OpAmp Measurement Series

[pstamler]

[quote author="JohnRoberts"]+1 for multi-tone IMD, I like the real world aspect of in band stimulus and in band distortion artifacts.

The power of Sam's effort is that he has applied a common or standardized test criteria across so many devices, so we can compare apples to apples. These discussions don't diminish that effort.[/quote]

Indeed not; his work is nothing short of magnificent. I suggest only that adding a HF IM test to the battery could exercise high-frequency distortion mechanisms in a way that provides remarkably graphic results. I'd add that these test have the advantage that they are doable with a simple, relatively cheap setup -- a competent sound card that will do full duplex recording and a DAW program with a spectrum analyzer option.

Peace,
Paul

 

[JohnRoberts]

I used two tone IMD at similar freq to what you used a few decades ago so I appreciate the merit for revealing flaws in circuit design. I am a little concerned that your simple test rig, may not reveal non-linearity floors in some of the better new opamps. My home brew IMD analyzer only looked at the in band products which was a lot easier to do 30 years ago.

I don't know the capability of Sam's test bench but I vote for one of the slicker 3 tone IMD test series, that is designed to reveal artifacts concealed by simpler series.

Perhaps it would be nice to have a wiki like website (or sub forum here?) where people could contribute their own test series results, but that could become babalicious without Sam's disciplined approach.

JR

 

[Samuel Groner]

I suspect that the data Paul presents does not show the difference between two measurement methods but rather the linearity difference of the DUTs at low and high frequencies. Running a standard THD+N measurement at 19 kHz would reveal a similar difference of the two DUTs.

Samuel

 

[JohnRoberts]

[quote author="Samuel Groner"]I suspect that the data Paul presents does not show the difference between two measurement methods but rather the linearity difference of the DUTs at low and high frequencies. Running a standard THD+N measurement at 19 kHz would reveal a similar difference of the two DUTs.

Samuel[/quote]

While I agree IMD is generally a different measure of the same nonlinearity, a weakness IMO of THD testing is that the products occur at higher harmonics, so HF distortion can fall outside the bandpass of the device being tested and understate this nonlinearity. I experienced this decades ago, in specific designs.

I also agree for devices with GBW much larger than audio, measured flat and wide band, simple THD+N is a useful metric.

I am not lobbying to make more work for you... I appreciate the work you did and that reference is useful as is for selecting opamps.

My interest in fully in-band testing is to characterize more complex complete audio paths. Simple THD at 19kHZ is IMO not adequate to fully characterize a digital CODEC designed for audio.

JR

 

[Samuel Groner]

Sure, bandwidth limiting is an issue. I agree that the distortion measurement around 10 kHz is not entirely convincing in the paper, particularly for low-GBW devices; at the 60 dB noise gain tests distortion might be attenuated by the low-pass action of the device bandwidth and for the high-frequency tests (at low noise gain) the System One residual starts to mask things. However, I'd rather prefer to run a THD+N measurement at lower noise gain (40 dB) for this than to use another measurement method (which would anyway need lower noise gain as well to not attenuate the test signal). If I get a more modern AP with their much lower residual I will consider redoing things (i.e. replacing all 60 dB noise gain tests with 40 dB ones), but currently this is no option.

BTW, I'm happy to announce that the next revision (scheduled for end of November) will include more discrete opamps; several manufacturers have kindly agreed to send samples--thanks!

Samuel

 

[JohnRoberts]

Again.. I am not being critical of your data set. It is useful as is for people who know how to interpret such things, and probably for those who don't.

Thank you for the effort, and be gentle with the tweaker DOAs. :roll: I look forward to see how the Deane Jensen DOA design holds up to modern ICs.

JR

 

[Animatic]

Sam and John. It is most enlightening to see your dual push pull discussion.
And SO much better to have it here in the forum.
Sam great work so far, and I am sure into the future.

My vote is for triple tones musically 1 3 5 but in different octaves such that
each can show it's affect above it's entry point before the next comes in above it.
C or A440 then it's 5th one or two octaves up , then 3rd up higher again.
I think measurements with some basis in the musical usage that
MOST of us will use the opamps for would be a valid and valuble additions.
And no doubt a heck of a lot of work... But still, for me this would be a
good addition. Even just a screen shot of a objectively standard triple tone
might be enlightening.

Preamps and EQ's are not used in a technical vacuum, as you well know,
but interacting with natures physics also.
Just some thoughts, and thanks gentlemen!

 

[bcarso]

For those of you with a bent for the maths, or if you just have a simulator with fourier capability, it is quite instructive to look at the effects of certain "pure" nonlinear operators on harmonic and IM distortion.

A pure cubing operator is quite interesting. A pure multiplier, connected as a squaring operator, is also good fun.

 

[Samuel Groner]

New revision is online, see link in the first post. The list has now grown to 37 tested amplifiers. There are several more interesting opamps (IC and discrete) waiting, will get to them at the end of the year.

Samuel

 

[Animatic]

Great job Sam,
you are doing our Xmas shopping for us!  ;D

 

[JohnRoberts]

Again, thank you for the good work..

