Measuring DAC nonlinearity vs. level

Greg Dunn

Member
Thread Starter
Joined
Oct 4, 2017
Messages
107
I know John is working on an update to allow REW to automate measurements over a range of input/output levels, but in the meantime are there any helpful shortcuts to measuring the linearity of a device over a pertinent range of signal levels? I'm looking at testing the bit nonlinearity of a few DACs so that I can share the data with a couple of online sites.

I do have a pretty precise interface (Scarlett 2i2) that will be sufficient for the measurement. Assuming there is no easy alternative, the plan is to set the signal generator to a certain value and then look at the return value with the "SPL Meter", with both of course set to use the Scarlett interface and (after doing calibration) insert the DUT. I'll probably also use REW's scope so that I can see when I'm approaching the noise floor. Is there a better approach at present?
 

John Mulcahy

REW Author
Joined
Apr 3, 2017
Messages
7,212
Probably better using the RTA than the SPL meter. You don't mention your OS but if it is Windows make sure you use ASIO drivers, Java drivers on Windows only support 16-bit audio data. MacOS and Linux support 24-bit data. Make sure the dither on the signal generator is set appropriately, or turned off if you want to use undithered signals.
 

Greg Dunn

Member
Thread Starter
Joined
Oct 4, 2017
Messages
107
Sorry, I am using Mac and 24-bit data (32 bits when running the DAC from USB). The only reason I thought to use the SPL meter is that it gives an easier to read display whereas the RTA only shows level in the little box in the upper right.

I actually did a test run yesterday and it seemed to work OK, though without an external bandpass filter the noise prevents me from seeing actual data at very low levels. That's something REW can't easily solve.
 

John Mulcahy

REW Author
Joined
Apr 3, 2017
Messages
7,212
without an external bandpass filter the noise prevents me from seeing actual data at very low levels.
The RTA will show the fundamental clearly when it has fallen far, far below the noise floor of the SPL meter display.
 

Greg Dunn

Member
Thread Starter
Joined
Oct 4, 2017
Messages
107
I guess that's right, isn't it! Well, I'll try using RTA for the next measurement and see how it goes.
 

Greg Dunn

Member
Thread Starter
Joined
Oct 4, 2017
Messages
107
OK, I used the RTA in Spectrum mode and was able to greatly improve the precision of the measurement. The noise floor was sufficiently low that I could see the DAC output clearly at 1 kHz way down into the mud. The device I was testing was a very cheap one that I ended up tossing in the "do not use" box because the FR and distortion were extremely poor. My guess is that it has a badly designed analog section because the DAC seemed reasonably linear to about 15 bits (best I could do with its internal noise). My test rig is at least 10 dB better (it's not in the quietest location). So as long as I don't need to go much beyond CD quality, my Scarlett 2i2 is well up to the task of characterizing DACs as well as measuring preamps and amps.
 
Top Bottom