SMSL SU-2: a dumb, wasted opportunity

[john61ct]

So are we saying this fade in lag time is at the start of every track?

or just per listening session ?

 

[dpturner]

Which begs the question: are there any songs with an sharp opening transient where this fade-in is detectable to a human listener?

Dunno, but how about Light My Fire by the Doors?

 

[PGAMiami]

From what I read, a well done master will have a very short preamble with dithered white noise just prior to the music starting. Not all are well done. I think this is very hard to notice if at all. And I go back to my prior question. Most modern chips have this built it. So how many of the finished products have it enabled so they don’t get clicks between tracks? I’m guessing almost all but that’s a guess. And a 1/60 second fade in does not sound like a fade in.

 

[PGAMiami]

So are we saying this fade in lag time is at the start of every track?

or just per listening session ?

From what I read the answer is it varies depending on how the master was done and how the firmware is setting the parameters for the DAC. But note just about all chips will do this if set up by firmware to do it.

 

[L-Train]

So are we saying this fade in lag time is at the start of every track?

or just per listening session ?

It seems just at the start of a stream of non-zero audio data. So, just per listening session as long as the player is gapless and there is no zero data between tracks (or perhaps some amount up to some threshold after which it fades/ramps subsequent data in).

I'll perform a few more tests in a few days to see how much zero data is needed for the fade in to occur. I'll also test the Khadas Tone Board V1 via its SPDIF input to see how it behaves in this situation. And I have a Fosi Audio DS3 (which also uses an ES9039Q2M) so we can also see how USB playback fares.

In general I think for regular music listening this fade/ramp is not much of an issue. However, for testing, measurement, or audio production I want the DAC to play the samples exactly as intended. If I generate a specific test signal with SoX, I want the DAC to play that back perfectly from the first to the last sample. If I'm working in a DAW and repeatedly play back at a specific sample point, I don't want there to be a slight fade/ramp that doesn't actually exist in the track. For my use cases I feel the SU-2 can be an unreliable tool if you're not aware of its quirks.

 

[PGAMiami]

Do you have other DACs that don’t do this

 

[L-Train]

Do you have other DACs that don’t do this

I took some time to measure all of the DACs I have available. All were measured with a 1 kHz tone at 48 kHz since the issue is most apparent at that sample rate with the SU-2:

Lynx E44 S/PDIF to Khadas Tone Board (aka Tone 1) (ES9038Q2M)

Douk Audio U2 Pro S/PDIF to Lynx E44 (CS4398)
- The Lynx E44 was configured to use the incoming S/PDIF signal as the clock source, and route the digital audio directly to its analog output DAC via hardware/FPGA routing

Fosi Audio DS3 USB playback (ES9039Q2M)
- Note that the DS3 inverts its unbalanced output

None of these DACs fade in or ramp from digital zero.

 

[PGAMiami]

Thank you for running these tests. I wonder how long the digital silence needs to be for this to be activated.

 

[NTTY]

From my tests, I saw that the DAC shuts down at a precise -122dBFS peak and will continue to play only if I add dither at bit 21. I’ll be documenting more in the review. This is certainly unique to this DAC.
I suppose it is a feature of the ESS DAC that SMSL is using to prevent other issues (?).

 

[Frunse]

I am not sure but this slide in is probably not an big Deal while listening to Music but i need to check how it will make the aktive Crossover designing more complicated.

But i used 2 off them with CamillaDSP and made an aktive 2 Way Crossover with 96dB @ 96kHz without notice this at all ???!!!
Also while Timealigning the Tweeter to the MidBas Driver LOL

Measured all with REW.
I guess its not an big Deal because when REW aplied Measuring Signal it took may be a ms longer till it comes from DAC Out but thats happen on Both Channel's the same so no problem at all.
And after alligning the Speaker you measure the whole Speaker at once and if the Stepresponse looks how it should and the Speaker Phase and Frequencyresponse looks also good, nothing to bother about.

Next Time i will look more into it while seting up an FIR Crossover.

Also my Signal runs thrue an HDMI to AES Extractor to an RME HDSPe Card the Signal already has some Latency there and after apllying the FIR Filter more Latency added!

Need to make some Measurings of that System Latency at all, also to know how much Libsync i need to Setup for Movies than.

I watched a Movie and didn't notice a lot off Asyncronity so need to Measure it to be precise.

But all still Project on the way, don't take it to serius what i noticed, experienced so Far.

 

[Roland68]

From my tests, I saw that the DAC shuts down at a precise -122dBFS peak and will continue to play only if I add dither at bit 21. I’ll be documenting more in the review. This is certainly unique to this DAC.
I suppose it is a feature of the ESS DAC that SMSL is using to prevent other issues (?).

