opamps vs transistors

[Sir H C]

From RG:

Actually, stability is not related to open loop gain so much as it is to phase shift and the Nyquist criterion of less than 180 degrees of phase shift before gain is under unity. High gain makes it harder to get gain down to unity sometimes, but more feedback makes it harder to maintain stability, not less. We were taught that stabilizing a gain-of-one opamp is harder than a gain of five. Some industrial opamps are internally stabilized for gains above X (often 5 or 10) but not unity. So could you explain needing to use more feedback to gain stability?

Amen, having done some of these, if  you know the application (or the op-amp is embedded in a bigger system) unity gain stability is not needed or wanted.

[alextheian-alex]

From RG:

Actually, stability is not related to open loop gain so much as it is to phase shift and the Nyquist criterion of less than 180 degrees of phase shift before gain is under unity. High gain makes it harder to get gain down to unity sometimes, but more feedback makes it harder to maintain stability, not less. We were taught that stabilizing a gain-of-one opamp is harder than a gain of five. Some industrial opamps are internally stabilized for gains above X (often 5 or 10) but not unity. So could you explain needing to use more feedback to gain stability?

Amen, having done some of these, if  you know the application (or the op-amp is embedded in a bigger system) unity gain stability is not needed or wanted.

Yes, that is all true, and not to keep this going, but -- correct me if i am wrong here -- the isn't it true Nyquist uses open loop response to predict what the closed loop behavior of the circuit.... and the higher the open loop gain is, the smaller the degree of the phase margin, ie a smaller margin of error, ie less stable without compensation?  Once you know that phase margin and predict the response, you can compensate by using internal feedback loops (Millet compensation, etc) for damping, etc.  Those introduce their own dominant pole to knock down the excess gain at the at that freq, while having the additional effect of slowing the slew of the chip, resulting in more high freq distortion (like I mentioned before... the upper harmonics just "hanging out") and also knocking down the open loop gain at freq above the pole, which is a tradeoff of using that feedback to stabalize the opamp because  there is less gain to correct he overall distortion within the circuit.


EDIT: that is all to explain why you need 'feedback to gain stability'

[Doug_H]

I still don't get the basic premise of this thread- "opamps (sic) vs transistors". Op amps are based on differential amplifier circuits which are made up of transistors. It's like saying "buses vs. carburetors", or "buildings vs. cinder blocks"... 

I now return you to the "Pedantic Op Amp Showdown of the Titans 2007".

[puretube]

ooops: I almost wanted to post: "and where is Barkhausen?", when I stumbled across this, found here;

guess I better keep relying on: "easi" : "ears/axe/scope/iron"... 

[alextheian-alex]

I still don't get the basic premise of this thread- "opamps (sic) vs transistors". Op amps are based on differential amplifier circuits which are made up of transistors. It's like saying "buses vs. carburetors", or "buildings vs. cinder blocks"... 

I now return you to the "Pedantic Op Amp Showdown of the Titans 2007".

HEH... pedantic eh?  i resemble that remark.  no, you are right though, it does sound silly.

i think that the semantic of the heading "opamps vs transistors"  gave the thread a kinda 'showdown' feel.  Sorry if i threw fuel on the fire.  And yes, opamps are just arangements of various active and passive elements in a specific circuit, and in the case of the IC variety that we generally think of, that circuit is on a single die.  Don't forget though that the first opamps were all tube --for years and years-- they were just not on the little IC dies that we are used to, so you can have an opamp without a single transistor.

Maybe it would be better to rearrange the subject into somthing more along the lines of "Discreet transistors and opamps: how can we best exploit the strengths of each, and which do you preferr for a given application?"   

