• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required as is 20 years of participation in forums (not all true). There are daily reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Topping D90LE - SINAD measurement (with E1DA Cosmos APU Notch)

OP
R

Rja4000

Major Contributor
Forum Donor
Joined
May 31, 2019
Messages
1,804
Likes
2,160
Location
Liège, Belgium
By the way, here is the same dashboard for 4V using only right channel
(ie not summing both channels)

Difference is below 1dB

2022-08-01 12_04_43-Greenshot_Crop.png


The left channel is "weaker" (Using "weak" word here looks like kind of a joke, actually :) )
The main difference being this 5th harmonic around... -135dBr

2022-08-01 12_09_22-Greenshot_Crop.png
 
Last edited:
OP
R

Rja4000

Major Contributor
Forum Donor
Joined
May 31, 2019
Messages
1,804
Likes
2,160
Location
Liège, Belgium
Note that the above is only if we strictly limit ourselves to 4V.
(Amir did not in his D90SE review)


If we push the 4V range to its limit (I measured 4.12Vrms), here is what we get (in Mono), which is the maximum SINAD I measured for the 4V range

2022-08-01 12_21_14-Greenshot_Crop.png
 
OP
R

Rja4000

Major Contributor
Forum Donor
Joined
May 31, 2019
Messages
1,804
Likes
2,160
Location
Liège, Belgium
My full Multi-Instrument screen contains much more details.

It looks like this

You see the measured "Gain" of the APU (-4.8dB)

On the bottom right, you also see the "Official" (Non AES-17 2015 standard "notch compensation") SINAD, SNR and THD

2022-08-01 12_21_14-Greenshot.png


If anybody's interested, I may share the "psf" (Pannel Settings) file.

Note that DMM fields (Vrms) have to be measured externaly.
I use my .Net program to read a multi-channel Digital Multi Meter and automate stuffs.
But you may do that manually as well...

Another side note:
In the SINAD vs Level plot, maximum SINAD is around 126dB.
Here we reach almost 127dB.
That's because the plot is using the RME AutoRef, which switches the ADC range automatically when level approaches the limit.
For the dashboard, I've manually decreased the range down from 24dBu to 13dBu, which has an impact on noise, and, therefore, on SINAD.
Automation is nice, but not magic...
 
Last edited:
OP
R

Rja4000

Major Contributor
Forum Donor
Joined
May 31, 2019
Messages
1,804
Likes
2,160
Location
Liège, Belgium
Interesting, and the performance actually improves with something like this? I would imagine it would get worse.
That's a classical trick to improve SNR.

If you sum twice the same signal including random noise, part of the summed noise will cancel each other.
Statistically, noise will therefore grow by square root of 2.
While the non random signal will double.
You therefore, statistically, get an improvement of square root of 2 of the SNR. (3dB)

That's in theory.
In practice, not all the "noise" is random, and distortion is of course not.
And both signals never have exactly the same amplitude and characteristics either.
So improvement is less than 3dB.
But, still, we have an improvement.

The same principle is used in other area:
With a tape recorder, doubling the tape width increases SNR.
For a digital camera, increasing the sensor area, increases noise performance and dynamic range.

And so on.
 

IVX

Major Contributor
Forum Donor
Joined
Mar 3, 2019
Messages
1,137
Likes
2,457
Location
South of China, SHZ area
Rja4000, not really like that. Noises aren't canceled but summed with no correlation Ns=(N1^2+N2^)^.5 with factor 2^.5=1.414. Two sines are correlated and the sum factor is 2.
2/1.414=1.414 i.e. 3db of a profit, we can see that 3db when DAC or ADC is used in the MONO mode.
 
OP
R

Rja4000

Major Contributor
Forum Donor
Joined
May 31, 2019
Messages
1,804
Likes
2,160
Location
Liège, Belgium
Rja4000, not really like that. Noises aren't canceled but summed with no correlation Ns=(N1^2+N2^)^.5 with factor 2^.5=1.414. Two sines are correlated and the sum factor is 2.
2/1.414=1.414 i.e. 3db of a profit, we can see that 3db when DAC or ADC is used in the MONO mode.
That's the same, isn't it ?

Similarly, when you sum 2 identical normally distributed signals, their variance (square of standard deviation) will sum, so std dev raises by square root of 2.
 