You measured that at 24-bit?

 

[PGAMiami]

From my tests, I saw that the DAC shuts down at a precise -122dBFS peak and will continue to play only if I add dither at bit 21. I’ll be documenting more in the review. This is certainly unique to this DAC.
I suppose it is a feature of the ESS DAC that SMSL is using to prevent other issues (?).

Thank you. This is very helpful. I was planning to use these with my surround and height channels via the AES outputs of an incoming Trinnov Altitude CI. My concern now is that the fade in will affect the Trinnov calibration measurements. I understand Trinnov fires three bursts and may ignore the first. But I don’t know if the pause between bursts will engage the fade in nor am I sure the first burst is ignored.

 

[Frunse]

May be i didn't noticed it at all because REW is playing Dither before the Signal Sweep is aplied not sure how long it is tho, need to test it soon!



Also the whole sweep (256k) is 2,7s long so 10ms for 96kHz isn't so much also the sweep starts with 2 Hz on low End so till 20 Hz it took longer than 10 ms probably with that Dither included

Even if i start the Measuring, let say at 200Hz for an 2kHz Crossover point, it took may be up to 250Hz or less and the Sweep has full Level already, that why i didn't noticed it.

Also don't measured my Subs till now so may be there i will notice something who knows.

My 2 Ways Starts from 60Hz.

 

[NTTY]

Considering the price of this DAC, we can assume that these are features of the ESS DAC, when used in "hardware mode", since there's basically no software, or say external control of the SU-2.

So, I had a quick look at the ESS datasheet, which indeed tells me that this is the AUTOMUTE function (described page 79):

And indeed when I tested, I saw that DAC was unmuted when sending 3dB more, the datasheet says 2.5dB, so it concurs. I guess the SU-2 is configured on the default setting, and the 3dB difference I witnessed (-122dBFS vs -119dB mentioned in the datasheet) might be RMS vs peak level, I can check that too. All that to say this is working as desgined per the ES9039Q2M datasheet.

As for the fade in, unless mistaken, it is also a configurable hardware mode described page 14 to 16, to prevent glitches when switching Sampling Rates (my guess):

I'll probably talk about it in the review, and again it concurs with what you measured @L-Train

 

[NTTY]

You measured that at 24-bit?

yep 24 bits

 

[PGAMiami]

yep 24 bits

Would this should up in the impulse response? Didn’t seem to be a problem with the PS200 Pro that you recently tested. Does ESS specify how long the low level signal needs to be to activate the mute?

 

[Frunse]

Only to be sure:

If i understand that right this Fade In aplied every Time an Signal went to Digital Silence (-119dB or lower) for an amound off Time and the ESS Chip goes in his Mute state so next Signal get already Fade in again right ?

 

[NTTY]

Would this should up in the impulse response? Didn’t seem to be a problem with the PS200 Pro that you recently tested. Does ESS specify how long the low level signal needs to be to activate the mute?

Yes:

Where Automute_time is defined per the below:

I understand it translates to the below durations, with reverse-engineering based on @L-Train measurements at the previous page:

Description	Hex value	Decimal	Duration at 44.1kHz FS	Duration at 48kHz FS	Duration at 96kHz	Duration at 192kHz
Disabled	11'h000	0	infinite	infinite	infinite	infinite
Slowest	11'h001	1	-5.94 seconds	-5.46 seconds	-2.73 seconds	-1.37 seconds
Default	11'h00F	15	-396.3 ms	-364.1 ms	-182.04 ms	-364.1 ms
SMSL Custom	11'h111	273	-21.77 ms	-20 ms	-10 ms	-5 ms
Fastest	11'h7FF	2047	-2.9 ms	-2.67 ms	-1.33 ms	-0.67 ms

 

[PGAMiami]

Yes:

View attachment 552530

Apologies but I don’t know how to interpret this to get to an amount of time for 48 or 96 kHz samples. That said, I’m guessing it does this very quickly, just a few samples? If so, then I don’t think I can use this DAC with a Trinnov that is trying to measure speaker distance, phase, amplitude etc. It would be very helpful if you could point out a few alternatives. Looking for something with balanced out and AES in. Could live with spdif coax in, but prefer all XLR. Many of the reviews I see don’t have impulse response.

 

[NTTY]

Would this should up in the impulse response? Didn’t seem to be a problem with the PS200 Pro that you recently tested. Does ESS specify how long the low level signal needs to be to activate the mute?

I did not see that with the PS200Pro indeed.
I updated my previous message to reply to your second question.