Modifying Existing Op Amp Circuits to Operate in Class A

[Aniol1349]

Good Afternoon everyone!

I was doing some research today on circuit modifications and I have run into this interesting concept of biasing the op amps to run in class A operation. Apparently Focusrite has used this on there platinum equipment.

I mostly work with console modules and I'm interested to find out if there is any chance to implement something like that into an existing op amp circuits.
If so what would be the real benefits of this configuration;  should this be applied to one key stage or could it be done to all the op amps carrying main audio signal.

Thanks for any input!

 

[JohnRoberts]

Think about it... if it made any sense at all the IC companies would already do it. It is not a new concept.

Some people take pleasure from modifying their gear to make it different than stock, just to be different (hey listen to this  ).

JR

PS: Since you cannot get inside the IC op amp, all you can do is shift the operating point of the op amp final output stage. Since the IC was not designed to operate that way, it could degrade thermal equilibrium in other parts of the IC.

 

[Deepdark]

I got an Octopre platinium at my studio and it became really hot, even Focusrite highly recommand to let an 1U free space. They claim it's because of the high end class A preamp running that hot

 

[Andy Peters]

I got an Octopre platinium at my studio and it became really hot, even Focusrite highly recommand to let an 1U free space. They claim it's because of the high end class A preamp running that hot

The preamp may be discrete.

 

[Gene Pink]

Some people take pleasure from modifying their gear to make it different than stock, just to be different (hey listen to this  ).

Alright, ya got me. I resemble that remark.

There was a certain power amp back in the '70s that the only way to completely clean up the crossover distortion was to set the bias to "room heater" mode. A 220R, 20W resistor from output to a rail (pick one, doesn't matter) shifted the crossover distortion point far enough away from quiet-passage listening, up to a level where the signal was loud enough to mask it. This allowed saner bias settings.

Gene

 

[JohnRoberts]

Some people take pleasure from modifying their gear to make it different than stock, just to be different (hey listen to this  ).

Alright, ya got me. I resemble that remark.

ding ding ding...

There was a certain power amp back in the '70s that the only way to completely clean up the crossover distortion was to set the bias to "room heater" mode.

and if there isn't enough heat sink, the output stage can go into thermal run away that never ends well. 

A 220R, 20W resistor from output to a rail (pick one, doesn't matter) shifted the crossover distortion point far enough away from quiet-passage listening, up to a level where the signal was loud enough to mask it. This allowed saner bias settings.

Gene

Just trading distortion for distortion...  (but crossover distortion sucks if you have it and know what it sounds like..mostly HF grundge.)

Power devices back in the '70s were not very fast but 25-35mA of class A bias in a class AB design should get it done for a decent design, even back then.

JR

 

[Samuel Groner]

Think about it... if it made any sense at all the IC companies would already do it. It is not a new concept.

I don't think that's a valid argument. IC companies know that class A bias reduces distortion, and that high power dissipation reduces sales. IC opamps are inherently a mass product, so the latter argument wins.

Having said that, I'm not overly optimistic about the benefit either. Audio companies in turn also know that the term "class A" increases sales, so that's likely the winning argument.

I've done some measurements years back but don't recall the actual distortion improvement. The Tektronix SG 505 and the AP System One use a related technique for the NE5534, which is a clear indication that there is a measurable benefit. IIRC Douglas Self shows some measurements in at least one of his books.

One thing to note is that the actual distortion improvement might not come from the opamp, but from reduced interference with the power supply. With class B operation, the power supply carries currents with relatively large harmonic distortion content. By means of common-mode (through bypass capacitors to ground) or magnetic coupling this can induce into the audio signal. With class A bias, the supply carries (ideally) only the undistorted signal current. A good PCB layout that reduces coupling is cheaper.

Samuel

 

[JohnRoberts]

Think about it... if it made any sense at all the IC companies would already do it. It is not a new concept.

I don't think that's a valid argument. IC companies know that class A bias reduces distortion, and that high power dissipation reduces sales. IC opamps are inherently a mass product, so the latter argument wins.

IC companies always strive to make better products, but within a context of being commercially profitable.

Increasing class A bias reduces crossover distortion but is different than slugging a DC current in one direction.  IC designers have the luxury of being able to add circuit complexity to reduce distortion relatively inexpensively considering silicon real estate, without the brute force hit to power consumption.   

