Stacking op-amps - pros and cons.

Status
Not open for further replies.
Jump to Latest
G

ginetto61

Member
Joined 2009
Paid Member
#1
Hi !
i was reading another 3D here when the op-amps stacking solution came along.
From what I understand if the power supply is strong enough at the place of one opamp i can put a pile of op-amps of the same model ?

ops.jpg


What are the major benefits from this move ?
What the drawbacks ?
It would not be better to go with a better specified op-amp instead ?
It seems to me a very uncommon solution anyway.
Thanks a lot and kind regards,
gino
 
#2
Hypothetically, suppose that in a stack of N opamp ICs, each one has a different input offset voltage. But all of them are within datasheet specs.

Opamp1 has Vos= +0.1mV
Opamp2 has Vos= -0.2mV
Opamp3 has Vos= +0.3mV

etc.

How well or how poorly would the combined stack work, as an opamp?
 
#3
I know of at least one commercial unit, a Behringer headphone amp, using this solution, with 3 or 4 i do not remember precisely of jrc 4558 opamps paralleled (in line package)
The sound is nice but there are some problem with the ground (i.e. noise)
kind regards,
gino
 
#4
My gut feeling is that naively stacking opamps is a seriously bad idea, unless you can guarantee that they will all have exactly the same input offset voltage. As far as I know, you can't guarantee that. They will pull the output in different directions.

On the other hand, properly paralleling opamps with all the required 'ballasting' resistors may work fine as a way to boost the output. Probably better to use a beefier opamp, or add a buffer.
 
#7
DF96 said:
My gut feeling is that naively stacking opamps is a seriously bad idea, unless you can guarantee that they will all have exactly the same input offset voltage. As far as I know, you can't guarantee that. They will pull the output in different directions.
On the other hand, properly paralleling opamps with all the required 'ballasting' resistors may work fine as a way to boost the output.
Probably better to use a beefier opamp, or add a buffer.
Click to expand...

Thank you very much for the valuable advice
i had the same feeling but just wanted to read the opinions of experts
Moreover today very powerful chips are available
No more doubts
Kind regards,
gino
 
#8
not with op amps

this was done with some early unity gain buffer chips

I believe it works with the OPA634 - because they are unity gain they don't amplify their input offset V, and they have ~10 Ohms internal in series with their outputs

I think some early external reference Iout DAC chips could also be pin for pin paralleled

having seen those very particular situations some may have thought to generalize

It won't work with op amps

due to the high DC gain, varying input offset V between devices each op amp has to have independent feedback, and series output R added


several op amp based headphone amps do parallel the output of several op amps – but use independent feedback, current sharing output R


http://benfeist.com/how-to-build-the-apheared-47-headphone-amplifier-for-grado-headphones/

https://docs.google.com/file/d/0B52...mLWIwM2MtNDI4ZWU4YWRjY2Y4/edit?hl=en_US&pli=1
 
Last edited:
#11
Just to show that it was not a my invention ...

Tube Screamer Modifications

But there are actually so many different op-amps around with very different performance that the idea of stacking many weak units instead of using just one more powerful seems weird enough.
There are exceptional op-amps available. The problem maybe is to be able to implement them correctly.
Regards,
gino
 
#12
Reduced Net Power Dissapation

Parallel vertical tight stacking will give 'n' times the output current, but something less than 'n' times the power dissapation capability of a single device. This is becuase such a tightly stacked array will have less total surface area for radiating heat than would the same number of devices not stacked. So then, the package-to-air thermal resistance of the array will also be greater than if not stacked, providing less heat transfer to the air than is assumed in the datasheet.
 
Last edited:
#14
ginetto61 said:
Just to show that it was not a my invention ...

Tube Screamer Modifications
Click to expand...

Gutiar players are always looking for new distortions, and apparently some of them like this one.

It seems like a very poor implementation to me. but evidently it must work because people are doing it. I thought it was crazy when I first saw it, but guitar players are always devising clipping circuits and exploiting diode nonlinearities to come up with a new "sound."