JR

 

[JohnRoberts]

For those of you with a bent for the maths, or if you just have a simulator with fourier capability, it is quite instructive to look at the effects of certain "pure" nonlinear operators on harmonic and IM distortion.

A pure cubing operator is quite interesting.  A pure multiplier, connected as a squaring operator, is also good fun.

Smarter folks than I have proposed unique series of tones to better reveal circuitry flaws. The attraction to me is the use of in all band test stimulus to excite in band errors. I have always been left wanting with simple THD, while for many circuits you can extrapolate from HF THD performance, and the alphabet soup of tim, sim, etc was mostly slew rate performance.

JR

 

[bcarso]

As I was trying to explain to a biz venture partner why a single THD number was not the whole story, I decided to formulate some examples.  After I sent the email I did some additional numerology on various hypothetical mixed signals, and was once again brought face-to-face with just how complicated the subject is.  An added twist is the clash between equal temperament and the overtone series.

 

[Animatic]

ooh, oooh,
ET vs JI, like being in the steel guitar forum.
NEVER a good agreement and mostly caused
broken heads and friendships.

I DID have those thoughts when I posted about I, III, V tone series.

I would be content with a perfect fifth and a barely beating third.
But others would go ballistic at the thought.
But I still think they are valid because the harmonics in real world
are not pure like a test signal. Even a perfect fifth can cause head aches.
Still give me a single good oboe 192k 24bit sample and see where it leads.

 

[bcarso]

ooh, oooh,
ET vs JI, like being in the steel guitar forum.
NEVER a good agreement and mostly caused
broken heads and friendships.

I DID have those thoughts when I posted about I, III, V tone series.

I would be content with a perfect fifth and a barely beating third.
But others would go ballistic at the thought.
But I still think they are valid because the harmonics in real world
are not pure like a test signal. Even a perfect fifth can cause head aches.
Still give me a single good oboe 192k 24bit sample and see where it leads.

The other effect that adds to the confusion is...fusion---when you have the absence of beats, of multiple components into what is perceived as a single one with a certain timbre.

One ex-friend (ex- not on account of arguments about tuning systems or distortion, but other erratic behavior) pointed out how much more sense the music of Frescobaldi made when played with the original temperament systems, and I had to agree.

There is also a good book out now:  How Equal Temperament Ruined Harmony (and Why You Should Care), by Ross Duffin (0393062279).

 

[Animatic]

ooh, oooh,
ET vs JI, like being in the steel guitar forum.
NEVER a good agreement and mostly caused
broken heads and friendships.

I DID have those thoughts when I posted about I, III, V tone series.

I would be content with a perfect fifth and a barely beating third.
But others would go ballistic at the thought.
But I still think they are valid because the harmonics in real world
are not pure like a test signal. Even a perfect fifth can cause head aches.
Still give me a single good oboe 192k 24bit sample and see where it leads.

The other effect that adds to the confusion is...fusion---when you have the absence of beats, of multiple components into what is perceived as a single one with a certain timbre.

One ex-friend (ex- not on account of arguments about tuning systems or distortion, but other erratic behavior) pointed out how much more sense the music of Frescobaldi made when played with the original temperament systems, and I had to agree.

There is also a good book out now:  How Equal Temperament Ruined Harmony (and Why You Should Care), by Ross Duffin (0393062279).

Yes, good book I read it twice!
I also agree that playing period music in the temperment it was written for is best.

One reason I keep coming back to the Oboe tone is because it's harmonics
are pretty natural in series, an organic sound stack so to speak,
as opposed to a pure electronic sine tone.

 

[Viitalahde]

I haven't thanked yet, so I'll be doing that now.

Thank you for this, Samuel!

I'm looking at op-amp alternatives for a new parametric EQ I'm working on, and it all seems pretty obvious that if you want to go above 5534's in performance, there are not many performers out there that would do it. For the state variable filters, AD797 I/O and OPA627 integrators still seem like a solid choice..

 

[Samuel Groner]

You're welcome. Some additional comments if you want to use the AD797 as integrator: the datasheet suggest the use of a 100 Ohm resistor in series with the feedback capacitor to maintain stability. What they did not mention is that you need to increase the value to 150 Ohm if the 50 pF distortion cancellation capacitor is used. An they don't mention either that alternatively you could place a 220 Ohm resistor in series with the inverting input--this makes the integrator much closer to ideal and would surely be the way to go for an EQ. If you don't like the additional noise/offset from this resistor you can experiment with a paralleled inductor (start around 22 uH).

Samuel

 

[bovox]

Hey, I´m not the right guy to make any comments,on your fine measuring. But I react of the results on the FF-993. I dont know the interior design of this amp,but i suppose it comes out of the Fet 990 (the millenia var.)
If we look at the FF-992 .It says that the open loop gain is 68 dB! If we suppose that the 993 has about the same, they are not intended ,without high THD, to work well at this high gain (60dB).
It really,had been nice to see the figures at 40 dB gain.
Is there any answers why this amp has that lower dist. at rails +-24v instead of those +-15V?
Cheers Bo

 

[Samuel Groner]

It really had been nice to see the figures at 40 dB gain.

You can simply divide the shown figures by 10 to get the results one would get (up to the small-signal frequency limit indicated) when running the measurements at 40 dB noise gain.

Samuel