How to - Distortion Measurements with REW

Just picked up a Motu M4 and REW and tried a loopback test of the M4. I'm getting weird results with some sample rates. I'm using the ASIO driver that came with the M4 and made sure the rates are the same in REW. The test at 44.1 looks right, whats with the other two? There 48k and 196k. 88.2k, 96k and 172.4k also look fine.
 

Attachments

  • 44.1.PNG
    44.1.PNG
    652.8 KB · Views: 526
  • 48k.PNG
    48k.PNG
    715.4 KB · Views: 521
  • 196k.PNG
    196k.PNG
    867.5 KB · Views: 528
Last edited:
Changed buffer, no difference. But I did find a clue. When I use the RTA real time at the bad sample rates the noise floor/harmonics jumps 100db for a fraction of a second every second or so. This doesn't happen otherwise.
 

Attachments

  • 196 rta1.PNG
    196 rta1.PNG
    400.8 KB · Views: 492
  • 196k rta 2.PNG
    196k rta 2.PNG
    405.7 KB · Views: 477
I can't directly compare but I wonder if Windows 10 is worse than Vista for this.

I have gone to paying for Windows Server Standard 2016 after I got annoyed with all the Windows 10 shove-down-your-throat cr@p.

You can get a 6 month evaluation copy free of charge and am looking trying out the Windows Server Essentials 2019 on my next computer.

Nice lean installation and worth spending the money to avoid all the Windows 10 aggravation. (need a reliable and smooth running PC for other activities).
 
Founder of XSA-Labs
Joined 2012
Paid Member
Just picked up a Motu M4 and REW and tried a loopback test of the M4. I'm getting weird results with some sample rates. I'm using the ASIO driver that came with the M4 and made sure the rates are the same in REW. The test at 44.1 looks right, whats with the other two? There 48k and 196k. 88.2k, 96k and 172.4k also look fine.

Try a basic test using default non-ASIO driver. If that works, then it might be some sort of interface issue.
 
Founder of XSA-Labs
Joined 2012
Paid Member
I built a very steep 100kHz low pass filter that was recommended by someone (I’ll have to dig through the forum to see who I got it from) in the Class D amp forum. It is supposed to have -130dB at 400kHz and is helpful to reduce any aliasing caused by HF PWM carrier frequencies used in Class D amps or SMPS. It helped to remove the hash I was seeing in my FFT spectra and dropped the noise floor by about 10dB (which I assume was there due to contamination from high speed switching stuff above 100kHz). The filter costs about $30 in parts for a balanced setup needed for BTL amps. Not hard to make and a worthwhile tool to add to your measurement toolbox.

Schematic and predicted transfer function:
attachment.php


Built using 5% film caps for nF values and 5% NP0 for pF values. Inductors are shielded CoilCraft rated for 300mA so should be good:
attachment.php


In action measuring a 350w Class D GaN amp:
attachment.php
 

Attachments

  • 0DF4692F-7C67-4271-AE20-76BC5CBACC56.jpeg
    0DF4692F-7C67-4271-AE20-76BC5CBACC56.jpeg
    519.1 KB · Views: 1,000
  • 8899107C-C688-4CD9-AAD0-EA0F065FDFDF.jpeg
    8899107C-C688-4CD9-AAD0-EA0F065FDFDF.jpeg
    570 KB · Views: 1,081
  • BFC40FD9-4B15-4ECB-9240-A6BB3BEF05A7.jpeg
    BFC40FD9-4B15-4ECB-9240-A6BB3BEF05A7.jpeg
    456 KB · Views: 1,011
Last edited:
You want to use bulk metal resistors - they are non inductive and make a huge difference. The gold metal shells ones add distortion. The wire wound ceramics do as well.

The best ones are made by EBG. You can get used UXP-300 for about $15 on eBay. Not sure how much new but probably pricey. These are 300w 10ohm designed to be used with a heatsink like a CPU cooler or large aluminum plate.

For example:
QTY:10, EBG (VISHAY) UXP/300 10RK 10 OHM/10OHM 300W FOIL NON-IND.POWER RESISTOR | eBay

Another way is the series parallel quality 1W metal thin film resistors. You can make circa 25W dummy loads this way 5x5 8ohm config for example. Avoid metal oxide as they are distortion generators as well.

The worst ones are the huge 6in long 100w wirewound heater element style coated with green ceramic.

Just changing to EBG UXP-300 resistors and using a balanced input on my Focusrite with quality MKP caps cleaned up the measurements quite a bit.

X - any thoughts on using these Caddock "Kool-Pak" resistors as test loads? They go up to 100 Watts and are non-inductive: http://www.caddock.com/Online_catalog/Mrktg_Lit/MP9000_Series.pdf
 
Given the distortion levels I have measured on the L20 V7, MX50x2 and my uPC1342V I wonder how bad the gold colored aluminum cased resistors really are. (I am using one 4 Ohm 100W gold colored aluminum cased non-inductive wire wound.)

The best I have seen on the best amplifier is 0.001% 2nd harmonic. The other "identical" channel is 0.003% and I wonder if I will ever figure out why. (Measured using E-MU 1820m which is 0.0001% loopback.)

At what level (of measured distortion) are the 100W gold colored aluminum cased non-inductive wire wounds a problem?
 
Last edited:
Metal resistors (thin film or wirewound) are not a problem for distortion, unless the wire is magnetic (never seen this) - then you can get distortion due to non-linear inductance.

Basically metal conductors are extremely linear (this is exactly a statement of Ohm's law), although at _very_ low frequencies you may be subject to thermal distortion due to temperature coefficient not being zero, but this is not normally observable at audio frequencies.

The metal encapsulated resistors are typically wirewound in ceramic paste inside a metal heatsink - i.e. no different from other wirewound except for having a heatsink. That's why they are known as "aluminium housed wirewound resistors"!
 
Load resistors form a voltage divider with the wiring, output inductor and Zobel network - especially at higher frequencies, so they need to be linear enough not to disturb the measurements of the actual amp - same reason speaker terminals should be brass not steel, so as not to intruduce non-linear element into the output circuit.


Its not a big point however, unless you are trying to reach ppm levels of distortion.
 
Member
Joined 2016
Paid Member
There was a thread discussing distortion in resistors a few years back, where several professionals also contributed:
Distortion in low resistance wirewound resistors

XRK971 also discussed earlier in the thread that he had found distortion in power resistors that became a factor when measuring THD for some of his low-distortion amplifier designs, for example #826 and #535.


Regards, Claas
 
AX tech editor
Joined 2002
Paid Member
Non-linearity in resistors is not disputed. But you are measuring distortion in resistors, you measure distortion in amplifiers.
If your amplifier distortion changes measurably because of a tiny change in load value, that's a pretty bad amplifier and I submit that this amplifier distorts pretty much even with a perfectly linear load.

We should be clear what it is we want to measure. If we want to measure the amp, measure at the amp output, not after some unspecified wire, solder joints and then on a load resistor. You really wouldn't know what it is you are measuring if you see distortion if you look across the load resistor.

And if you are after the amp performance without the influence of output cables, loads and their solder joints, measure at the feedback pick-off point. As far as the amp is concerned, that is the point where the output is.

You can also make the case that the distortion of a solder joint with its dissimilar metal connections would swamp any load resistor distortion anyway.

Jan