Coupling Caps for Near-Zero DC Offset Op-amps?

[thermionic]

Hi,

Many modern op-amps have virtually no DC offset at their outputs (and inputs) when wired in certain configurations. Obviously, one has to use a coupling cap when interfacing with the outside world to prevent external units potentially being harmed if a fault condition were to develop. However, what about internal coupling between circuits? Would you go on the assumption that, whilst there may be only a few uV of offset today, you'd prefer a cap in front of, say, a pot, in case offset were to drift over time?

When looking at the schematics for certain OEMs, it appears that they're quite liberal with coupling caps, even with op-amps that have little appreciable DC offset. In fact, 'tweakers' may choose to bypass these caps... So, who's right? Do you take a conservative approach? Why?

TIA

Justin 

 

[Nora]

I prefer to use caps when pots are involved.

But I'm curious to hear some opinions...

 

[sahib]

I would say, again it all depends on what you are designing. A few uV off-set in a single stage opamp circuit may not mean much but in multi stage amplification where the later stages deliver high gain things will be different. 

 

[ppa]

Yes, a cap must be used to protect the opamp stage from the DC voltage that might be at its input, even if this stage has not the need to having a decupling cap for overall feedback. Infact, even if true that a modern opamp can have a high DC gain with a very low offset , if the previus stage has a low offset voltage, this DC gain can amplify this offset voltage to a high value.
So sometime it's important to have the same a decupling cap or in the feedback network or in the input for this reason.

 

[clintrubber]

While at it here, shall we try to understand the curious combination in the API 325 line-amp of an output-cap (400uF) in parallel with a 47 Ohms resistor ?  They're between the output of the 2520 opamp & the output-TX primary. As we know the cap is for AC-coupling, and the resistor in parallel spoils all this...  ???

I could come up with some possible explanations of what they might have had in mind, but they'd be far-fetched...

 

[abbey road d enfer]

While at it here, shall we try to understand the curious combination in the API 325 line-amp of an output-cap (400uF) in parallel with a 47 Ohms resistor ?  They're between the output of the 2520 opamp & the output-TX primary. As we know the cap is for AC-coupling, and the resistor in parallel spoils all this...  ???

I could come up with some possible explanations of what they might have had in mind, but they'd be far-fetched...

Very simple. The combination of a cap and an xfmr make a resonant circuit that presents an extremely low impedance to the output stage at a few Hz, and a big hump at the same frequency. Adding a resistor adjusts the damping factor of the resulting 2nd-order hi-pass for a much better response.
Check the attached graph. The blue one is without the 47R across the cap, the green one with it.

 

[PRR]

> the curious combination in the API 325 line-amp of an output-cap (400uF) in parallel with a 47 Ohms resistor?

I thought that was pretty clear.

Well, there's two possibilities.....

1) For lowest distortion the transformer wants low source impedance and low DC current.

The output DC offset voltage may be low, but never zero. The transformer DC impedance is as low as possible, to reduce loss and to reduce distortion. Since this is a 150r(?) winding, say 15 ohms DC resistance.

15mV across 15 ohms is 1mA, which just might be enough off-center on the iron to cause some sonic change.

The 400u+47r is 8 ohms at 50Hz, 20 ohms at 20Hz, rising toward 47r at 8Hz and lower. 15mV in 47+15= 62 ohms is 0.25mA. Perhaps it was learned/predicted that 0.25mA DC current was not a problem. Meanwhile above 50Hz the THD is as low as direct-connection.

2) {Abbey types/graphs faster} The L of the winding and the C of 400u will form a 2nd order high-pass which just might bump-up the natural bass-droop of a transformer. It could be the difference between -2dB at 10Hz and +/-0.5dB at 10Hz, +/-0.25dB at 20Hz. For one stage, this is small; on the radio network run from London to Scotland the dozens of network stations along the way could shave several dB and we Scots don't like short bass.

 

[abbey road d enfer]

Attached graph shows current that output stage has to deliver.
Green without 47R, red with 47R.
Blue is a 20 mAmp reference (pretty much the max current most common audio opamps can deliver - think 5534).
You can see that at the bump, the current is about 3x more.
This could happen if you bump a mic stand; with the added resistor, the loud VLF can be filtered out in post, without it, the signal will probably be too distorted to be redeemable.

 

[ppa]

its important to tell that's with these new opamps possible makes good trimmerless servo for DOAs stages to reduce to zero the DC voltage to the xfmer.

 

[MikeClev]

Very simple. The combination of a cap and an xfmr make a resonant circuit that presents an extremely low impedance to the output stage at a few Hz, and a big hump at the same frequency. Adding a resistor adjusts the damping factor of the resulting 2nd-order hi-pass for a much better response.
Check the attached graph. The blue one is without the 47R across the cap, the green one with it.

Is this case specific to that particular transformer, or does this occur with a lot of other transformers? When would you know to use such an arrangement?

-Mike

 

[thermionic]

its important to tell that's with these new opamps possible makes good trimmerless servo for DOAs stages to reduce to zero the DC voltage to the xfmer.

Very interesting. Do you have a link to an example?

Thanks.

J

 

[QUEEF BAG]

API uses an LF 411 for their servos in the legacy consoles
and i can't recall if they use 'em in the 550B eq or not...
but i think so, very little difference between the 550B and 550L

as for as the 47R in the output of their designs, my understanding
(or should i say my gullable impression) is it is also sort of a protection
device for the tranny. if the op amp ***** the bed, the resistor will burn
thus protecting the tranny. 
with this in mind, i will not put in a higher wattage resistor in that spot.
i have taken out a number of 1 W resistors that people have put in.

So...this resonant circuit thing makes me wonder what Abby's graph would
look like with a 470uf cap and 47R resistor since those are what seem to
get put in that spot these days.

