REW measurements Problem with slow PC-laptop

[putko]

Hi

I was making normal acoustic measuments with TBone Microphone and sound "tube" whic is USB connected. The results showed unnatural zeros.

Later I noticed that the location of these dips depends on the sweep length and it happened with both java and Flexasio interface.

I woud like to know has anybody had similar experiences and is there a possible solution

Best wishes HPu

PS
The longer the sweep the less are the zero frequencies

 

[John Mulcahy]

Make sure monitoring is not active, the input signal must not be routed back to the output.

 

[putko]

how can I find out is it or not. i havr my first windows 11. tomorrow fedora, lets see.!

 

[John Mulcahy]

Make a loopback measurement, output connected to input.

 

[putko]

I have done itwith another soundcard. all ok.
but with this (tbone-microphone) it neef a little cable tinkering.

more afttr that

 

[phofman]

Flexasio merges two soundcards with independent clocks into one callback. IMO technically that is not correct, and works ok only due to short measurements sessions where the buffers do not have time to go apart.

IMO the asynchronous Wasapi exclusive is the proper API for this.

 

[putko]

this sounds great. in finlad now is night. I`ll test.

 

[MichaelBrown]

Replying for the update.

 

[putko]

Morning
I have checked that the listening is not on or off. it is OFF.
Following are the results this far
The comfbeffect depends on the swap length and sample rate.
with 44khz it starts on lower frequences but the shape is same as in 48kht.
The appears with all interfaces and loudspeakers (I have java, flexasio and steinberg, asio4all does not work)
I tried to change the java allocated memory betwee 512 1024 and 2048, no effect.

It does not show in ARTA. i'll try next HolmImpulse (Besically I'm interested only lower basses, but it woud be nice to use REW

 

[phofman]

The appears with all interfaces and loudspeakers (I have java, flexasio and steinberg, asio4all does not work)

Does it happen with both capture and playback using java exclusive in REW too?

 

[putko]

Phofman
I did not understand your question. I make REW measurment and the resurts are not looking about the same with different sweep lengts. When I listen these zero>SPL frequences (with REW signal generator) there are no difference. It means there is no drop in hearins sensation

Here is a screenshot of HOLMinpulse neasurement. It looks too good, but its about correct the MAOP10 had a F0 39Hz. Closett box quite a lot of filling

 

[putko]

REW has its own java-runtime I have mo other javas installed

 

[phofman]

I make REW measurment

Do you use wasapi exclusive (i.e. selected in REW https://www.roomeqwizard.com/help/help_en-GB/html/calsoundcard.html#audioconfig ) for that measurement?

 

[putko]

I have noticed the EXCL devices but paid no attention. In the picture are four measuments with these devices. the first number is the device buffer length from preferences. (with bigger thar 128 REW goes crazy)
and the second number is sweep lebgth

the low freq measusements are consistent maybe up to 2kHz. (the 1kHz dip is not infinity and migth be real)

 

[phofman]

I would recommend to keep using wasapi exclusive (EXCL), to eliminate the potential issue with duplex ASIO to two different audio devices.

 

[putko]

yes. that souds goot. anyhow I will install fedora and test also with faster laptop, but thats next week

 

[John Mulcahy]

Use the EXCL devices, buffer length 16k for input and output, then make measurements with the output looped back to the input. Use the same input channel you use with the mic. That will show whether there are any issues with the interface settings.

 

[phofman]

Well, it's possible you are getting buffer glitches. But they would likely not occur always at the same moment of the sweep (like at 3kHz with the 256k buffer), the timing would be more or less random, as the OS does not keep up. Unlike buffer glitches caused by ASIO merging the two devices - there the glitch is a deterministic function of the difference between the two device clocks (rather stable) and the buffer size (fixed for the same setup).

Do you get an RTA with averaging (which visibly catches any buffer glitch), running stable/no glitches for several minutes?

 

[putko]

Hi
It is not easy to get a loop with tbone microphone. BUT no problem
the 16k buffers and EXCL devices corrected the issue.
Results of sweeps of different lengths are very same. no extra dips etc

Thank you all for help.
Now back to work (to get a listenable pair and corresponding measures)

 

[phofman]

BUT no problem
the 16k buffers and EXCL devices corrected the issue.

The chart at https://www.avnirvana.com/threads/rew-measurements-problem-with-slow-pc-laptop.14573/post-111536 suggests exclusive at 16k, with the peaks. What was changed afterwards that it's ok now?

 

[putko]

Hi
Big news. I have beeb able the repeat when the anomalu happens.
If I take just one sweep, everything is OK, but with multiple sweeps thera appears thos strange dips.
i include here a rew saved measurements. the names tell what it is. Extra info available if needed,

 

[John Mulcahy]

with multiple sweeps thera appears thos strange dips

That means the soundcard is not maintaining time alignment between the sweeps. It might be because the card is internally resampling to a different rate, so you could try at 44.1 kHz to see if that is any different, but easiest would be to use single sweeps. If you need greater signal-to-noise use a longer sweep.

 

[putko]

I have tested the system with 48 and 44.1. Behaviour is the same.

In my computer (Windows 11, Slooow) everything works good with one sample.

My goal is to write a beginners guide to finnish hobbyists how to measure loudspeakers that you have planned and build with possible frugal (cheap) equipment. It is not so very important to get the decibels right, but to find out big errors

attachment is a MAOP-10.2 driver 13,5l closed box with half a pillow fillings in a room near field measurement

Souds good in my ears.
Best wihes Hpu

 

[phofman]

@John Mulcahy : Do the mdat serializations contain the actual recorded stream?

 

[putko]

I did the "save all" and in my PC they show the graphs. I am not sure about the measuments.
I can make new ones if needed.