op-amps in parallel

[A3rd.Zero]

in order to keep the idea simple. What would happen if you stacked the OPamps in a CMOY simple amp circuit, in parallel?

 

[morsel]

Without an output resistor for each opamp they would fight each other and probably burn up.

 

[Budgie]

I found out about that the hard way- my attempt at quad 5532 opamps, in parrallel, resulted in a small room heater. And a burnt finger tip.

 

[A3rd.Zero]

thanks, for the first time in my life i thought id ask first

 

[guzzler]

the Apheard #47 design is not stacked, but it is effectively paralleled opamps...

check out 'Apheared's Project Scrapbook' on the Headwize libary for schematics!

g

 

[pooooo]

Quote:

Originally posted by A3rd.Zero in order to keep the idea simple. What would happen if you stacked the OPamps in a CMOY simple amp circuit, in parallel?

Speaking of this reminds me of meta42. If stacking OPamps is not a good idea, why should stacking buffers be a good one?

 

[Budgie]

I don't think anyone was suggesting to not parrallel opamps, just to be sure to do it properly. It works well with opamps and buffers.

 

[pooooo]

Quote:

Originally posted by morsel Without an output resistor for each opamp they would fight each other and probably burn up.

Sorry I've quoted the wrong post. How do you prevent buffers fighting each other while "stacking" buffers? I thought stacking means that there are not output resistors.

 

[tangent]

The difference is that open-loop buffers don't have their own feedback control loop like op-amps do. An op-amp can cope with almost any behavior in its feedback loop as long as it behaves in a predictable manner. A buffer's output voltage is controlled by its input, so there is no fight.

If you directly parallel two op-amps (such as by stacking them) you have two output stages to double the current, but you also have two control stages each with slightly different "goals" due to component tolerance issues. The two chips strain against each other trying to force the output to their idea of the ideal.

In a design like the A47, one chip is set up as a slave to the other, and the current sharing resistors ensure that one chip doesn't take over and try to fight the other. The main downside of this is that this increases the output impedance of the whole, which eliminates one potential advantage of paralleling output stages. Another downside is that it erodes phase margin, reducing stability. I've had several A47s that oscillated. I don't think I've ever had a buffered headphone amp that oscillated unless provoked.

 

[aos]

Opamps have typically much lower output impedance than buffers. So for a given variation in output voltages (which should ideally be the same) it takes higher current through that resistance to bring voltages to be equal. Adding resistor in series with output swamps that output impedance, reducing those currents flowing from one output to another.

I was stupid enough to try parallelling a few TLE2426 - they have extremely low impedance (so low that it's advertised on the front of the datasheet). They get very hot very fast (luckily the classic rule of never turning power on for more than a few seconds when first starting up any circuit paid off - nothing got destroyed even after repeated tries).

 

[Glassman]

just want to let you know that on M-Audio Revo, there are paralleled opamps too.. one 5532 for one channel, resistors are on their outputs..

 

[rickcr42]

I have successfully stacked opamps in existing gear but it means clipping the output leads of the present opamps (unless socketed) , desoldering the output leads from the pcb holes and then running a 10 ohm resistor from the existing opamp and the new opamp (soldered directly across the first but with the output pins bent outwards) to the pcb pads

no easy task but poassible and the benefits with some gear is worth it

rick 'n roll