OpAmp Measurement Series

[Samuel Groner]

Hi

Over the last months I finally found enough spare time to make considerable progress with a project begun some years back--the systematic measurement of opamps with respect to distortion. Personally I found doing these measurements very enlightening, they told me much about IC opamp limitations and the suitability of opamp topologies. Even the best IC opamps showed deficiencies with one or the other respect which need careful attention in implementation; good to have another confirmation that even today designing quality audio circuits is more than just applying text book knowledge.

Here we go: opamp_distortion.pdf (large file!)

This ain't the end of the research; I'll add more measurements as time permits and the appendix will include more information hopefully soon. Updates of the document will be announced in this thread.

Samuel

 

[audiox]

Great job! And very good looking document. Made by Latex?

You probably know that Douglas Self has done something similar (not as in-depth as you however).
http://www.dself.dsl.pipex.com/ampins/webbop/opamp.htm

 

[radiance]

Thanks Samuel :thumb:

 

[bcarso]

Very happy to see this public. It is high time the manufacturers who claim low distortion are held accountable for real-world perfromance, especially the common-mode behavior.

I'm dealing with a S&M type right now (that's Sales and Marketing you dirty-minded ones out there) who goes nutz over single THD number specs. I am slowly trying to persuade him to look at more than that, but also to peer deeper into claims on datasheets. This AM he asserted that National had an amplifier that did 0.0003%. Well, they have a driver chip that supposedly does this...and so what?

Meanwhile he likes teeny and high-efficiency, and it is difficult to get him to understand that switchmode designs are waaaay harder to extract performance from at class A or AB-achievable levels.

 

[clintrubber]

[quote author="bcarso"]who goes nutz over single THD number specs..[/quote]
Her/she/it must be from Japan then, not ? Been involved in a project where certain irrelevant specifications had become a holy grail for them, and then try to persuade them it's not about that - at all....
... for which we have the description 'uphill battle' :roll: :? :wink:

[quote author="bcarso"] I'm dealing with a S&M type right now (that's Sales and Marketing you dirty-minded ones out there)[/quote]M/S makes me just think of a micing-setup, but thanks for the hint to call the decoder box the S/M-thingie :wink:

Thanks Samuel for sharing this. To be honest, I'm not yet sure where this will fit into the kind of music I'd love to make & the gear I'm using for it to record it, but from a technical point of view it's potentially mighty interesting and that's not in jest.

Regards,

Peter

 

[JohnRoberts]

Good job Sam.. That looks like a formidable investment in time. Standardized 3rd party testing is hugely useful when manufacturers may selectively provide specs that make them look good while ignoring ones that don't.

JR

PS: Back when I was going through my analytical phase I wanted to invent a spec for slew linearity, but you can impute that from other existing tests.

 

[syn]

Thank you Samuel :guinness: :thumb:

 

[Samuel Groner]

Thanks for your interest.

Made by Latex?

Sure.

You probably know that Douglas Self has done something similar.

Yes, but as you noted a single sweep just ain't gone tell the whole story...

I'm not yet sure where this will fit into the kind of music I'd love to make & the gear I'm using for it to record it.

It's funny that personally the knowledge doesn't bother me at all; I got the chance of currently devoting perhaps 80% of my live to music (both as active musician and recording producer), and I simply don't feel bad about using my equipment which at least partially ain't up to the highest standards. But of course it doesn't stop me from trying to progress on the technical side of things.

Samuel

 

[clintrubber]

[quote author="Samuel Groner"]

I'm not yet sure where this will fit into the kind of music I'd love to make & the gear I'm using for it to record it.

and I simply don't feel bad about using my equipment which at least partially ain't up to the highest standards. But of course it doesn't stop me from trying to progress on the technical side of things.[/quote]
That's the perfect situation I'd say, this way it can be fun on both sides, not the one getting spoilt by some awareness that things should actually be better and can't really work since this&that can be better because of etc etc :thumb:

FWIW, see an analogy for being happy (-enough) using DAW-plugins and substituting when the real hardware version is finished.

Regards,

Peter

 

[Animatic]

Great paper.
I have a couple of AD797's coming to test,
and was really happy to see them top of list.

Does anyone have any good PCB's for these as buildin blocks
or PCB layouts for making them?

Bcarso with those marketing types in you life you need to read

The Tao of Opamps, Sysiphus version ...
:grin:

I just got two AD chips for testing this afternoon.
SSM2141P and SSM2142P
As longer lines, driver and receiver.

 

[emrr]

Very nice work, Sir!

 

[jdbakker]

Awe-inspiring is all I can say.

Great stuff,

JDB.