Last edited:
OP
R

Rja4000

Major Contributor
Forum Donor
Joined
May 31, 2019
Messages
1,804
Likes
2,160
Location
Liège, Belgium
RME ADI-Pro fs R DAC (Mono mode)

SINAD Measurement for each range using Cosmos APU 1kHz Notch
All measurements are done using the RME ADI-2 Pro fs R ADC in "Mono" (M/S) and AutoRef

SINAD is >114dB from 0dBu to 28dBu (0.78Vrms to 20Vrms)

2022-08-14 16_33_36-Greenshot.png




If you use the level in Volt rms (like Amir is doing), you get this

I've added the Topping D90LE for 4V and 5V - also in mono.
As you see, and as expected, its SINAD is higher... for a limited level range, from approx. 1.2V up to 5V
The RME performs better from 0 to 1.2V and, of course, from 5V to 20V.

2022-08-29 12_52_56-Greenshot.png
 
Last edited:
OP
R

Rja4000

Major Contributor
Forum Donor
Joined
May 31, 2019
Messages
1,804
Likes
2,160
Location
Liège, Belgium
the resulting noise isn't canceled, not reduced but increased by a factor 1.414. Fortunately, 2 sines sum increasing more.
Non random signal (signal and distortion) just sum up.
Random noise don't exactly sum up, because of (random) partial cancelation. Statistically, it grows by sqrt of 2 instead of 2 factor.

That's basic statistical law.
And it translates in real life (fortunately).
 
OP
R

Rja4000

Major Contributor
Forum Donor
Joined
May 31, 2019
Messages
1,804
Likes
2,160
Location
Liège, Belgium
Here-attached is the Virtins Multi-Instrument PSF (Pannel Settings File) I created for the measurements above.

This is optimized for 1920 x 1080 resolution, with automaticaly hidden Task Bar (to allow an easy screen snapshot export)

Sans titre-1-Récupéré_2.png


Voltage measurements are foreseen to cover most of my measurements scenarii.
They have to be entered manually.
(I do it automatically through .Net automation though)

I manually drag the relevant DMM (Digital Multi Meter) Voltage measurement to the Dashboard area, according to the test scenario.
(here we see it for a DAC measurement)

NB: You may get the Voltage from the ADC level, if the signal shape is matching a sine wave (which you may check using the Crest factor measurement, just below the generator window - typically, it works above -50dBFS), but you should NEVER use the DAC dBFS level to guess the voltage, since it will vary with the load impedance.


In the bottom right, you'll find the "standard" SINAD, THD and SNR measurements.
(ie the ones computed on a much narrower fundamental frequency bandwidth)
Note that THD figures are the same. The difference lies in the SNR.
Also note that, as you may read in posts above, SNR and SINAD are also giving the exact same figures when the APU notch is not used.

Behind the Signal Generator (top right), you may find more measurements (essentially Noise Levels) using both methods.


I use Dolph-Chebyshev 200 to match AP's proprietary window function.
But since I always tune frequency to avoid spectral leakage, "Rectangle" (ie "NO") window function gives the exact same results.

Note that I'm using the AES-17 standard 997Hz frequency for tests.
This allows the 20th harmonic to show in the THD and THD+N (SINAD) results.
I tuned the APU accordingly.


EDIT: I forgot to mention that, to use the APU notch, you need to select a corresponding compensation file, in the "Spectrum analyzer" configuration screen.
I first performed a measurement of the notch with REW, then converted the data to the right format with Excel and saved it in the compensation file.
This is a one shot operation.

If requested, I may describe this process in details.

RE-EDIT: The Panel Settings file was optimized for "Hot Panel Settings Toolbar" disabled
You may uncheck it in the "Setting" menu.

Some DDPs (measurement values) will show in front of the generator window.
You may either close them or just click on the generator window to push them behind.

The PSF file requires 24 DDPs, which is the maximum number available since Multi-Instrument version 3.9.6.1
I now use 3.9.6.2 (latest to date)

Questions and comments are, of course, welcome.
 

Attachments

  • JAR Dashboard Test with DMM - AES17 Notch - 24DDP.zip
    2.4 KB · Views: 19
Last edited:
Top Bottom