Intermodulation distortion tests - thoughts?

Hey folks. I was messing around today with three different preamps: my own tube pre, a SCA C84, and an AML ez1073. Did the 19kHz/20kHz signal IMD test and came up with these three different results. (See photo.)

All three preamps were calibrated with this same input and output levels.

So, what does all this mean? All three preamps exhibit extra bands to the left and right (evidence of IMD, right?) The C84 (Millennia) showing no 1k "beat" frequency? The extra bands around the Neve?

Just trying to understand this better.

BluegrassDan said:

Hey folks. I was messing around today with three different preamps: my own tube pre, a SCA C84, and an AML ez1073. Did the 19kHz/20kHz signal IMD test and came up with these three different results. (See photo.)

All three preamps were calibrated with this same input and output levels.

So, what does all this mean? All three preamps exhibit extra bands to the left and right (evidence of IMD, right?) The C84 (Millennia) showing no 1k "beat" frequency? The extra bands around the Neve?

Just trying to understand this better.

Click to expand...

The 1kHz is THE DISTORTION, not a beat.  Only the first path looks decent.

Spurs at 2Khz etc, seems even less linear.

Back in the 70s I rolled my own 19:20Khz IMD tester and found it very useful...  19k and 20Khz are both in band signals, and 1Khz is definitely inband audible distortion. Harmonic distortion for 19/20kHz is generally not audible. 


JR

BluegrassDan said:

Hey folks. I was messing around today with three different preamps: my own tube pre, a SCA C84, and an AML ez1073. Did the 19kHz/20kHz signal IMD test and came up with these three different results. (See photo.)

Click to expand...

It just shows that the Millenia deserves its reputation for pristine performance, that the Neve is undoubtedly "coloured", and that your preamp shows all the signs of a healthy tube design that has intrinsically low distortion and thus doesn't need heavy NFB.

Do you have the other common IMD test signals available (SMPTE 60/7kHz, DIN 250/8kHz, AES17) ?

Distortion characteristics are determined principally by the transfer function of the active devices used  in the preamp. In transistors this is basically a log law and in tubes it is a two thirds power law. This is reflected both in the harmonic and inter-modulation  distortion produced. Tubes are generally much more benign in their production of both types distortion and hence need much less NFB to obtain acceptable performance.

Cheers

Ian

IMO,  measurement of IMD using  SMPTE method (60Hz/7kHz 4:1) as proposed from trobbins gives  much more useful results.

Also IMO, if you are using ITU-R method (19/20kHz) you should know what's the level of THD for each frequency and you need to have a measurement equipment capable to measure in frequency range up to 100kHz. 
BTW, this method shows linearity problems at high frequencies, but which source you will ever record has so high energy at 19 or 20kHz?

From your results I can see that there is obvious that C84 is PP design, there is no 1kHz product, but 18 and 21kHz exist.
Neve 1073 is most probably  over-driven  with a test signal (output transformer is saturated?) at that frequencies.
Your tube preamp is fine, has measurement results as a good REDD47 at lower gain (higher NFB).

moamps said:

IMO,  measurement of IMD using  SMPTE method (60Hz/7kHz 4:1) as proposed from trobbins gives  much more useful results.

Click to expand...

I stopped use SMPTE IMD testing in the 70s because it was too easy for even circa 70s circuitry. 7kHz may be HF relative to 60Hz but is only roughly 33% the slew rate of 20kHz already considered desirable in the 70s. 

Typical active electronics with dominant pole compensation will have rising nonlinearity with frequency due to falling open loop gain.  20kHz is notably harder than 7kHz and it shows up in testing.

Also IMO, if you are using ITU-R method (19/20kHz) you should know what's the level of THD for each frequency and you need to have a measurement equipment capable to measure in frequency range up to 100kHz.

Click to expand...

International Telecommunication Union?  As I tried to say before, but maybe I wasn't clear, the beauty of two-tone IMD at 19k:20kHz is that both tones are arguably in band and the distortion is also very much inband (1kHz).

Your advice to use a much wider measurement bandwidth applied more to THD measurements that involve higher harmonics that for a 20kHz tone quickly rise to 80kHz and higher.

The linearity of the 19k and 20k stimulus is not very strict... Less than pure sources can cause some extra sidebands but not much error. Back in the 70s when I modified my heathkit SMPTE IMD bench unit to 19k:20k it was relatively straight forward and I found it very useful for revealing relative circuit path performance.

