T-S Measurement - "Insufficient Data"

[WingNut]

Hello, I'm a recent new user. When running free-air measurements to calculate the T-S parameters using the added mass method, I get a message when I start the calculation telling me that there's insufficient high frequency data. I'm setting the upper end of the sweep to 20 kHz. What am I doing wrong? Also, when I attempt to apply smoothing to the curves, with some measurements the impedance curve disappears but not with others. I don't know whether this is related...

Thanks for your help!

 

[John Mulcahy]

Which REW version are you running? If you aren't using the latest beta from here worth trying that first.

 

[WingNut]

Which REW version are you running? If you aren't using the latest beta from here worth trying that first.

I was using V5.19 "with JRE" (my OS is Windows 7 Pro). I have now installed V5.20 Beta 2, which seems to have up-dated the original installation. However, when I check levels prior to an impedance measurement, I now get a "Ref Input Level Low" message. My connections are the same as previously.

 

[John Mulcahy]

It's a bug in that beta I think, open the REW signal generator first and it should be OK.

 

[WingNut]

It's a bug in that beta I think, open the REW signal generator first and it should be OK.

Yes, if I open the signal generator first, the low input level message doesn't appear. Thank you.

The "Insufficient High Frequency Data" message has gone too and the T-S parameters are now making sense.

I've run the T-S measurements on my woofer and get a message "DC resistance figure may be too high". It goes on to say that the blocked impedance deltaR value is close to zero and suggests that I pick a lower Re. However, the message continues to appear and I just chase the figure lower and lower. I've also run the T-S measurements on my midrange with no problem. If I go back to the woofer, the same message appears...

 

[John Mulcahy]

Can you attach example measurements for the Re problem? Generally safe to ignore the message, but would be useful to see the data.

 

[WingNut]

Can you attach example measurements for the Re problem? Generally safe to ignore the message, but would be useful to see the data.

These are the woofer numbers (with the Re problem):

For good measure, here are the midrange numbers:

Although the T-S parameters have improved, they are still sometimes a long way from the manufacturer's figures. For example, Fs for the midrange is showing as 297 Hz - more than double the manufacturer's figure of 130 hz. Same for Qes, while F3 and Vas are just silly!

 

[John Mulcahy]

Those are very noisy measurements, for something more typical take a look at the SPH170 measurements in the REW Sampledata.mdat. Drive units act as microphones, so they need to be measured in a quiet location. If it is noisy where you are measuring that might account for some of what you are seeing, but the measurement setup needs more investigation. If you measure a resistor do you get a clean, flat trace?

 

[WingNut]

Those are very noisy measurements, for something more typical take a look at the SPH170 measurements in the REW Sampledata.mdat. Drive units act as microphones, so they need to be measured in a quiet location. If it is noisy where you are measuring that might account for some of what you are seeing, but the measurement setup needs more investigation. If you measure a resistor do you get a clean, flat trace?

Here's the trace for a 10 Ohm resistor:

Ambient noise is 20 dB this morning with the resistor. It's 32 dB when measuring a driver - I think the difference is my computer's cooling fan.

Edit: I should have mentioned that ambient noise is usually around 50 dB during the day when measured with the REW meter.

 

[John Mulcahy]

It should be perfectly flat and it also shouldn't be affected by the input volume setting, unless the input channels end up with different gains when the level is changed (in which case calibration should be run again after changing input volume). I'd suspect there may be some feedback occurring, perhaps some monitoring is active or "Listen to this device" is selected. Whatever the cause, until you see a clean, flat line with a resistor measurement you won't be getting meaningful results.

 

[WingNut]

It should be perfectly flat and it also shouldn't be affected by the input volume setting, unless the input channels end up with different gains when the level is changed (in which case calibration should be run again after changing input volume). I'd suspect there may be some feedback occurring, perhaps some monitoring is active or "Listen to this device" is selected. Whatever the cause, until you see a clean, flat line with a resistor measurement you won't be getting meaningful results.

Thanks for the diagnosis...you are spot on. "Listen to this device" was checked in the microphone recording levels dialogue box in Windows Control Panel > Sound. I'm neither a computer guy nor an electronics guy, so I much appreciate your help.

The resistor now looks like this:

Re-testing my woofer, I now get this:

I'm still getting the Re message, "DC resistance figure may be too high", but at least the T-S parameters are looking better and are modeling better in Unibox. I'll re-test the midrange next...

Edit: I've just tested a second woofer (this is a speaker with a pair of identical woofers) and didn't get the Re message. Here are the resulting T-S parameters in case it helps to clarify this Re issue:

I've retested the midrange and got a much smoother curve. However, with this driver some of the T-S parameters are still way off:

 

[John Mulcahy]

The resistor measurement is cleaner now, good to get that feedback corrected, but there is a big droop at low frequencies and a lot of noise above a few kHz. The measured value also looks to be a long way off if the resistor actually is 10 ohms since it is measuring a little over 13 ohms (the dark green trace is the impedance magnitude, the lighter green is the phase, right hand axis). The accuracy might improve if you go through the impedance calibration process, but that low frequency roll-off might be an indication the soundcard isn't really suited to use for impedance measurement. What soundcard are you using? If you have a headphone output that will usually work better than a line output.

 

[WingNut]

The resistor measurement is cleaner now, good to get that feedback corrected, but there is a big droop at low frequencies and a lot of noise above a few kHz. The measured value also looks to be a long way off if the resistor actually is 10 ohms since it is measuring a little over 13 ohms (the dark green trace is the impedance magnitude, the lighter green is the phase, right hand axis). The accuracy might improve if you go through the impedance calibration process, but that low frequency roll-off might be an indication the soundcard isn't really suited to use for impedance measurement. What soundcard are you using? If you have a headphone output that will usually work better than a line output.

How's this...the green is the previous test and the brown in the new one:

I increased the microphone input level to 100 in Windows Control Panel > Sound and then reduced the REW sweep level to -30 dBFS to eliminate clipping. This has had the same effect on my driver sweeps: cleaned up the high frequencies and brought the impedance levels down in line with the manufacturers' curves. The woofer modeling has improved in Unibox, but the T-S parameters for the mid still seem to be off.

I did do an impedance calibration, but I should probably do it again now that I've changed the input level.

My sound card is an Asus Xonar DGX PCIE5.1 card and I have been using the headphone output.

 

[John Mulcahy]

That certainly looks a great deal better.