Qualities of opamp output signal?

[Svart]

I've been doing a lot of opamp simulations lately. Yes simulations, but i'm finding that the spice models are close enough to really work and are fairly close in real life.

that being said, I'll get into my questions.

I've noticed that a few models have a tendency to round the leading edge of the output signal(1khz Square) and the rise/fall times are slightly longer than the input signal. While this is more pronounced at high frequencies, they can be made to exhibit this at much lower frequencies through various methods(more output capacitence and such).

I have not explored much further but it seems that some opamps are much more prone to this behavior than others. The AD810 is more prone to do this with more output capacitence, while the ne5534 tends to have a crazy overshoot and ring with the same load.

So, other than testing with prototypes, would there be any interest in actually tailoring a certain signal responce from an opamp to change the qualities of the signal itself? I could see this giving the "color" of the opamp, something we talk about a lot around here.

any thoughts?

 

[Svart]

by the way I found some AD810s in some old video gear in the junk bin. These are pretty damn good. hint hint..

 

[tk@halmi]

The leading/trailing edge rounding is a related to slew rate and charging of intrinsic and/or compensation capacitors. If the circuit is stable at all frequencies you see the slope and rounding at the edges. Sometimes the rounding is masked by the instability (overshoot).
The NE5534 has external compensation so you could increase the capacitance from the recommended 22pF to something larger to decrease slew rate and give it more color. How about a variable capacitor for adjustment to taste?

 

[Svart]

In unity/+1 gain setups I see the effect when driving capacitence of a few hundred pf to a few thousand pf, essentially out pin to ground through a cap.

the feedback capacitor circuit causes the ne5534 to have a little ring on the leading edge of a 1mhz square, getting worse with load.

the AD810 is current feedback, gain of +1= 1k feedback resistor. pretty damn stable too. the slew for this part is 1kv/us, usually used for video. i am attempting to use it for audio.. :shock: I tried putting a feedback cap even though currentfeedback opamps don't like this and the rounding of the leading edge is much more pronounced, almost a triangle wave.

Really I am just playing around and studying the effect of changes around the opamp and it's feedback circuit, finding optimal values and the like.

I guess I am also looking for some of the trademarks of certain opamps and what gives them their "sound". I noticed that the AD810's rise time is slightly longer than the ne5534 at most frequencies but more pronounced at higher freqs, even though it's 70 times "faster".

 

[bcarso]

If the models are correct the 810 ought to be blazingly faster than the 5534.

 

[Svart]

14v/us vs. 1000v/us


a little bit faster..

 

[tk@halmi]

[quote author="Svart"]14v/us vs. 1000v/us
a little bit faster..[/quote]

So much for models...

 

[PRR]

> rounding is a related to slew rate

In simple classic amplifiers: rounding at all levels is gain-bandwidth, slanty lines at high level are slew.

> a few hundred pf to a few thousand pf, essentially out pin to ground through a cap.

I believe you can find app-notes which explain why that is usually a bad idea.

We can't help driving capacitors, disguised as cables. Take feedback directly from the output pin, put 22 to 470 ohms from output pin to any capacitive load more than a few pFd. The resistor should be big enough that, even when the cap is a near-short, the chip has near-normal gain to the output pin and feedback loop. Beasts like 5532 can go under 100 ohms, wimps like TL072 need over 300 ohms. This can also be your cable termination and short-proofing resistor. Gain-bandwidth will be reduced: 100 ohms and 10,000pFd (300', 100m of cable) will be down 3dB at 170KHz, even if your chip is flat to cell-fone frequencies.

Modern models may simulate MHz settling effects well enough to pre-qualify chips for a specified application before you order samples. But, unless the model is very computationally expensive, it is unlikely to give the exact shape of the response: practical models don't use the "architecture" of actual chips.

 

[Svart]

another, likely simple, question..

The AD810 has an identical footprint for +in,-in, out pins. It has offset null as pin#1. The NE5534 has comp as pin#1. according to the data sheet for AD810, you can bias pin#1 for DC offset in a similar fashion as the NE5534..

nevermind I answered my own question. :green:

How about a 1kv/us opamp with DC servo? Does the speed of the servo amp matter in a case like this? I would think it would need to react as quickly as possible..