Question on making cal file and uncorrected phase shift

[JonPike]

Hello John and all...

Been quite a while since I used your excellent software, as well as done any acoustical measuring. But, time to get back into things.

Something I'm seeing, is after I do the soundcard cal, I do a measure with the same loopback that I used, and I see the phase tilting up at the end of the range to about +30deg.
I get a nicely corrected frequency response, but I would think I should see the phase corrected to nearly 0deg, like I see at the low frequency end, shouldn't I? I believe I have a loopback connection, have got L and R on input and output connected, unless I have a cable problem. I see equal signal on both input channels, and I have Adjust clock with Loopback checked. Have I missed something else?

 

[John Mulcahy]

The cal procedure removes any delay in the signal, placing t=0 at the associated position. The measurement may have t=0 at the peak, hence the small time shift causing the phase change. If you use Estimate IR Delay and remove the delay it should end up flat.

 

[JonPike]

OK, more learning experience achieved.... I also learned that Win 10 seems to be behaving differently after all these updates. Had to figure out that one has to set the sample/data rate for the sound card in shared mode, despite menus implying that REW wouldn't be affected. Setting to 16b/96k allowed me to run at 96k again, solving that little head scratcher. Don't think I had to do that before?

On to the timing issue. I've used ARTA a lot in the past, and am used to looking at the impulse response, and setting the point visually. I haven't gotten used to the IR display in REW, assumably its possible to zoom way in there and see what delay is needed and set it?

Found and played with the IR Delay. I'm using the loopback timing again. An odd thing was, the Estimate feature told me there was zero delay, and I assume removed zero. Bumping it up and down manually, I found that it needed -0.0006ms to get close to flat. Getting within +0.75 degree of flat at 44khz is pretty good. But, I couldn't get it perfect. Might be a need for one more decimal place, -0.0005ms was about perfectly -0.75 deg below flat, so -0.00055 should have done it. Seems you can't enter a smaller time delay number value, though. Again, pretty good, for just about any work.. so don't worry if it's a big hassle to add a decimal place in the number scheme.

I feel like I have a calibrated instrument now! Thanks!

 

[sm52]

I feel like I have a calibrated instrument now

Hi. What sound card do you have for measurements?

 

[Breeman]

Hi John,

Interesting discussion and sorry for hijacking the thread. I noticed too that the phase of my soundcard calibration is only flat in the mid-frequency range and slopes upwards on either frequency extremes. I didn't worry about it much because I thought it gets applied equally to all the measurements taken with the same soundcard. In that regard, is the correct procedure to correct for phase deviation at higher frequencies as described above? Is using 96k (if the soundcard supports it) instead of 48k better when doing these calibrations? Below is my calibration measured at 48k, for the Steinberg UR22 MkII soundcard. I get flatter phase with adjustment of -0.0060ms. Which one would be better to use or does it even matter?

 

[John Mulcahy]

It depends what the measurements are being used for. In an acoustic context delays corresponding to the time it takes sound to travel a couple of mm (0.006 ms) or a couple of tenths of a mm (corresponding to the delays in the preceding post) are pretty much irrelevant.

 

[JonPike]

Hi. What sound card do you have for measurements?

Its a Creative Labs Soundblaster X-fi HD, a USB card. Was also playing with my motherboard's internal sound card, which was supposed to be pretty good. I'm pretty happy with them, but had been using them with different software.

Let me know John, if I'm running astray... sm52, your pics seem to be of your soundcard's raw reponse, the measurment that gets used to become the "cal file". In other words, thats the card's response, that is used to compensate out where it isn't flat. My pics were of a regular measurement, after the cal, of the same direct loopback setup that the cal was made with. Ideally, that would be perfectly flat in both FR and phase, with everything corrected out. I usually take one after a cal to verify it's a good one.

Yes, it does depend on what you're doing measurement wise. John's point is a good one, not much of an issue in, say a room measurement situation. Me, I'm being picky (since I can get that way) and thinking about if I was going to make detailed measurements of drivers for speaker crossover design. There, phase accuracy becomes more important. Even there, the phase towards the middle of the frequency range (near the crossover point) is more critical than at the high end, though I guess if you want to do advanced things like remove test baffle response for a "raw" driver data set, any phase error might influence results. I was trying to see how "perfect" I could get it.

 

[sm52]

sm52, your pics

it's not mine. They were posted by Breeman

 

[JonPike]

Oops...

 

[Breeman]

Its a Creative Labs Soundblaster X-fi HD, a USB card. Was also playing with my motherboard's internal sound card, which was supposed to be pretty good. I'm pretty happy with them, but had been using them with different software.

Let me know John, if I'm running astray... sm52, your pics seem to be of your soundcard's raw reponse, the measurment that gets used to become the "cal file". In other words, thats the card's response, that is used to compensate out where it isn't flat. My pics were of a regular measurement, after the cal, of the same direct loopback setup that the cal was made with. Ideally, that would be perfectly flat in both FR and phase, with everything corrected out. I usually take one after a cal to verify it's a good one.

Yes, it does depend on what you're doing measurement wise. John's point is a good one, not much of an issue in, say a room measurement situation. Me, I'm being picky (since I can get that way) and thinking about if I was going to make detailed measurements of drivers for speaker crossover design. There, phase accuracy becomes more important. Even there, the phase towards the middle of the frequency range (near the crossover point) is more critical than at the high end, though I guess if you want to do advanced things like remove test baffle response for a "raw" driver data set, any phase error might influence results. I was trying to see how "perfect" I could get it.

Hi JonPike,

Those pics are indeed raw measurements of the soundcard used as its calibration file. I thought the measurements you were referring to were also raw measurements, I misread your post, sorry. I have never bothered re-measuring with the cal file applied.