ACMP-73 tests

[antichef]

OK - my time to do this testing is somewhat rare, but I got everything back out just now. In Nuemes' thread, MSHilarious posted a test file to run through the ACMP-73 and see what happens. Tarnationsauce2 did it with some preamps, but I never did:

Hey MSH, I performed your test for the heck of it.
I did a direct patch of the converter In to Out (DAW at 0dB).
With all the preamps I did have to vary the DAW output a little bit within a couple dB of eachother. They were all around -24dB out of the converter biggest difference was the SCA A12 needing about 6dB more than the N72 and ACMP.

In all cases I used MIC input, High impedance used, input gain all the way down, output all the way up. And on the ACMP the EQ button was out.

I realized after the fact I recorded all these in 32-bit float. Hopefully you can support that format. I didn't want to downsample because I didn't want aliasing or anti-aliasing to throw off the noise floor. I did no sort of post digital gain matching (normalizing etc.), it was all done while recording. These are the raw untouched files, don't want to contaminate the test!

Using an FFT, the ACMP proved to be pretty good in these tests. Better than equal to the N72 in all cases. Noise, harmonics, 60Hz hum, frequency response, and intermodulation look good to me.

Converter
ACMP-84
SCA N72
SCA A12

until now -- here's the test file bounced through Logic completely in the box (so hopefully very similar to what was originally posted):

http://www.recursor.net/testfile-no73.wav

and here it is through one of my 73s, using the *line* input - input gain about 9 o'clock (about 75% of max) and output gain maxed - no attempt at normalization or anything:

http://www.recursor.net/testfile-73.wav

I *still* haven't figured out how to do a FFT in Logic, but I'm working on it.

 

[mshilarious]

How did you turn a 2.2MB file into 14.3MB? I'm downloading now to have a look . . .

 

[mshilarious]

OK, wow, that is a giant amount of distortion. The 1kHz tone shows THD approaching 3%, and it's nearly all odd-order. The signal is not clipped, but it's very . . . . um . . . colorful. The intermod distortion is about the worst I've ever seen. Do any of the clip LEDs on the unit light up at this level? How about the VU (or does this model not have the VU)?

So if I understand this test correctly, you are feeding your converters a signal that peaks around -1dBFS? No normalization of the file? How about repeating the test at -6dBFS, just to make sure nothing funny is happening at your converter? Also, you think you can do 44.1kHz mono rather than 96kHz stereo? It will save a lot of time on the download . . .

If this unit has similar distortion behavior at -6dBFS, then it's a partial explanation of the hi-hat test done by . . . um, whoever did that, I forgot.

Also, instead of the ITB bounce, do a converter loop as a comparison instead. Match the level of the converter loop with the ACMP.

 

[antichef]

OK - I think I did it:

http://www.recursor.net/testfile-loop.wav

and

http://www.recursor.net/testfile-73.wav

The 73 file peaks about about -6.2 on my box. These are 44.1k mono, but still about 6 megs

edit: and the hihat test was me!

editedit: wait - indeed the output from my interface was pushing the VU meter on the 73 into the red -- I'm re-doing them now

http://www.recursor.net/testfile-loop-lowoutput.wav

http://www.recursor.net/testfile-73-lowoutput.wav

This is with the output from the interface reduced by 8db (and the peaks are therefore reduced even more) -- thanks for your patience, everyone

 

[mshilarious]

Actually, the -6dBFS samples are fine, so further gain reduction is not required. Why isn't the loop file unity gain though I'd like to see what happens when your converter is hit with a -1dBFS signal from the loop.

 

[antichef]

Actually, the -6dBFS samples are fine, so further gain reduction is not required. Why isn't the loop file unity gain though I'd like to see what happens when your converter is hit with a -1dBFS signal from the loop.

I used the same I/O plugin settings for both the loop and the 73 - and since the 73 adds gain and the loop doesn't, does that 'splain it? Sorry for the thick-headedness. It's my recurring ignorance problem. At the risk of still not hitting the mark, here's another file through the converter loop with no gain adjustments:

http://www.recursor.net/testfile-loop2.wav

by "fine" do you mean that you're not seeing the extreme distortion?

BTW, before I broke down my extended shop last night (putting away the 73s, but not the main converter), I rotated and mu-metal-shielded the toroidal transformers in the 73s, which greatly reduced the hum. I was liking the vocal sound by the end of the night (but don't have samples, sorry) - the learning curve is an unsung barrier for those of us who have little/no experience with these kind of units.

 

[mshilarious]

The key is if you are testing a signal at a given level (say -6dBFS) with your converter, you have to know how much distortion your converter is generating at that level. With the latest loop file I can see the converter does have some distortion--not a lot, like 0.01% THD, so try to retest the ACMP with the 1kHz sine wave hitting at -6dBFS. In fact, try different combinations of the input and output gain on the ACMP, all still hitting at -6dBFS. That should represent various typical operating conditions of the ACMP. Of course, turn down the converter output to -24dBFS or whatever you need given the minimum gain of the ACMP.

I suspect that the ACMP performs reasonably if the output control is always max, because that keeps the signal level low throughout the whole path. But as you increase input gain, you will distort and eventually clip the gain stages, which will then be attenuated by the output control (so that you don't clip the converter). It will be interesting to see if any desirable distortion characteristics are present before clipping, which I think is why people wanted these units in the first place.

In contrast, your typical IC opamp circuit doesn't do that--it will tend to stay very clean until the signal approaches its power rails, then it will clip suddenly.