• 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.

Sabaj A20d Review (Balanced DAC)

ModDIY

Senior Member
Joined
Mar 26, 2020
Messages
333
Likes
160
This cube-like design starts to grow on me, now I want a DAC in a form of perfect cube.

Check this, with the amplifier A20a and A20h (headphone amplifier).
 

Attachments

  • Sabaj A20a.jpg
    Sabaj A20a.jpg
    22 KB · Views: 119

PeteL

Major Contributor
Joined
Jun 1, 2020
Messages
1,500
Likes
1,428
Coming from doing pro audio/studio work, installed audio/video for churches and light electrical, I can tell you standards for audio/video/electrical are very important, I just can't imagine installing a connector without a standard, very esoteric.
Sure, but I2S itself is a standard, there is just no firm specifications on how to transport it. In this specific case, It's a DAC, so it's the receiving end. It's the driving end, and cable manufacturers that worries about that. When there is such standards, this DAC will support it because all it does is just connecting the necessary signals, they don't have to worry about this.
 

milan616

Active Member
Forum Donor
Joined
Feb 25, 2021
Messages
135
Likes
104
Check this, with the amplifier A20a and A20h (headphone amplifier).

The frustrating thing about the A20 series is that you have to choose between RCA and XLR for the headphone amp and speaker amp. There's no pass through on either of them, so the DAC can only feed balanced to one and single ended to the other.
 

garbz

Member
Joined
Sep 17, 2021
Messages
50
Likes
75
Isn't that what a standard effectively is?
No. Two companies copying off each other is not a standard. A standard is a defined specification in a written document which a company can agree to adopt and has performance characteristics against which it can be judged.

Take a classic audio example: XLR vs RCA.
XLR is a standard. The connector is defined in IEC61076-2-103. All dimensions are given and consistent. Performance is consistent.
RCA is not a standard. It's a plug made by a company that was never properly defined. Some connectors are hollow some are not, some mate on a single point, others on a ring. Amphenol specifies a 15Nm insertion force, Kobi a 4Nm one. I have plugs in the back of my equipment which fall out when bumped. I have others where I think I may break something I have to yank them so hard.

Why does that matter? Works well when you are prototyping and breadboarding. And many evaluation modules or OEM streaming or DSP modules already give access to the I2S interface trough a Pinout. What's wrong with HDMI cable transfer?
I hear you, but really, I'm too lazy to calculate this, but a bit too much fuss is given to this, we are talking data in the kHz range, not in Mhz, it's in the grand sheme of things, low frequncies

Both of these are related to each other. Answering the second point should tell you the problem with the first. You're *not* in the kHz range, not even close. In fact you're usually in the 192kHz *32bit*2ch = 12.288MHz range. That's the rate for both SD and SCK. MCLK is typically double that at 24.768MHz, and LRCK is fs/256 or fs/128 so 96/192kHz.
All of this a 5V TTL without proper termination means it flat out would not work simply plugging in a HDMI cable, and if it did work you're likely to end up with jitter so bad amir may actually complain about it ;-).

And that's also why it isn't done in any of these DACs. I2S is first converted to LVDS (Low voltage differential signaling - 8 pins) before being connected to a HDMI cable. It's the only way you have any hope of getting the signal more than a few centimeters in the first place.

But ... remember what I said about fs/256 and then I used the word "or"? It's more complicated than that. I2S doesn't define a standard, it just defines ratios. Your MCLK may be 36.864MHz and your LRCLK fs/192, some DACs can autodetect this, some expect to be told. The audio interface bitdepth can't be auto negotiated so your DAC needs to understand in advance if it is getting 16 20 24 or 32bit, though many will work if they receive less than expected they will fail if they receive more.

Just having I2S capability doesn't guarantee two chips can talk to each other even if they receive the data correctly, and even if the HDMI cable is connected the same way at both ends (again, no standard).
This is why it's appropriate to shout "It's alive! It's alive!" when connecting two devices using this I2S over HDMI Frankenscheme.
 

