(Unofficial) Topping D50 III DAC review and measurements

[CedarX]

Thanks, so eventually it might indeed be possible to route the SPDIF through the XMOS, but for good reasons developers hesitate to try to get this going.
In case XMOS would develop an example application that people can copy things may be different. This clearly would be added value for the XMOS.

In the case of the D50-III, I don’t even know if SPDIF is routed through the XMOS chip.

 

[kipik]

Too bad that this model will most probably show OLED display burn in like D50s, since the same screen is displayed permanently even on AUTO mode

 

[Bleib]

Too bad that this model will most probably show OLED display burn in like D50s, since the same screen is displayed permanently even on AUTO mode

It's not an OLED display, it's LCD. You can see the screen on even when the unit is OFF just to display a dot in the lower right corner.
This doesn't have the contrast of an OLED display.

 

[DLS79]

Too bad that this model will most probably show OLED display burn in like D50s, since the same screen is displayed permanently even on AUTO mode

Sorry to be rude, but did you read the manual? If you want the D50s screen to turn off after 30 seconds you don't put it in AUTO!

 

[DLS79]

It's not an OLED display, it's LCD. You can see the screen on even when the unit is OFF just to display a dot in the lower right corner.
This doesn't have the contrast of an OLED display.

The Topping manuals for the D50s & D50 III both claim its an OLED screen.

I've read a few places this is the screen for the D50s.

Nouveau Et Original Écran Oled 0.96 Pouces 13 Broches 64x128 Vg-6428tswag03 - Buy 0.96 Inch 13pins 64x128 Vg-6428tswag03,Lcd Display,Original Product on Alibaba.com

Nouveau Et Original Écran Oled 0.96 Pouces 13 Broches 64x128 Vg-6428tswag03 - Buy 0.96 Inch 13pins 64x128 Vg-6428tswag03,Lcd Display,Original Product on Alibaba.com

french.alibaba.com

 

[Bleib]

The Topping manuals for the D50s & D50 III both claim its an OLED screen.

I've read a few places this is the screen for the D50s.

Nouveau Et Original Écran Oled 0.96 Pouces 13 Broches 64x128 Vg-6428tswag03 - Buy 0.96 Inch 13pins 64x128 Vg-6428tswag03,Lcd Display,Original Product on Alibaba.com

Nouveau Et Original Écran Oled 0.96 Pouces 13 Broches 64x128 Vg-6428tswag03 - Buy 0.96 Inch 13pins 64x128 Vg-6428tswag03,Lcd Display,Original Product on Alibaba.com

french.alibaba.com

There is no way an OLED has this poor black levels. All the lights are off in the room, only some faint light comes from further away.
Unit is off:

The whole screen emits some light just because of the dot in the corner is active. This is typical LCD behavior with no mini-LED zones.

 

[staticV3]

There is no way an OLED is has this poor black levels. All the lights are off in the room, only some faint light comes from further away.
Unit is off:
View attachment 370806View attachment 370807
The whole screen emits some light just because of the dot in the corner is active. This is typical LCD behavior with no mini-LED zones.

could also be a poorly designed GUI which doesn't make use of OLED's per-pixel illumination by having a gray background

 

[Bleib]

New firmware with special instructions, safest is to install 1.28 first then 1.3

D50 III version V1.30 firmware update - TOPPING AUDIO

www.toppingaudio.com

 

[Rja4000]

New firmware with special instructions, safest is to install 1.28 first then 1.3

D50 III version V1.30 firmware update - TOPPING AUDIO

www.toppingaudio.com

From what I read, the only functionalities changed since v1.21 are
---------------------------------------------------
**V1.23 HV:2.5 (May 13, 2024)**
---------------------------------------------------
1. Added the option to switch between preamp mode and DAC mode.
2. Added the option to switch between UAC1.0 and UAC2.0.
3. Added the function of saving PEQ settings in user-defined settings C1 and C2.

Rest is bug fixing.

