REW - Impedance measurement - accuracy?

[LeaUK]

Hi All

New to REW so please forgive me for a while ;-)

So this weekend I created the REW impedance measuring jig including a 0.1% 100R resistor to be used with my two channel Focusrite Scarlett 2i2 (headphone out) and it was a reasonable success, however I have a couple of questions/observations where maybe I'm doing something wrong? I followed the REW guide on impedance measurement and calibration for both shorted measurement 100R and shorted RLeads (croc clips at load). See here for my setup: https://1drv.ms/u/s!AgmKo7c0CB2xjS235af9K7K34W89

The 100R resistor is close to the Scarlett and you'll see croc clips used to short the 100R for cal purposes - this I found to be a must. The croc clips at the end of the lead are for speaker connection. The cables used aren't the best quality (Chinese imports) however I noted the cal process so this would hopefully compensate. I performed repeated tests to provide confidence in the cable and clips construction and all joints throughout were soldered.

Observations:

1. Fiddly - Using the Scarlett (which has two input gain potentiometers), I found it tricky to align both input channels to within 2%, I got there by using the Generator and Levels controls by eye, then switching to impedance measurement mode and running the 100R short cal process (RSense = 0) - some further slight adjustment was needed at this too, tiny movements required.

2. After carrying out the above rig calibration, I noticed the headroom may also need adjusting depending on load R, my test sample was a 0.1% 100R resistor, so maybe a little too large for real world, but it's the only one I can depend on the accuracy for (don't have a 4 wire resistance meter to hand). Don' touch the input pots, simply increase/decrease the headphone out volume.

3. For each speaker measurement it seems to make sense to calibrate both, leads and rig, before hitting go, as any variation of the input pots coupled with internal Scarlett variations (due to temperature/voltage regulation) could seriously affect results.

4. Accuracy - so RLead short gave 766mR, sounds perfectly sensible considering the quality and length of the test leads, I double checked by shorting the leads close to the 100R measurement R and this yielded 0R, so can confidently assume 766mR is good for my leads.

This was entered into the RLeads field.

However - my 0.1% 100R test resistor measured at 99.3R.

So, as i can't measure the exact R for both the measurement R and the R undertest, we could assume worst case of both resistors being 0.1% (0.1R) yielding 0.2% error, however that only allows for 0.2R i.e. a measurement worst case at 99.8R
.
Of course that's probably too simplistic so I'll check the math...

I'm wondering if the accuracy increases with lower R AND for real world impedance plots, do we really care if the measurement is out by a fraction of an R?

Any thoughts most appreciated.

Cheers
Lea

 

[luegotelodigo]

I've never used REW to measure impedance or haven't read the guide so excuse me if I'm saying something stupid but I suppose it uses a shunt resistor with a known resistance and the measurements will be more accurate if the DUT has a similar impedance, in your case 100 Ohm

 

[LeaUK]

I've never used REW to measure impedance or haven't read the guide so excuse me if I'm saying something stupid but I suppose it uses a shunt resistor with a known resistance and the measurements will be more accurate if the DUT has a similar impedance, in your case 100 Ohm

You can find the guide here: impedance measurement

And that's the catch, the measurement R & DUT R are both at 0.1% 100R and it's out by more than 0.2R...

 

[LeaUK]

Here's a thought especially considering that 0.1% resistors aren't cheap and values <100R are rare, use 10 x 1% in parallel and create various known loads to check the accuracy between 0R - 100R.

If my memory serves me correctly, even if every single resistor was at the max 1%, the maximum tolerance of 10 in parallel would be 0.1%. The inductance is also reduced by a similar x10 ratio, i.e. /10, but capacitance increases. Wish I had thought of that before ordering 0.1% 100R devices.

However, I still wonder what's causing the REW measurement inaccuracy.

 

[John Mulcahy]

I'd suspect the stability of the input gain pots on the soundcard, even a tiny change after calibration would shift things. There is a 5% limit for gain matching between channels to allow calibration to be accepted, though that could be fiddly to achieve if inputs have individual gain pots with a wide gain range. In practice the potential deviation you are seeing wouldn't be of any concern for measuring drive unit parameters if that is what you are planning.

 

[LeaUK]

Hi John

I really appreciate your response and indeed I also initially suspected stability of pots/Scarlet, although they feel mechanically very stable, but realising this could lead to error ensured no movement between measurements and cal etc. I repeated the tests to include calibration each time for certainty and the results were identical, measuring a consistent 99.3R which may indicate internal measurement error rather than stability of pots or temperature coefficients within the Scarlet - temperate also remained identical between tests.

I removed the 766m Ohm RLead bias and the 100R measured at exactly 100.0R

I wonder if we're seeing error as a result of some type of system noise?

I also seem to recall a low level signal warning message when measuring the 100R load (which I ignored), I shall crank up the headphone output/both inputs in order to overcome the warning to see if this increases accuracy, but I seem to recall this may not be achievable due to clipping of the reference channel and the minimum signal level requirement for a high 100R load. Again this may suggest that there's not enough range between clipping and the minimum needed to measure the 100R accurately - I'll try again tonight.

Thanks for confirming the practicalities, as yes I'm planing on measuring drive parameters.


Great application John, thanks you so much :-)

Lea