Working with opamps and Coupling caps.

[diy didi]

Hi
I have been designing numerous opamp based audio circuits for years.
Recently I built a circuit using inverting amplifiers using opamps with gain no larger than 5. I used ±15 supplies.
I realised that this created slight DC voltages on the opamp outputs. I tried different opamps to no avail.
I accredited this to using a gain of 5. (as opposed to unity gain)
The only solution was to use coupling caps at a specific spots in the circuit to block DC.
My question is:
Is there a rule of them one can use to strategically place coupling caps without actually having to build the circuit and fixing it on the fly.
I have always assumed using ± split supplies, negates the use of coupling caps.
Your input appreciated.

 

[dknguyen]

Have you tried nulling? split supplies and dc blocking caps are completely unrelated. also beware caps blocking bias currents from flowing into op amp inputs or the ciruuuuwilfails minutes after starting up

 

[AnalogKid]

I have always assumed using ± split supplies, negates the use of coupling caps.

Nope. The input stage of an opamp has two error sources, offset voltage and bias current. They are almost invisible in unity-gain circuits, but are very apparent in higher gain things like a photo preamp with 40 dB of gain. The error is multiplied right along with the signal, so 5 mV offset error at the input turns into 0.5 V at the output.

Of course, there is no clear answer with your schematic. Guessing, you can see in many non-inverting circuits that the shunt leg of the negative feedback divider has a capacitor in series for GND. This reduces the gain to unity at DC, so that 40 dB preamp has only 5 mV of offset at the output.

The other error is input bias current. This is the (very small) base current needed by the input stage differential pair to do anything. For an inverting circuit with a 10K input resistor, every uA of bias current equals 10 mV of error voltage added to the input signal for amplification.

There are ways around this that do not require coupling capacitors.

ak

 

[Nigel Goodwin]

There are ways around this that do not require coupling capacitors.

But generally it's FAR easier to just use capacitors

And it doesn't have any drifting problems then either.

 

[AnalogKid]

And it doesn't have any drifting problems then either.

Neither does a DC servo.

ak

 

[KeepItSimpleStupid]

You have offset nulling, but you might have to do it the hard way.

As long as your within your device you may not care about the offset, so remove it at the end

I have an audio pre-amp that has two outputs; one is DC to 100 kHz and the other does not include DC. My amp has a 0.5 Hz filter.

The FM tuner in the same series had an output to 15 KHz and then one closer to 18 KHz because 18kHz could interfere with tape erasing.

 

[AnalogKid]

The FM tuner in the same series had an output to 15 KHz and then one closer to 18 KHz because 18kHz could interfere with tape erasing.

I'd be very surprised if that were the official reason from the company. In the US, commercial stereo FM is bandwidth limited to 15 kHz so it does not interfere with the 19 kHz pilot tone. I forget the rolloff requirement, but it is pretty steep. I can see the 18 kHz output bandwidth for a mono signal, but I doubt it adds much fidelity to a stereo source.

Working from memory, analog tape bias oscillators are usually above 50 kHz, and with the same oscillator driving the erase heads. Curious, what is the make/ model of the amp and tuner?

ak

 

[KeepItSimpleStupid]

The tuner is the ST-9030 and the pre-amp SU-9070 pre-amp.

Info avalable here: https://www.fmtunerinfo.com/gsearch.html?search=st-9030 and https://www.vintageshifi.com/repertoire-pdf/Technics-5.php

Cassette Deck was the RS-B-100 https://www.vintageshifi.com/repertoire-pdf/Technics.php
which is pretty impressive.

I also have a Carver Tx1-11 Asymetrical Charged-Coupled Decoder which makes multipath go away.

Antenna is aimed using the scope outputs of the tuner as Audio signals. Trick I learned way back.

Then you follow that with a dbx 4bx dynamic range expander.

The amp is home built and is a version of the Leach Amp. https://leachlegacy.ece.gatech.edu/lowtim/ with some enhancements like metal film resistors when possible, 10 Turn bias pot, Log audio ramp, Thump suppression, protection (It killed one resistor as it was supposed to when i switch the NPN and PNP outputs by mistake), dual bi-polar power supplies, star ground

 

[Nigel Goodwin]

Neither does a DC servo.

ak

Only by a LOT more effort, and usually considerable extra complexity. If it's an AC circuit, there's no real point in making it a DC one.

