Topping D90SE Measurements (DAC)

[yanm]

But what is the practical difference? I can certainly see a ~1-2 dB SNR difference between analog and digital volume control at lower volume positions (depending on how low noise your analog volume control is) but is this worth worrying about when noise floor is already so low?

I built a simple spreadsheet to model amp + DAC noise to help me better understand. Here is a DAC with ~1.4 uV noise over 22 kHz BW (supported by both @nagster's and @Amir's measurements of the D90SE) and an amplifier with 25.6 dB gain, 35 uV noise over 22 kHz BW (roughly the best I've seen from NC252MP measurements). This assumes 4 V at 0 dB volume position which is why the plots are limited to -8 dB and below to avoid amplifier clipping. Orange is an analog volume control with 0.5 uV noise over 22 kHz BW and blue is a digital volume control.

At -30 dB volume position you are looking at -94.6 dB SNR for digital and -96.3 dB SNR for analog, is this difference worth worrying about?

24 bit data - SNR
View attachment 197209

16 bit data - SNR
View attachment 197210

Michael

Well said, I think that digital volume control is fine in many/most situations. One should look at the sound pressure produced by just the noise of the DAC (that is, after the amplifier and transducer). If it’s well below 0 dB SPL with digital volume control, you’re good to go and analogue volume control will make no audible difference whatsoever. For example, I did those calculations for my system there: https://www.audiosciencereview.com/...ng-e50-review-balanced-dac.26219/post-1087704

Some systems with very efficient transducers (e.g. IEM headphone) or with strong amplification may be better off with analogue volume control as the noise floor of the DAC / digital volume control may cross the audibility threshold.

 

[jae]

What software are you using for the non-multitone tests/graphs?

 

[VintageFlanker]

What software are you using for the non-multitone tests/graphs?

All clearly stated in the review. Look closer.

 

[Rja4000]

Topping D90SE Measurements (DAC)

View attachment 197253
​

Hi folks,

I just got my hands on the mighty Topping D90SE and wanted to share my objective analysis. Some might wonder what would be the point to post this, when we already have a complete test at ASR, with our host's skills and his APx555... Well, at least three reasons, I would say:

1) I need an undisputed reference to set the benchmark for my further gears analyses.
2) That way, I am curious to find if I may reach the limits measurable by my ADC (no doubt I should).
3) In any case, repeatability is always a good thing ! This should remove some doubts about Amir receiving a cherry-picked product from the manufacturer... In case you're wondering, this unit is brand new from Amazon FR in March 2022, so nine months after Amir tested it. By today, MSRP is 899€.

Disclaimer: Measurements you are about to see are not intended to be as precise or extensive than what you get from a 30k€ AP. There is obviously both hardware and software limitations here, so not quite apples to apples comparison with Amir's testing. For example, I estimated SINAD (AKA THD+N) to be usually 5 to 6dB worse with my measurements rig, when compared to ASR reviews of the exact same products. Still, this data is enough to have a pretty good idea if the gear is bad or not, stellar, broken, or sub-par...

- ADC : E1DA Cosmos (Grade B), set to 4.5Vrms input for measuring XLR, or 2.7Vrms for RCA. Minimum phase filter.
- Software : RMAA 6.4.5 PRO and Multitone Loopback Analyzer 1.0.13.
- Method : 8 runs for each test, then I chose the closest to the average. All tests are running 24bits / 44.1Khz.
- Topping D90SE : Fixed XLR outputs (4.2Vrms) and USB input for most tests, unless specified otherwise.​

RMAA Summary​

Test	Results​	RMAA rating​
Frequency response (from 40 Hz to 15 kHz), dB​	+0.04, -0.02​	Excellent​
Noise level, dB (A)	-125.0​	Excellent​
Dynamic range, dB (A)	124.5​	Excellent​
THD, %	0.00005​	Excellent​
THD + Noise, dB (A)	-115.7​	Excellent​
IMD + Noise, %	0.00031​	Excellent​
Stereo crosstalk, dB	-125.9​	Excellent​
IMD at 10 kHz, %	0.00023​	Excellent​

Wow. Both THD and THD+N are the best I've measured so far and would translate, per my own estimations, to at least 120dB SINAD in a proper measurements rig. Needless to say, RMAA's rating is somehow outdated to evaluate this stunning level of performance. We reach -125dB Noise, which is the limit of my Cosmos ADC (Grade B)... at least in Stereo Mode. It is now pretty clear that I have to set my Cosmos in Mono Mode to find out the real D90SE abilities. To push the Cosmos to its absolute limits, I also measured the "5V mode" (really 5.2Vrms):​

