1 (edited by weme 2021-02-26 00:43:01)

Topic: ADI-2 DAC (AK4493) with sound problems - THD+N + IMD+N change strongly

While probably 99% of ADI-2 DAC buyers are happy about their device, we are disappointed.

We now have a second ADI-2 DAC with the same problems. We will send the unit to RME via the dealer for repair. If the unit needs to be replaced, the replacement should be done by RME. In any case, we would like to know the cause of the fault.

For both units the values for THD+N / IMD+N in the output signal are too high. Our measuring system see further under *1).

The first ADI-2 DAC (AK4493) was only briefly listened to, briefly measured (RMAA only) and sent back. We can inform RME about the serial number.

The second ADI-2 DAC (AK4493) was measured more intensively. The old ADI-2 was used in parallel as a reference.


(A) The results with the old ADI-2 with Hardware 1.4 (ADI-2-HW1.4) - measured with RMAA:
The values of THD+N and IMD-N remain constant.
The values of THD and IMD remain constant.
The remaining measured values are either constant or deviate only minimally.


(B) The results with the ADI-2 DAC FS (ADI2DAC-AK4493) - measured with RMAA:
The values of THD+N and IMD-N change strongly.
The values of THD and IMD remain constant.
The remaining measured values are either constant or deviate only minimally.

Other 12 V power supplies or the use of a laboratory power supply (HM8142) do not bring any changes.

The analogue outputs XLR, RCA, Phones (IEM not measured) behave the same. The fault should therefore be in the area of the power supply or in the area of the DAC IC AK4493 with its first subsequent stage OPA1602 (?) or in the DSP area.


(C) Measured IMD values [%] with REW - RTA after cold start in operating minutes:

The RMAA measured values show that the measurement of the values of IMD-N is sufficient to assess the error. Therefore, only an IMD signal SMPTE (60 Hz & 7 kHz 4:1 -1.7 dBFS) with REW (RTA: Spectrum 256k Point FFT with 4 Averages) was measured. Especially because the SMPTE signal partially correlates with a DIM30 measurement (https://www.ti.com/lit/ds/symlink/opa1602.pdf - Figure 12.).

Very noticeable are strong interference frequencies in the spectrum of the REW - RTA. At high values of TD+N these become visible: Two interference frequencies with ~ -101 dBu at a distance of ~ +- 4 Hz each next to the signal frequency of 60 Hz.

This is the reason for the changing values of IMD+N ("TD+N" in the tables).

Undoubtedly, this is not audible. But it is not perfect.


(C.1) ADI-2-HW1.4
Input........SPDIF Coax
VOL.........max
Output.....11,28 dBu (XLR)
Ref...........+4dBu

MIN IMD-DIN TD+N
1    0,0063    0,0021
2    0,0062    0,0028
3    0,0061    0,0020
4    0,0060    0,0020
5    0,0060    0,0022 (between 4-5: max. 0,0036)
6    0,0060    0,0020
7    0,0060    0,0020
8    0,0060    0,0020
9    0,0060    0,0020
10    0,0060    0,0020
11    0,0060    0,0020
12    0,0060    0,0020
...
120    0,0058    0,0020


(C.2) ADI2DAC-AK4493
Input........SPDIF Coax
VOL.........0 dB
Output.....11,29 dBu (XLR)
Ref..........+7dBu
DAC-F.....Sharp
(Linear: No B/T - EQ - Loudness)

