Measuring crossover distortion

[Martin B. Kantola]

How would you do it? Is there a good way? We're not talking power amps, but OP-amps...

Martin

 

[NewYorkDave]

Inject a sine wave into your amplifier, take the output to channel 1 of a scope and to the input of a THD analyzer. Connect the THD analyzer output to channel 2 of the scope, trigger off channel 1 and superimpose the waveforms. If you see little "spikes" at the zero-crossings, that's crossover distortion. Be sure to check it at low output levels; that's where it usually shows up.

 

[JohnRoberts]

Crossover distortion will be most apparent at high frequency, because it is related to the transition time from upper and lower devices in class B output. It will be more apparent at low level but above the class A bias threshold if class AB. It will also be somewhat aggravated by load.

JR

 

[bcarso]

Looking at the analyzer residual per NYD is definitely the best way IMO. I saw something recently where someone asserted that the rising distortion at lower levels on a plot of THD vs. level was evidence of crossover distortion. But it looked more like something due to the increasing contribution of noise. A look at the residuals would have made things clear.

 

[JohnRoberts]

I used to run my distortion analyzer product output into my spectrum analyzer input.. useful since that old POS analyzer only had a 50 dB dynamic range...

You do what you gotta do, when you're paying for your own test equipment.

JR

 

[pstamler]

Do like Dave says. And make sure you test it into the lowest possible impedance the opamp would be expected to drive (don't forget the feedback network).

Peace,
Paul

 

[Samuel Groner]

Do you want to investigate a specific device in general or a specific application? If it's the former you want to make sure that you use inverting mode to avoide CM errors and to run the part at a noise gain of ~40 dB to get well ahead of analyzer residual (see OpAmp_Test_Jig_r1.pdf, test 4, and make R403 100 ohm). With modern quality parts it might still happen that you see nothing more than just noise, at least below 5 kHz or so. As already suggest, spectral analysis will bring you further. In addition to this I might suggest the use of a digital scope which allows you to average the residual.

Samuel

 

[Martin B. Kantola]

Thanks everyone, this is so helpful! Since my old POS analyzer doesn't work with these low levels of distortion I didn't cross my mind that I could use an even older POS, which is simply a passive notch and then some (SE) amp to amplify my residual... right?...

...and then look at it on the scope triggered as Sir NYD suggested! Even have a look at the spectrum of it, but I'm only interested in how much cross-over distortion there is at this point. With music playing at full spectrum I believe I can hear it clearly, but it would be great to confirm it with some basic measurements on a single frequency. Will definitely try higher than 1kHz too!

Samuel, thanks for the test rig schematic, still working with the OP-amps in my console, so I'll keep them in the circuitry for now, but if I can't find one that works as a straight plug-in I might have to dig deeper. Much deeper.

Martin

 

[Samuel Groner]

Just for the records two notes:

Be sure to check it at low output levels; that's where it usually shows up.

Absolute level of crossover distortion is highest at high amplitudes; just relative distortion level usually increases with decreasing amplitude, although there might be a point where this changes again.

Crossover distortion will be most apparent at high frequency, because it is related to the transition time from upper and lower devices in class B output.

Switch-off-distortion due to finite transistor switching times is indeed a problem with power transistors (at least with standard parts). However I doubt that with todays GHz processes it is a serious IC opamp problem within the audio range. Crossover distortion will increase with frequency nonetheless as loop gain decreases.

Samuel

 

[Guest]

I whisper in front of a microphone gradually increasing a distance. The higher is level of overall crossover distortions, the sharper is dependence of a loudness on a distance. However, it requires some skills, but the main idea is, I made it for listening, but not for measuring of specs. To compare OpAmps it is better to have a dozen of them connected in a chain, with unity gain each, to hear differences, but different types of OpAmps in different strips of the same console is enough to compare and select them. Experimenting this way I've found some surprising results...

 

[JohnRoberts]

Didn't Raytheon suffer from a "purple plague" (?) or something like that process problem... I never used other than signetics 5532 which IIRC licensed that design from Phillips originally. IIRC Raytheon had a quad OA (4136) that was popular for a while, before bifets quashed them like a bug...

I believe a special house part number at my last day job might have been a raytheon 5532 that was selected for noise (spectra). These parts were pretty good, but this was years after the process problem, and the noise selection would probably weed out problem parts.

JR

 

[SonsOfThunder]

All I can tell ya is how I do it on power amps...

Low level signal (start with 100mV move to 50mV then to 10mV) at 10kHz. Look at the output on a scope, it will be plainly visible if you have one or two complete cycles on the display.

You can look at the residual from a THD analyzer on a scope, but if you don't know what you are looking for...?

Hope that is helpful...at least a little.

 

[Martin B. Kantola]

[quote author="Wavebourn"]I whisper in front of a microphone gradually increasing a distance. The higher is level of overall crossover distortions, the sharper is dependence of a loudness on a distance. [/quote]

Think I understand what you mean... Had a similar experience, with all instruments suddenly getting more sustain when I swapped to better pre-amps in my studio. Almost spooky.

But I really want to try to measure too. Your idea with a chain of them is good, that's why I put 10 of those quads in each channel for listening tests. Best (worst) case the signal is almost passing 40 op-amps!

Martin