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.
"Why do designers use polar electrolytic capacitors in the signal path of an amplifier?"
I have used truckloads of polar electrolytic capacitors in the signal path of audio circuits,,, mainly because they are cheap. If the poles are tuned low enough below the audio pass band there will be almost no terminal voltage across the capacitors.

For audio poles inside the audio pass band I prefer to use film caps. In value products polyester caps don't suck too bad.

Most mic preamps use polar phantom voltage blocking capacitors.... sorry.

JR
 
I have used truckloads of polar electrolytic capacitors in the signal path of audio circuits,,, mainly because they are cheap. If the poles are tuned low enough below the audio pass band there will be almost no terminal voltage across the capacitors.

For audio poles inside the audio pass band I prefer to use film caps. In value products polyester caps don't suck too bad.

Most mic preamps use polar phantom voltage blocking capacitors.... sorry.

JR
The tests in the linked documents showed significantly higher distortion for 1 khz test signals with polarized vs. unpolarized electrolytics, with no DC present.
 
MCI used bipolars and SSL used two polars in series with a biasing resistor at the midpoint. HOWEVER, most desks (with IC opamps and bipolar PSUs) used regular polar 'lytics. One would tend to think that after the audio passed through probably dozens on a large desk, the distortion should be....say....33 1/3%. <g>

Bri
 
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?"
At the risk of repeating myself, Cyril's tests were conducted with significant ac voltage across the capacitor. Even then, the distortions measured are of the order of 0.001%. A properly sized coupling capacitor will have an ac voltage at least 10 times less than Cyril used so you can stick another zero in there.

There is no disputing that film capacitors produce less distortion than electrolytics. Nothing new there. Wouldn't touch them in an EQ circuit where there are significant ac voltages across the capacitors. The other problem with electrolytics is leakage current which makes them useless for tube circuits where they have to sustain very high dc voltages. I use 10uF 250V film capacitors as the output coupling capacitor in my tube preamps. Have you see the size of those things?

Cheers

Ian
 
Also, something that is (easily) overlooked is that EVERY stage that has some amplification (each op amp section for example) amplifies the DIFFERENCE between what it 'thinks' is two input terminals, one of which may be a 'ground' and usually the more obvious 'input' signal so as John almost said, a mixering desk is one hell of a complicated collection of interacting circuit elements where the 'complexities' of a single channel mic (or any amplifier stage) are realities that exist on a different planet. Anyone can assemble a single mic amp but even then connecting it up to the next stage and a power amp so you can hear the supposed quality is already a game of luck unless the input to the 'next' stage is completely 'balanced' (high CMR at ALL frequencies) and is actually connected to the correct points of your simple mic amp stage. Bear in mind that any conducting path more than a few inches is a good aerial at microwave frequencies so since we all like to carry a microwave transmitter in our pockets this fact and measures to prevent demodulation need to be factored into any design or just pure luck! The world is a complicated place and too many people try to simplify things and expect to be able to throw some money at 'resolving' something that has probably not been properly evaluated whether it is a cause OR effect. Dirty ground is simply a conductive path that carries a signal that you prefer not to have as part of your final output.
 
At the risk of repeating myself, Cyril's tests were conducted with significant ac voltage across the capacitor. Even then, the distortions measured are of the order of 0.001%.
Sorry, not true. For instance, even at 0.2V AC distortion reached 0.0026 with no DC bias in a test. My point is, with a bipolar cap you're automatically at less than half the distortion of the polarized cap, so that's what I'll use.