 

[hifizen]

Wow. Great paper, that must have been a lot of work! :shock: This one's definitely going into my archives for future reference. Sincere thanks for posting this! :thumb:

Doug Self's work is very interesting. His series of articles on power amplifier distortion, published in Electronics World & Wireless World around '93-'94 was my first deep dive into the inner workings of the conventional amplifier circuit. Seminal work, I believe (for me, anyway). Apart from exposing some of the more subtle distortion mechanisms within the amplifier circuit, it was for me, a great way to simply learn about how amplifier circuits work... He breaks the amplifier circuit down into pieces and discusses in some depth how each piece functions. Those articles, along with a few others, were later compiled and published together as "Self on Audio"... I would recommend it if you can find a copy. Dr. Self has a more recent book, which probably covers much the same material though I haven't read it.

On the subject of IC opamps, I have been interested in 'transient' distortion mechanisms for a few years now. As Samuel notes, static THD measurements certainly have their value, but modern high-performance amplifiers have all but eradicated THD as a factor in audio performance, and "a single sweep just ain't gone tell the whole story". IC's in particular may be subject to thermal distortions, given the tiny mass (milligrams) of an opamp wafer, and the fact that the power-dissipating output transistors are integrated onto that die. The thermal time constant can be very short... as low as 1ms from what I've read.

But information on this topic seems a bit scarce, and I lack the means to appropriately measure these effects. What I would love to try, is a burst-stimulus measurement, where a low-amplitude distortion measurement is taken on the 'cool' die. Then apply a high-amplitude burst into a low-Z load, followed immediately (within milliseconds) by a very quick low-amplitude distortion measurement to compare with the initial measurement. In this way, one might observe the effects of the wafer's thermal tail - changes in the distortion level / spectra due to parametric changes in the on-die transistors, etc. Seems to me this kind of effect would be highly relevant to music signals, which are typically high dynamic range. If true, it might help explain the apparent subjective superiority of discrete circuits vs. IC opamps. One day, I'll get to this.

Yet another interesting area is microphonics. IC wafers can indeed be piezoelectric, and I've seen at least anecdotal evidence that IC opamps can be highly microphonic. One report posted to a forum somewhere was illustrating an example where an opamp was generating substantial electrical output (tens of millivolts, if memory serves), with the tap of a pencil eraser. The author of that post solved his problem by soldering the leads to short flexible wires which suspended the chip up off the PCB. Yet another fascinating area to explore.

I'm curious what your thoughts are on this (Sam, or whomever else is reading)...

Anyway, sorry to ramble... I love this stuff! Thanks again, Sam.

 

[hifizen]

[quote author="Animatic"]The Tao of Opamps, Sysiphus version ...
:grin: [/quote]

Link? Me searchie, no find. :?

 

[jensenmann]

Wow, Samuel, thanks a lot. What an interesting paper. :thumb:

 

[Animatic]

Sysiphus is the greek guy doomed to for eternity,
to roll a large rock up a mountain top,
only to see, at the last minute, it roll to the bottom,
and have to start over again.

Which sounded like Bcarso's thing with marketing people.
And then someone else's thing Japanese, who are partly Taoists,
with their distortion specs mono-mania.
Sysiphus, he certainly could use some taoism to keep calm
with his day job, so it all sounded like a perfect book title.

But it was strictly humor.

 

[JohnRoberts]

I thought more of Diogenes than Sissiphus...(seeker of truth) but Sam not Brad.

On the subject of transient distortions... this is somewhat complicated by definitions. What is bandpass, etc. I kicked around obscure distortion measure to parse out unknown distortions but I'm not really sure there's unfound stuff there.

Simple sine wave testing has always bothered me since common error terms are sinusoidal merely phase shifted so concealed by most testing.

One idea I have kicked around for years as an in-band and very appropriate test signal is real music, vocals, instruments whatever.

Use a digital recorded source file that is robust. Then compare that digital source file to the playback from the device under test. Any and all deviation, other than pure delay, is distortion. For years there were questions about resolution of such a test but modern digital is getting pretty good, and we can listen to a null product to learn where a circuit has trouble keeping up.

Or not...

JR

 

[hifizen]

heh... Some things just don't sink in at 1am. :roll:

 

[Animatic]

John good point.

Digital cross analysis of the purest Oboe tones we can find
sampled REALLY high, might make for some interesting test results.
A low C from a Bosendorfer and some cymbals too.

 

[pstamler]

The problem with input-vs-output tests, whether done by digital recording or direct comparison in the analog world, is that any phase shift in the device under test swamps other differences. Peter Walker of QUAD once proposed a "crooked wire bypass test", where the output of an amplifier was compared with the input, but the latter had been passed through a passive circuit with phase-shift characteristics identical to the amplifier. Designing the passive circuit proved to be quite a challenge; he did the test on one amplifier then seemingly dropped the matter.

Peace,
Paul