Replacing console fader amp with 2520 - circuit changes necessary?

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
JohnRoberts said:
abbey road d enfer said:
JohnRoberts said:
Input bias current is a non-issue for BiFET op amps. There is reportedly a subtle distortion mechanism associated with imbalanced DC resistance at + and - inputs.
With BiFET's? Never heard about this... Do you have a reference?

http://www.proaudiodesignforum.com/forum/php/viewtopic.php?f=6&t=703

http://www.ti.com/lit/an/slyt595/slyt595.pdf

Sorry for the veer... and I wouldn't lose too much sleep over this for your post fader gain stage.

There is a subtle affect related to input capacitance that changes with input voltage. Apparently the old BiFets (like TL07x) were susceptible to this mechanism.

JR
OK. I was aware of that. What made me scratch my head was "imbalanced DC resistance".
 
JohnRoberts said:
abbey road d enfer said:
living sounds said:
Soundcraft never put a resistor between fader wiper and op amp input since input DC offset is not a problem with a FET op amp like the TL071, right?
When Soundcraft designed the 500/600, they decided TL0's weren't good enough for post-fader amp, so they used 5534 in that position. They obviously were aware of the risks of noise due to bias current, so they put a capacitor between the wiper and the + input and a resistor to ground, but customers did complain because the faders were more noisy than the previous designs. They thought the faders were of poorer quality; some replaced them with P&G's, to no avail.  The cause was noise modulation. In BiFET's input noise current is almost inexistant but with bipolars it's significant. Varying the source impedance (which is what happens when you move the fader) changes the noise level. Although this variation is almost unmeasurable, it is clearly audible.

The .4-1.5 pA of noise current in a 5534 is all of 1-3 nV/rt Hz of noise contribution worst case (2.5k from 10k fader) .  This is in addition to the roughly 3k impedance at the - input.

This might (?) be audible in a WFO listening test (I try not to argue about what other people hear),  but I would expect other noise sources in that complete audio path to dominate.

JR
That's what I said: "almost unmeasurable". But there was a clear emergence when moving the fader.
 
abbey road d enfer said:
JohnRoberts said:
abbey road d enfer said:
JohnRoberts said:
Input bias current is a non-issue for BiFET op amps. There is reportedly a subtle distortion mechanism associated with imbalanced DC resistance at + and - inputs.
With BiFET's? Never heard about this... Do you have a reference?

http://www.proaudiodesignforum.com/forum/php/viewtopic.php?f=6&t=703

http://www.ti.com/lit/an/slyt595/slyt595.pdf

Sorry for the veer... and I wouldn't lose too much sleep over this for your post fader gain stage.

There is a subtle affect related to input capacitance that changes with input voltage. Apparently the old BiFets (like TL07x) were susceptible to this mechanism.

JR
OK. I was aware of that. What made me scratch my head was "imbalanced DC resistance".

Supposedly the distortion nulls or is minimized for equal resistance at + and - inputs.

JR
 
abbey road d enfer said:
.. but customers did complain because the faders were more noisy than the previous designs. They thought the faders were of poorer quality; some replaced them with P&G's, to no avail.  The cause was noise modulation. In BiFET's input noise current is almost inexistant but with bipolars it's significant. Varying the source impedance (which is what happens when you move the fader) changes the noise level.
But you WILL have noise changing when you move the fader.

A 10k fader will present 2k5 at -6dB ... and 0r at -zillion or 0dB (assuming its fed from LoZ).  And of course as its opened up, the noise from previous sections will be louder.

Are you saying the 5534 noise current adds sufficient extra noise in this varying source R to be audible above the 'usual' varying fader noise and other stuff?

IMHO, the operating levels in the desk would have to be serious wonky for that to be a problem.

My (1980's Calrec) experience is that 5534 was plenty good to buffer a fader.