My philosophy is to use WHATEVER given active or passive element where i think it will work best... I have no prejudice against any just for the sake of hating it.  I would personally never put an opamp in a situation where it was not abslutely needed, or a tube or transistor for that matter unless i was just doing it for personal amusement.  Some guys just beat an opamp or discreet into a circuit out of principle.  i definately favor tubes and FETs for signal amplification and nonlinear applications like clipping, and discreet BJTs and MOSFETs in power supplies/CCS and anywhere i need a lot of Gm or to swing a bunch of current around.  i like IC opamps in the DC realm (servo loops etc) oscillators, and to simplify squeeky clean linear audio stages where they are in no risk of being driven into saturation, and wherever a RIDICULOUSLY high voltage gain or speed is required.

[Jay Doyle]

Steben,

Thank you for your reply, but unfortunately I can't see where half of your answers respond to my comments in any way that I can properly respond. It may be a language thing, but at any rate I apologize. For example when you talk about ignoring feedback are you talking the 'standard' output to input resistor type or do you mean the source/emitter resistors as well? Because that can change an argument completely.

My point was only that an opamp isn't better than a bipolar or better than a JFET or MOSFET or any of them any better than any other: if you don't know the situation in which they are being used.

For example, your challenge of the 5532, driving a 600 ohm load with gain? Sure, you win (although, off the top of my head, a CCS loaded MOSFET input diff amp, driving a 2N5089 CCS loaded gain stage into a complimentary darlington buffer output might do the trick rather nicely and you could probably reduce the noise and get the same input impedence as the 5532 if you used 2N5088s). But a unity gain buffer into a nominal 10k load? A 2n2222 wins hands down in my book.

It is all in the situation.

[brett]

Hi
thanks for the posts concerning the Hidrosis (below).

I see you reduce the first gain back with at max settings a resulting gain of 1.25, giving a total of around 580. Yet you bring back a first gain (yet added noise from opamp) and you do it with noise-inducing resistors, in combination with a following high gain. If that doesn't give noise, nothing does.

Does it doesn't really make for a super-noisey circuit to reduce gain at an early stage with a volume/gain control?  Most tube amps seem to do that (?)  As far as I'm aware, that decreases both the signal and the noise from the previous stage (proportionally).  Sure, it adds a little bit of it's own noise. but I've used a metal film resistor and a small value pot (10k) to minimise thermal noise (amps often use 1M pots, which would be much noisier, wouldn't they?).  I guess I could go to a 3.3k resistor and 1k pot for even less thermal noise, but I figured that these components weren't adding much noise anyway. (I use the pedal with this pot set to 50 to 80%, if that makes any difference).

Don't forget that a guitar's own noise is audible at those gain levels, it is even giving the main source of noise.

That's kinda obvious.  My guitars are a separate issue. I can live with their noise.
I simply want to quieten this noisey op-amp circuit.  Racedriver205 indicated that this was easy because quietness was a charactistic advantage of op-amps over discrete circuits.  I'm hoping that it true.

If you bring that 1M resistor back to 100k to 150k, the noise will be gone without a lot of difference.

I wish this were true.  I tried a 330k in that position and the circuit lost many of its great characteristics (gain was slashed for one!).  Of course, there's some interactions with the feedback cap, too, but the 1M can't be simply swapped out.  

There seems to be some interaction with the internals of the op-amp that prevents me from down-scaling the 120pF smoothing cap and the resistances around it.  Otherwise, everything could change by an order of magnitude and the feedback resistor could drop to 100k.

Any help will  bbe much appreciated.  Except for the noise, this is a great circuit.


 

[Ardric]

There seems to be some interaction with the internals of the op-amp that prevents me from down-scaling the 120pF smoothing cap and the resistances around it.  Otherwise, everything could change by an order of magnitude and the feedback resistor could drop to 100k.

If you changed the 2.2k/2.2u to 220/22u, won't the feedback network will be too low an impedence to drive correctly with a TL072?

I'm also a bit suspicious of the 2nd opamp's biasing through the pot.  I'd give another coupling cap and bias resistor a try.

