REW offline measurements, what should I put caution on ?

[vair]

Hi,

I've been lurking this forum for some months, thank you everyone for your fantastic data. Just bought a hidiz DAC and I'm super happy with it.
I'm a professional sound engineer (music recording/mix/mastering, website), and just started a job in car audio, diving more into acoustic measurements. This new job drove me to your forum.

Most of the time we have to do measurements in offline mode (aka open loop), meaning the computer isn't sending the measurement signal.
To access more information than whata frequency response can offer, I started to dive into room eq wizard.

Room eq wizard has this fantastic feature "import sweep recording" allowing exactly that. I was just wondering what could appear using this method ?
Is phase/group delay information reliable ? When playing on devices that have a bad quality DAC, how does this translate into the measurements ?

For example, I'm frequently seeing "drifting" in waterfall plot and don't know how to interpret those. For example, a strong resonnances (room mode ?) decaying at 150Hz that drift to 135Hz after 100ms. Can this be related to the measurement method ? Is it a phenomenon related to the (most of the time really crappy) speakers or is it purely acoustic ?


Not sure if I'm in the right section of the forum.. I have so many questions for you guys haha

Thanks for your answers !
Nicolas

 

[AnalogSteph]

When playing on devices that have a bad quality DAC, how does this translate into the measurements ?

Usually not at all. You may be able to make out something like 0.2 dB of periodic ripple in a gated measurement, and distortion may show if it's truly off the charts (like Amir's bottom 10 for SINAD)... not anything you'd normally encounter.

For example, I'm frequently seeing "drifting" in waterfall plot and don't know how to interpret those. For example, a strong resonnances (room mode ?) decaying at 150Hz that drift to 135Hz after 100ms. Can this be related to the measurement method ? Is it a phenomenon related to the (most of the time really crappy) speakers or is it purely acoustic ?

99% sure something like this would be acoustic.

Not sure if I'm in the right section of the forum.. I have so many questions for you guys haha

We do have a room acoustics section which may be a more appropriate spot.

 

[JohnPM]

You can also use file playback if you are capturing on a computer that has REW. When the playback and capture devices are different clock rate differences can cause some odd effects, the Analysis option to "Adjust clock with acoustic ref" can correct that.

 

[k525]

Hi,

I've been lurking this forum for some months, thank you everyone for your fantastic data. Just bought a hidiz DAC and I'm super happy with it.
I'm a professional sound engineer (music recording/mix/mastering, website), and just started a job in car audio, diving more into acoustic measurements. This new job drove me to your forum.

Most of the time we have to do measurements in offline mode (aka open loop), meaning the computer isn't sending the measurement signal.
To access more information than whata frequency response can offer, I started to dive into room eq wizard.

Room eq wizard has this fantastic feature "import sweep recording" allowing exactly that. I was just wondering what could appear using this method ?
Is phase/group delay information reliable ? When playing on devices that have a bad quality DAC, how does this translate into the measurements ?

For example, I'm frequently seeing "drifting" in waterfall plot and don't know how to interpret those. For example, a strong resonnances (room mode ?) decaying at 150Hz that drift to 135Hz after 100ms. Can this be related to the measurement method ? Is it a phenomenon related to the (most of the time really crappy) speakers or is it purely acoustic ?


Not sure if I'm in the right section of the forum.. I have so many questions for you guys haha

Thanks for your answers !
Nicolas

Hi, I have some experience in aerospace acoustic an vibration measurements when I was young, and I am playing with REW measurements at home,
If you are making spectrum analysis with an external source and not using REW internal generator it could be better you use white noise and measure spectrum response applying an FFT windows such Hanning or Hamming for example, because it is important to make FFT over samples buffers that begin and end with which 0 values (near to be periodic) see: https://towardsdatascience.com/brie...reprocessing-of-discrete-fourier-8b87fe538bb7, If your input signal is an externally generated sweep it can be a burst short enough to be contained into the buffer and pre-triggered in order to live zeros at the beginning, otherwise you may acquire a very log sweep using Hanning.

I do not have experience in time domain analysis but it seem sensed that you acquire impulses in synchronous triggered way, try to read this:
http://pcfarina.eng.unipr.it/Public/Presentations/NordicSound-Farina_presentation.pdf

Hope the help and not confuse.. I some guru here find mistakes don't hesitate to kill!

 

[vair]

Thanks for your replies ! I'm discovering the REW beta with a lot of joy !

You can also use file playback if you are capturing on a computer that has REW. When the playback and capture devices are different clock rate differences can cause some odd effects, the Analysis option to "Adjust clock with acoustic ref" can correct that.

What kind of "odd effects" can we observe ? That's exactly what I was trying to understand.

