• 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). Come here to have fun, be ready to be teased and not take online life too seriously. We now measure and review equipment for free! Click here for details.

Measurements of Classé Sigma SSP Mk2

LTig

Senior Member
Joined
Feb 27, 2019
Messages
474
Likes
543
Location
Europe
#1
The Classé Sigma SSP is an AV-preamp targeted for audiophile music lovers who don't want to setup 2 different systems (one for audio, one for home cinema). It is optimized for very high quality on the stereo channels and somewhat less performance for the remaining surround channels: the stereo channel has both one balanced input and one balanced output (XLR) while the remaining analog IOs (2 x stereo in, 7.1 out) are single-ended (RCA). The specs are as follows (selected):

Frequency response: 8 Hz - 200 kHz < 1 dB (stereo digital bypass), 8 Hz - 20 kHz < 0.5 dB (all other sources)​
Distortion (THD+noise): 0.0005% (digital source / bypassed analog source), 0.002% (processed analog source)​
Maximum input level (single-ended): 2 Vrms (DSP), 4.5Vrms (bypass)​
Maximum input level (balanced): 4 Vrms (DSP), 9 Vrms (bypass)​
Maximum output level: 9 Vrms (single-ended), 18 Vrms (balanced)​
Input impedance: 100 kΩ (single-ended), 50 kΩ (balanced)​
Output impedance: 100 Ω (single-ended), 300 Ω (balanced)​
Signal-to-noise ratio (ref. 4Vrms input, unweighted): 104 dB (bypassed analog source), 101 dB (processed analog source), 105 dB (digital source)​

The unit I am testing is in fact a Sigma SSP (Mk1) which had been upgraded to Mk2. This involves the replacement of the board with the HDMI I/Os and the DSP doing the audio processing. Otherwise both models are identical. It runs with firmware version 2.0.0.29 (currently the newest).

I used a brand new RME ADI-2 PRO fs to feed the SSP with digital signals (SPDIF Coax and Toslink, USB and HDMI) and connected the XLR output of the SSP to the XLR input of the RME (one channel only). All internal processing in the SSP was disabled (no EQ, no tone control, nothing) and the inputs were set to pass through (no volume control). In this configuration the SSP delivered almost 14 dBu (ca. 3.8 V rms, very close to the SN input spec) output level with a 0 dBfs (-3dB rms) sinus. The analog input of the RME was set accordingly to +19 dBu with 5 dB higher gain, resulting in a full scale sensitivity of 14 dBu.

The RME was connected via USB to a Dell Latitude E7250 (Windows 10, RME ASIO driver) running on battery. I used RoomEq wizard v5.19 (REW) for all measurements, but for the jitter test and multitone I had to use foobar2000 to play wav files bit correct (verified by playing the specific test signals supplied by RME). The J-Test and multitone wavfiles were created with a DIY program running on Linux.

Just for the records: here is the loopback performance of the RME alone (XLR out to XLR in) at 14 dBu output level and 14 dBu input sensitivity. At 44.1 kHz samplerate and 24 bit, THD with 1 kHz sinus -3 dB rms (=0 dBfs), 32 FFT size, no averaging:
THD 1kHz -3dB RMS +19dBu-5dB 32k.jpg

REW's distortion panel says this:
32768-point spectrum using Hann window and no averaging​
Input RMS -3,1 dBFS, -3,1 dBC, -3,1 dBA​
Distortion at 1.000,0 Hz, -3,1 dBFS based on 8 harmonics:​
THD: 0,00016 %​
N: 0,00041 %​
THD+N: 0,00044 %​
2nd harmonic 0,00011% 92 deg​
3rd harmonic 0,00011% 8 deg​
4th harmonic 0,000035% 91 deg​
5th harmonic 0,000007% -123 deg​
6th harmonic 0,000015% 94 deg​
7th harmonic 0,000012% -166 deg​
8th harmonic 0,000011% 63 deg​
9th harmonic 0,000005% 154 deg​
We cannot expect better performance than this, so we must at least reduce the results of the SSP by the corresponding results of the RME.