Going back a LONG, LONG time there was an audio design called the 'Texan' (as it came from Texas Instruments) using 741/748 opamps, and a later revision apologised for the original design where they spent too much time trying to eliminate capacitors, and refitted them where it made far more sense to have them.

 

[AnalogKid]

Only by a LOT more effort, and usually considerable extra complexity.

I have some experience in this area, and I'm familiar with the history of audio circuits design. I was waiting to see if the TS showed any interest before going into the details of other options.

And, I don't think the added complexity is all that great. It's basically an opamp integrator and a couple of resistors. Speaking of using the 741 in audio circuits, this is one of the few places where they do a good job. For many years it was a commonly used part for this application across many designers in many countries.

ak

 

[Nigel Goodwin]

Speaking of using the 741 in audio circuits, this is one of the few places where they do a good job. For many years it was a commonly used part for this application across many designers in many countries.

AG will be on momentarily, to tell us how utterly useless the 741 is for audio use

 

[AnalogKid]

I know. I throw it out there sometimes just to make sure his heart rate gets up into the exercise (or exorcised) range.

ak

 

[unclejed613]

I have always assumed using ± split supplies, negates the use of coupling caps.

you won't need coupling caps because the signal is able to be referenced to 0V instead of 1/2 of a positive or negative rail voltage

You have offset nulling, but you might have to do it the hard way.

if you use a cap to DC block the op amp feedback (between Rf' and ground, you end up with a DC gain of 1, and the AC gain is whatever is set by the feedback resistors. this is commonly done on power amplifiers. or, you can avoid the cap, and match the resistors that are used to ground reference the inputs. in other words, if your feedback network(Rf and Rf` ) is 100K and 10k, then the resistor that ground references the noninverting input should be (Rf || Rf` ) about 9k. this balances the input currents, which should (because the diff amp transistors are on the same piece of silicon) cancel, and eliminate the input currents as a source of offset. if you want to make it "the hard way" (one of several hard ways of minimizing offset ) you could use 1% tolerance resistors, and then experimentally determine the value of the noninverting input's ground resistor (if you tried this with the inverting input's ground resistor Rf`, you would be trimming the offset and changing the gain as well). if i'm looking for a close trim that might need to be changed in the future, i use a 10 or 20 turn potentiometer. if i want a really close fixed trim, the multiturn pot can be used the same way a mechanic uses calipers, put the pot in circuit, and null out the offset, then measure the pot, and use that value to choose the permanent resistor.

 

[crutschow]

I realised that this created slight DC voltages on the opamp outputs.

The opamp input offset voltage and input bias current generate a total offset voltage (as determined by the input equivalent resistance) that is amplified by the circuit gain.

I tried different opamps to no avail.

There are auto-zero type opamps that have essentially zero offset ( a few microvolts).

Is there a rule of them one can use to strategically place coupling caps without actually having to build the circuit and fixing it on the fly.

If the total DC offset isn't large enough to compromise the desired maximum AC voltage swing, then you can use just one capacitor at the output.

 

[unclejed613]

even audio power amplifiers don't have to be perfectly settled at 0V. anything below 100mV is usually ok, and doesn't produce a turn-on thump. the only time offset needs to be tightly controlled is if the circuit is used for measurement, and depending on the application, even then some small value of offset is allowable. now if you are driving a multiplier as an RF modulator, there's an application where any DC is generally unacceptable, because it causes RF leakage through the modulator.

 

[unclejed613]

ok, so here's some info you might find useful, not only does it demonstrate how the input balancing technique works, but also what happens when you use an op amp that auto-nulls.


this is an OP113 which is a general purpose op amp. notice the DC offset when R3 = 0. this is due to the same value input currents creating different voltages across differing resistances.

this is the OP113 with R3 = 6.9k. notice the offset is now 0.03mV this result required trial and error, placing different values of R3 until the result was less than a millivolt. this type of nulling could use a potentiometer or resistance decade switch and "DFT" (Determined at Final Test) placement of trimmer resistors if it were being made in a production environment.


this is an OP27 which auto nulls, and that changing R3 has no effect on the offset.