So I side with JR in this  :)

The TL07x problem was known in the early 80's to some.  I wrote an internal note about this at Calrec .. mainly about TL07x vs 553x
 
abbey road d enfer said:
JohnRoberts said:
Supposedly the distortion nulls or is minimized for equal resistance at + and - inputs.

JR
Not resistance, impedance.

Offset is "minimized" with matching resistance, or at least made more stable, you could compensate for the offset changing the resistance at one input (by a small amount) but it will be unstable with temp, since bias current changes. Then you could compensate for it in the usual way with opamps which have the feature, if you are making your own DOA you could trim that at "production" selecting the resistor or with (as JR loves) a trimpot.

JS
 
ricardo said:
abbey road d enfer said:
.. but customers did complain because the faders were more noisy than the previous designs. They thought the faders were of poorer quality; some replaced them with P&G's, to no avail.  The cause was noise modulation. In BiFET's input noise current is almost inexistant but with bipolars it's significant. Varying the source impedance (which is what happens when you move the fader) changes the noise level.
But you WILL have noise changing when you move the fader.
Indeed.
A 10k fader will present 2k5 at -6dB ... and 0r at -zillion or 0dB (assuming its fed from LoZ).  And of course as its opened up, the noise from previous sections will be louder.
Customers noted the issue with Mute On, which would disconnect the previous stages. I know it's not a realistic test, but it's what happens.
Are you saying the 5534 noise current adds sufficient extra noise in this varying source R to be audible above the 'usual' varying fader noise and other stuff?
Yes. The emergence may be very low, when it is repeatedly correlated with fader movements, it becomes an evidence.
IMHO, the operating levels in the desk would have to be serious wonky for that to be a problem.
Operating level in the 500/60 is -6dBu, which proved to be a practical issue, not only in terms of noise, but also in terms of interfacing with outboard.
My (1980's Calrec) experience is that 5534 was plenty good to buffer a fader.
Indeed it is, but there are arguments for BiFET opamps, especially with the existence of much better than TL0, such as OPA134.
 
abbey road d enfer said:
Indeed it is, but there are arguments for BiFET opamps, especially with the existence of much better than TL0, such as OPA134.
As OPA134 has more than 6dB extra voltage noise than 5534, this should hide any fader noise modulation  ;)

I haven't done the detailed sums but you probably need something like OPA627, AD745 to improve on 5534 for this application.  These also win on $$$ and 'hand carved from solid Unobtainium by virgins' cred.  8)
 
ricardo said:
abbey road d enfer said:
Indeed it is, but there are arguments for BiFET opamps, especially with the existence of much better than TL0, such as OPA134.
As OPA134 has more than 6dB extra voltage noise than 5534, this should hide any fader noise modulation  ;)
The point is that with OPA134, there would be NO noise modulation, all the more in regard of the increased steady noise. It's a little similar to the debate over class-B VCA-s vs. class-A. Although class-A VCA's produced more constant noise at moderate gain, they didn't exhibit the noise vs. gain variation of class-B, which was deemed unacceptable by euphonists. The debate is moot today, since classA VCA's are an extinct species, to my knowledge. And anyway, digital mixing has made it pointless. 
I haven't done the detailed sums but you probably need something like OPA627, AD745 to improve on 5534 for this application. 
At the time, there were not much practical choices, so experimentation was limited. In fact, I don't remember having had ANY choice; I guess AD743 would have bettered the 5534,  but its cost was prohibitive - probably didn't exist at the time.
 
joaquins said:
abbey road d enfer said:
JohnRoberts said:
Supposedly the distortion nulls or is minimized for equal resistance at + and - inputs.

JR
Not resistance, impedance.

Offset is "minimized" with matching resistance, or at least made more stable, you could compensate for the offset changing the resistance at one input (by a small amount) but it will be unstable with temp, since bias current changes. Then you could compensate for it in the usual way with opamps which have the feature, if you are making your own DOA you could trim that at "production" selecting the resistor or with (as JR loves) a trimpot.

JS

No Abbey is correct, the phenomenon is AC so the "impedance" at both inputs is what matters not DC resistance. Dealing with DC offsets is a completely different issue.