Sigma SSP via SPDIF Toslink

Let's start with THD, 44.1 kHz samplerate, 24 bit, 1 kHz sinus @ -3 dB rms (=0 dBfs), 32k FFT size, no averaging:
THD 1kHz -3dB RMS 32k.jpg

REW's distortion panel says this:
32768-point spectrum using Hann window and no averaging​
Input RMS -3,3 dBFS, -3,3 dBC, -3,3 dBA​
Distortion at 1.000,0 Hz, -3,3 dBFS based on 8 harmonics:​
THD: 0,00073 %​
N: 0,0029 %​
THD+N: 0,0030 %​
2nd harmonic 0,00057%​
3rd harmonic 0,00017%​
4th harmonic 0,000047%​
5th harmonic 0,00021%​
6th harmonic 0,000035%​
7th harmonic 0,00032%​
8th harmonic 0,00013%​
9th harmonic 0,00012%​
This is not bad at all:
  • when we subtract the RME's THD (0.00016 %) from the SSP's THD (0.00073 %) we get a THD of 0.00058 % which is pretty close to the specified 0.0005 % and certainly within our error margin.
  • Subtracting the RME's noise (0.00041 %) from the SSP's noise (0.0029 %) results in a noise of 0.0025 %. I am not sure if the subtraction is correct (regarding the stochastic nature of noise) and how to convert % into dB (a simple conversion, e.g. with Sengpiels dB calculator gives 92 dB for the SSP and 107 dB for the RME - values which I think may be too bad; the RME should deliver ca. 10 dB more, which then would also apply to the SSP). Maybe someone can shed some light here ...
Strange though is the nature of the noise, it's somewhat "wavey". Lets see what averaging (32x) and a higher FFT size (128k) reveals:
THD 1kHz -3dB RMS 128k avg32.jpg

Beats me - but certainly not audible. Here is the same measurement with RME loopback:
THD 1kHz -3dB RMS +19dBu-5dB 128k avg32.jpg


This looks very clean, just some very tiny hash north of 10 kHz.

Next is IMD, 44.1 kHz samplerate, 24 bit, 19 + 20 kHz sinus @ -10 dB rms, 32k FFT size, no averaging:
IMD CCIF 19+20kHz -10 dB RMS 32k.jpg

REW's distortion panel says this:
32768-point spectrum using Hann window and no averaging​
Input RMS -7,6 dBFS, -18,6 dBC, -16,6 dBA​
IMD is 0,00017 % for f1 = 19.000 Hz, f2 = 20.000 Hz​
d2L: 0,00017%​
The RME has IMD of ca. 0.0001 % in loopback, so this is pretty good. But again we see the wavey nature of the noise floor.

Now J-Test, 44.1 kHz samplerate, 24 bit, 32k FFT size, no averaging:
JTest 32k.jpg

Nothing to complain about, but let's see if averaging (32x) and higher FFT siźe (128k) reveal something:
JTest 128k avg32.jpg

Indeed there are some spurous signals but definitely not audible at all.

Sigma SSP via SPDIF Coax

THD, 44.1 kHz samplerate, 24 bit, 1 kHz sinus @ -3 dB rms (=0 dBfs), 32k FFT size, no averaging:
THD 1kHz -3dB RMS 32k.jpg

REW's distortion panel says this:
32768-point spectrum using Hann window and no averaging​
Input RMS -3,3 dBFS, -3,3 dBC, -3,3 dBA​
Distortion at 1.000,0 Hz, -3,3 dBFS based on 8 harmonics:​
THD: 0,00073 %​
N: 0,0030 %​
THD+N: 0,0031 %​
2nd harmonic 0,00057%​
3rd harmonic 0,00021%​
4th harmonic 0,000070%​
5th harmonic 0,00025%​
6th harmonic 0,000038%​
7th harmonic 0,00027%​
8th harmonic 0,000045%​
9th harmonic 0,00016%​
The numbers are the same as Toslink.

