Summing amp?

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
The VCA-channel is worse on paper in the addition of the cap to ground (C78), creating low end loss and additional phase shift there... I guess it also eliminates some DC... Why would he do is on one channel and not on the other?
 
I wrote about the transamp as a common base sum amp in my 40 year old console article, but the transamp was not 1nV/rt HZ.

JR

PS sorry about your keyboard....

When I first read that article many years ago I focused on the current output "noise-less" approach you suggested.
Engineering the last whisper of noise out of a sum bus amp is what engineers do for entertainment.
Now I realize you were that engineer. LOL.

I didn't particularly pay attention to the Trans-Amp summing topology until now. You wrote:

Summing Amplifiers

Once again the Trans-Amp [current feedback] topology has merit. Grounding the input and connecting the buss directly to the emitter of the input gain stage reaps the same open loop to closed loop gain tracking with the possible qualification that the Trans-Amp will not excel when a small number of inputs are being summed the Trans-Amp will outperform the previous examples for large buss structures. [comment added]
Roberts_Console_Design_Figure_13.JPG
I wasn't sure what you were previously referring to when you wrote "gain tracking." I then found the following:

Microphone Preamplifiers

The closed loop gain of the circuit in Figure 10 is 1+(RF/RG).
The open loop gain becomes: Gol = Aol + Rol(1/RG+1/RF).

If we vary RG to change our closed loop gain, the open loop gain changes also since RG is common to both equations.This topology is capable of very large ±160 dB open loop gains since the feedback factor equal Rol/RF the amplifier is stable at all gain settings.

Note that Rol in both figures is the collector load resistor and that Aol is the op amp open loop gain.

Roberts_Console_Design_Figure_10.JPG

So it looks like the advantage of using a common base current feedback topology is that the open loop gain increases with the noise gain. Or, to put it another way, the open loop gain increases as RG, the summing resistor value, is lowered.
 
Forgot to mention in my post that the benefit is more corrective feedback and the potential for reduced distortion, particularly at high frequencies where the open loop gain falls with rising frequency.

Correct me if I'm wrong but:

The open loop gain becomes: Gol = Aol + Rol(1/RG+1/RF)
Applies when gain is expressed in dB.
If not in dB it would be Gol = Aol * Rol(1/RG+1/RF)?
 
Forgot to mention in my post that the benefit is more corrective feedback and the potential for reduced distortion, particularly at high frequencies where the open loop gain falls with rising frequency.
+1... improved loop gain margin corrects several flaws of open loop transfer function. Distortion is one obvious one.

Another is that the open loop transfer function of op amps are effectively one pole integrators, (i.e. 90' phase shift). The dominant pole in 5534 starts in the low hundreds of Herts.

Of course negative feedback reduces all open loop errors to below significance as long as there is adequate loop gain margin. High closed loop noise gain diminishes that ability of NF to provide linear output.

JR

PS: when op amps are used at silly high noise gains, it is possible that differences in the open loop transfer function of different op amps might become audible.
 
Odd that the more complex route should have lower distortion. OTOH it may be that trimming the VCA cancels some distortion from the previous stage leading to an overall reduction in harmonics distortion.. This is a technique that was often used on early tube mice pres like the REDD 47.

Cheers

Ian
Are you writing about local NFB at the output, global NFB, maybe 2nd coupling cap value?
 
Are you writing about local NFB at the output, global NFB, maybe 2nd coupling cap value?
I was referring to designs where the distortion in one stage is deliberately at least partially cancelled by equal but inverse polarity distortion in the next stage.

Cheers

Ian
 
So it looks like the advantage of using a common base current feedback topology is that the open loop gain increases with the noise gain. Or, to put it another way, the open loop gain increases as RG, the summing resistor value, is lowered.
Another advantage is that because the open and closed loop gains track, the amount of NFB is almost constant which greatly simplifies maintaining stability. This technique does not have to be limited to VE summers; it is also very useful where you want to vary the gain over a wide range as in a mic pre for example.

Cheers

ian
 
Another advantage is that because the open and closed loop gains track, the amount of NFB is almost constant which greatly simplifies maintaining stability. This technique does not have to be limited to VE summers; it is also very useful where you want to vary the gain over a wide range as in a mic pre for example.