MIN IMD-DIN TD+N     
1    0,0033    0,1500
2    0,0033    0,0015
3    0,0033    0,0087
4    0,0033    0,0190
5    0,0033    0,0170
6    0,0033    0,0110
7    0,0033    0,0530
8    0,0033    0,0016
9    0,0033    0,0020
10    0,0033    0,0041
11    0,0033    0,0065
12    0,0033    0,0088
13    0,0033    0,0120
14    0,0033    0,0150
15    0,0033    0,0180
16    0,0033    0,0220
17    0,0033    0,0200
18    0,0033    0,0180
19    0,0033    0,0160
20    0,0033    0,0140
21    0,0033    0,0130
22    0,0033    0,0110
23    0,0033    0,0098
24    0,0033    0,0082
25    0,0033    0,0068
26    0,0033    0,0058
27    0,0033    0,0049
28    0,0033    0,0039
29    0,0033    0,0029
30    0,0033    0,0025
31    0,0033    0,0017
32    0,0033    0,0015
33    0,0033    0,0015
34    0,0033    0,0015
35    0,0033    0,0015
36    0,0033    0,0015
37    0,0033    0,0017
38    0,0033    0,0020
39    0,0033    0,0022
40    0,0033    0,0024
41    0,0033    0,0027
42    0,0033    0,0029
43    0,0033    0,0031
44    0,0033    0,0033
45    0,0033    0,0035
46    0,0033    0,0036
47    0,0033    0,0038
48    0,0033    0,0039
49    0,0034    0,0040
50    0,0034    0,0042
51    0,0034    0,0043
52    0,0034    0,0045
53    0,0034    0,0046
54    0,0034    0,0046
55    0,0034    0,0047
56    0,0034    0,0048
57    0,0034    0,0048
58    0,0034    0,0048
59    0,0034    0,0050
60    0,0034    0,0052
65    0,0034    0,0054
70    0,0034    0,0055
75    0,0034    0,0056
80    0,0034    0,0056
85    0,0034    0,0058
90    0,0034    0,0059
100    0,0034    0,0062
110    0,0034    0,0060
120    0,0034    0,0060

After 120 minutes of operation, the units remained on the power supply and the SMPTE signal was switched off.
After 140 minutes, the SMPTE signal was switched on again:


(C.3) ADI-2-HW1.4
...

MIN IMD-DIN TD+N
140    0,0057    0,0019
160    0,0057    0,0019
170    0,0057    0,0019


(C.4) ADI2DAC-AK4493
...

MIN IMD-DIN TD+N
140    0,0034    0,0220 (RMAA: IMD-N: 0,012)
160    0,0034    0,0170 (RMAA: IMD-N: 0,0091)
170    0,0034    0,0110 (RMAA: IMD-N: 0,007)


*1)
Our measuring system:

Case Travla C138 with build-in 120W DC power board and external DC power adapter
Intel Desktop Board D2700DC (Mini-ITX)
Dual-core Intel Atom processor with integrated graphics and memory controller
Chipset Intel NM10
RME HDSP 9632 (Driver 3.38)
FAN: 120 mm with constant ~ 650 rpm

The measuring system is calibrated, very stable and thus provides reproducible results!

Of course, HDSP 9632 and the PC-HW are no longer good enough to evaluate an ADI-2 / DAC. But if units are measured under the same conditions and not at the end of the range, these results can be compared well and thus interpreted.


Our software - used with ASIO drivers:

OS XP

REW 5.20 Beta 60 (RTA, ...) - (https://www.roomeqwizard.com/index.html)

ARTA 1.9.3 (SPA) - (https://www.artalabs.hr/)
calculates THD+N but not IMD+N
Values of IMD ~= values of IMD DIN (REW - RTA)

RMAA 6.3.0 (6.4.5 has too many useless decimal places) - (https://audio.rightmark.org/index_new.shtml)
Values of THD-N / IMD-N are too low
Values of THD-N are missing in the overview
Resolution of the graph is too coarse in the lower frequency range. The interference frequencies described above are not visible.

Re: ADI-2 DAC (AK4493) with sound problems - THD+N + IMD+N change strongly

Hi

what is RMAA?

I am absolutely not clear how you measured?

ADI2-Pro to be measured needs an analyzer which is better, a newest Audio Precision analyzer or better.

Peter

3 (edited by weme 2021-02-24 03:02:44)

Re: ADI-2 DAC (AK4493) with sound problems - THD+N + IMD+N change strongly

Thanks for the hint. Bold print and underlining added for quick readers.

Linking added

4 (edited by KaiS 2021-02-24 10:25:36)

Re: ADI-2 DAC (AK4493) with sound problems - THD+N + IMD+N change strongly

I prefer measuring THD with simple 1 kHz sinewave and passive analog prefiltering.
This way the THD of the measuring system doesn’t cover the DUT’s.

IMD measurement I do with 1 kHz spaced high frequency pairs like 17 kHz / 18 kHz.

The spectrum below 1 kHz is simply too spoiled with hard to avoid mains voltage artifacts for constant results with the extremely low distortions that are to expect.

Re: ADI-2 DAC (AK4493) with sound problems - THD+N + IMD+N change strongly

If you like, look for my corrections:

Underlining at points (A) and (B)

Text addition to point (C):

"This is the reason for the changing values of IMD+N ("TD+N" in the tables).
Undoubtedly, this is not audible. But it is not perfect."