IMD, 44.1 kHz samplerate, 24 bit, 19 + 20 kHz sinus @ -10 dB rms, 32k FFT size, no averaging:
IMD CCIF 19+20kHz -10 dB RMS 32k.jpg

REW's distortion panel says this:
32768-point spectrum using Hann window and no averaging​
Input RMS -7,6 dBFS, -18,6 dBC, -16,6 dBA​
IMD is 0,00019 % for f1 = 19.000 Hz, f2 = 20.000 Hz​
d2L: 0,00019%​

The numbers are (within the error margin) same as Toslink.

J-Test, 44.1 kHz samplerate, 24 bit, 128k FFT size, 32x averaging:
JTest 128k avg32.jpg

No real difference to Toslink.

Multitone, 192 kHz samplerate, 24 bit, 30 sinus signals, 128k FFT size, no averaging:
Multitone 0dB 32k.jpg

Nothing to worry about. Averaging does not reveal any more spurious signals.

Sigma SSP via USB

THD, 44.1 kHz samplerate, 24 bit, 1 kHz sinus @ -4 dB rms (=-1 dBfs), 32k FFT size, no averaging
I had to reduce the output level by 1 dB because at 0 dB there was significan distortion through clipping. This is probably a problem in the Windows 10 audio driver.
THD 1kHz -4dB RMS 32k.jpg

REW's distortion panel says this:
32768-point spectrum using Hann window and no averaging​
Input RMS -5,3 dBFS, -5,3 dBC, -5,3 dBA​
Distortion at 1.000,0 Hz, -5,3 dBFS based on 8 harmonics:​
THD: 0,00073 %​
N: 0,0037 %​
THD+N: 0,0038 %​
2nd harmonic 0,00052% -102 deg​
3rd harmonic 0,00012% 51 deg​
4th harmonic 0,00017% 73 deg​
5th harmonic 0,000045% -107 deg​
6th harmonic 0,000085% -172 deg​
7th harmonic 0,00035% 177 deg​
8th harmonic 0,00015% -146 deg​
9th harmonic 0,00024% 174 deg​

A bit more noise than SPDIF, otherwise the same.

IMD, 44.1 kHz samplerate, 24 bit, 19 + 20 kHz sinus @ -10 dB rms, 32k FFT size, no averaging:
IMD CCIF 19+20kHz -10 dB RMS 32k.jpg

REW's distortion panel says this:
32768-point spectrum using Hann window and no averaging​
Input RMS -12,4 dBFS, -23,4 dBC, -21,4 dBA​
IMD is 0,00068 % for f1 = 19.000 Hz, f2 = 20.000 Hz​
d2L: 0,00068%​

There is seemingly more hash and noise, let's see with 32x averaging and 128k FFT size:
IMD CCIF 19+20kHz -10 dB RMS 128k avg32.jpg

This is not noise. Performance is clearly not on the same level as SPDIF, but are we sure that the sum of all those spurious signals is audible?

J-Test, 44.1 kHz samplerate, 24 bit, 128k FFT size, 32x averaging:
JTest 128k avg32.jpg


This also is not noise, it's too regular. Lets spread the frequency axis:
JTest 128k avg32 spread.jpg

Jep, not noise. Probably not audible though.

Sigma SSP via HDMI

A warning up front: The following measurements should be taken not with a grain, but with a bucket of salt:
  • I had to use 48 kHz samplerate because Windows 10 resampled all audio to 48 kHz. Hence no multitone (requires 192 kHz).
  • I had to reduce the output level of all 0 dBfs signals by 1 dB because at 0 dB there was significant distortion through clipping. This is probably a problem in the Windows 10 audio/video driver.
  • Even with 1 dB less signal level the distortion is much higher compared to SPDIF and USB, and it does not get smaller with smaller signal level. Maybe a result of truncation withput dithering? Who knows, maybe Mr. Gates ...