Cheers

ian
The "Transamp" was designed and marketed as a mic preamp. The sum amp application was shared in their application notes as something extra you could use it for (I never did that either).

Paul Buff was a very smart guy (RIP).

JR
 
Another advantage is that because the open and closed loop gains track, the amount of NFB is almost constant which greatly simplifies maintaining stability. This technique does not have to be limited to VE summers; it is also very useful where you want to vary the gain over a wide range as in a mic pre for example.

Cheers

ian
Ian - You may have missed the post where I show the SSM2017 re-purposed as a balanced summing amp.

The "Transamp" was designed and marketed as a mic preamp. The sum amp application was shared in their application notes as something extra you could use it for (I never did that either).

Paul Buff was a very smart guy (RIP).

JR

Trans_Amp_Frequency_Response_vs_Gain.JPG

Checkout the noise level and NF for 100 channels summed using 10KΩ:

Trans_Amp_Balanced_Summing_Amp.JPG
Dynamic range with ±15V supplies and 100 inputs is about 112 dB for a 20 kHz measurement BW. (RThevenin=200Ω, noise gain 40 dB.)

Paul mis-spelled "buss" too.

I have a collection of Valley People docs I pulled these from: Valley People TransAmp, ECG-101 VCA, TA-103 Documents - Pro Audio Design Forum
 
Last edited:
Problem solved. :cool: It turned out there was insufficent local decoupling. The negative supply got a small, bad electrolytic (thanks again Ricardo, for this piece of vital advice in the big decoupling thread) and now distortion is as low as the VCA chain and everything sounds and measures all right regardless of chain or op amp manufacturer.

I found this out with the help of a Signetics NE5534 that showed much higher distortion than the other NE5534 and a TL071 that suspicously worked better than the designated op amps.

Now all I have to do is apply the fix to every channel. :p
 
The negative supply got a small, bad electrolytic (thanks again Ricardo, for this piece of vital advice in the big decoupling thread)
The thread seemed a bit vague to me as to what exactly is a dirty ground iirc ...Any chance you can link to where that explains it clearly or maybe some general description?
Also, are these grounds generally very close by or are we talking running jumpers sometimes?
 
No, the console allows you to switch signal paths. In the default position the rec channel goes through the smaller fader (no VCA) and the tape channel uses the big VCA fader. Press the FDR button and it gets switched around.

View attachment 115331

Looking at this, I'd be tempted to replace the everything between fader wiper and op-amp with a wire and swap the 5534 to a (J) FET input amp (OPA828?).
Pot's and faders really, really, really dislike the wiper loaded, unless they use Cermet (metal film) or wire wound tracks.

The variable base current of the 5534 input will cause H2, while general loading tends to increase H3. They might also interact with the electrolytic cap, but the relative impedance's are very high, so that is unlikely.

I'm seeing 235k wiper load, which is not terrible, but switching it up to GOhm is likely a good idea.

If you insist on having a resistor on the wiper - bootstrap it.

It may also be worth to clean switches and check if relays have gone bad. All these can increase HD.

Thor
 
The thread seemed a bit vague to me as to what exactly is a dirty ground iirc ...Any chance you can link to where that explains it clearly or maybe some general description?
Also, are these grounds generally very close by or are we talking running jumpers sometimes?
I think there a descriptions in that thread and others about hierarchical grounding structures and how to look at grounds correctly. I also think the experts did agree that it gets harder to apply these insights in something as complex as a large format console. Me, I'm not one of the experts, I only try to apply their knowledge and wisdom.

That said, when I added dirt grounds, like in the GSSL, I used thick copper wire underneath the PCB that connect to the decoupling caps as a net seperate from the audio ground, only joining the grounds at a point near the regulator ground. It might also be good practice to couple audio ground and dirt ground together at points far away from the place where the grounds actually join with a small cap, I think John explains the rationale for it somewhere in the grounding thread.

My most important takeaway was Ricardo's advice to use "bad" high ESR electrolytics for decoupling to counter trace inductance and prevent resonances in the PSU impedance response.

