Igor Sontec - chanels doesn't "null"...

[tonedude]

Trying here as well, I hope it's ok.

----------

I wonder if my Igor Sontec is working as it should?

When I send in a mono signal (eg white noise...) in to both channels of my Igor Sontec, and try to cancel the signal out through phase shifting one of the returning channel, it doesn't null, not even close (I get about -30 dB or something).  Filters are bypassed of course.

When doing the same with my Behringer T1951, modified to something "close" to BP Net EQ spec, it almost nulls.

How can I understand this? Is there something wrong with my Sontec i/o board?

 

[PRR]

I do not understand the technique.

-30dB null is hardly "not even close".

You can't phase-shift hiss.

If you feed true white-noise, with hiss out to MegaHertz, through any audio device, the higher frequencies present in the source and weakened in the audio box will not-null.

Try a one-tone or few-tone source.

 

[tonedude]

PRR,

I don't understand my technique either so that's really quite understandable...    ;D

My "problem" was that my T1951 cancelled out white noise to a much higher degree than my Sontec did, which made me think there was something wrong with the sontec. (I was repairing the T1951 and tried to match channels with white noise. After fixing the behringer I thought that I'd try to do the same with the sontec, just for "fun", and maybe I'll learn something new on the way...)

After writing this I looked at the sontec schematic and opened up the EQ, I then found out that the grounding of the i/o boards was loose. Tightening that upp did fix it!

Thanx PRR, so I learned something new.

I do not understand the technique.

-30dB null is hardly "not even close".

You can't phase-shift hiss.

If you feed true white-noise, with hiss out to MegaHertz, through any audio device, the higher frequencies present in the source and weakened in the audio box will not-null.

Try a one-tone or few-tone source.

 

[audiomixer]

well did tightening the ground fix the A-B channel subtraction as well?

feeding the same signal into both channels and subtracting the two should  yield in a null - independent of the source type, or do I miss a point. of course you might get a easier to read response with something other then white noise, like multi-band sinuses or sine sweep, but this technique has been used in the past to assess differences in two signal paths as far as I know....

if the reference signal takes a different path, let's say without passing the equalizer, then a different HF filtering would produce significant level after the subtraction.

- Michael

 

[tonedude]

well did tightening the ground fix the A-B channel subtraction as well?

- Michael

Yes, it got alot better! Just as good as my modded T1951, or as good as bypassing the unit all along connecting the L/R xlr cables directly to one another.

 

[JohnRoberts]

As long as the inputs fed to both channels are identical they should null...  The depth of the null will be degraded by how closely the levels match.

A 30 dB null suggests a level mismatch of roughly 3%.  Further the null quality can be degraded by phase shift so if the source signal has content outside the bandpass of the two paths being nulled, differences in the two channels HPF and LPF can degrade null quality for the out of band material. 

JR

PS: I like null testing, it doesn't care what the differences are, and if they null well they are the same....

 

[PRR]

> inputs fed to both channels are identical they should null... 

I missed the fact that he was feeding two "identical" channels.

Yes, that should null for any signal.

The phase-shifter should not be necessary if the two paths are identical.

(And the fact it did not null lead him to the ground problem.)

I was thinking nulling against a "straight wire", the more common proof-test. On very good equipment, if all of the source is inside the equipment bandwidth, the null should be very good. However if any part of the signal goes outside the equipment bandwidth, you can not get a null for both the in-band and the out-band signals. For few-tone signal at band edge you may be able to shift-phase and get better band-edge null, but what are you proving? Eventually your "straight wire" has to be so twisted that it in fact duplicates the imperfection of the equipment. While this is one way to "measure" equipment (find what has to be done to get the null and then measure what you did), it sure seems like the hard way.

 

[Gustav]

A 30 dB null suggests a level mismatch of roughly 3%. 

How are you reaching this number?

Gustav

 

[PRR]

30dB of voltage is 3.16%.

 

[JohnRoberts]

A 30 dB null suggests a level mismatch of roughly 3%. 

How are you reaching this number?

Gustav

Yes, as PRR already answered  two 0dBu signals that null to -30dBu means that the error signal is roughly 1/33 of the full signal.

This is a relatively small error, demonstrating the sensitivity of null testing for comparing audio paths.

JR