 

[SloPoke]

New firmware with special instructions, safest is to install 1.28 first then 1.3

D50 III version V1.30 firmware update - TOPPING AUDIO

www.toppingaudio.com

Probable mistake in the firmware instruction notes:

Important Note: For D50 III product users with serial numbers starting with 2310, 2403, or 2405 and firmware version V1.28 or earlier, it is crucial to upgrade to V1.30 before updating to V1.28. Directly upgrading to V1.30 might cause UAC1.0 to fail to start.​

Says that its crucial to upgrade to V1.30 before updating to V1.28 and this is likely (?) revearsed. Rather it should say (I think) "....it is crucial to upgrade to V1.28 before updating to V1.30".

 

[Bleib]

Rest is bug fixing.

Well, that's great. Who wants bugs?

1.28 will obviously be needed for future versions to expand how much can be written in the firmware.

1.3 fixed DSD sound problems as well

Also note:
"To upgrade the firmware, the product must be set to DFU mode: Power on while holding down the button on the left side of the front panel to enter DFU mode."

 

[FrenchFan]

Hello everyone

I add here a personal measurement on the RCA output of the D50III.

I implemented the reduction of measurement noise by cross correlation in my tool.

I use an E1DA cosmos ADC coupled with an E1DA cosmos scaler.

The noise floor is ~ -122dBFS / -124dBFSA for the ADC the E1DA scaler
does not bring additional noise so much it is low at the gain used.

The principle of cross correlation is to measure the same signal on the 2 channels of the ADC
and to correlate these 2 channels this functionality is implemented by the analyzers QuantaSylum.

https://forum.quantasylum.com/t/mono-and-aes-17-notch/787

Attached is the measurement of the 2 channels and the measurement by correlation.

The average of the cross correlation of 1000 samples of 32K samples reduces the noise of the ADC by about 15dB.
This type of measurement is long and must be done on a prior recording. No real-time measurement possible

We see today that DACs and ADCs have practically the same noise level and it is quite complicated
to carry out measurements, the cross correlation brings a possibility of being a little more precise,
although with this noise level I doubt that we can tell the difference between -120dB and -124dB of noise.

Bye

 

[Rja4000]

Hello everyone

I add here a personal measurement on the RCA output of the D50III.

I implemented the reduction of measurement noise by cross correlation in my tool.

I use an E1DA cosmos ADC coupled with an E1DA cosmos scaler.

The noise floor is ~ -122dBFS / -124dBFSA for the ADC the E1DA scaler
does not bring additional noise so much it is low at the gain used.

The principle of cross correlation is to measure the same signal on the 2 channels of the ADC
and to correlate these 2 channels this functionality is implemented by the analyzers QuantaSylum.

https://forum.quantasylum.com/t/mono-and-aes-17-notch/787

Attached is the measurement of the 2 channels and the measurement by correlation.

The average of the cross correlation of 1000 samples of 32K samples reduces the noise of the ADC by about 15dB.
This type of measurement is long and must be done on a prior recording. No real-time measurement possible

We see today that DACs and ADCs have practically the same noise level and it is quite complicated
to carry out measurements, the cross correlation brings a possibility of being a little more precise,
although with this noise level I doubt that we can tell the difference between -120dB and -124dB of noise.

Bye

A few comments, if you allow me.

1. For low noise, you may as well connect each Scaler output to each ADC input and use ADC in mono mode. You'd get around 3dB less ADC noise. (The Scaler is meant for that: a mono signal on left input is duplicated on both outputs when the right input is not connected)

2. Dynamic range - which is how we measure noise for a DAC - is usually done following the AES-17 method.

You first check the maximum output level that gives less than 1% THD, then measure noise with a signal 60dB below that.
Signal should be a 997Hz sine wave and measurement done wIth a bandwidth limited to 20kHz.
(Standard weighting here is CCIR-2k, but you may measure using no weighting or A weighting. Just specify what you've been using, to allow comparison.)

