate the output with a resistor equal to the
input impedance of the following stage.
Those impedances might be difficult to define exactly in the
case where the DUT did not find a real live application.
I guess an input termination of 1K-10K and an output
termination of 100K-1M would be a good values to start with.

Bear in mind the input impedance of your (mV)meter when
determining the output termination.
Ideally it should have high Rin and (very) low Cin*.

And what signal levels are normally used? If anybody
has a simple recipe I'll make some measurements...

Don't worry about levels as cross-talk is about ratio
between signal carrying channel and channel under test.
Instead, take care of a signal level on the output of the
signal carrying channel of, say, 80dB above the noise floor.
E.g., with no signal applied, measure the noise on both
channels, for argument sake f.e. -70dBV (~0.25mV).
Apply a signal on input of the L-channel such that the
output of the L-channel is -70 + 80 = +10dBV (~2.5V).
Now, measure the R-channel output, f.e. -50dBV.
You now know the cross-talk is -50 - (+10) = -60dBV.
Also you know S/N difference is -50 - (-70) = +20dBV.
The S/N difference tells that the R-channel output is well
above the noise floor and, because of that, the value
found for cross-talk can be regarded as a reliable figure.

Another parameter, relevant for the performance of the
attenuator, might be Power Supply Rejection Ratio.
For that superimpose AC voltage on the relay DC supply
and determine PSRR in the same way as cross-talk.

*My good old GM6012 has 4M/20pF for 3V and below.
With the standard screended cable connected, input
capacitance will go up with ~130pF to ~150pF in total.
I.e., the meter's Zin will be down to ~50K @ 20KHz!
Bear this in mind when connecting meters to outputs!
Use either an active probe or a passive (if sensitivity
allows) 10:1 frequency compensated probe for low Cin.


Plan to give it a shot using a rotary 5 bit BCD encoder I happen to
have
handy.

Did you give it a shot?

After all, when I dug up the encoder, it turned out to
be Gray encoded i.s.o. BCD. So, I left it alone for a
while. But it's still itching. I'll report when I've got results.

Regards,
Jan.


Best regards,

Mikkel C. Simonsen

JKoning wrote:

Cross-talk is frequency dependant. At what f. did you obtain -63dB?

The -63dB was calculated not measured. If my formalas are correct the
attenuation at the min. setting should be -63dB. I have measured the
attenuation at -31 and -39dB, and at those two settings it was correct.

As Robert suggested already, terminate the input, of the
channel to be measured, with a resistor equal to the output
impedance of the preceeding stage.

If a normal op. amp. output is driving the input, then I guess that
would be around 100 Ohms. If a cathode follower is used it's probably in
the 300 Ohms to 1k range. I chose 680 Ohms which should be about the
same as my old HP oscillator.

But also terminate the output with a resistor equal to the
input impedance of the following stage.
Those impedances might be difficult to define exactly in the
case where the DUT did not find a real live application.
I guess an input termination of 1K-10K and an output
termination of 100K-1M would be a good values to start with.

The output load has to be correct to get the correct attenuation. In
this case the load should be 50k (same as the input impedance), so I
chose 56k (I didn't have any 51k).

Bear in mind the input impedance of your (mV)meter when
determining the output termination.
Ideally it should have high Rin and (very) low Cin*.

My meter isn't sensitive enough (1.5V full scale), so I used the scope
instead.

These are the results. For the first test I used a 1kHz 4Vpp signal. At
0dB attenuation I measured 0.8mV at the "other" channel, making the
crosstalk -74dB. At -31dB I measured 110mV and 0.4mV - 49dB difference.
But as these readings were made using a scope, the actual figure could
be better - it was difficult to see the signal at all...

I repeated the same tests at 10kHz, where the figures were -69dB and
-43dB. Is that good or bad? I guess I should try to find a better meter
for more accurate measurements.

Another parameter, relevant for the performance of the
attenuator, might be Power Supply Rejection Ratio.
For that superimpose AC voltage on the relay DC supply
and determine PSRR in the same way as cross-talk.

As the powersupply for the relays is on the same board, that is a bit
difficult...

After all, when I dug up the encoder, it turned out to
be Gray encoded i.s.o. BCD. So, I left it alone for a
while. But it's still itching. I'll report when I've got results.

So just use an EPROM as a Gray to BCD decoder...

Best regards,

Mikkel C. Simonsen

"Mikkel C. Simonsen" schreef in bericht
...
JKoning wrote:

Cross-talk is frequency dependant. At what f. did you obtain -63dB?

The -63dB was calculated not measured. If my formalas are correct the
attenuation at the min. setting should be -63dB. I have measured the
attenuation at -31 and -39dB, and at those two settings it was correct.

I mixed up, I thought you were talking about -63dB cross-talk in
the previous post while in fact you were talking about attenuation.


As Robert suggested already, terminate the input, of the
channel to be measured, with a resistor equal to the output
impedance of the preceeding stage.

If a normal op. amp. output is driving the input, then I guess that
would be around 100 Ohms. If a cathode follower is used it's probably in
the 300 Ohms to 1k range. I chose 680 Ohms which should be about the
same as my old HP oscillator.

Fine, but don't forget to terminate the "other channel" too.


But also terminate the output with a resistor equal to the
input impedance of the following stage.
Those impedances might be difficult to define exactly in the
case where the DUT did not find a real live application.
I guess an input termination of 1K-10K and an output
termination of 100K-1M would be a good values to start with.

The output load has to be correct to get the correct attenuation. In
this case the load should be 50k (same as the input impedance), so I
chose 56k (I didn't have any 51k).