THD, 48 kHz samplerate, 24 bit, 1 kHz sinus @ -4 dB rms (=-1 dBfs), 32k FFT size, no averaging
THD 1kHz -4dB RMS 32k.jpg

I'm not showing the numbers since they are not correct.
On a next shot I could try to burn the test signals on a CD and play them with my Bluray player (HDMI and SPDIF Coax).

IMD, 44.1 kHz samplerate, 24 bit, 19 + 20 kHz sinus @ -10 dB rms, 32k FFT size, no averaging:
IMD CCIF 19+20kHz -10 dB RMS 32k.jpg

IMD is good but there is also a lot of hash all over the full frequency range. Again no numbers.

J-Test, 44.1 kHz samplerate, 24 bit, 128k FFT size, 32x averaging:
JTest 128k avg32.jpg

Oh oh. Not sure whether the J-Test signal was played bit correct at all, so no judgement possible.

Conclusion
  • respectable performance both on SPDIF Coax and Toslink
  • almost same performance via USB. Those results which are worse compared to SPDIF could be a result of the Windows OS.
  • performance via HDMI is not clear - the bad results can be a result of the Windows OS.
    However it is very easy to circumvent audio through HDMI. The HDMI inputs can be configured such that HDMI is used for video only and any other input can be used for audio. You may loose multi channel though - I'm not sure which surround audio formats can be transported through SPDIF.
Let me know please if you find any flaws or if there is anything I can do better.

And one last point: Now that I've made my first full set of measurements I know how much work this is. I really don't know how Amir manages to publish almost one review per day. Lets all thank Amir for his daily work!
 

Attachments

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
20,048
Likes
23,034
Location
Seattle Area
#2
And one last point: Now that I've made my first full set of measurements I know how much work this is. I really don't know how Amir manages to publish almost one review per day. Lets all thank Amir for his daily work!
Well, you did a lot more work than my lazy fingers do :).

Great post. I will promote to home page.
 

DDF

Active Member
Joined
Dec 31, 2018
Messages
238
Likes
220
#3
LTig,
Thanks, this is very comprehensive!
Would you be able to share your jitter test method in more detail? I'm testing using RMAA Pro and RTX6001 which works for many things but not jitter.
 

DuxServit

Active Member
Patreon Donor
Joined
Apr 21, 2018
Messages
202
Likes
163
#4
@LTig This is a great review. Thanks.

Would love to see you annotate the graphs with notes (“what this means”).
 

restorer-john

Major Contributor
Joined
Mar 1, 2018
Messages
2,249
Likes
3,453
Location
Gold Coast, Queensland, Australia
#5
A thorough investigation of distortion for sure. Well done. :)

But, as this is a preamplifier, not just a D/A converter, I'd like to see the following tests done:

Frequency response.
Channel balance, volume tracking etc
Input levels and overload characteristics
Input impedance/s
Maximum output level
Analog THD vs Level & THD vs Freq or spot frequencies and THD numbers.
Output impedance/s

In the interests of investigating the D/A converter performance at 0dBFS without making having to make excuses for Windows10, use a CD source 1KHz/0dBFS 16/44 test track to determine whether the D/A converter has issues or not at full level.
 

LTig

Senior Member
Joined
Feb 27, 2019
Messages
474
Likes
543
Location
Europe
#6
LTig,
Thanks, this is very comprehensive!
Would you be able to share your jitter test method in more detail? I'm testing using RMAA Pro and RTX6001 which works for many things but not jitter.
I used foobar2000 to play the wav file with the J-Test signal. In REW I just opened the RTA window and clicked on the record button.
 