The phenomenon is an input device associated capacitance that changes with voltage inputs so matching "impedance" keeps the effect better behaved .

Again sorry about the veer...

JR
 
abbey road d enfer said:
ricardo said:
abbey road d enfer said:
Indeed it is, but there are arguments for BiFET opamps, especially with the existence of much better than TL0, such as OPA134.
As OPA134 has more than 6dB extra voltage noise than 5534, this should hide any fader noise modulation  ;)
The point is that with OPA134, there would be NO noise modulation, all the more in regard of the increased steady noise. It's a little similar to the debate over class-B VCA-s vs. class-A. Although class-A VCA's produced more constant noise at moderate gain, they didn't exhibit the noise vs. gain variation of class-B, which was deemed unacceptable by euphonists. The debate is moot today, since classA VCA's are an extinct species, to my knowledge. And anyway, digital mixing has made it pointless. 
I haven't done the detailed sums but you probably need something like OPA627, AD745 to improve on 5534 for this application. 
At the time, there were not much practical choices, so experimentation was limited. In fact, I don't remember having had ANY choice; I guess AD743 would have bettered the 5534,  but its cost was prohibitive - probably didn't exist at the time.

OK, going completely down the rabbit hole, I like to joke that console design is one of the most complicated simple circuits to design. Superficially all of the circuit blocks are relatively simple and manageable, but the devil is in the details.

The scenario that Abbey is talking about, falls under what I call console ergonomics, or how control behavior like gain laws (or noise change) influences consumer perception.  A 10 db gain stage left in place after the channel mute might be expected to exhibit a very faint noise that increases with fader level, but instead as a consequence of the faders source impedance interacting with bipolar input noise current, it rises to a peak at 50% fader resistance (actually 50% in parallel with 50%), then falls again.  In use with the channel un-muted this noise profile would never be heard underneath the louder input strip noise floor, but once the tire-kicker gets this in his craw he will not be happy with the design. As I think I mentioned earlier, a few K resistor in series with the wiper will reduce how quiet this gets during the quiet part of the fader travel and would probably be more acceptable to a finicky customer than the literally quieter design.  (another tweak might be putting the mute switch at the wiper to ground but mutes often have to kill  several signal feeds and electronic mutes could leak switching noise if applied closer to the gain stage... See it isn't ever simple.)

These kind of customers would drive me crazy, and there is a great deal of engineering effort invested into managing these subtle user interface perceptions.  Tiny things like how an EQ pot responds. Scratchy pots and switch clicks are obvious failures. At Peavey we had a special house part number for a selected 5532 that was graded for input noise. But this noise spec was not a simple noise maximum but a specified acceptable noise frequency response contour or ratio between the mid band noise and the LF (1/F) noise. This was purely to address customer perceptions about the sound of a mixer noise floor when listening with the gain WFO ("wide full open"  or something like that  ;D)... A quieter op amp can always be turned up loud enough to hear it's noise floor. If that noise floor is dominated by 1/F noise it sounds like something is broken or about to fail, while the whiter noise floor sounds normal and what you expect noise floors to sound like.

I spent decades pondering such subtle details and am glad I don't have to any more. It's not just about making the design better, but managing the perception of better.

JR

 
I just wanted to follow up on this:

The 2520 wasn't working properly, but I got it to work like it should, and out of the 16 I built only one still has gain lower (between ca. 1 db and 2 db, it changes sometimes) and higher distortion than it should. It's a great thing to be able to put a discrete op amp in the position of a DIP op amp like this. Sounds wonderfull, and THD is lower than with a 5534.

John's tip about putting the 75 ohm resistor after the fader wiper was spot on, works flawlessly. I did not like what electrolytic caps did in this position, even with added bias. A 4.7uf poly cap sounded good, but it's too low in capacitance and already too large in size.

Thanks again to everyone.  8)
 

Latest posts

Back
Top