Having said that, I'm not overly optimistic about the benefit either. Audio companies in turn also know that the term "class A" increases sales, so that's likely the winning argument.

ding ding ding...

I've done some measurements years back but don't recall the actual distortion improvement. The Tektronix SG 505 and the AP System One use a related technique for the NE5534, which is a clear indication that there is a measurable benefit. IIRC Douglas Self shows some measurements in at least one of his books.

If desirable the IC makers could beef up their output stage in one direction or the other, and account for expected internal dissipation increase.  Perhaps use one of the modern high current output op amps, for slugging.

As you know op amps are 5 terminal devices, and everything matters. There were even AES papers written about PCB layout of power rails and grounds inside power amps (Cherry) because that affects residual distortion too. 

One thing to note is that the actual distortion improvement might not come from the opamp, but from reduced interference with the power supply. With class B operation, the power supply carries currents with relatively large harmonic distortion content. By means of common-mode (through bypass capacitors to ground) or magnetic coupling this can induce into the audio signal. With class A bias, the supply carries (ideally) only the undistorted signal current. A good PCB layout that reduces coupling is cheaper.

Samuel

Amen... If one has a high enough resolution test bench you could probably parse out several subtle circuit interactions, but at some point you are just moving around the lump of dust under the carpet..

JR

PS: I have killed more than a few brain cells designing alternate output stage topologies and biasing schemes, to gain the benefits of class A without the huge efficiency cost. I think there were one or more patents covering this technology from the Japanese a few decades ago. In pretty much every case that delivered near class B efficiency, the output still alternated between push and pull, while the transition was slowed down and stretched out to reduce distortion. An actual hard class B transition always occurred but was shifted up-stream into lower power electronics. I would be very surprised if some of this is not already being done inside the uber performance SOTA IC op amps.  I remember the good old days when they published schematics for inside ICs.

 

[sparko311]

Hey guys.  One can hang a resistor from an op amps output to V- (or even V+)  to force it into class A mode.  The IC guys dont do this for thermal / power dissipation reasons.  The drawback to this is that this method will only give you single ended class A operation, which is only 25% efficient.  Push pull class A operation is more desirable and more (50%) efficient.  If you want good class A push pull output, you may have to look at discrete op amps.  Something like this

http://sparkoslabs.com/discrete-op-amps/

 

[JohnRoberts]

Hey guys.  One can hang a resistor from an op amps output to V- (or even V+)  to force it into class A mode.  The IC guys dont do this for thermal / power dissipation reasons.  The drawback to this is that this method will only give you single ended class A operation, which is only 25% efficient.  Push pull class A operation is more desirable and more (50%) efficient.  If you want good class A push pull output, you may have to look at discrete op amps.  Something like this

http://sparkoslabs.com/discrete-op-amps/

Umm that looks like class A/B but with 8 mA of class A quiescent current.

JR

 

[PRR]

> single ended class A operation, which is only 25% efficient.

Can you show this with numbers?

No fair running DC through the *load*; that's usually not-done in audio.

> hang a resistor

IMO, efficiency of a resistance-loaded SE amp is more like 6% at best. And that is with a specific load, and output at not-half supply. Values under 3% are common.

 

[pucho812]

Biasing an opamp into class a operation can be done a few ways.
The resistor Method is one of them.

http://tangentsoft.net/audio/opamp-bias.html

 

[joaquins]

Talking about efficiency in signal stages seems not very useful, at least not nowadays with bridging standards. 10k loads with 0dBu is what? Emmm 0.06mW!!!

ok, let's say we want 50% efficiency we need an opamp which uses at 30V rails (±15V, pretty standard, right?) the not so reasonable amount of 2µA while idling! This includes not only the output stage bias but all the other biasing for the 5+ stages inside an IC op amp...

My point is not to go crazy about the efficiency of the stage but to the total power consumption, for a single preamp in a rack you can probably get away with almost anything without running into thermal or supply issues. For a 500 series you have 130mA per module, there is the number you can't run over. Single opamp for a preamp you could use 100mA of bias current for the output stage and everybody is happy, extra 30mA for a few intermediate stages, you probably want 3 or 4 for the first stage and after that even  lower. That's about 3W, with our nominal 0.06mW @ 0dBu/10k load is 0.002% Eficency!!!

So, 25% or 3%, doesn't really matter if you ask me, talking about signal stages.