Max level is at 0dBFS here: the DAC never reaches 1% THD, even at max output level.

If you do the noise measurement manually, you could increase the Scaler gain to maximum when measuring noise with the -60dB signal.
You'd then have the DAC noise raised significantly above the ADC noise, and measurement will then be more relevant, without averaging or correlation tricks (and faster too).

Oh, and "Rectangle" or "no window" FFT is more accurate than Blacmann Harris or any other, for accurate noise level measurement. You'd need to aligned your 997Hz frequency to an exact bin center frequency, though. REW (or most FFT softwares) would do that for you.
(In case it doesn't work, that's 997.56 Hz for 48kHz/32768 FFT. )

I use a similar method to get my Dynamic range figures in this review.

 

[FrenchFan]

I think I explained myself badly.

I use the E1DA scaler as you indicate,
however I need the same 2 signals to do the cross correlation,
so E1DA adc in stereo mode.

I'm not talking about -3db of noise that the mono mode
of the E1DA adc brings but about -15db of noise with cross correlation.

Concerning the noise of the E1DA adc I don't even need to measure it,
it is marked behind the ADC box (-124dBA) so ~ 122dbFS for an E1DA adc iso B.

I just wanted to show that we can reach practically -124dB (A weighted) on an RCA output of the D50III.
Each output measured without correlation reaches ~ -120dBFSA.with a notch of THD+N = one octave (AES17 2015 standard notch)

I wanted to show here that we have another method of noise reduction
than the notch of the E1DA apu, certainly less effective, but also implemented in the QuantaSylum.

I don't know how you made the measurement concerning the RCAoutput and it's good that I can contact you,
did you use the notch filter of the E1DA apu??, why is the foot of the fundamental so wide in your measurements,
and in the measurement of the BluetoothLDAC I find a much narrower fundamental foot ,identical to what I measure,
the jitter of the D50III is really very low and gives a very narrow fundamental foot.

If you only find -120db with the notch of the apu there is a problem.

Look at the attached signal under REW, direct measurement with the E1DA adc stereo + E1DA scaler ,
we find N+D ~~ THD+N(A weighted) =-120.xx dBFSA with or without "AES17 2015 standard notch".

So we directly measure THD+N(A weighted) ~ -120dBFSA without APU for an RCA output.

Correlation give -123.48 dBFSA

 

[nanook]

Cross correlation is an interesting method to "dig deeper into the noise floor". I have not yet fully understood this method though.
Thanks a lot for pointing this out.

The APU is a very powerful add-on to Ivan's ADC(iso) if it comes to measuring S/N and THD at 1 KHz or 10kHz, but it is limited to these frequencies. Cross-correlation should work for all sorts of stimuli, even for Multitone.

REW offers a feature "Coherent averaging" that helps to bring the noise floor down as well; it is only suited for single tone measurements as far as I understood, since it phase-aligns the captured data to the fundamental.
https://www.roomeqwizard.com/help/help_en-GB/html/spectrum.html (see section "Coherent Averaging".)

Edit: A friend of mine uses "time domain averaging" extensively to get random noise down. This is basically like having 100 or 1000 ADC channels instead of 2 channels only (mono mode). It needs proper alignment of the captured data such that it is not trivial. This is a wide and interesting field.

You were talking about "my tool". You have written a tool to do this? I have decided to go with E1DA and REW (not quantasylum) so I'm interested in this feature.
Would you mind comment on this tool and on how it is used? Or would you even share this tool?
Edit: This discussion should probably take place in a new thread.

 

[Rja4000]

This discussion should probably take place in a new thread.

Indeed

 

[FrenchFan]

Cross correlation is an interesting method to "dig deeper into the noise floor". I have not yet fully understood this method though.
Thanks a lot for pointing this out.

The APU is a very powerful add-on to Ivan's ADC(iso) if it comes to measuring S/N and THD at 1 KHz or 10kHz, but it is limited to these frequencies. Cross-correlation should work for all sorts of stimuli, even for Multitone.