My console has no seperate "dirt" grounds. There's an external linear PSU supplying DC, which then gets dropped a few volts by linear regulators on every channel/module. All the grounds go to the bottom of each module and are connected by a steel bar on every module and jointly go to the bus bar, which is made of lot's of copper. There is no problem with hum in this console. Also not very much PSU decoupling in general. The 10uf high ESR electrolytic I added to the positive rail a the op amp in question went to a ground nearby, and I saw no negative effects from it (no visible hum/noise increase). It certainly fixed the stability problem.
 
Last edited:
Looking at this, I'd be tempted to replace the everything between fader wiper and op-amp with a wire and swap the 5534 to a (J) FET input amp (OPA828?).
Pot's and faders really, really, really dislike the wiper loaded, unless they use Cermet (metal film) or wire wound tracks.

The variable base current of the 5534 input will cause H2, while general loading tends to increase H3. They might also interact with the electrolytic cap, but the relative impedance's are very high, so that is unlikely.

I'm seeing 235k wiper load, which is not terrible, but switching it up to GOhm is likely a good idea.

If you insist on having a resistor on the wiper - bootstrap it.

It may also be worth to clean switches and check if relays have gone bad. All these can increase HD.

Thor
I'm still wondering what the designer was thinking when he layed out these schematics. Why the differences between chains when they can be switched around?

Switches have been cleaned and I have replaced every single reed relay. I know these can cause distortion even when they're working properly...
 
Last edited:
That said, when I added dirt grounds, like in the GSSL, I used thick copper wire underneath the PCB that connect to the decoupling caps as a net seperate from the audio ground, only joining the grounds at a point near the regulator ground. It might also be good practice to couple audio ground and dirt ground togather at points far away from the place where the grounds actually join with a small cap, I think John explains the rationale for it somewhere in the grounding thread.

My most important takeaway was Ricardo's advice to use "bad" high ESR electrolytics for decoupling to counter trace inductance and prevent resonances in the PSU impedance response.
Thanks!
Yeah it was pivitol when I was reading it years ago. Damn Kingston.... Too bad I read it after years of "modding" some gear. In haste, I just started removing all the ceramic bypasses I had + to ground and - to ground and just put them + to - until I could understand more of the dirty ground thing to get into the electros . Many things are still sitting unchanged in lieu of building.
 
Thanks!
Yeah it was pivitol when I was reading it years ago. Damn Kingston.... Too bad I read it after years of "modding" some gear. In haste, I just started removing all the ceramic bypasses I had + to ground and - to ground and just put them + to - until I could understand more of the dirty ground thing . Many things are still sitting unchanged in lieu of building.
I tried that in my console - rail-to-rail decoupling, and it seems the regulators didn't like it. It made performance worse. In some DIY gear I got rid of ceramic bypass caps altogether, or only used them with a small series resistor, and always in addition to the high ESR electrolytic.

Also interesting: Stability problems got worse with small film caps added to the signal path electrolytics in parallel. My best practise is now to always use quality bipolars in the signal path (again, see Cyril Bateman's measurements) rather than non-polarized electrolytics and not to add film caps in parallel.
 
I half joke that consoles are the most difficult simple designs... Looking at individual circuit blocks are relatively simple in isolation but combining hundreds of these blocks inside a single chassis can result in odd interactions.

Good luck and do what works...

JR
 
My best practise is now to always use quality bipolars in the signal path (again, see Cyril Bateman's measurements) rather than non-polarized electrolytics and not to add film caps in parallel.
Cyril's measurements were all made with 3V rms across the capacitor. A properly sized signal path capacitor never sees a voltage across it anywhere near that. Doug Self's much more recent measurements provide a more accurate picture.

Cheers

Ian
 
Cyril's measurements were all made with 3V rms across the capacitor. A properly sized signal path capacitor never sees a voltage across it anywhere near that. Doug Self's much more recent measurements provide a more accurate picture.

Cheers

Ian
https://linearaudio.nl/cyril-batemans-capacitor-sound-articles
See document Nr. 5- Comparative test results for 1uF electrolytic and film capacitors

Even with no bias applied the polarized electrolytics were doing far worse than the nonpolar electrolytics. And with significant bias applied this still held true.

Quotes CB from the aformentioned document:

"In every distortion test, the Bi-polar capacitor produced much lower distortions than were measured on similar value and voltage polar capacitors."

"Why do designers use polar electrolytic capacitors in the signal path of an amplifier?"
 

Latest posts

Back
Top