[Steben]

Steben,

Thank you for your reply, but unfortunately I can't see where half of your answers respond to my comments in any way that I can properly respond. It may be a language thing, but at any rate I apologize. For example when you talk about ignoring feedback are you talking the 'standard' output to input resistor type or do you mean the source/emitter resistors as well? Because that can change an argument completely.

My point was only that an opamp isn't better than a bipolar or better than a JFET or MOSFET or any of them any better than any other: if you don't know the situation in which they are being used.

For example, your challenge of the 5532, driving a 600 ohm load with gain? Sure, you win (although, off the top of my head, a CCS loaded MOSFET input diff amp, driving a 2N5089 CCS loaded gain stage into a complimentary darlington buffer output might do the trick rather nicely and you could probably reduce the noise and get the same input impedence as the 5532 if you used 2N5088s). But a unity gain buffer into a nominal 10k load? A 2n2222 wins hands down in my book.

It is all in the situation.

The buffer situation might be true. In the other situation you make an alternative solution which is in fact a discrete build-up of another opamp, no?

But you're right: nothing is better if there is no context. JFET's and CMOS seem to be superior if music/distortion is required. Bipolars and OPAMPS are superior when HiFi is required. This "division" has grown because of a lot of reasons of which I mentioned some.

[Steben]

Hi
thanks for the posts concerning the Hidrosis (below).

I see you reduce the first gain back with at max settings a resulting gain of 1.25, giving a total of around 580. Yet you bring back a first gain (yet added noise from opamp) and you do it with noise-inducing resistors, in combination with a following high gain. If that doesn't give noise, nothing does.

Does it doesn't really make for a super-noisey circuit to reduce gain at an early stage with a volume/gain control?  Most tube amps seem to do that (?)  As far as I'm aware, that decreases both the signal and the noise from the previous stage (proportionally).  Sure, it adds a little bit of it's own noise. but I've used a metal film resistor and a small value pot (10k) to minimise thermal noise (amps often use 1M pots, which would be much noisier, wouldn't they?).  I guess I could go to a 3.3k resistor and 1k pot for even less thermal noise, but I figured that these components weren't adding much noise anyway. (I use the pedal with this pot set to 50 to 80%, if that makes any difference).

Not convinced, brett. The gain up front trick has its value, but only if you USE that gain, without reducing it. Let me try to explain.
Let's take an analytical look with very exagerated numbers.

Let's say your input is 1 for example, with 0.01 noise (1%). Imagine there are 3 amplifiers with each three stages. The first has three stages with same gain 10 (1000 total). the second has stages with a gain of 2, 5, and 100 (total 1000). The Last one has 100,5 and 2. Each stage gives 0.01 extra noise too (resistors+opamp).
then we have:

Amp 1
input:  1 + 0.01
1:      10 + 0.10 +0.01
2:    100 + 1      +0.1  +0.01
3:  1000 + 10     + 1   + 0.1  + 0.01
total 1000 + 11.11 noise (1.111%)

Amp 1
input:  1 + 0.01
1:       2 + 0.02 +0.01
2:     10 + 0.1   +0.05  +0.01
3:  1000 + 10    +5      +1     + 0.01
total 1000 + 16.01 noise (1.601%)

Amp 3
input:  1 + 0.01
1:    100 + 1     +0.01
2:    500 + 5     +0.05  + 0.01
3:  1000 + 10   +0.1    + 0.02     + 0.01
total 1000 + 10.31 noise (1.031%)

As you can see it is, with the same components, always bad to put gain at the end.
the best is to put the most gain up front and step down with each stage.

In your case it is tough, since you have the TS style amp coming after. Maybe you can use a higher treshold (LED's, zener) in combination with a decent gain in the first stage.

[brett]

Hi
thanks for the suggestions.
I didn't explain at the start that the first pot in the Hidrosis isn't just a gain pot.  I'm not sure whether you noticed that.  It simultaneously affects tone and gain.  While it's not unique, it is a  feature of the circuit that is really good when combined with the high-gain clipper in stage 2.
cheers