A Different Opamp Compensation Technique.

[Mooly]

I wondered whether to start a new thread for this or tag this on to a (very) old thread. I've decided on new but will cross link the other to this as it touched on the subject...

So what's this all about. Well a comment in the linked thread by Nelson Pass grabbed my attention. Nelson commented that when using opamps in low or unity gain configurations that he liked to "throw" some of the available open loop gain away in the interests of "stability" as many opamps under unity gain are running close to their stability margin. The result of doing that was a subjective improvement in sound quality.

The whole of that thread is interesting reading and so I would recommend anyone reading this to go and have a read at the other one too.

opamp inverting input sounds better?
-----------------------------------------------------------------------------------------------------------------------------------------------


There are 4 groups of shots below, each with the relevant circuit. The opamp was the OPA134 in all cases running on -/+12 volts. The input frequency was 40Khz and the circuit tested at two levels, first 10 volts pk-pk output, and then the input was reduced by 40db to give 10 mv pk-pk output. This was done as non-linearities and differences can occur as levels increase/decrease.

Given a choice, which "compensation" technique would offer the best results do you think... and I don't know, I haven't tried this for real. I was wondering whether it would be worth trying on my amp which uses two opamps in inverting configuration.

You'll notice that the "normal" method with the 22pF cap is very very similar to the one with no cap and a value of 620 ohms for Rx. Would it be subjectively better ? Or might a combination of the two techniques give the best subjective results ?

I don't even know what you call this method. I've seen it referred to as "increasing the noise gain" and it's a well known technique (with lower values of Rx) for increasing noise/distortion levels of the best opamps to make those parameters more easily measurable.

[Mooly]

Here's the normal compensation,

[Mooly]

And here is the first of the alternatives,

[Mooly]

And here is the final one with 620 ohms. Should just add that the two traces in all are input and output for comparison.

[bear]

Interesting.

Wondering if the output Z of the sig gen plays a role in this? (seems not, but...)

Also could you tweak the pf and get the same looking result?

Also why would a lower R result in over compensation??

Was there a load on the opamp?

I have no answers, btw...

_-_-bear

[kipman725]

signal generator terminated correctly?
instead of scope probe use resistor divider with 50ohm characteristic impedance loading the opamps output and set scope to 50ohm mode, use BNC soldered to output of divider. I don't believe scope probes on switching transients.

[Mooly]

Generator has a 50 ohm output impedance.

The top trace is measured across "In" on the circuit.

No load on the opamp output other than 10/1 probe.

Tweaking the compensation cap gives the same "looking" result at a given frequecy. You can see that in circuit 2 and circuit 4.

Quote // "I don't believe scope probes on switching transients."

If it were high frequency and high current/inductive loading etc then I would agree but I don't see a problem here.

[knutn]

You are not compensating anything by adding a resistor between the inverting and non-inverting input. In fact you are adding a capacitor between these inputs. Just look at your scope probe specifications to see how large this capacitor is.

[DF96]

One question to ask is to what extent is the 'ringing' seen in the first plot due to verging on instability, rather than some approximation of brick-wall filtering of a square wave. Have you measured the frequency response? Does it have a significant peak, or merely a sudden drop?

Putting a resistor between the inputs will shunt the stray capacitance there and so reduce loop HF phase shift. It may also help equalise the source impedance seen by the two inputs which may increase open loop linearity, while at the same time reducing loop gain which will reduce closed-loop linearity. The net result might be a reduction in higher order products from feedback mixing.

[Mooly]

Originally Posted by knutn

You are not compensating anything by adding a resistor between the inverting and non-inverting input. In fact you are adding a capacitor between these inputs. Just look at your scope probe specifications to see how large this capacitor is.

The scope probes don't come into contact with the opamp input pins, only the output pin and the input to the circuit (the generator output)

Originally Posted by DF96

One question to ask is to what extent is the 'ringing' seen in the first plot due to verging on instability, rather than some approximation of brick-wall filtering of a square wave. Have you measured the frequency response? Does it have a significant peak, or merely a sudden drop?

Putting a resistor between the inputs will shunt the stray capacitance there and so reduce loop HF phase shift. It may also help equalise the source impedance seen by the two inputs which may increase open loop linearity, while at the same time reducing loop gain which will reduce closed-loop linearity. The net result might be a reduction in higher order products from feedback mixing.

Yes there is a peak. With the output set to 10 volts pk/pk midband as the frequency rises from around 600khz there is a gentle rise followed by a sharp "switch" to around 17volts pk pk output at around 870khz followed by a roll off. The point at which this increase occurs is sudden and there is hysteresis in that the frequency has to be reduced a little before it "recovers" back down.

I'm going to post some more shots that might be of interest shortly...