BTW, this method shows linearity problems at high frequencies, but which source you will ever record has so high energy at 19 or 20kHz?

Click to expand...

The obvious IMD trigger is close miked cymbal crashes... listen for LF crud riding under the cymbal hits. An old studio trick to parse out HF linearity is to jangle a key chain in front of a mic. Keys can easily have frequency content up at 40-50 kHz.

JR

From your results I can see that there is obvious that C84 is PP design, there is no 1kHz product, but 18 and 21kHz exist.
Neve 1073 is most probably  over-driven  with a test signal (output transformer is saturated?) at that frequencies.
Your tube preamp is fine, has measurement results as a good REDD47 at lower gain (higher NFB).

Click to expand...

JohnRoberts said:

The obvious IMD trigger is close miked cymbal crashes... listen for LF crud riding under the cymbal hits. An old studio trick to parse out HF linearity is to jangle a key chain in front of a mic. Keys can easily have frequency content up at 40-50 kHz.

Click to expand...

I do most of my recordings at 88.2k SR. Quite often I use Spectral Cleaning, a native Samplitude plug-in, that allows cleaning tracks by showing and editing their spectral content. Indeed many tracks show significant content up to 40kHz. IMO it's a good reason for using Double Speed; since the energy is rolling off, aliasing tones are much lower in amplitude.

JohnRoberts said:

....International Telecommunication Union?..

Click to expand...

https://www.rane.com/note145.html

"IMD -- ITU-R (CCIF). Intermodulation Distortion -- ITU-R Method
How is it measured? The common test signal is a pair of equal amplitude tones spaced 1 kHz apart. Nonlinearity in the unit causes intermodulation products between the two signals. These are found by subtracting the two tones to find the first location at 1 kHz, then subtracting the second tone from twice the first tone, and then turning around and subtracting the first tone from twice the second, and so on. Usually only the first two or three components are measured, but for the oft-seen case of 19 kHz and 20 kHz, only the 1 kHz component is measured."

  As I tried to say before, but maybe I wasn't clear, the beauty of two-tone IMD at 19k:20kHz is that both tones are arguably in band and the distortion is also very much inband (1kHz).

Click to expand...

Not always.

Audio Measurement Handbook - Bob Metzler (Audio Precision)

"This “twin tone” test permits stressing band-limited systems at their highest frequencies while still measuring an in-band IMD product. The most severe limitation of the simplified analysis technique is that it measures only the second order product. The simplified technique is thus not useful to measure distortion produced by non-linear transfer functions which are symmetrical about zero, such as the BH curves of magnetic tape; this method should not be used to measure tape recorders. The same twin-tone signal can be used with an FFT analyzer or other selective analyzer to measure IMD products of all orders and thus yield more information, but analysis of such information requires more skill than the simple “one number” readout of the simplified analysis technique."

The obvious IMD trigger is close miked cymbal crashes.

Click to expand...

There should be pad used in the first place.  If this signal is the cause of the high IMD in preamp, what the mess happens in the mic itself?

moamps said:

https://www.rane.com/note145.html

"IMD -- ITU-R (CCIF). Intermodulation Distortion -- ITU-R Method
How is it measured? The common test signal is a pair of equal amplitude tones spaced 1 kHz apart. Nonlinearity in the unit causes intermodulation products between the two signals. These are found by subtracting the two tones to find the first location at 1 kHz, then subtracting the second tone from twice the first tone, and then turning around and subtracting the first tone from twice the second, and so on. Usually only the first two or three components are measured, but for the oft-seen case of 19 kHz and 20 kHz, only the 1 kHz component is measured."

Not always.

Audio Measurement Handbook - Bob Metzler (Audio Precision)

"This “twin tone” test permits stressing band-limited systems at their highest frequencies while still measuring an in-band IMD product. The most severe limitation of the simplified analysis technique is that it measures only the second order product. The simplified technique is thus not useful to measure distortion produced by non-linear transfer functions which are symmetrical about zero, such as the BH curves of magnetic tape; this method should not be used to measure tape recorders. The same twin-tone signal can be used with an FFT analyzer or other selective analyzer to measure IMD products of all orders and thus yield more information, but analysis of such information requires more skill than the simple “one number” readout of the simplified analysis technique."

Click to expand...

Thanks, that did not come up in my search...

