Feedback capacitor to minimize DC offset

[Fred Dieckmann]

ESR is not a linear resistance and low ESR is usually one of the indicators of capacitor quality. Does anyone else actually research these things or is just throwing opinions out adequate effort for the forum? Many of the things that seem to be controversial have been known for a decade or more. Spend 10 or 15 years trying some of this stuff and I will happy to discuss honest opinions based on experience.

Jocko and I both came into high end audio with healthy skepticism and degrees and work experience as EEs. Nobody has asked my degree back for trying many the mods discussed on the forum, even some that conflict with first order engineering. Looking into subtleties has made me a better engineer when troubleshooting problems in Telecom design. Nothing ventured nothing gained.

The difference in price for higher voltage caps in some of the specialty caps is substantial, as anyone who has priced them will tell you. Size is also an issue when a layout is tight or when modifying existing audio products. The penny pinching I had in mind concerning using a good cap as opposed to a cheap electrolytic cap.

I just got off the phone with Jam and mentioned your post about capacitor quality in this circuit locating. After he stopped laughing he told me that in his experience that this cap was very important to sonics. I would be quite content to agree to disagree though rather than argue about a simple circuit change that the truly curious can try without getting a second mortgage on the house.

[Peter Daniel]

This cap is as important to sonics as the series coupling cap installed at amps input (or maybe even more). That's why some people decide not to use it at all and connect resistor directly to ground, risking more DC offset in this way. In my experience, the offset is usually not more that 60mV and it's fine with me.

Looking at prices, at partsconnexions, we see that BG N 33/16 is $4.40 and similar cap in 50V range is about twice as much (47u). The second row represent current new price (which is greatly reduced comparing to previous prices)

BLKGATE- 60091 NX 0.1 50v 4 x 7 $2.75 $2.20
BLKGATE- 60092 NX 0.47 50v 4 x 7 $2.95 $1.95
BLKGATE- 60097 N 1 50v 5 x 11 $3.95 $2.90
BLKGATE- 60098 N 4.7 50v 5 x 11 $4.95 $3.15
BLKGATE- 60099 N 10 50v 8 x 11 $6.50 $4.50
BLKGATE- 60205 NX 22 6.3v 5 x 7 $3.50 $2.80
BLKGATE- 60093 N 33 16v 6 x 11 $8.95 $4.40
BLKGATE- 60206 NX 47 6.3v 6 x 7 $3.75 $2.55
BLKGATE- 60100 N 47 50v 13 x 24 $14.95 $9.95
BLKGATE- 60102 NH 68 350v 30 x 25 $47.50 $35.95
BLKGATE- 60207 NX 100 6.3v 6 x 12 $8.95 $5.35
BLKGATE- 60094 N 100 16v 13 x 20 $15.95 $10.30

[Rafal]

will it make a big difference is I use 33uF cap instead of 22uF?

[Peter Daniel]

It could only be better. This cap shouldn't be lower than a certain value, but it can surely be bigger. I've seen one guy using 150u.

[SY]

ESR is not a linear resistance and low ESR is usually one of the indicators of capacitor quality. Does anyone else actually research these things or is just throwing opinions out adequate effort for the forum?

If the ESR varies by a factor of 5 (which it doesn't) over the audio band, it still doesn't vary the effective resistance of that shunt leg by more than 0.005% or so, given what actual esrs of these caps are. And that's a linear frequency response error. Now, tan delta can cause nonlinearities, and that's yet another reason not to cut the corners you suggest and use a low voltage cap; all else equal, lower voltage caps show higher tan deltas.

<fred imitation>Do you ever bother to do order-of-magnitude calculations, or do you think it's just fine to throw out onto the forum any sort of audio urban legend that you have chosen to believe?</fred imitation>

[SY]

Originally posted by Rafal
will it make a big difference is I use 33uF cap instead of 22uF?

No. It will extend the lf response a little bit, and slow down offset correction a little bit. The latter is unimportant assuming you're not planning to thermally shock the amp in actual operation.

[GregGC]

I also think that the quality of the cap in the feedback loop is very important (the feedback network has to transfer the output signal to the -in without any distortions in order for the amp to get proper feedback signal). That's why I don't use one (20mV DC offset on my amp). If you decide to use one use as reasonably large value as possible. That way the voltage across the cap would be smaller, thus less ditortions (based on some tests on cap. distortions). As Peter said if you're going to use one use 100.0 for 680 ohm FB resistor.

/Greg

[Rafal]

Thank you all for your help.

I think I'll order a 100uF / 50v Blackgate (series N) since the cost difference between 50v and 16v is only $2.65us)

Thanks again.


Rafal

[wrenchone]

There are two basic feedback networks for reducing output offset. They both have unity gain at DC, so the intrinsic input offset of the amp is passed on without amplification, assuming resistor values are used that are appropriate for the amp in question (bipolar or FET input). The circuit in the attached file labeled "A" is the one discussed in this thread so far. The disadvantage of this approach is that C1 must be relatively large, and it does see some reverse bias when the output swings negative. I would try using an Oscon for C1 in circuit A, as they are supposed to tolerate substantial reversal without ill effects.

I first learned about the Circuit labeled "B" over 20 years ago in Audio magazine. It appeared in an analysis of a Nakamichi amplifier. In this circuit, unity gain is provided by feedback resistor R3, and the gain network(R1-2) is coupled to the summing node via C1. For FET input amplifiers, R3 can be about 1M, and the coupling capacitor can be a high quality film unit of 1-2uF. Bipolar input amps will reqire an R3 of several 10s of k to reduce interaction with the input bias current. I have tried 33-47k and ~5-10uF with good results. I use this circuit ("B") in my projects almost always instead of the conventional one ("A").

[GregGC]

Originally posted by wrenchone
There are two basic feedback networks for reducing output offset. They both have unity gain at DC, so the intrinsic input offset of the amp is passed on without amplification, assuming resistor values are used that are appropriate for the amp in question (bipolar or FET input). The circuit in the attached file labeled "A" is the one discussed in this thread so far. The disadvantage of this approach is that C1 must be relatively large, and it does see some reverse bias when the output swings negative. I would try using an Oscon for C1 in circuit A, as they are supposed to tolerate substantial reversal without ill effects.

I first learned about the Circuit labeled "B" over 20 years ago in Audio magazine. It appeared in an analysis of a Nakamichi amplifier. In this circuit, unity gain is provided by feedback resistor R3, and the gain network(R1-2) is coupled to the summing node via C1. For FET input amplifiers, R3 can be about 1M, and the coupling capacitor can be a high quality film unit of 1-2uF. Bipolar input amps will reqire an R3 of several 10s of k to reduce interaction with the input bias current. I have tried 33-47k and ~5-10uF with good results. I use this circuit ("B") in my projects almost always instead of the conventional one ("A").

Thanks for sharing it. Interesting aproach. I quite like it. I gues the advatage is that you don't have to use a high value for the FB resistors in order to decrease the value of the FB cap. In the clasic case of 1k/22k AC FB res of the NIGC and 200k DC FB res. you can use just 1uF or even less. And the res from +in to GND have to be also 200k (to keep the output DC offset low) which means that the input impedance of that thing can be pretty High, thus your input cap can be also very small (1uF or less). Nice. I'd like to try it out of curiosity (though no caps is the best).

/Greg