I use 10uF 250V film capacitors as the output coupling capacitor in my tube preamps. Have you see the size of those things?
I use those in my tube compressors. :)
 
Also, something that is (easily) overlooked is that EVERY stage that has some amplification (each op amp section for example) amplifies the DIFFERENCE between what it 'thinks' is two input terminals, one of which may be a 'ground' and usually the more obvious 'input' signal so as John almost said, a mixering desk is one hell of a complicated collection of interacting circuit elements where the 'complexities' of a single channel mic (or any amplifier stage) are realities that exist on a different planet. Anyone can assemble a single mic amp but even then connecting it up to the next stage and a power amp so you can hear the supposed quality is already a game of luck unless the input to the 'next' stage is completely 'balanced' (high CMR at ALL frequencies) and is actually connected to the correct points of your simple mic amp stage. Bear in mind that any conducting path more than a few inches is a good aerial at microwave frequencies so since we all like to carry a microwave transmitter in our pockets this fact and measures to prevent demodulation need to be factored into any design or just pure luck! The world is a complicated place and too many people try to simplify things and expect to be able to throw some money at 'resolving' something that has probably not been properly evaluated whether it is a cause OR effect. Dirty ground is simply a conductive path that carries a signal that you prefer not to have as part of your final output.
In my console, I conducted more tests with a few added high ESR decoupling caps in places furthest removed from the PSU and other decoupling caps, and these made a very obvious difference. So what's the risk of these injecting noise into the signal that actually hampers op amp performance (besides possibly raising the noise floor)? In one of those places (small sub-pcb with input / channel amps) according to the schematics the next generation of my console had these caps in place by design, so I think those are safe. The other one was at the buffer amp where distortion was clearly lowered by the addition of decoupling on the negative PSU rail.
 
Remember that a 'difference' is not necessarily what other users want to hear. The original design, components used and layouts were using what was available at the time and no one spent days analysing every individual stage to check that it performed to it's theoretical best but simply put a load of sub circuits together which then resulted in the product that was sold. Remember that the outputs of op amps and (SSM2017 series devices need to 'see' essentially resistive 'loads' to maintain stability, so yes the 50pF 'limit' as mentioned on the data sheet for a SSM2017 IS very important.
A 'ground loop' OUTSIDE a piece of gear can inject hum into interesting places such as fader automation but it does not appear in the 'audio' path but makes the automation run 'badly'. Known affectionately as 'twinkling faders' in France!
 
I wonder if (marginal) stability problems are actually caused in this console by excessive decoupling with small parallel caps. The main (outboard) PSU has film caps added after the fact in parallel to the electrolytics by the service engineer, contrary to the schematic. And the per-channel regulation has small ceramics in parallel at the input as well as the output of the regulator (below).

From measurements I have seen, the combination of modern electrolytics with small low-ESR caps (ceramic or film) can (and often does) produce antiparallel resonances in the impedance response. Example: paralleling film caps with electrolytic caps

I don't think I've ever seen a regulator datasheet that calls for parallel small caps to the electrolytic on the input of the regulator, and I seem to recall to have read more than once that it would not only be unnessary but counterproductive.



Channel_PSU.jpg
 
Last edited:
I shared this story before, but back in the 1980s I had a console (of my design) end up installed less than one mile away, inside the main beam of a AM radio broadcast tower. I found my "clean ground" swinging 1V p-p wrt the dirty power ground. I vividly remember that the radio station was broadcasting at 960kHz. :rolleyes: I stopped the singing ground by coupling the two grounds together with low impedance at HF ceramic disc caps, at multiple places inside the console.

My console wasn't the only RF problem that studio encountered.
====
I wouldn't expect regulator stability to be impacted much by input capacitance while low impedance at HF should be desireable.

Output capacitance is another story, so read the manufacturers data sheet/application notes. I have encountered subtle differences between different brands of even the same regulator PNs.

JR
 
I wouldn't expect regulator stability to be impacted much by input capacitance while low impedance at HF should be desireable.

Output capacitance is another story, so read the manufacturers data sheet/application notes. I have encountered subtle differences between different brands of even the same regulator PNs.

JR
The question is at which HF and how the different caps interact. Furthermore, different brands, ages, size of electrolytics and even film or ceramics may vary significantly in the parameters relevant for impedance.

From what I understand adding the small parallel cap may have the effect of, for example, lowering impedance in a range so high that it doesn't matter at all for the circuit in question, while at the same time lowering it further down (creating a resonance) where stability of the regulator or op amps are impacted.
 
The question is at which HF and how the different caps interact. Furthermore, different brands, ages, size of electrolytics and even film or ceramics may vary significantly in the parameters relevant for impedance.

