Home Theater Forum and Systems banner

Calibration Correction at High Frequencies

2K views 7 replies 3 participants last post by  Randy114 
#1 ·
Over the past several months, I have been using REW5.01Beta17 to sweep and measure distortion of amplifiers, crossovers, and other electronic components. REW is a terrific program for doing Amplitude/Phase and Distortion measurements! I am using an EMU-404 (24-bit 48-kHz external soundcard) whose Calibration File always overcorrects the high frequency region from the uncorrected -17 degs at 20-kHz to the corrected +7 degs at 20-kHz (see attachment). The over-correction starts around 5-kHz and continues up to 24-kHz. The corrections to the low frequency regions below 5-kHz for phase are perfect and the amplitude correction is perfect across the entire 2-Hz to 24-kHz band. These over-corrected phase effects also happen at 24-bit and 96-kHz sampling rates. The EMU is run at 0.25 Vrms (Input and Output) to corresopond to the REW -12 dB settings. Any suggestions as to the problem?
 

Attachments

See less See more
1
#2 · (Edited)
There is no problem here.

The phase at the top end (20kHz) should be placed at 0° instead of 7°. Normally, if REW preferences are set to "Set T=0 at IR Peak", it does just that. If you have that box unchecked that may the reason.

Either way you could manually move the IR location slightly as needed to get it properly aligned.

Oops, I just remembered:
The "Sub-sample Timing Adjustment" may also be the timing alignment issue. It should also be checked for accurate automatic alignments.
 
#5 ·
JohnM, first -- thankyou for getting back to me. I have repeated several times the Loopback Calibration of my EMU 404 at 24-bit, 48-kHz with Input and Output at -12 dB (and verifed the levels are indeed 0.25 Vrms). I have also let the unit warm up and that makes no difference in results. The problem of not having accurate correction at high frequecies still exists.

Attached is the *.mdat file. My *.cal file is 435 kB and the Forum does not allow that size -- how do I get you the *.cal file?

Randy

PS -- I will try Java with 16-bits and 48-kHz and see what that produces
 

Attachments

#6 ·
Randy114,
I looked at your .mdat. I just thought that while you are waiting for JohnM to take a look. I would point out a couple of things to think about.

I opened your file It doesn't appear to have any significant issues. [Things can always be better, but all appears very typical for this level hardware setup.]

I then deleted the cal measurement (the first one) and copied the loopback measurement so that I then had 2 identical versions of your loopback measurement. All the measurement panels confirmed this.

I then made a small manual shift to the IR position of the copy (in the IR panel). I shifted it -0.0006 ms. This is calculates to about -5° at 22 kHz. Below is the close-up of the 2 IRs after the shift. The shift is so small that that it is difficult to see even at maximum magnification.

Text Line Green Plot Slope


Below is the impact on the phase chart of the original IR overlaid with the shifted copy. You can see that the Phase at 22 kHz is now brought down about 5° so that it is now flat in line across the entire range. Had I shifted it further it would start to appear to be tilting down at 22 kHz so I stopped there.

Text Line Design Parallel Slope


Also note that the phase across the entire range is now at about 1° rather than at 0°. This small offset from zero is a characteristic of your setup and I am not sure why. Possibly whatever causes the slight phase offset from zero is related to the minor error in setting the IR to zero. Maybe a little DC offset in the balanced I/O's circuitry? Are you using a balanced to unbalanced cable for loopback? [These are just wild guesses.]

JohnM will likely have some further comments to help explain/improve the calibration.
 
#7 ·
JTalden – I appreciate your help in this matter. Although you show that a very slight shifting of the Impulse Response reference (i.e. -0.0006 mS) can make the Phase Response near flat, this cannot be done in a “real” measurement scenario. A “real” measurement of an unknown device (for instance an amplifier under test) produces a phase plot whose actual response was not known before the test. The purpose of the Calibration File in “Loopback” is to provide a correction to the soundcard swept measurement thus revealing the unknown device’s true amplitude and phase response. No “tweaking “ of the Impulse Response can be allowed since one does not have prior knowledge of what the result should be.

In looking at the typical Calibration File, it seems to be, in text form, the default measured response (under Soundcard Calibration routine) and then I presume that REW just uses this to subtract from any ensuing measurements. You have mentioned soundcard DC Offset and the EMU404 does indeed have a -1.3 mV DC offset with and without “loopback” to the input which is AC coupled. The connections are Mono ¼” TRS plugs in unbalanced mode. I also “checked” the box under soundcard for “High Pass” filter and that did not change any results.

Since the timing of the Impulse Peak as pertains to phase clearly requires sub-sampling and reproducibility for accurate phase at high frequencies, it is not clear to me as to “how well” REW can actually do this. Amplitudes are always corrected flat but not phase, which appears to be more sensitive to Impulse Peak timing. Maybe I am asking too much of REW to make accurate high frequency measurements?

As you stated, comments from JohnM will be much appreciated when he gets the chance.

Randy
 
#8 ·
High Frequency Calibration Problem Solved! Thinking more about Jtalden’s exercise in shifting the T=0 Reference time of the Impulse response that did indeed correct the calibration for high frequencies, I unchecked the box “Set t=0 for IR Peak” under the Analysis Tab. It turns out that the soundcard Calibration Routine works fine in this situation. My conclusion is that REW must use its own determination of Impulse Peak timing for generating correct Calibration Files. Although this soltuion works, I am presuming that this is what JohnM intended.

I thank JohnM and Jtalden for their input, which made me continue to explore this problem until solved.
 
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top