Tks

Major Contributor
Joined
Apr 1, 2019
Messages
2,903
Likes
4,572
[EDIT] Looks like I was color blind about F2 and F7, so I would just point it out on the graph to avoid confusion.


If you want to see the most detailed and precisely named plots (apart from the ones from the datasheet), visit @WolfX-700 's website:

As a rule of thumb, ESS linear phase fast filter has the best imaging artifact suppression (120dB, as specified on the datasheet) so it is often the best one appears on the 90kHz BW THD+N plots, but the price is to slightly violates Nyquist.

The blue, dotted-line one is from the Khadas Tone Board which always show up as a template, uses the apodizing filter. Also, the filter numbers (F1-F7) can potentially be shuffled in different products, so ideally, products display and manual should use filter name instead of number.


The brickwall and apodizing filters, at the expense of not violating Nyquist, have poorer imaging artifact suppression, and they have poorer passsband ripples as well, clearly shown on the plots below:

Okay, so let me ask you something just quickly since I heard John, amir, and someone else claim something of the sort that a proper brickwall can never have the flat and smooth suppression of artifacts as seen in the filters you see with minimum/linear ones you call them.

Is this true? All brickwalls must by definition have this jagged performance unequivocally ? And this is a fact that is observed as a brickwall filter tries to respect Nyquist, and the cost is paid in this way?
 

PeteL

Major Contributor
Joined
Jun 1, 2020
Messages
1,500
Likes
1,428
No. Two companies copying off each other is not a standard. A standard is a defined specification in a written document which a company can agree to adopt and has performance characteristics against which it can be judged.

Take a classic audio example: XLR vs RCA.
XLR is a standard. The connector is defined in IEC61076-2-103. All dimensions are given and consistent. Performance is consistent.
RCA is not a standard. It's a plug made by a company that was never properly defined. Some connectors are hollow some are not, some mate on a single point, others on a ring. Amphenol specifies a 15Nm insertion force, Kobi a 4Nm one. I have plugs in the back of my equipment which fall out when bumped. I have others where I think I may break something I have to yank them so hard.




Both of these are related to each other. Answering the second point should tell you the problem with the first. You're *not* in the kHz range, not even close. In fact you're usually in the 192kHz *32bit*2ch = 12.288MHz range. That's the rate for both SD and SCK. MCLK is typically double that at 24.768MHz, and LRCK is fs/256 or fs/128 so 96/192kHz.
All of this a 5V TTL without proper termination means it flat out would not work simply plugging in a HDMI cable, and if it did work you're likely to end up with jitter so bad amir may actually complain about it ;-).

And that's also why it isn't done in any of these DACs. I2S is first converted to LVDS (Low voltage differential signaling - 8 pins) before being connected to a HDMI cable. It's the only way you have any hope of getting the signal more than a few centimeters in the first place.

But ... remember what I said about fs/256 and then I used the word "or"? It's more complicated than that. I2S doesn't define a standard, it just defines ratios. Your MCLK may be 36.864MHz and your LRCLK fs/192, some DACs can autodetect this, some expect to be told. The audio interface bitdepth can't be auto negotiated so your DAC needs to understand in advance if it is getting 16 20 24 or 32bit, though many will work if they receive less than expected they will fail if they receive more.