​

Test​	Topping D90SE XLR​	Topping D90SE XLR (Mono)​	Topping D90SE XLR 5V (Mono)​
Frequency response (from 40 Hz to 15 kHz), dB	+0.02, -0.02​	+0.02, -0.02​	+0.02, -0.02​
Noise level, dB (A)	-125.0​	-127.4​	-127.7​
Dynamic range, dB (A)	124.5​	126.6​	126.7​
THD, %	0.00005​	0.00004​	0.00003​
THD + Noise, dB (A)	-115.7​	-117.7​	-119.3​
IMD + Noise, %	0.00031​	0.00028​	0.00027​
IMD at 10 kHz, %	0.00023​	0.00019​	0.00016​
Stereo crosstalk, dB	-125.9​	/​	/​

I have no words. With setting to 5V, THD is simply out of this world at only... 0.00003% (-130dB!), but the biggest surprise was to realize that, yes, I can measure up to 119dB SINAD at home... About 127dB for both SNR and DR are still the limitations here, being the absolute limits of my Cosmos. Before setting back to Stereo/4V, for a more apples to apples comparison with others DACs, let's have a look at this masterpiece:​

​

THD + Noise (at -3 dB FS)​

View attachment 195761​

...Now, let's run the usual bunch of tests :​

Frequency response​

View attachment 196025​

	Left​	Right​
From 20 Hz to 20 kHz, dB	-0.03, +0.03​	-0.04, +0.01​
From 40 Hz to 15 kHz, dB	-0.02, +0.02​	-0.03, +0.01​

Noise level​

View attachment 196029​

	Left​	Right​
RMS power, dB	-122.9​	-123.8​
RMS power (A-weighted), dB	-124.5​	-125.5​
Peak level, dB FS	-87.1​	-87.1​
DC offset, %	-0.0​	-0.0​

​

THD + Noise (at -3 dB FS)​

View attachment 196030​

	Left​	Right​
THD, %	0.00005​	0.00005​
THD + Noise, %	0.00019​	0.00017​
THD + Noise (A-weighted), %	0.00017​	0.00016​

​

Dynamic range​

View attachment 196043​

	Left​	Right​
Dynamic range, dB	+123.1​	+124.0​
Dynamic range (A-weighted), dB	+124.1​	+124.9​
DC offset, %	-0.00​	+0.00​

​

Intermodulation distortion (swept tones)​

View attachment 196044​

	Left​	Right​
IMD + Noise at 5000 Hz,	0.00024​	0.00022​
IMD + Noise at 10000 Hz,	0.00024​	0.00020​
IMD + Noise at 15000 Hz,	0.00025​	0.00021​

Intermodulation distortion​

View attachment 196045​

	Left​	Right​
IMD + Noise, %	0.00032​	0.00030​
IMD + Noise (A-weighted), %	0.00021​	0.00018​

Stereo crosstalk​

View attachment 196046​

	Left​	Right​
Crosstalk at 100 Hz, dB	-127​	-126​
Crosstalk at 1000 Hz, dB	-126​	-124​
Crosstalk at 10000 Hz, dB	-124​	-116​

What about RCA outputs? I expect a logical drop in performance (whatever E1DA says, Cosmos ADC is not optimal for single ended measurements), the question is : by how much? Let's find out:

THD + Noise (at -3 dB FS)
View attachment 196068

XLR VS RCA​

Test​	Topping D90SE XLR​	Topping D90SE RCA​	Topping D90SE RCA 5V​
Frequency response (from 40 Hz to 15 kHz), dB​	+0.02, -0.02​	+0.02, -0.02​	+0.02, -0.02​
Noise level, dB (A)​	-125.0​	-121.6​	-122.0​
Dynamic range, dB (A)​	124.5​	121.6​	121.6​
THD, %​	0.00005​	0.00007​	0.00007​
THD + Noise, dB (A)​	-115.7​	-112.5​	-113.2​
IMD + Noise, %​	0.00031​	0.00039​	0.00038​
Stereo crosstalk, dB​	-125.9​	-116.5​	-116.5​