LTig

Senior Member
Joined
Feb 27, 2019
Messages
474
Likes
543
Location
Europe
#7
A thorough investigation of distortion for sure. Well done. :)

But, as this is a preamplifier, not just a D/A converter, I'd like to see the following tests done:

Frequency response.
Channel balance, volume tracking etc
Input levels and overload characteristics
Input impedance/s
Maximum output level
Analog THD vs Level & THD vs Freq or spot frequencies and THD numbers.
Output impedance/s
Let me see what can I think I can do with my current equipment:
  • Frequency response: yes
  • Channel balance: probably yes
  • volume tracking: no
  • Input levels: what do you want to know?
  • Input overload characteristics: I can only measure the output level and its distortion
  • Input impedance/s: no
  • Maximum output level: yes
  • Analog THD vs Level: not with REW, at least not with a simple measurement.
  • THD vs Freq: not with REW, at least not with a simple measurement.
  • spot frequencies and THD numbers: which frequencies
  • Output impedance/s: no
What I definitely want to do is:
  • THD/IMD using the analog input without digital processing (digital pass-through)
  • THD/IMD using the analog input with digital processing (fixed at 96/24)
  • frequency response for digital sources
  • frequency response for analog sources
However it may take some time until I'm able to do this. Got a problem with a broken water exhaust pipe in the floor above; water dripping on bed (wife not amused), water dripping on my LP12 (me not amused), three rooms in need of renovation :(
In the interests of investigating the D/A converter performance at 0dBFS without making having to make excuses for Windows10, use a CD source 1KHz/0dBFS 16/44 test track to determine whether the D/A converter has issues or not at full level.
This is not necessary, the DAC is OK as the measurements of both SPDIF inputs show.
 

DDF

Active Member
Joined
Dec 31, 2018
Messages
238
Likes
220
#8
I used foobar2000 to play the wav file with the J-Test signal. In REW I just opened the RTA window and clicked on the record button.
Got it, thanks. This method can't measure jitter in the time domain but provides frequency domain averaged with an FFT.

Is the j-test signal public domain? Would you be able to share a link for it (16/44.1 & 24/96)? I could Google but don't trust random sources and want to make sure that the signal is high quality and correct.
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
20,048
Likes
23,034
Location
Seattle Area
#9
Is the j-test signal public domain? Would you be able to share a link for it (16/44.1 & 24/96)? I could Google but don't trust random sources and want to make sure that the signal is high quality and correct.
Send me your email address and I will share out my copy with you. It is hard to find a copy online.
 

LTig

Senior Member
Joined
Feb 27, 2019
Messages
474
Likes
543
Location
Europe
#10
Is the j-test signal public domain? .
.
I don't think that it is patented. A thorough description is here at ASR where Amir describes how he measures stuff. It was good enough that I was able to program it in C++.

Would you be able to share a link for it (16/44.1 & 24/96)? I could Google but don't trust random sources and want to make sure that the signal is high quality and correct.
See this link at wetransfer (expires April 16th).
 

JohnPM

Active Member
Technical Expert
Joined
Apr 9, 2018
Messages
112
Likes
287
Location
UK
#12
You can include the distortion panel on the graph capture to save typing the figures out separately. If REW is the signal source you can lock the generator to the FFT length and use a rectangular window for the FFT to maximise the frequency resolution and avoid window spreading effects. Might want to update to the latest V5.20 beta to have more options for graph captures.

rta.png

As a PS: You can't subtract the test equipment distortion or noise from your measurements. Per AES17 the test equipment needs to be at least 3 times better than the equipment being measured, so about 10 dB lower THD, for example. As long as that is the case the THD results would be within 0.5 dB of the 'true' figure.
 