REW offers a feature "Coherent averaging" that helps to bring the noise floor down as well; it is only suited for single tone measurements as far as I understood, since it phase-aligns the captured data to the fundamental.
https://www.roomeqwizard.com/help/help_en-GB/html/spectrum.html (see section "Coherent Averaging".)

Edit: A friend of mine uses "time domain averaging" extensively to get random noise down. This is basically like having 100 or 1000 ADC channels instead of 2 channels only (mono mode). It needs proper alignment of the captured data such that it is not trivial. This is a wide and interesting field.

You were talking about "my tool". You have written a tool to do this? I have decided to go with E1DA and REW (not quantasylum) so I'm interested in this feature.
Would you mind comment on this tool and on how it is used? Or would you even share this tool?
Edit: This discussion should probably take place in a new thread.

Hello Nanook

Yes I am making several tools
with "octave", the free matlab.

My main tool is the distortion measurement + IMD

Putting my script online right away risks drowning me under
questions, and like everyone else I am not always in front of my PC, I can do it for a few people but no more.
In addition I still consider myself a beginner under "octave" I am not a
computer scientist, I get by, so the software is full of small
BUGS.

I am mainly trying to understand the BF measurements and I am not there yet.

On the other hand I am going to put online under Discord the
FFT calculation scripts with the weighting windows + the THD+N
calculation script + the calculation and launch of the correlation, to try
the concept.

Discord: https://e1dashz.wixsite.com/index/cosmos-adc

You go to the end of the page, you will find the link to
Discord, you better register, it is the source of
all explanations for E1DA products, Ivan is
always there to answer.

Under record, I am "kiki".

Concerning a new discussion thread, I agree but I am
a little clumsy in ASR to create one, If you want to start, I follow you.

I leave you the links of my sources concerning the theory.

Reducing the Noise Floor and Improving the SNR with Cross-Correlation Techniques

Ultra-sensitive voltage measurements are a crucial requirement for applications such as quantum device characterization [1]. Weak signals become easily obscured by thermal noise, especially that coming from the measurement apparatus...

www.zhinst.com

https://www.edn.com/measuring-small-signals-accurately-a-practical-guide/

I just got the D50III and haven't tried it yet.

Regarding the coherent average in REW, I get a similar result by Autocorrelating the signal
with a shift on the sample blocks, the noise is never correlated, so the noise reduction
both towards infinity and the number of correlations time towards infinity.

 

[Rja4000]

Concerning a new discussion thread, I agree but I am
a little clumsy in ASR to create one, If you want to start, I follow you.

Here you go:

Using Cross Corelation to lower influence of ADC for DAC measurements

I start this thread to allow @FrenchFan to explain his noise reduction technique and a discussion on this subject One of the refetence links he provided: https://www.edn.com/measuring-small-signals-accurately-a-practical-guide/ UPDATES V5.40 beta 49 is now available with the cross...

www.audiosciencereview.com

Suggestion to
@Jimbob54
Maybe you could move posts related to this topic from #172 into this new thread ?

 

[Palladium]

could also be a poorly designed GUI which doesn't make use of OLED's per-pixel illumination by having a gray background

I doubt it. I have the D50 III, the backlight leakage isn't uniform on the screen which is characteristic of a LCD.
OLED greys don't do this, and I also had the older D50S which was clearly OLED just from the black levels. (That never had any burn-in under auto mode even when 24/7 on anyway and I got it used too)

That said LCD is clearly the superior choice over OLED for this application, nobody needs the strengths of OLED in exchange for burn-in risk just to display basic information.

 

[FrenchFan]

Hello @Rja4000

1)

I think I'm a bit clumsy,
I can't find "Topping Tune" where
is it?

2)

In the D50III setup there is
the "Bandwith setting" option which modifies
the Jitter very significantly.

In position 15 the jitter is really horrible.
I set it to 1 when I did the cross-correlation
measurements.

1 = best quality.

Do you know the interest of this setting by chance.