Apart from crosstalk being logically inferior, I am pleased to see that THD is still exceptionally low at 0.0007%. Using 5V Mode (2.6Vrms), THD+N is still 6dB better than my Matrix Mini-i Pro 3's balanced outputs, which is quite an accomplishment! For my future measurements of unbalanced DACs, I am finally glad to know the reference being that high !​

USB VS S/PDIF​

Test​	Topping D90SE USB​	Topping D90SE S/PDIF​
Frequency response (from 40 Hz to 15 kHz), dB​	+0.02, -0.02​	-0.00, -0.03​
Noise level, dB (A)​	-125.0​	-124.4​
Dynamic range, dB (A)​	124.5​	123.9​
THD, %​	0.00005​	0.00005​
THD + Noise, dB (A)​	-115.7​	-115.8​
IMD + Noise, %​	0.00031​	0.00031​
Stereo crosstalk, dB​	-125.9​	-124.7​

Performance seems to drop by a little using S/PDIF. Ignoring run-to-run variations and averages, I would call both results to be similar. In any case, nothing to worry about.​

Multitone Analyzer (100 tones, 5 averages)

RCAView attachment 196057

XLRView attachment 196058

XLR 5VView attachment 196060​

Multitone test of unbalanced outputs, while not perfect, is still a hair better than (again) my Matrix balanced. XLR performance is by far the best I saw in this test. I set back the D90SE to 5V for the last run and got this stellar picture you see above.


One last thing...

When running my first measurements, I noticed quite a few variations in overall frequency response. I then realized that I accidentally changed the filter with the remote. The D90SE has 7 filters called "Modes" from 1 to 7 (default is Mode 3).
View attachment 196063
None of them measured the same as far as FR is concerned (44.1Khz). Mode 1, in particular, is acting weird, really... My recommendation is for Mode 5 (Fast roll off linear), being the flattest :

View attachment 196062​

... Well, that's all, folks !

Flanker rating : State Of The Art.

​

Very nice

I'll try and mimic those with my
RME ADI-2 Pro fs R.

I can say already I'm able to reach 119dB SINAD with it as well at around 19dBu in Mono mode, since I've measured that multiple times.

But the other measurements will be interesting too.
If I had more time...

 

[IVX]

Actually, I'm not sure if RMAA calculates THD+N correctly, I didn't test that but all FFT SW that I tested were wrong, excluding REW and Arta.
THD+N-115db looks too bad for that DAC and ADC, I think REW will show -120db+. Also, the THD+N result of RMMA is quite low due to the reference level isn't -.5dbfs, or the ADC unit has a mrginal SNR for grade-B.

 

[pma]

@IVX , RMAA calculates THD+N wrong, I agree.

 

[IVX]

actually, I did not test RMAA just suspect because many others are wrong too. I don't like RMAA because it is auto-test, maybe that's nice for some kinda bloggers who want spend to one test no more than 3 min without understanding the result.

 

[VintageFlanker]

@IVX, @pma. It's"wrong" anyway since it normalised at -3dBFS.

THD+N-115db looks too bad for that DAC and ADC, I think REW will show -120db+

I will check that. Do you think it's achievable in stereo mode for a Grade B?

I don't like RMAA because it is auto-test, maybe that's nice for some kinda bloggers who want spend to one test no more than 3 min without understanding the result.

Non taken. You would be plain wrong if you think these tests took me 3 minutes with no verification.

 

[IVX]

I know a few reviewers who told me they gonna use RMAA only because they have no idea how to measure without results marked as good/average/bad. BTW, one of them even do not realize that Cosmos ADC has a lot higher precision vs his motu )

 

[pma]

I have done several tests where RMAA shown higher THD+N (1kHz) A-weighted than unweighted, with a “normal” spectral distribution. This is close to impossible.

 

[IVX]

I have done several tests where RMAA shown higher THD+N A-weighted than unweighted, with a “normal” spectral distribution. This is close to impossible.

it is possible if the 3rd harmonic is dominating and a lot higher than the noise floor, so the peak of the AW-filter gain is matched with 3000Hz(+1.2db I just checked), but noise BW limiting doesn't affect the result.

 

[IVX]

I want to give you guys another example when non-pro, non-blogger but just one Korean guy Hee found that ES9068 has noticeable worse IMD 19k/20k performance vs ES9038Q2M. I'm pretty sure that nearly nobody here even heard about that. Hee found that with Cosmos ADC and his curiosity

 

[VintageFlanker]

Actually, I'm not sure if RMAA calculates THD+N correctly, I didn't test that but all FFT SW that I tested were wrong, excluding REW and Arta.

