• 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.

Does Quality of Coax Input Matter for DACs?

eliash

Senior Member
Joined
May 29, 2019
Messages
345
Likes
164
Location
Bavaria, near lake Ammersee
The title is much more lofty than this little test but I thought I share the results of testing a DAC with super high quality Coax S/PDIF vs a lower quality one. Question came up in the review of FX Audio DAC-SQ3: "how good is its coax output?" I have not done many measurements of these "digital to digital bridges" so I thought I expand it into an article by itself.

As I showed in the review of DAC-SQ3, it is one of those odd birds with USB input and Coax/Toslink output:

index.php


So if you happen to have a DAC or AVR that you like otherwise but lacks USB input, you can use this box for that (and get a DAC for free). It was not long ago that companies were selling and promoting such bridges. Getting one question out of way, no, the volume control on DAC-SQ3 does NOT change the output of coax output. It is strictly a bit-exact conversion from asynchronous USB to synchronous S/PDIF.

Coax Output Measurements
My Audio Precision analyzer has a clever mode where on digital input it can extract clock jitter and then analyze it using the rest of its tools. So let's use that to measure its own S/PDIF output to input, compared to output from DAC-SQ3:

View attachment 158528

This is a wideband measurement going to 110 kHz or so. The signal being played is the j-test which has 12 kHz square wave in it. That shows up as odd multiples of 12 kHz in both measurements. It also has a small square wave running at 250 Hz. This too is showing up in both.

The vertical scale is unusual: it is in seconds. The value though is quite small with the top of the display being around 25 nanoseconds (billtionth of a second). As a way of reference, if you have as sinusoidal jitter that has 250 picoseconds of jitter (trillionth of a second), it can damage the low order bit of 16 bit PCM data! So even small numbers matter.

Given the above, you may be alarmed then that the level of jitter out of DAC-SQ3 would deteriorate even your 16-bit audio. But such is not the case because the clock extraction logic in the DAC receiver acts like a flywheel to filter out high frequency jitter (think of heavy platter of a turntable for you analog heads). That filtering is not perfect though and will depend on quality of the implementation. To test the extent of that, I grabbed the Topping DX3 Pro+ DAC which I just reviewed and fed it both ways: through Audio Precision and then FX-Audio DAC-SQ3. Let's start with our dashboards:

View attachment 158529

This is basically the performance we got with USB input for DX3 Pro+. Let's now feed it data over USB to FX-Audio DAC-SQ3 and the Coax to DX3 Pro:

View attachment 158530

As you see, nothing is out of place despite the DAC-SQ3 having worse quality Coax output. If anything, it eliminated power supply spikes that were visible when AP was driving the DX3 Pro+.

Jitter however is frequency sensitive. The higher it is, the more it is visible. Our dashboard is run at just 1 kHz. So let's up the stakes by running the standard J-test where the primary tone is at 12 kHz (so 12 times magnification of jitter):

View attachment 158531

We see that the response is identical. There are essentially no jitter components visible down to whopping -150 dB. All that jitter that we saw has been filtered by Topping DX3 Pro+.

Granted, your DAC may not have such good filtering so there is a bit of unknown there. In all the jitter measurements I have made though, hardly any show levels that rise to audibility so generally speaking, you should be OK. You will certainly be fine if you buy highly rated DACs that I measure.

Conclusion
Digital audio is sensitive to clock jitter due to its high bandwidth and bit depth. Thankfully this has been a known problem for decades and DACs routinely have filters that get rid of much jitter. So within reason, I would not worry about the level of jitter we see out of devices like FX Audio DAC-SQ3.

----------
As always, questions, comments, recommendations, etc. are welcome.

Any donations are much appreciated using: https://www.audiosciencereview.com/forum/index.php?threads/how-to-support-audio-science-review.8150/
...the real nice thing about these USB to SPDIF converters is that many feature a built-in signal transformer, which breaks ground loops from digital (Streamer, NAS, PC, etc.) to analogue (DAC, amp) world. Without such (i.e. an USB DAC with identical digital and analog ground) there will always be some hum or noise under certain conditions (grounded amp, grounded devices connected via RCA, phono input/amp...).
 

Bombadil

Senior Member
Forum Donor
Joined
Jan 28, 2017
Messages
415
Likes
458
Location
Davis CA
I appreciate the review of this topic very much, many of us use the digital output of streamers, like the Node products, and I have assumed that the digital output is neutral, i.e. it does not degrade or enhance the signal. However Amir's review of the Audioengine streamer showed that its toslink output did indeed degrade the signal:
Of course that was not a coax output but it showed that every component can do harm if given the opportunity by the manufacturer lol.
I'm going forward with the belief that the coax output of my Node 2i moves the signal forward with no harm and provides a hi res stream for those who wish to have it. I'm personally good with CD quality output.
 