99% sure something like this would be acoustic.

Does this behavior have a name or something ? How could I understand better what's happening there ?
I'll come back this week with screenshots.



(may an admin move this post in te appropriate section ? Or can I do it myself ?)

 

[JohnPM]

What kind of "odd effects" can we observe ? That's exactly what I was trying to understand.

The impulse response becomes stretched, in extreme cases (sample rate mismatches rather than clock errors) starting to look like a short sweep itself. There is no real impact on the magnitude response but the phase is affected. You can see the effect by manually applying a clock rate correction using the controls provided.

 

[vair]

Thank you for your quick answer ! I'll try it out soon !



Here are 2 screenshots of the phenomenon I'm trying to understand. Sorry if it is obvious, I'm new to this. It is measured in a car.

For example here, I don't understand why a peak at 45Hz "drift" as two resonances at 40 and 48Hz, two frequencies that were not appearing in the short term analysis. It does not ressemble a typical room mode decay does it ?

How to interpret this behavior ? Is it purely acoustic and usual ?
Is it related to several modes causing interferences ?
Does the 40 and 48Hz take longer time to settle ?

Same phenomenon here :

Any ideas of keywords or books to investigate those behavior ? I've been thinking about this for months !

Thanks a lot !

 

[Ge0]

I can't help answer your question about frequency drift. But, I do quite a bit of car audio measurement. Doesn't your car audio system have a useable auxiliary input? You can route the REW signal generator to your audio interface output --> the car stereo's input.

 

[JohnPM]

Here are 2 screenshots of the phenomenon I'm trying to understand. Sorry if it is obvious, I'm new to this. It is measured in a car.

For example here, I don't understand why a peak at 45Hz "drift" as two resonances at 40 and 48Hz, two frequencies that were not appearing in the short term analysis. It does not ressemble a typical room mode decay does it ?

How to interpret this behavior ? Is it purely acoustic and usual ?
Is it related to several modes causing interferences ?
Does the 40 and 48Hz take longer time to settle ?

That's quite usual. There are often many overlapping modes with different rates of decay and there can also be delayed contributions from other coupled spaces (such as a door into another room or any other connected volume) and, in the lower parts of the waterfall, system or external noise sources. Overlapping modes can beat against each other and their differing rates of decay can make it appear that there is a shift in frequency with time. What is changing is the summed effect of the various modal contributions, not the individual modes.

 

[vair]

I can't help answer your question about frequency drift. But, I do quite a bit of car audio measurement. Doesn't your car audio system have a useable auxiliary input? You can route the REW signal generator to your audio interface output --> the car stereo's input.

Most of the time I have to work with radios that don't have an analog auxiliary input.. Most of them have bluetooth but I then faced non-constent latency that made it unusable.. I'm searching for a solution that might involve "audio over usb" because carplay/android auto use a digital input via USB, I just don't know where to look for to be able to stream audio from a computer that way..

That's quite usual. There are often many overlapping modes with different rates of decay and there can also be delayed contributions from other coupled spaces (such as a door into another room or any other connected volume) and, in the lower parts of the waterfall, system or external noise sources. Overlapping modes can beat against each other and their differing rates of decay can make it appear that there is a shift in frequency with time. What is changing is the summed effect of the various modal contributions, not the individual modes.

Thanks a lot, really, that fully answer my doubts !

Refering to the previous screeshot and thinking about equalization to enhance the system behavior, would it be correct to state that equalizing the peak around 120 Hz would give a better rendering short term and that equalizing the resonnance at 100Hz would enhance the "long term" response / decay ? Or is it more complicated ?

(I understand that the resulting resonance is significantly lower, I understand that equalizing won't enhance the time response and that a "flatter" balance doesn't always means a more pleasant response, but I'm trying to understand how to interpret those results.)

 

[JohnPM]

You can use the predicted waterfall plot on the EQ window to see the effect of filter settings on decay, there is an option in the graph controls to have it update live as you alter the settings. Our sensitivity to slow decay is far lower than our sensitivity to peaks in the response, so generally it is best to prioritise evening out the bumps in the low end response.

 

[Ge0]

You can use the predicted waterfall plot on the EQ window to see the effect of filter settings on decay, there is an option in the graph controls to have it update live as you alter the settings. Our sensitivity to slow decay is far lower than our sensitivity to peaks in the response, so generally it is best to prioritise evening out the bumps in the low end response.

I haven't quite figured this out but our sensitivity to decay is a function of frequency, amplitude, and time. I believe the math exists. I just can't wrap my head around it.

You are correct though. We are much more sensitive to anomalies in frequency response that have a wider Q factor