Last edited:

restorer-john

Major Contributor
Joined
Mar 1, 2018
Messages
2,249
Likes
3,453
Location
Gold Coast, Queensland, Australia
#13
Let me see what can I think I can do with my current equipment:
  • Frequency response: yes
  • Channel balance: probably yes
  • volume tracking: no
  • Input levels: what do you want to know?
  • Input overload characteristics: I can only measure the output level and its distortion
  • Input impedance/s: no
  • Maximum output level: yes
  • Analog THD vs Level: not with REW, at least not with a simple measurement.
  • THD vs Freq: not with REW, at least not with a simple measurement.
  • spot frequencies and THD numbers: which frequencies
  • Output impedance/s: no
What I definitely want to do is:
  • THD/IMD using the analog input without digital processing (digital pass-through)
  • THD/IMD using the analog input with digital processing (fixed at 96/24)
  • frequency response for digital sources
  • frequency response for analog sources
The frequency response and channel balance can be done with two L/R sweeps layered on one plot.
Volume channel tracking is really a bit fiddly. Amir made a motorised device IIRC to turn the volume pot. I do it a different way. I take ten spot readings of the absolute voltage between the channels (A-B) throughout the rotation and plot that in dB.
Input sensitivity. x mv for full rated output. How close to spec etc. Convert to gain in dB.
Input overload is more important with RIAA stages as they will overload long before the line stages do (generally).
Input impedance can be done with a finite source resistance frequency generator and a millivolt meter and calculating it.
Spot frequencies are fine IMO. I take several spot frequency readings of THD at various levels for my testing as I can't easily do high powered THD vs Level. I use 50,100,1K,3K,10K and 20KHz. Power outputs I use 1W, half rated power and full rated power.
Output impedance is the same way as input, except loaded vs unloaded voltage. As a bonus, you can run a loaded vs unloaded FR plot and layer them to see the change in output impedance vs frequency.
 

LTig

Senior Member
Joined
Feb 27, 2019
Messages
474
Likes
543
Location
Europe
#15
You can include the distortion panel on the graph capture to save typing the figures out separately. If REW is the signal source you can lock the generator to the FFT length and use a rectangular window for the FFT to maximise the frequency resolution and avoid window spreading effects. .
Thanks. I've barely touched the basic stuff in REW, using Amirs introduction to room measurements. Have to study the manual now to avoid asking stupid questions.

Do I have one shot free though? Is it possible to let the generator play a wav file on disk (some kind on signal import)? This makes it easier to do jitter test (with J-Test signal) and THD with multi tone, or generally play whatever I might need. Frees the generator from supplying seldom used signals.
Might want to update to the latest V5.20 beta to have more options for graph captures..
will do.
As a PS: You can't subtract the test equipment distortion or noise from your measurements. Per AES17 the test equipment needs to be at least 3 times better than the equipment being measured, so about 10 dB lower THD, for example. As long as that is the case the THD results would be within 0.5 dB of the 'true' figure.
Good to know.

One more: will averaging improve the precision of the numbers (THD, noise) given in the distortion panel, or will it make them smaller than they are?
 

trl

Addicted to Fun and Learning
King of Mods
Joined
Feb 28, 2018
Messages
634
Likes
407
Location
Iasi, RO
#16
Send me your email address and I will share out my copy with you. It is hard to find a copy online.
Isn't the audiocheck.net files good for jitter test anymore? Or the Stereophile Test-CD?
 

JohnPM

Active Member
Technical Expert
Joined
Apr 9, 2018
Messages
112
Likes
287
Location
UK
#18
Is it possible to let the generator play a wav file on disk (some kind on signal import)? This makes it easier to do jitter test (with J-Test signal) and THD with multi tone, or generally play whatever I might need
The generator includes multitone and dual tone signals, it has been substantially revised in the V5.20 beta versions. There isn't an ability to play a WAV file, I can look at that. I could add the J-test signals as well.
will averaging improve the precision of the numbers (THD, noise) given in the distortion panel, or will it make them smaller than they are?
Averaging will reduce the variance, it should not change the level.
 
Top Bottom