I found my crude DIY two-tone IMD very revealing for my design work back in the 70s. At the time I was designing things like phono preamps, that have LPF built into the typical equalization... The RIAA poles roll off the HF harmonic distortion and HF noise making the THD+N results look better than they should. Alternately the RIAA equalization actually boosts the 1kHz IMD distortion relative to 20kHz making a night and day difference between the same circuit measured with THD+N versus two tone IMD....  IMO the IMD test better correlated with how the circuits sounded in use playing music, but it was nice to be able to see and measure in band IMD.

There should be pad used in the first place.  If this signal is the cause of the high IMD in preamp, what the mess happens in the mic itself?

Click to expand...

Indeed but it is not always intuitively obvious (it wasn't to me) that the LF crud, was caused by HF nonlinearity.

I would not expect a passive microphone to exhibit HF nonlinearity, like a typical NF circuit, of course poorly designed active mic electronics could. High frequencies are naturally attenuated by distance from the source, thus my caveat about close miked HF sources.

JR

PS: I had an old Spectrum analyzer on my bench that I purchased used for $700,  but it only read up to 30kHz with 50 dB dynamic range.  Kids today don't appreciate how good they have it.

JohnRoberts said:

Kids today don't appreciate how good they have it.

Click to expand...

True. I made my first audio spectrum analyzer for my BScEE thesis (about XX years ago) consisting of LED arrays and 1/3 octave filters. I was very happy whit it, it was useful tool for tape recorders adjustment.
Nowdays is million dollar question what's Time Code and why it was recorded into first or last channel. 8)

moamps said:

True. I made my first audio spectrum analyzer for my BScEE thesis (about XX years ago) consisting of LED arrays and 1/3 octave filters. I was very happy whit it, it was useful tool for tape recorders adjustment.
Nowdays is million dollar question what's Time Code and why it was recorded into first or last channel. 8)

Click to expand...

Don't remind me about time code... one of my early gigs at Peavey was supporting using SMPTE time code synchronization to lock up a (Peavey) 4T cassette deck to (consumer) VCR...  a massive can of worms, when the software developer for the synchronizer was not great (was not even good). I kept a technician busy checking it for bugs, and every time he fixed a reported bug, he created two new ones. I was the inside guy who was the face of the program to the outside world, not fun.

JR

ruffrecords said:

Distortion characteristics are determined principally by the transfer function of the active devices ...

Click to expand...

In Dan's test, two of the systems have significant reactance by 19kHz.

In JR's chip work the clean bandwidth extended to 100kHz, so 19k is a fine frequency.

Stuff with wideband transformers may strain at 19kHz yet do fine with typical (non-cymbal, non-keys) speech/music where most energy is below 5kHz.

None of Dan's plots looks "bad". Obviously if he demanded ppm IM out to his sample rate he would not be using antiquated 3-transistor or tube transformer coupling in this 21st century. Wouldn't even plug that old stuff in! OTOH there can be a charm in old gear. Don't let mere numbers tell you otherwise.

Thanks, everyone. I appreciate learning as much as I can from those with decades of experience and education.

Always interested in finding out the science regarding why things sound the way they do. In this case, I like how all three of them sound for different reasons. Redhead, brunette, and blonde.

I would like to know why a 6J7 has less IMD than a 6SJ7.................gm related?

RDH4 Page 510

DaveP

DaveP said:

I would like to know why a 6J7 has less IMD than a 6SJ7.................gm related?

Click to expand...

Yes, 6J7 has smaller dgm/dVg.

Maybe a weaker stream of electrons has less chance of interaction/collision and therefore less IMD?

Many designs carried on using the 6J7, long after the 6SJ7 appeared in Oct 1938.

DaveP

The very first electronic project I did back in 1961 was a 6J7 and 6V6 audio amplifier. Its only issue in audio circuits was the top cap grid. The valve museum says:

Although originally designed as a detector its linear characteristic and good internal shielding (cf. 6Q7GT) made it a natural choice for audio.
It was for two decades the international standard valve for audio amplifier input stages and audio systems generally. The Mullard-Philips type EF37A, for many years the standard British low-noise audio pentode, is really a 6J7G under another name.

Click to expand...

Cheers

Ian

DaveP said:

Maybe a weaker stream of electrons has less chance of interaction/collision and therefore less IMD?

Click to expand...

It's not problem in higher gm per se. The problem is the gm changing vs. Vg or Ia for high output signal.

It's not problem in higher gm per se. The problem is the gm changing vs. Vg or Ia for high output signal.

Click to expand...

That makes sense

DaveP