From what I understand adding the small parallel cap may have the effect of, for example, lowering impedance in a range so high that it doesn't matter at all for the circuit in question, while at the same time lowering it further down (creating a resonance) where stability of the regulator or op amps are impacted.
Prudent design often involves modest resistors in series with op amp output to protect stability in the presence of capacitive loads. One popular design where negative feedback is taken from after the build out resistor results in HF peaking due to excessive capacitive loading.
===
Speaking of 3 terminal regulators back several decades ago I opined that the active circuitry inside cheap 3 terminal regulators was 741 era performance so not very much gain bandwidth. Since the low output impedance of these regulators depends on NF and loop gain margin, the old 3 terminal regulators have rising impedance at HF (generally above human hearing).

I found from bench testing that the falling impedance of a 1,000uF aluminum electrolytic, nicely complemented the rising regulator impedance. In combination this delivered nice mOhm source impedance to at least a couple octaves above the audio band.

The big electrolytic capacitors still have some internal ESL and that inductance protected against instability.

Caveat, this bench testing was done several decades ago so modern parts are likely to behave differently. YMMV

JR
 
Prudent design often involves modest resistors in series with op amp output to protect stability in the presence of capacitive loads. One popular design where negative feedback is taken from after the build out resistor results in HF peaking due to excessive capacitive loading.
I think every single 5532/5534 in the console has a 10 ohm resistor on the output, always as part of the feedback loop, as pictured in the schematics in posts 1 and 34. Are you saying this may be detrimental?

===
Speaking of 3 terminal regulators back several decades ago I opined that the active circuitry inside cheap 3 terminal regulators was 741 era performance so not very much gain bandwidth. Since the low output impedance of these regulators depends on NF and loop gain margin, the old 3 terminal regulators have rising impedance at HF (generally above human hearing).

I found from bench testing that the falling impedance of a 1,000uF aluminum electrolytic, nicely complemented the rising regulator impedance. In combination this delivered nice mOhm source impedance to at least a couple octaves above the audio band.

The big electrolytic capacitors still have some internal ESL and that inductance protected against instability.

Caveat, this bench testing was done several decades ago so modern parts are likely to behave differently. YMMV

JR
The console only uses decades old 3 terminal regulators, no modern low drop parts. I think I'll try and see what removing small caps can do, starting with the PSU.
 
I think every single 5532/5534 in the console has a 10 ohm resistor on the output, always as part of the feedback loop, as pictured in the schematics in posts 1 and 34. Are you saying this may be detrimental?
That depends on how much capacitance to ground it sees at that output node. If you look closely at the topology with the added capacitance to ground, it resembles a common 2 pole active LP filter, but it is under damped, which means that the HF response rises before it falls off. In typical console applications there is not enough capacitance hanging on internal signal nodes to be a problem.
The console only uses decades old 3 terminal regulators, no modern low drop parts. I think I'll try and see what removing small caps can do, starting with the PSU.
I do not advocate for "what if" circuit design. Perhaps start with conventional old school bench measurements. If there are HF peaking issues, that will show up in a simple frequency response sweep. If the frequency response is well behaved, relax.

I can measure stuff that I can not hear... I could always measure things that I heard. Have you identified what you are hearing?

JR
 
That depends on how much capacitance to ground it sees at that output node. If you look closely at the topology with the added capacitance to ground, it resembles a common 2 pole active LP filter, but it is under damped, which means that the HF response rises before it falls off. In typical console applications there is not enough capacitance hanging on internal signal nodes to be a problem.
I simulated that fader buffer, and HF doesn't rise in the simulation. There are 2nd order Butterworth filters at some other points in the console (two series resistors before the non-inverting input with a cap between them going to the inverting input) that do create resonances with the values originally used (but not in the schematic, so I changed the caps in the console to values that gave a flattened response).


I do not advocate for "what if" circuit design. Perhaps start with conventional old school bench measurements. If there are HF peaking issues, that will show up in a simple frequency response sweep. If the frequency response is well behaved, relax.