Re: ADI-2 DAC (AK4493) with sound problems - THD+N + IMD+N change strongly

I think your measurement system is simply not capable enough.

ADI-2 FS is ~10 times better than the figures you get.

Plus- it‘s well possible that ADI-2 DAC is set wrong, or you have a ground loop or whatsoever might go wrong.

If 2 different specimen of a known good product gives such results I would question the measurement configuration first.

Don‘t get me wrong - I‘m a measurement guy and really like to help.

7 (edited by weme 2021-02-25 15:16:40)

Re: ADI-2 DAC (AK4493) with sound problems - THD+N + IMD+N change strongly

The measuring system would indeed not be suitable for measuring the better DAC out of two faultless ADI-2 DACs. I also wrote: "Of course, HDSP 9632 and the PC-HW ...".

But the measuring system shows me
that my old ADI-2 gives extremely stable readings,
and that the ADI-2 DAC only (!) gives unstable readings when the "noise component" is measured. This was confirmed by different software:
The measured values of the software REF: "TD+N" in the table above,
the measured values of the software RMAA: "THD + Noise, dB (A)" as well as "IMD + Noise, %" (not listed).

However, there is one ambiguity with REW:
(IMD+N) cannot be less than (IMD). With the old ADI-2, these values are smaller. I have not yet found an exact description of "IMD-DIN" "TD+N".


The used soundcard RME HDSP 9632 (I like it very much) has a transformer-balanced, galvanically isolated, according to AES3-1992 SPDIF output. So a ground loop is not possible. Of course I also compared the results with the Toslink output.

According to the manual, the AD - Line input is supposed to have 24 bits:
SNR: 109 dB RMS unweighted, 111 dBA @44.1 kHz
THD @-3 dBFS: -101 dB, < 0.001 %
THD+N @-3 dBFS: -99 dB, < 0.0015 %
- SR from 44.1 - 192 kHz

We can therefore accept measured values from ~ > 0.001 %. The max value of the ADI-2 DAC was 0.0220 %.

HDSP 9632 works correctly. Measured values of the HDSP 9632 - analog Out -> analog In @+4dBu (measured with ARTA):
THD: 0,00096% THD-N: 0,0014 @+12,4 dBu
THD: 0,00078% THD-N: 0,0013 @+10,0 dBu
THD: 0,00064% THD-N: 0,0014 @+7,0 dBu

The measuring system delivers an SPDIF signal with 0 dBFS and processes the analogue input signal via 2 x 5 m microphone cable. I have given the analogue input values in the table. For measuring, only 3 cables are exchanged and the measured values are read from the monitor. So far I cannot detect any systematic error.
I still have a measuring system from the last millennium in the cupboard. I should try that out ...


The frequency 1 kHz is known to be in a range with the lowest distortion values.

A measurement with REW
    RTA: Spectrum 512k Point FFT with 16 Averages,
    SPDIF output 0 dBFS,
    analogue input -0.59 dBu
shows two generated signals with ~ -94 dBu: ~ 995.6 Hz and ~ 1004 Hz. Reducing the SPDIF signal brings no change.

These signals are missing in the spectrum of the old ADI-2!

8 (edited by weme 2021-02-26 13:28:17)

Re: ADI-2 DAC (AK4493) with sound problems - THD+N + IMD+N change strongly

KaiS wrote:

If 2 different specimen of a known good product gives such results I would question the measurement configuration first.

Plus- it‘s well possible that ADI-2 DAC is set wrong, or you have a ground loop or whatsoever might go wrong.


It was only a functional test:
These are IMD values and they are significantly worse than THD values. Also measurements were taken with 0 dBFS input and max output: VOLUME max (old ADI-2) / VOLUME 0 dB (ADI-2 DAC).



B/T: no
EQ: no
LD: no

DAC-Filter: Sharp
Ref..........+7dBu

Input........SPDIF Coax
Output.....11,29 dBu (REW)
XLR via 2 x 5 m microphone cable

VOL.........0 dB

9

Re: ADI-2 DAC (AK4493) with sound problems - THD+N + IMD+N change strongly

weme wrote:

shows two generated signals with ~ -94 dBu: ~ 995.6 Hz and ~ 1004 Hz. Reducing the SPDIF signal brings no change. These signals are missing in the spectrum of the old ADI-2!