The classic "sound" is an overdriven triode. It's the "sound" us older folks knew 30 years ago. The "sounds" in some new music sounds like something is burning up in your amplifier. 😀
 
#15
It seems like a very poor implementation to me.
Click to expand...
You are very kind.
I just call it nonsense.

but evidently it must work because people are doing it.
Click to expand...
Not so sure about any of those premises.😛
Never saw anybody (besides these guys) doing it.🙄

And one thing is sure:
a) it *might* lower output internal impedance and/or increase current capability.
Useless because the distorting Op Amp drives as not-so-low-impedance-load , so it "solves" a non existent problem and
b) it **definitely** lowers input impedance 😱 , increases input biasing current, parasitic capacitance (whatever value it had, it will have more) and worst of all: multiplies battery current consumption by the amount of Op Amps involved (even if their outputs are left open as suggested in the article, another nonsense).

Of course, to be able to offer "something different" about the most studied and worn circuit of them all, you have to retort to "wild" implementations, just to avoid being another carbon copy.

What's wrong about taking a clean empty protoboard and just try to actually design something different?:shhh:
 
#16
JMFahey said:
Not so sure about any of those premises.😛
Never saw anybody (besides these guys) doing it.🙄
Click to expand...

I've heard of it before. People put the "stacked" chips into commercial equipment sometimes 😱 to get the "sound."

And one thing is sure:
a) it *might* lower output internal impedance and/or increase current capability.
Useless because the distorting Op Amp drives as not-so-low-impedance-load , so it "solves" a non existent problem and
b) it **definitely** lowers input impedance 😱 , increases input biasing current, parasitic capacitance (whatever value it had, it will have more) and worst of all: multiplies battery current consumption by the amount of Op Amps involved (even if their outputs are left open as suggested in the article, another nonsense).
Click to expand...

Of course there's all kinds of things wrong with it. But you're preaching to the choir here.

Of course, to be able to offer "something different" about the most studied and worn circuit of them all, you have to retort to "wild" implementations, just to avoid being another carbon copy.

What's wrong about taking a clean empty protoboard and just try to actually design something different?:shhh:
Click to expand...

You can't beat the sound of an overdriven triode for guitar in my book. There's so much a musician can do with it if they know what they're doing.

Now they use diodes in the feedback loop and stuff 😛 to generate an "overdriven" sound. BJTs sound terrible when overdriven, an op amps are even worse.
 
#17
I learned the hard way not to stack chips. You are MUCH better paralleling them with their own NFB loop. (Instead of the + and - inputs all connected) I did this with op-amp chips, and they almost all get hot, (especially OPA4227/8) and this did work with a MosFet Rail-2-Rail chip, but I had reduction in the sound quality in the treble, my guess being from NFB trying to cancel out HF oscillations???

Datasheets usually show one op-amp running as "Master" then driving other "Slave" amps, where they all parallel with a resistor. Look in the OPA4227 Datasheet for the Headphone amp.

Also a Web Article on paralleling amps Get increased output from an improved Howland current pump
 
#18
Here is a link using 32 NE5532 paralleled for power amp app.
The 5532 OpAmplifier, part 2 - ELEKTOR.com | Electronics: Microcontrollers Embedded Audio Digital Analogue Test Measurement

2 JRC4556 paralleled for h/p amp app.
NwAvGuy: O2 Design Process

I've tried paralleling just 2 NE5532 & the result was 5V DC at o/p(with 1 Ohm series R)If more than 2 opamps are used in parallel, higher series R is req. at o/p to prevent opamps fighting each other according to NwAvGuy. Simply piggy backing opamps is a disaster(waiting to or already happened)
 
#20
If I were to parallel opamps, whether stacked or alongside, I would incorporate an output resistor on EVERY output pin.
Much the same as adding an emitter/source resistor for paralleled output devices.

I may also look to inserting an input resistor to EVERY -IN pin.
 
Status
Not open for further replies.
Top Bottom