Just having I2S capability doesn't guarantee two chips can talk to each other even if they receive the data correctly, and even if the HDMI cable is connected the same way at both ends (again, no standard).
This is why it's appropriate to shout "It's alive! It's alive!" when connecting two devices using this I2S over HDMI Frankenscheme.
OK, yes I guess I needed an other coffee and was completely off with the frequency. Lot of things in what you say make sense, but not enough to say "cant work". Cables can transport much higher speed, and well, as I said, Developpers do this all the time, connecting 2 boards together with I2S. As for Jitter, sorry I didn't get your point. Why exactly it's intrinsyncly worst? You don't have to define the incoming I2S clock as the master clock as for autodetect nothing that can't be done. I agree this is Frankensteinish as of now but I don't see any of this as a fatality.
 

alekksander

Active Member
Joined
Aug 6, 2018
Messages
115
Likes
67
Location
Poland
anyone has an idea why lower frequencies in multitone tests often tend to be lifted?
 

bennetng

Major Contributor
Joined
Nov 15, 2017
Messages
1,380
Likes
1,311
Okay, so let me ask you something just quickly since I heard John, amir, and someone else claim something of the sort that a proper brickwall can never have the flat and smooth suppression of artifacts as seen in the filters you see with minimum/linear ones you call them.

Is this true? All brickwalls must by definition have this jagged performance unequivocally ? And this is a fact that is observed as a brickwall filter tries to respect Nyquist, and the cost is paid in this way?
It all depends on the filter length of a specific chip. The article below is specifically written to answer these kinds of questions.
I'd recommend you download SoX and my test files and scripts and see for yourself. The filter lengths of my "virtual" DAC chip are based on my observation of a number of real chips, and therefore not particularly short or long.

You can also scroll down and see I used a real music file (instead of white noise or impulse) to test the filters, and see the distribution of artifacts in the two different filters.

Some older DAC chips like WM8741 actually have longer filter length to support 0.5fs stopband with low ripple and high artifact suppression, but it also means higher latency when the filter is linear phase, and higher intersample peaks regardless of phase. The 9038 chips while "programmable", has a shorter filter length limit, the AK449x are slightly shorter than ESS, and therefore unable to provide a "brickwall" filter with low enough passband ripple and stopband attenuation. Of course, whether the modulator and subsequent analog stage are capable of producing low distortion or not is another thing.

To conclude, if the DAC itself is able to produce -120dB THD+N within all frequency up to Nyquist, then a filter of 120dB attenuation is required to eliminate the visible wiggling in the wideband THD+N plots.
 

beefkabob

Major Contributor
Forum Donor
Joined
Apr 18, 2019
Messages
1,096
Likes
1,346
If the port works, the port works. If there is a formal standard or just a casual agreement, what does it matter? I haven't seen posts saying, "I2S doesn't work!" I have been seeing posts complaining about some violated aesthetic engineers hold dear.

I've deal with plenty of well-documented standards that have issues. USB and Thunderbolt are messes. Wifi? Oof. SIM cards? Double oof!
 
Last edited:

Nathan Raymond

Active Member
Joined
Feb 5, 2018
Messages
146
Likes
132
If the port works, the port works. If there is a formal standard or just a casual agreement, what does it matter? I haven't seen posts saying, "I2S doesn't work!" I have been seeing posts complaining about some violated aesthetic engineers hold dear.

I've deal with plenty of well-documented standards that have issues. USB and Thunderbird are messes. Wifi? Oof. SIM cars? Double oof!
SIM cars and Thunderbird, hmm, let me guess - you been spending a lot of time in your garage, or perhaps behind a Fanatec wheel! :D
 

GimeDsp

Senior Member
Joined
Oct 30, 2020
Messages
409
Likes
341
Location
Earth
SIM cars and Thunderbird, hmm, let me guess - you been spending a lot of time in your garage, or perhaps behind a Fanatec wheel! :D
If you are going to pirate a connector then the locking display port would have been a much better choice.

Standards are created for a reason, safety, compatibility, backward compatibility, and most of all standards lead to advancement. Just look at ethernet cable, HDMI, usb, etc. Having known standards and performance allows users, designers, manufacturers to not only ensure those things but to also make steps to improvement. Without a standard none of this is possible.

"Did you make it better?"

"maybe. lets just say it's better"

"What if it's worse"

"They'll never know"

"But what about us?"

"We'll never know either"

"OK, as long as nobody knows..."
 