These signals are low frequency jitter and have nothing to do with your other measurements. They don't change values for IMD or THD, and are (of course) not audible. You can easily remove them by changing clock source to INT at the ADI. They are caused by a change in firmware 34/34, and are already fixed (release in the next firmware update). If you need SPDIF sync for your measurements then you could flash the former firmware 27/34:

http://www.rme-audio.de/downloads/fut_m … _27_34.zip

Regards
Matthias Carstens
RME

10 (edited by KaiS 2021-02-25 09:33:58)

Re: ADI-2 DAC (AK4493) with sound problems - THD+N + IMD+N change strongly

weme wrote:

... the ADI-2 DAC only (!) gives unstable readings when the "noise component" is measured.

However, there is one ambiguity with REW:
(IMD+N) cannot be less than (IMD). With the old ADI-2, these values are smaller. I have not yet found an exact description of "IMD-DIN" "TD+N".

You did not measure the residual noise without signal.

THD+N is NOT the most significant value to measure because of the mix of distortion and noise.
I prefer to measure both individually.


Your ADC inside the computer is prone to catch all kinds of noises and interferences, especially ground related ones.

Even balanced connection has its limits in suppressing those.

weme wrote:

The used soundcard RME HDSP 9632 (I like it very much) has a transformer-balanced, galvanically isolated, according to AES3-1992 SPDIF output. So a ground loop is not possible. Of course I also compared the results with the Toslink output.

... analogue input signal via 2 x 5 m microphone cable.

These two cables is where a ground loop happens for sure, with 10 m size. Measure one channel at a time and unplug the other one.
Stay with Toslink, just to be sure.

weme wrote:

The frequency 1 kHz is known to be in a range with the lowest distortion values.

And it’s the common standard frequency.
For comparison with the factory specs it’s the one to measure.

Re: ADI-2 DAC (AK4493) with sound problems - THD+N + IMD+N change strongly

It is a great pleasure for me when the engineer and co-founder of RME intervenes here.

MC wrote:

These signals are low frequency jitter ... You can easily remove them by changing clock source to INT at the ADI.

They are caused by a change in firmware 34/34, and are already fixed (release in the next firmware update). If you need SPDIF sync for your measurements then you could flash the former firmware 27/34:

http://www.rme-audio.de/downloads/fut_m … _27_34.zip

But, with the ADI-2 DAC I have no possibilities to change anything - even according to the manual (EN - Version 34/34). Or have I overlooked something?

Should I still change to flash the former firmware 27/34 to understand parts of my "problems"?

12

Re: ADI-2 DAC (AK4493) with sound problems - THD+N + IMD+N change strongly

INT refers to USB with the ADI-2 DAC. So why not flash it? You can revert to the current or a newer version any time.

Regards
Matthias Carstens
RME

13 (edited by weme 2021-02-26 13:50:33)

Re: ADI-2 DAC (AK4493) with sound problems - THD+N + IMD+N change strongly

We had these problems:

The values of THD+N and IMD-N change strongly.
THD+N / IMD+N in the output signal are too high.
ADI-2 DAC only (!) gives unstable readings when the "noise component" is measured.

Very noticeable are strong interference frequencies in the spectrum of the REW - RTA. At high values of TD+N these become visible: Two interference frequencies with ~ -101 dBu at a distance of ~ +- 4 Hz each next to the signal frequency of 60 Hz.

The frequency 1 kHz shows two generated signals with ~ -94 dBu: ~ 995.6 Hz and ~ 1004 Hz. Reducing the SPDIF signal brings no change. Another measurement later also showed ~ -78 dBu! These signals are missing in the spectrum of the old ADI-2!


Downgrading from firmware 34/34 to 27/34 (see above) solved these problems. Great!

How nice also that I was completely wrong with my error analysis (too little knowledge about digital "DSP technology").


It would be great if people would investigate if there are differences in sound:

E.g.: "ADI-2 DAC AK4493 Sound FW34/34 vs FW27/34"






Here now some more data for HDSP 9632 users:

RMAA - 24-bit - 44.1 kHz

ADI-2 DAC AK4493  ---  old ADI-2 --- (ADI-2 DAC AK4493 with FW 34/34 - worst case)

Noise level, dB (A)...................114.8   ---   114.0      ---  (either constant or deviate only minimally) 
Dynamic range, dB (A)............114.7   ---   113.9      ---  (either constant or deviate only minimally)
THD, %.......................................0.0002  ---   0.0008   ---  (either constant or deviate only minimally) 
THD + Noise, dB (A)...............-100.9  ---   -95.8        ---  (-81.2)
IMD + Noise, %........................0.0011  ---   0.0016    ---  (0.013)
                   
Stereo crosstalk, dB...............114.0   ---   112.8       ---   (either constant or deviate only minimally) 
IMD at 10 kHz, %....................0.0009  ---   0.0016    ---   (either constant or deviate only minimally)

Re: ADI-2 DAC (AK4493) with sound problems - THD+N + IMD+N change strongly

Interesting, does the ADI-2 FS R have the same issue with firmware 238/98?