I can measure stuff that I can not hear... I could always measure things that I heard. Have you identified what you are hearing?

JR

In the audio band FR looks fine, and with all the bandwith limitation there isn't very much to measure far beyond the audio band in the console. I couldn't find anything of note with the scope and signal generator.
 
The late Bob Pease, in his book "Troubleshooting Analog Circuits" (a truly excellent read BTW) had some wise words regarding the output capacitors on 3-terminal voltage regulators on pages 135 and 136:
"But in all cases, on all the parts [3-terminal regulators] I know, an electrolytic capacitor will work, and a film or ceramic capacitor won't work-its series resistance is just too small. Now, if you put a 1 Ω resistor in series with a 1 µF ceramic capacitor, the filtering will probably be adequate around room temperature; the loss factor is then similar to a tantalum capacitor. But if you take it to -40 or +100 °C, the ceramic capacitor's value will shrink badly (refer to Chapter 4 on capacitors) and the regulator will be unhappy again. It may start oscillating, or it might just start ringing really badly."

Remember, this was back in the day when ordinary electrolytic caps has ESRs of an ohm or so - not the low ESR parts currently common in switching regulators. Also realize that the output capacitor is the loop compensation for the feedback/gain in the regulator. Technically, it needs a "zero" in the compensation - normally provided by the ESR of that output capacitor. So putting a film or ceramic capacitor across the electrolytic (as so many audiophiles are fond of doing to make the caps "better") can be a serious mistake at the output side of a 3-terminal regulator. And Bob Pease would know, he designed many of the analog parts for National Semiconductor that have become long-lived standards in the industry, like the LM7800 series!
 
The late Bob Pease, in his book "Troubleshooting Analog Circuits" (a truly excellent read BTW) had some wise words regarding the output capacitors on 3-terminal voltage regulators on pages 135 and 136:
"But in all cases, on all the parts [3-terminal regulators] I know, an electrolytic capacitor will work, and a film or ceramic capacitor won't work-its series resistance is just too small. Now, if you put a 1 Ω resistor in series with a 1 µF ceramic capacitor, the filtering will probably be adequate around room temperature; the loss factor is then similar to a tantalum capacitor. But if you take it to -40 or +100 °C, the ceramic capacitor's value will shrink badly (refer to Chapter 4 on capacitors) and the regulator will be unhappy again. It may start oscillating, or it might just start ringing really badly."

Remember, this was back in the day when ordinary electrolytic caps has ESRs of an ohm or so - not the low ESR parts currently common in switching regulators. Also realize that the output capacitor is the loop compensation for the feedback/gain in the regulator. Technically, it needs a "zero" in the compensation - normally provided by the ESR of that output capacitor. So putting a film or ceramic capacitor across the electrolytic (as so many audiophiles are fond of doing to make the caps "better") can be a serious mistake at the output side of a 3-terminal regulator. And Bob Pease would know, he designed many of the analog parts for National Semiconductor that have become long-lived standards in the industry, like the LM7800 series!
Very interesting.

I have learned the hard way not to use low-ESR caps in audio circuits.

But I do wonder why so many commercial designs seem to ignore crucial lessons in good practice. Rule-of-thumb parallel decoupling schemes with electrolytics and ceramics are ubiquitous, and what I was able to find from people who have actually measured and looked deeper into the circuit interactions, they rarely make any sense for audio circuitry.

I did remove the 10 ceramic decoupling caps on a channel in my console and this channel had lower harmonics at the higher end of the spectrum than the others. I do have to test it with at least another channel yet to find out if it was actually the cause. It did sound clearer and less dynamically constricted than the others.
 
I'm not a fan of onboard regulators in mixers, particularly 78/79's. They inject noise in the "ground" path that can too easilly pollute the whole system. My first mixern when I was wet behind the ears, had a 7824 on each channel (yes, no bipolar PSU at the time). I was not pleased with the noise performance.One day, as an experiment, I replaced the 7824 with a 10 ohm resistor. The noise improvement of the summing amp was significant. After that, no more on-board regs. Saves heat too.
I never tried with 317/337 though.
 

Latest posts

Back
Top