OP
amirm

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
35,336
Likes
135,186
Location
Seattle Area
I agree this is Frankensteinish as of now but I don't see any of this as a fatality.
It is a disaster. They couldn't even get the simple signaling standardized for the connector in such a simple application. Here is *part* of the documentation in A20d on I2S signalling:

1633724787561.png


Get it wrong and you get all kinds of weird results.

On top of this, there is no benefit to this interface. It is just adding cost and confusion.
 

GWolfman

Senior Member
Forum Donor
Joined
Nov 3, 2020
Messages
383
Likes
444
Audio data is sent in the invisible pixels in video over HDMI. Again, it has no relationship with this I2S signaling supported in this DAC.
Thanks. Wonder if it's DoP (DSD over PCM), though DoP appears to be USB-transport related.
 

PeteL

Major Contributor
Joined
Jun 1, 2020
Messages
1,500
Likes
1,428
It is a disaster. They couldn't even get the simple signaling standardized for the connector in such a simple application. Here is *part* of the documentation in A20d on I2S signalling:

View attachment 157960

Get it wrong and you get all kinds of weird results.

On top of this, there is no benefit to this interface. It is just adding cost and confusion.
Ha OK, I didn't know there was 4 versions already, well at least they offer them all? As I said, in my view it is not very useful now, if at all and I know full well there is no benefits if we mean in term of performance, but there is no drawback neither. Standardisation ACTUALLY adds cost, you know that a licence to have a USB port is about 4K a year? It is what it is and you are right nobody needs this, until they do, it's a new interface, new possibilities, it may never be found useful but in the future if it gets used in more upstream product, well, they'll have it. You know you can't send DSD over SPDIF right? well that's one use case because, for a non-computer device, a "non-CPU based" streamer for exemple, dvelopping a audio class 2 usb OUT is much more complicated and costly, than that, because that you already use this signal internally. Sure nobody really need DSD, but it's out there. If you are a DAC manufacturer today, and your DAC don't do DSD, many wouldn't buy it just because this is what it is, FOMO. It adds cost, but they WANT to add cost, DACs are cheap to make, all companies competes with offering exactly the same thing. They have to be creative in order to differentiate. Nobody needs a 1000$ DAC, still some will buy it for that tiny little performance bump likely inaudible. If you ask me I would much rather have DSP than a connector like that, but why not, it's still only a 400$ Box, in ratio to a full hi end system, it's nothing, I think that's the point.
 

musicforcities

Active Member
Joined
Sep 18, 2021
Messages
146
Likes
137
That NOT hdmi interface sounds like a terrible idea. First there is no reason this could not be done With high speed USB-C or cat 5/6/7 with a controller programmed into a chip or even firmware perhaps. Second, chances are the complexity of the non-standard means it will never be adopted. Third, the hdmi plug and socket is a clumsy one, way too much strain is too easily put on the socket and it does not seat tightly in part to deal with the mechanical stress. It’s like all the problems s micro-usb 20 times the size! Display port at least had a latching mechanism that took some strain of the board connectors and onto the case/socket.

What a kludge.


Funky stuff.
 
Last edited:

musicforcities

Active Member
Joined
Sep 18, 2021
Messages
146
Likes
137
They boil down to Topping/SMSL/Sabaj etc not having 'a sound'

Their transparency is why I will never own one of these brand’s DAC. I NEED a DAC THAT WILL TRANSFORM the signal so my ABBA tracks sound like AC/DC and my Klipsch speakers sound less bright, more like Revels! Cause I can’t stand those speakers as they are and I hate ABBA!

Seriously though, the same audiofools have disdain for tone controls and eq/dsp is sacrilege. Because they pay good money for components to have the eq hardwired in as “house sound.”
 
Last edited:

elvisizer

Member
Joined
Sep 16, 2018
Messages
94
Likes
77
only thing it's missing for me is 10 volt output! if it had that i'd buy one.
 
Top Bottom