What about 1Khz Sine in Multitone Analyzer? Just ran a few tests with my Matrix (Stereo mode for both, but only R channel recorded in MA)

RMAA :

THD + Noise (at -3 dB FS)​

​	Right​	Left​
THD, %	0.00037​	0.00020​
THD + Noise, %	0.00049​	0.00036​
THD + Noise (A-weighted), %	0.00054​	0.00038​

 

[pma]

it is possible if the 3rd harmonic is dominating and a lot higher than the noise floor, so the peak of the AW-filter gain is matched with 3000Hz(+1.2db I just checked), but noise BW limiting doesn't affect the result.

This measurement, for example. 3rd is not dominating and there is nothing special.

 

[CinDyment]

I noticed 1 main complaint about it on many subjective reviews.
Narrow soundstage... Amir said its crosstalk is enough, but why does competing sabre dacs like Gustard X26 Pro and Stuff from Matrix guys said to sound wider and measurements also show significantly better crosstalk...
Topping a90 same story, all complaints mainly about it sounding narrow.
Is small box is not enough to properly separate right and left channels from each other to allow better crosstalk?
Both Matrix and Gustard top tier models have bigger boxes.

Brains make up all kinds of stuff absent actual information.

That or people didn't know what filters they were using or were default. Nothing like some high frequency noise and ultrasonics distorting into audible frequencies (speakers) to create "air". All supposition, but nothing in those numbers will affect sound stage. Sound stage is time of arrival 200-1500Hz and frequency response. Most comments about sound-stage are total BS.

 

[Rja4000]

E1Da makes for a better measurement instrument than the RME ADI-2 PRO fs, at least regarding THD

Well, yes and no.

If you set the E1DA to the right range, to optimize its performance, it's actually able to measure lower THD and noise.

But there are several limitations :
Those range changes require DIP switches changes, and they will not last forever.
Also, each range change will modify the ADC input impedance, which may have an impact on the DUT performance.
And, most important in my eye, you can't sync the E1DA clock to the DAC clock, which makes some measurements way more difficult (especially for frequency response).

So, in short:
for the very best DACs, at their most performing output level, the E1DA provides more accurate SINAD figure than the RME, and will give you those last 2-3dB of SINAD, but for most other measurements, the RME remains a more efficient solution.

And for 99% of DACs, the RME is enough, anyway, to check for transparency (one can measure un-weighted SINAD up to slightly above 116dB in loopback mode with the RME)

At least that's my experience, using both.

Beware though that Ivan is preparing some magic additions that may push the limits even further, like a high performance impedance adapter and a pre-amp with notch filter.

Of course, the E1DA ADC is way cheaper.
While the RME can do so many more things...

 

[Rja4000]

What about 1Khz Sine in Multitone Analyzer? Just ran a few tests with my Matrix (Stereo mode for both, but only R channel recorded in MA)
View attachment 197678

RMAA :

THD + Noise (at -3 dB FS)​

View attachment 197680

​

​	Right​	Left​
THD, %	0.00037​	0.00020​
THD + Noise, %	0.00049​	0.00036​
THD + Noise (A-weighted), %	0.00054​	0.00038​

Is there a way to increase the line thickness in these graphs ?
They are hardly readable on a smartphone.

 

[IVX]

What about 1Khz Sine in Multitone Analyzer? Just ran a few tests with my Matrix (Stereo mode for both, but only R channel recorded in MA)
View attachment 197678

RMAA :

THD + Noise (at -3 dB FS)​

View attachment 197680

​

​	Right​	Left​
THD, %	0.00037​	0.00020​
THD + Noise, %	0.00049​	0.00036​
THD + Noise (A-weighted), %	0.00054​	0.00038​

REW, and Arta I did test both, and they matched well with AP.

 

[VintageFlanker]

Is there a way to increase the line thickness in these graphs ?

For RMAA, nope. It already has poor enough resolution, IMHO.

They are hardly readable on a smartphone.

Strange. I have no issue at all reading these with my Pixel 5 (Google Chrome) from where I do most of my posts here, BTW.

In any case, you may just click on pictures and zoom-in as you please:

 

[Ajax]

Very nice review Vintage Flanker, many thanks for posting. I also found the follow up discussion in this thread interesting and informative.

Thank you.

Ajax