15

Re: ADI-2 DAC (AK4493) with sound problems - THD+N + IMD+N change strongly

Yes.

Regards
Matthias Carstens
RME

16

Re: ADI-2 DAC (AK4493) with sound problems - THD+N + IMD+N change strongly

We just released firmware 249/98 for the Pro and 35/34 for the DAC. I will not do an extra thread for these right now, so just as a small note: if one measures jitter attenuation over frequency the attenuation will be quite high at higher frequencies, and quite low at lower ones. The simple reason is that one needs to follow the incoming clock, and if that one varies slowly or a bit quicker (intentional detuning, varipitch) you don't want to loose sync. But that would usually happen when the jitter filter does not let through such slow changes.

In version 249 and 35 we finished what we had begun in version 238 and 34. In those versions the jitter attenuation was 4 x times higher than before, but a bit unstable in the low frequency area (which caused these low level nearby sidebands reported above). This is now resolved. The new SteadyClock works flawlessly with no difference to before, but jitter attenuation now has a corner frequency no longer around 200 Hz, but at around 1 Hz. That means more than 50 dB jitter reduction also at 50 Hz and similar real-world jitter frequencies. Absolutely top-notch for any kind of incoming clock signal.

Later on I will show the difference in behaviour with measurements on my APx555B.

Regards
Matthias Carstens
RME

Re: ADI-2 DAC (AK4493) with sound problems - THD+N + IMD+N change strongly

This is why I buy and support RME, always pursuing technical perfection. Keep it up!

18

Re: ADI-2 DAC (AK4493) with sound problems - THD+N + IMD+N change strongly

It's a real pity that the planned firmware updates for Pro and DAC with many improvements and new features are delayed by Corona, global chip crisis, component allocation and and and.... This all blocks an unbelievable amount of time and resources...

Regards
Matthias Carstens
RME

19 (edited by weme 2021-03-19 17:17:19)

Re: ADI-2 DAC (AK4493) with sound problems - THD+N + IMD+N change strongly

We have now also joined the club "99% of ADI-2 DAC buyers" (see #1).

With firmware 35/34 everything is great (sound). Also a wonderful room depth - even at low volume - with our very special setup (... Neumann KH310A - more about later in this forum).
Try firmware 35/34!
Thanks to RME!

The direct download links of firmware 35/34 (2021-03-19):
https://rme-audio.de/downloads/fut_madiface_mac.zip (MAC)
https://rme-audio.de/downloads/fut_madiface_win.zip (WIN)

Graphs of the above 1 kHz measurements with the RME HDSP9632 in comparison:
ADI-2 DAC FS AK4493 (firmware 27/nn vs 34/nn vs 35/nn) vs ADI-2 (old)
https://www.audiosciencereview.com/foru … ved.21136/

Re: ADI-2 DAC (AK4493) with sound problems - THD+N + IMD+N change strongly

hello, i just updated with the latest firmware. my Rme still indicates version 34 with the date 12/2019. is this normal?

merci !

Re: ADI-2 DAC (AK4493) with sound problems - THD+N + IMD+N change strongly

the-carbu wrote:

hello, i just updated with the latest firmware. my Rme still indicates version 34 with the date 12/2019. is this normal?

merci !

I think the 34 refers to the dsp and 35 to the fpga. So all good.

Re: ADI-2 DAC (AK4493) with sound problems - THD+N + IMD+N change strongly

MC wrote:

It's a real pity that the planned firmware updates for Pro and DAC with many improvements and new features are delayed by Corona, global chip crisis, component allocation and and and.... This all blocks an unbelievable amount of time and resources...

Bummer, so this is why it's just Steadyclock being improved, which is still incredible on it's own of course. Can you give us a sneak peek of what is in the making?

As for the delays, it can't be helped with the weird times we live in right now. Looking forward to the ADI-2 becoming even better than it already is.