DIfferences in ways of measuring Thiel/Small parameters

[davidallancole]

Hello,

I have read a few different ways of measuring Thiel/Small parameters. In Rod Elliotts article he uses a 10 ohm resistor in series with the speaker. I think it was in Vance Dickasons book, Vance used a 10k ohm resistor and I have seen on the epanorama website a 1k ohm resistor.

My question is, is any one way more accurate or better?

Rod Elliotts method requires a power amp in the chain to get the voltage levels right, but it seems with Vances method you could use the output of you sound card on a computer to measure it.

Thanks.

Here Rods way: http://sound.westhost.com/tsp.htm
Epanorama: http://www.epanorama.net/documents/audio/speaker_parameters.html

 

[Jack Hidley]

David,

There are two important things to remember when measuring TSPs.

One is that you must be measuring the actual impedance of the driver. The first method you posted measures the actual impedance. The second method only measures an estimate of it. Note where the writer says that this method forms a decent current source? An ideal current source has an infinite output impedance. His current source uses a 1k output impedance. This could cause a 10% error in Q measurements with a high BL driver. The only reason to use it, is that it makes the math slightly easier. Use the first method with whatever resistor value you want to.

Two is drive level. The TS model assumes that the driver being modeled is linear. They are not. Changing the drive level will therefore affect the measurements. What drive level should you use to take the measurements? That depends mostly on the drivers Xmax. We usually shoot for an excursion at Fs of about 0.2Xmax. That keeps the driver in its mostly linear range, but keeps it out of the low level noise and hysteresis. How much voltage or current does this correspond to? That completely depends on the output resistor you use, the drivers Qes, Qms, Mms and Xmax. Each driver will be different.

 

[davidallancole]

Thanks for the reply Jack.

Is there any benefit then to using a constant current source to do these measurements instead of the resistor method?

Is there a calculation to figure out how much current or voltage is required to hit the 0.2 Xmax at Fs? Does this require knowing the BL of the speaker? This all sounds a little more complicated than I thought.

I will follow Rod Elliotts proposed method then to do my speaker measurements.

 

[soongsc]

The way to measure is the way closer to normal operation signal levels since that is the final operating mode. So Jacks 0.2 xmax seems a good point to select.

The constant current method may have some problems when using ferro fluid cooling.

 

[Jack Hidley]

There is no real difference between using a constant current source or a constant voltage source (normal amplifier) or anything in between.

To calculate the current needed to achieve a given displacement, you need to know BL, Mms and Cms. But if you already know these, then you don't need to take the measurement Just connect the driver to an amplifier with a 50ohm resistor in between, drive it with a sine wave and adjust the frequency until you get maximum visible amplitude. Then adjust the drive level until you get 20% of Xmax by eye. Measure the voltage directly at the driver terminals. This should be your drive level during the TSP measurements.

 

[soongsc]

Trying to use current to get the xmax range using the constant current method requires a much larger amp than normally available. Plus the load on the amp would be so high that would deviate from its original design intent. Hard to say what will affect the test results.

I do wonder if anyone is going to get into the delta mass versus delta volume issue.

 

[davidallancole]

Thanks guys.

 

[pinkmouse]

I do wonder if anyone is going to get into the delta mass versus delta volume issue.

All I will say is that I have used both methods and found a pretty good correlation.

 

[richie00boy]

Why does using constant current have problems with ferrofluid?

 

[Ron E]

Why does using constant current have problems with ferrofluid?

Constant current (with a 200-1k series resistor) loses accuracy for higher impedance peaks, and ferrofluid lowers the impedance peaks, so the comment about ferrofluid is mostly spurious. Both methods may have difficulty with small impedance peaks due to the nature of their calculations.

Note that most of these methods assume zero inductance, and there is a way to reduce the effects of inductance on the Qms calculation.

I used to use a 10 ohm series resistor and measure the voltage across the resistor, the speaker and the combo, then you can calculate impedance phase as well as magnitude, and compensate for meter frequency response issues. This is in fact what speaker workshop does. The spot measurements were tedious, so now I use speaker workshop.

Most meters have a poor frequency response. You will get better measurements using speaker workshop or Praxis using a jig.

 

[davidallancole]

I am unfamiliar with Speaker Workshop. Is it free on the Internet? Does it use the computers sound card to emit the frequencies and measure the impedance at the certain frequencies?

 

[pinkmouse]

Yes.

 

[soongsc]

I am unfamiliar with Speaker Workshop. Is it free on the Internet? Does it use the computers sound card to emit the frequencies and measure the impedance at the certain frequencies?

http://www.speakerworkshop.com/SW/Download.htm

 

[soongsc]

All I will say is that I have used both methods and found a pretty good correlation.

I think this is kind of interesting. I've ran tests using both on a driver once, and ended up with quite different Vac values. Can't remember how different or whether other values changed or not.

 

[pinkmouse]

I lost a lot of stuff when my old PC died, but some results are here on diyAudio, I'll see if I can dig them out.

edit: Yup. Here we go:

http://www.diyaudio.com/forums/showthread.php?postid=938522#post938522

 

[soongsc]

I lost a lot of stuff when my old PC died, but some results are here on diyAudio, I'll see if I can dig them out.

edit: Yup. Here we go:

http://www.diyaudio.com/forums/showthread.php?postid=938522#post938522

Can't tell what the added mass method looks like, but it seems the calibration in SW was not done for interchannel latency calibration.

 

[pinkmouse]

Now it's the morning, and I'm awake, that's the wrong thread! I'll do some more digging...

As for the latency issue, that was done with my old PC and soundcard, and it was set as good as it would get. However, I've just bought Soundeasy S/H from another member, as SW refused point blank to run on my new motherboard/soundcard combo. I'm only half way through the manual, so haven't fired it up yet to test properly, but will soon. I have lots of projects in the pipeline!

 

[richie00boy]

Constant current (with a 200-1k series resistor) loses accuracy for higher impedance peaks, and ferrofluid lowers the impedance peaks, so the comment about ferrofluid is mostly spurious. Both methods may have difficulty with small impedance peaks due to the nature of their calculations.

Thanks. I was aware of the pseudo constant current source accuracy issue, and you have confirmed what I thought about ferrofluid.

 

[forr]

Hi,
There is good discussion about measuring parameters in D'Appolito's book.
There are Audua and Audiotester software with demo versions available.
As audio generators, there are freeware software for PC. Compared with old hardware sine generators, frequencies and levels are very very precise.
I recently made many measurements of two drivers. It is quite puzzling of how parameters are not so stable. For example, measuring the resonance frequency, I had a difference more than 5% if the driver was horizontal or vertical.

 

[davidallancole]

Thanks for all the responses guys.

What have you found to be the best way to mount the driver in free air to do the test measurements on Fs and the Qs of the driver? I am having a hard time thinking of how to hold them in the air.