Parameterization of DAC measurements using REW RTA and their impact on the results

[johny_2000]

Hi all,

I need help understanding the parameterization of DAC measurements using REW RTA.
I ran a series of measurements on a test DAC. In this case,
DAC = RME ADI-2/4 Pro SE, XLR output +19 dBU = 6.904 Vrms.
ADC = E1DA Cosmos ADCiso, Mono mode, 3.500 Vrms input = 13.1 dBU.
AES17 Notch = E1DA Comsos APU, +0 dB gain.

I settled on the following measurement method in REW RTA (latest stable version):
- Generator 1 kHz RTA FFT (64k), 0 dbFS, "FS Sine Vrms" = 6.904 V.
- Distortion settings: 'High pass' / 'Low pass' 20Hz...20,000Hz.
- Checkbox "Manual fundamendal" = 6.904 V.
- Checkbox "Use AES17-2015 Standard notch"
- RTA window top right side "FS Sine Vrms" = 3.500 V.
- RTA input "Cal Data" has selected the APU Notch frequency response calibration file.

So here are the results:

Does everything look correct in this measurement?
I am confused by the over-optimistic value (-128 dB) of "THD+N" from this DAC.

 

[johny_2000]

Next step is I'll take a True RMS multimeter and measure the following voltages:
- DAC output voltage,
- AES17 Notch output voltage = ADC input voltage,
- Actual ADC 0 dBFS voltage

I have a feeling that there are some (minor) differences in the I/O characteristics of the devices.
Here's why I think so:

Signal path amplitudes: +19 dBU (DAC) - 30 dB + 20 dB (Notch) = +9 dBU at the ADC input

ADC: +13.1 dBu (3.5 Vrms input) = 0 dBFS; +9 dBu signal applied = -4.1 dBFS measured.
The actual RTA shows a measured ADC value of -5.28 dBFS, which is -1.18 dBFS less than the devices' spec settings.
So we have either a different output voltage from the DAC, or more / less gain from the AES17 Notch filter, or other actual value Vrms = 0 dbFS for ADC.
Or a combination of all of these factors.

 

[johny_2000]

Also, I'm curious about these two settings in the Analysis Preferences tab of the REW:
"Limit cal data boost to 20 dB" and "Apply cal files to distortion"
My AES17 Notch calibration file reaches 30 dB values at certain frequencies.

I don't have a PC nearby to see what they are currently set to, but how do I set them up correctly in my case?
Should I enable both of them?

 

[AnalogSteph]

Does everything look plausible when measuring without the notch?

Your results almost look like you may have had the notch calibration enabled in REW but not the notch filter itself.

Did you get your cal file by measuring the response of your own notch? Something looks really amiss. It thinks your ca. -5 dBFS input signal is at +5.9 dBFS... you should be seeing a much lower 1 kHz peak if that's a 30 dB notch, like about -24 dBFS. Maybe try unchecking the "AES17-2015 standard notch" and check "apply cal files to distortion" instead.

 

[johny_2000]

Does everything look plausible when measuring without the notch?

Your results almost look like you may have had the notch calibration enabled in REW but not the notch filter itself.

Did you get your cal file by measuring the response of your own notch? Something looks really amiss. It thinks your ca. -5 dBFS input signal is at +5.9 dBFS... you should be seeing a much lower 1 kHz peak if that's a 30 dB notch, like about -24 dBFS. Maybe try unchecking the "AES17-2015 standard notch" and check "apply cal files to distortion" instead.

Thank you Steph!

I spent the last day checking several parameters and talking to @JohnPM about my specific case. So far the Notch was in the circuit, that's for sure. Then I calibrated the voltages on the signal generator and RTA. And also checked the options to apply more than 20dB calibration file and apply them to the distortions. Neither of these made any difference to the RTA's distortion report. Then I realized that the only parameter that was shifting all the numbers was the "Manual fundamental" voltage field. I understand the description of this parameter in the REW manual and tried to use it, but unfortunately I have to uncheck it because I don't understand the correct value to enter. I have +19dBu (6.904Vrms) before the notch and +9dBu (1.956Vrms) after it, measured with a True RMS DMM. Without the manual fundamental measurement option, I get THD+N readings that are closer to reality to the DUT's specifications and other software (Multitone, RightMark Audio Analyzer) results.