 

[abbey road d enfer]

Is this case specific to that particular transformer, or does this occur with a lot of other transformers? When would you know to use such an arrangement?

-Mike

It is just a normal consequence of the xfmr being inductive, so yes it happens with ALL transformers, to more or less degree. Some transformers with very high inductance and relatively high DCR have only a small bump at lower infrasonic frequency; in that case, they don't need the arrangement.
Obviously, if the output stage has significant DC offset, the resistor should not be used.
After 30 years of dealing with transformers, I came to the conclusion that a transformer should be included in a specific stage that provides some kind of distortion compensation, be it by using negative resistance or feedback.
I think there's nothing worse than sticking a transformer at the output of a stage that has not been specifically designed for it.
In particular I see many manufacturers offering a transformer option on EBOS (electronically balanced output stage) that can be retrofitted without further ado; IMO, it's utterly wrong.

 

[ppa]

as for as the 47R in the output of their designs, my understanding
(or should i say my gullable impression) is it is also sort of a protection
device for the tranny. if the op amp ***** the bed, the resistor will burn
thus protecting the tranny. 
with this in mind, i will not put in a higher wattage resistor in that spot.
i have taken out a number of 1 W resistors that people have put in.

yes, it's a good idea to use a low wattage resistor.

 

[mitsos]

Obviously, if the output stage has significant DC offset, the resistor should not be used.

I've built a few 312s with that 47R in parallel and I liked the sound, but I recently had a problem when trying to add a home-etched Bo hansen "312 DI mod". Not sure if you're familiar with it, but it goes basically directly into the opamp.  The PCBs I used already have a spot to use a shorting TS jack as a DI direct to the opamp, in this case I used a TRS jack, relay and PNP switch to switch from mic to DI when the cable was inserted(like Fabio and Dan kennedy have).  It worked... for about 10 seconds, then I got crazy hum, and the 47R started to smoke. Now I know that this DI increases DC offset on the 2520 (on another preamp -without the 47R- I measured 88mV), so I was wondering if that was the cause of the problem, and then I saw you post what I quoted above...  Hoping you can explain a bit why you wouldn't use the resistor in parallel with the cap in the case of high DC offset?  The DC goes direct to ground through the resistor and trafo winding? Is that what did in my 47R?

thanks!

 

[abbey road d enfer]

yes, that's exactly it. Whatever offset voltage goes through the 47R and then through the xfmr primary. If the offset voltage is moderate (<1V) the primary will be magnetised, which reduces drastically its inductance and generates 2nd-order THD. If the offset voltage is higher, the output stage may overheat and the resistor may smoke.

 

[gentlevoice1]

Hi,

In my experience caps - like most all components - imprint on the sound. So, everything else identical, and the circuit working, I would be careful including caps. See e.g.:

http://www.humblehomemadehifi.com/Cap.html

Best,

Jesper

 

[mitsos]

well, I removed this DI and stuck in my normal switching DI straight to the opamp and everything went back to normal. But I did not test without the 47R (stupid me, I was in a hurry) so I will try that again..
thanks!

 

[FunkyB]

Very simple. The combination of a cap and an xfmr make a resonant circuit that presents an extremely low impedance to the output stage at a few Hz, and a big hump at the same frequency. Adding a resistor adjusts the damping factor of the resulting 2nd-order hi-pass for a much better response.
Check the attached graph. The blue one is without the 47R across the cap, the green one with it.

> the curious combination in the API 325 line-amp of an output-cap (400uF) in parallel with a 47 Ohms resistor?

I thought that was pretty clear.

Well, there's two possibilities.....

1) For lowest distortion the transformer wants low source impedance and low DC current.

The output DC offset voltage may be low, but never zero. The transformer DC impedance is as low as possible, to reduce loss and to reduce distortion. Since this is a 150r(?) winding, say 15 ohms DC resistance.

15mV across 15 ohms is 1mA, which just might be enough off-center on the iron to cause some sonic change.

The 400u+47r is 8 ohms at 50Hz, 20 ohms at 20Hz, rising toward 47r at 8Hz and lower. 15mV in 47+15= 62 ohms is 0.25mA. Perhaps it was learned/predicted that 0.25mA DC current was not a problem. Meanwhile above 50Hz the THD is as low as direct-connection.

2) {Abbey types/graphs faster} The L of the winding and the C of 400u will form a 2nd order high-pass which just might bump-up the natural bass-droop of a transformer. It could be the difference between -2dB at 10Hz and +/-0.5dB at 10Hz, +/-0.25dB at 20Hz. For one stage, this is small; on the radio network run from London to Scotland the dozens of network stations along the way could shave several dB and we Scots don't like short bass.

Abbey road d enfer's and PRR's answers are very clear. You don't want too much dc current through the primary and you don't want the LF resonance bump.
All API 2520 opamp / 2503 trafo schematics that I've seen have this R//C combination in series with the primary.

But... Strangely enough the 525C doesn't have such a cap:
https://www.sowter.co.uk/schematics/API 525C.pdf
The 525C uses a 2623 transformer and a 2520 opamp. I just measured the "primary's" DCR of my EA2623 clone, and that happens to be 14 Ohm. So it's not like the 2623's "primary" has much more DCR than the 2503.

Of course my question is: why do all API circuits have a dc blocking cap or a R//C in series with the primary except the 525C and maybe more 2623 circuits?

Does anyone have any idea?

 

[Bo Deadly]

Of course my question is: why do all API circuits have a dc blocking cap or a R//C in series with the primary except the 525C and maybe more 2623 circuits?

Does anyone have any idea?

The gain cap (the cap in series with the gain setting resistor like C7 in your schematic) stops the gain from multiplying the amp offset.