Do I understand correctly the attenuator as shown at :
http://stiftsbogtrykkeriet.dk/~mcs/6bitatt.gif
is designed for termination with a 50K load?


Bear in mind the input impedance of your (mV)meter when
determining the output termination.
Ideally it should have high Rin and (very) low Cin*.

My meter isn't sensitive enough (1.5V full scale), so I used the scope
instead.

Your meter's sensitivity can be cranked up quite easy.
Make sure the meter's frequency response is 20kHz.
Attach a suitable instrumentation amp to the meter input
such as an AMP01 and you'll have 1.5mV f.s. sensitivity.
According paper specifications expect the following.
With gain set to 60dB, frequency response (-3dB) is
~ 26kHz, CMRR 125dB and output noise is 0.9mV.
Such an "active probe" is very handy for scope work too!


These are the results. For the first test I used a 1kHz 4Vpp signal. At
0dB attenuation I measured 0.8mV at the "other" channel, making the
crosstalk -74dB. At -31dB I measured 110mV and 0.4mV - 49dB difference.
But as these readings were made using a scope, the actual figure could
be better - it was difficult to see the signal at all...

Your cross-talk signal "drowns" in the noise.
There should a difference of preferably 20dB, or 10dB
at least, between noise floor and cross-talk.
Don't worry about input voltage, only output from the "other"
channel in reference to the output of the signal carrying
channel is important. Refer to my other post for "how to".

For a first measurement, results sound quite promising.


I repeated the same tests at 10kHz, where the figures were -69dB and
-43dB. Is that good or bad?

At least results show there is some coupling, due to
parasitic capacitance/inductance, between L and R.
The coupling might be due to use of DPDT relays as
L and R signals, by nature, are close together.
If you have perfected your instrumentatation set-up,
than compare SPDT to DPDT cross-talk figures, as,
in my opinion, it does'nt make sense to construct an
attenuator with perfectly equal steps, no wiper noise,
etc., and as a result gain "poor" channel separation.


I guess I should try to find a better meter
for more accurate measurements.

See above for a simple way....


Another parameter, relevant for the performance of the
attenuator, might be Power Supply Rejection Ratio.
For that, superimpose AC voltage on the relay DC supply
and determine PSRR in the same way as cross-talk.

As the powersupply for the relays is on the same board, that is a bit
difficult...

Depends on the layout of the board....
If the copper-tracks of the relay supply, at some point,
are "separate" from the IC supply than just cut that relay
supply track and superimpose AC with a transformer.
Measuring PSRR is just a way to determine how clean
the relay supply must be in order not to induce any
supply noise in the L/R channels, as, again by nature,
channels and relay coils are very close together.


After all, when I dug up the encoder, it turned out to
be Gray encoded i.s.o. BCD. So, I left it alone for a
while. But it's still itching. I'll report when I've got results.

So just use an EPROM as a Gray to BCD decoder...

Maybe, maybe not...
My interest is in the attenuators audio performance...
Not in logic, to test even relay control by switch works fine...


Regards,

Jan.



Best regards,

Mikkel C. Simonsen

JKoning wrote:

As Robert suggested already, terminate the input, of the
channel to be measured, with a resistor equal to the output
impedance of the preceeding stage.

If a normal op. amp. output is driving the input, then I guess that
would be around 100 Ohms. If a cathode follower is used it's probably in
the 300 Ohms to 1k range. I chose 680 Ohms which should be about the
same as my old HP oscillator.

Fine, but don't forget to terminate the "other channel" too.

One channel was terminated with the 680R resistor. The other channel
should be terminated by the output transformer of the signal generator I
guess.

The output load has to be correct to get the correct attenuation. In
this case the load should be 50k (same as the input impedance), so I
chose 56k (I didn't have any 51k).

Do I understand correctly the attenuator as shown at :
http://stiftsbogtrykkeriet.dk/~mcs/6bitatt.gif
is designed for termination with a 50K load?

Yes. Each stage in the attenuator "expects" a 50k load. That load is
provided either by the following stage or the 50k termination. The
output terminating resistor should be adjusted to suit the input
impedance of the following stage of course - or you could use a 51k
resistor at the input of the following tube amp

The attenuator will work just fine without the 50k termination, but the
attenuation will not be exactly as calculated. But as both channels will
be "off" by an equal amount that may not matter.

Your cross-talk signal "drowns" in the noise.
There should a difference of preferably 20dB, or 10dB
at least, between noise floor and cross-talk.
Don't worry about input voltage, only output from the "other"
channel in reference to the output of the signal carrying
channel is important. Refer to my other post for "how to".

For a first measurement, results sound quite promising.

I repeated the same tests at 10kHz, where the figures were -69dB and
-43dB. Is that good or bad?

At least results show there is some coupling, due to
parasitic capacitance/inductance, between L and R.
The coupling might be due to use of DPDT relays as
L and R signals, by nature, are close together.

Yes. There should be some coupling between the channels, but we'll need
more accurate measurements to know the exact amount...

If you have perfected your instrumentatation set-up,
than compare SPDT to DPDT cross-talk figures, as,
in my opinion, it does'nt make sense to construct an
attenuator with perfectly equal steps, no wiper noise,
etc., and as a result gain "poor" channel separation.

Then you can just use seperate relays for the two channels. In that case
you could parallel the contacts of the relays - but I guess that might
increase the capacitive coupling between the contacts decreasing the
attenuation at high frequencies...

Best regards,

Mikkel C. Simonsen

"Mikkel C. Simonsen" schreef in bericht
...

snip

Hi Mikkel,

Have you checked your website mail?

Regards,
Jan.