About the topic of classing-A a class-B stage... well, there are better ways to skin the cat, for 10k loads 15V rails you only need 1.5mA, plus the original bias of the stage, probably 5mA is enough for many opamps. NE5534 asks for 4mA typ, so only worsen by a factor of 2, very reasonable. By now my efficiency point is pretty clear and calling them room heaters can stop. 

The one idea I've seen around and I liked is about using the compensation pin5 of the NE5534 as output since it's the last stage before the class AB output stage, the driving capability is not the best for this pin, but is possible to build a discrete output stage with the biassing configuration of your preference. Nice hack for an existing pin on an IC, never tried it nor seen numbers but should worth to try it, at least to have some fun. That way you get the input specs, open loop gain and some other things from the NE5534 with an output stage you do like (if you didn't already the one inside the chip) and you don't push it's output outside the designed working conditions.

JS

PS: Happy 1k5 post for me!!!  ;D ;D ;D

 

[CesarIII]

The one idea I've seen around and I liked is about using the compensation pin5 of the NE5534 as output since it's the last stage before the class AB output stage, the driving capability is not the best for this pin, but is possible to build a discrete output stage with the biassing configuration of your preference. Nice hack for an existing pin on an IC, never tried it nor seen numbers but should worth to try it, at least to have some fun. That way you get the input specs, open loop gain and some other things from the NE5534 with an output stage you do like (if you didn't already the one inside the chip) and you don't push it's output outside the designed working conditions.

It is a good idea. So we can use a low-power class A output stage. I'll try it.

 

[JohnRoberts]

It is a good idea. So we can use a low-power class A output stage. I'll try it.

This thread is not quite 10 years old (only 8 years old).

"low-power class A" is a contradiction in terms. Class A by definition wastes power.

JR

 

[kyronmusic]

This thread is not quite 10 years old (only 8 years old).

"low-power class A" is a contradiction in terms. Class A by definition wastes power.

JR

Not to be a Smart “A..” (pun intended). Low power Class A is not a contradiction in terms. Class A is INEFFICIENT... (wastes power)... but can be low power ie... a 250mW amp... or high power... a 50 Watt amp.

Furthering the subject... the crossover distortion in a class AB output is fixed in amplitude. That is the transition point when the NPN turns off and the PNP turns on (and vis-versa). As such... the relative percentage of crossover distortion reduces as the amplitude of the signal increases.
Biasing the output stage of an op-amp to class A need only been done to handle very small signals. Once the amplitude gets high enough the crossover distortion become negligible and the op-amp can transition back to class AB. Furthermore... using a FET constant current source is apparently a better option than using a resistor.
My two cents worth.
Cheers.

 

[JohnRoberts]

Not to be a Smart “A..” (pun intended). Low power Class A is not a contradiction in terms. Class A is INEFFICIENT... (wastes power)... but can be low power ie... a 250mW amp... or high power... a 50 Watt amp.

Furthering the subject... the crossover distortion in a class AB output is fixed in amplitude. That is the transition point when the NPN turns off and the PNP turns on (and vis-versa). As such... the relative percentage of crossover distortion reduces as the amplitude of the signal increases.
Biasing the output stage of an op-amp to class A need only been done to handle very small signals. Once the amplitude gets high enough the crossover distortion become negligible and the op-amp can transition back to class AB. Furthermore... using a FET constant current source is apparently a better option than using a resistor.
My two cents worth.
Cheers.

And that is precisely the rationale behind class A/B power amps to shift the power device cutoff threshold away from 0V and small signal levels.

JR

 

[Voyager10]

Samuel Groner's encyclopaedia of op-amp distortion (PDF linked from https://groupdiy.com/threads/opamp-measurement-series.29265/post-354360) includes plenty of relevant information.

Crossover distortion will show up as wonky-looking 'residual' signals, generally in the 600R output loading tests. LT1007 is very poor (p132, top left) but many others show little or no observable misbehaviour, even under extreme conditions.

Where the op-amp isn't driving a low impedance load, crossover distortion is a bit of a non-issue IMO.

 

[k brown]

"Crossover Distortion": When you take your SUV off-road, and discover it was never actually engineered to do that.

 

[Cqwet Dbdfte]

Class A must be better than Class B. Class C, D and F are definitely out, maybe Class AB is acceptable for the working classes.
Ugh.