PeteL

Major Contributor
Joined
Jun 1, 2020
Messages
1,524
Likes
1,442
75 ohm BNC connectors are available, but judging from pictures of AP equipment, it looks like they use 50 ohm BNC connectors.
Here is a picture, which shows the difference between the two types:
50 and 75 ohm BNC

For SPDIF it probably isn't terribly important what connector it is.
I find the picture difficult to judge but the specs are 75 ohms.
 

mansr

Major Contributor
Joined
Oct 5, 2018
Messages
3,903
Likes
8,765
Location
Hampshire
The SPDIF spec calls for 75 ohm impedance for the signal chain. While the cable—probably—is very close to the nominal value, the RCA connector is nowhere near the correct impedance value.

RCA connectors—by virtue of their inner conductor to outer conductor diameter ratio and dielectric—are closer to 40 - 50 ohms. There are some companies that claim to make compression fittings in RCA style that are 'true' 75 ohm impedance, but that's simply not possible given the dimensions of the RCA connector. There will always be an impedance mismatch at both ends of the cable.
The spec calls for an interconnect with a characteristic impedance of 75 Ω ± 35%. That means 50 Ω is still within the acceptable range. In practice, even larger deviations usually work just fine.
 

pseudoid

Addicted to Fun and Learning
Forum Donor
Joined
Mar 23, 2021
Messages
545
Likes
270
Location
SoCal
The spec calls for an interconnect with a characteristic impedance of 75 Ω ± 35%. That means 50 Ω is still within the acceptable range. In practice, even larger deviations usually work just fine.
May work fine but...
Most EEs (at least, in the pre-1980s colleges) were imbued in 50Ω systems for their RF/uW theory/design/use.
This 50Ω 'standard' was born in the 1930s when hi-power RF communications were the 'thing'.
Commercial CATV later established 75Ω impedance for [we could all guess] cost-cutting purposes
There is a nice read for the nerds in the group, titled "Why 50Ω" here.
-174dBm/Hz is probably etched in every RF/uW engineer, like a duck imprint but...
 

KSTR

Major Contributor
Joined
Sep 6, 2018
Messages
1,592
Likes
3,437
Location
Berlin, Germany
However Amir's review of the Audioengine streamer showed that its toslink output did indeed degrade the signal:
I'd say the TOSLINK output is totally fine but is fed with a broken signal to begin with. The signal is already degraded (truncated to 16bits) before it even enters the device.
In the context of this discussion (how much influence does the physical connection has) this is a big difference of definitions.
 

garbz

Member
Joined
Sep 17, 2021
Messages
55
Likes
77
The spec calls for an interconnect with a characteristic impedance of 75 Ω ± 35%. That means 50 Ω is still within the acceptable range.
It's a great spec isn't it. Written at a time when 48kHz was not only standard, but the maximum, never took into account bitrates higher than 3Mbps, considered reusing existing crappy cables, never considered isolation requirements, specified a stupidly low voltage transition and written at a time before jitter was thought to be relevant for a DAC. I suspect if it were still the 80s and jitter were still an audible problem we'd have a hell of a time sending 192kHz over 10m terminated into a crappy "coax" input. I think we should all be happy that we've worked around the shortcomings of what is considered "acceptable" :)

In practice though most I/O I've seen on DACs don't actually follow the S/PDIF spec but rather the AES/EBU spec in all aspects except for impedance. Stands to reason though as most commercially available receivers / transmitters are advertised as the latter, and implement the former by simply grounding one of the lines.
 

mansr

Major Contributor
Joined
Oct 5, 2018
Messages
3,903
Likes
8,765
Location
Hampshire
In practice though most I/O I've seen on DACs don't actually follow the S/PDIF spec but rather the AES/EBU spec in all aspects except for impedance. Stands to reason though as most commercially available receivers / transmitters are advertised as the latter, and implement the former by simply grounding one of the lines.
AES/EBU allows signal levels up to 7 V, enough to fry a receiver expecting the S/PDIF standard 0.5 V. Now the lower limit is the same for both variants, 0.2 V, so a receiver designed for AES/EBU levels will also accept S/PDIF signals. The impedance is typically set by an external resistor. It is probably no coincidence that AES/EBU signalling is identical to RS-485, meaning such transceivers can be repurposed. Since the S/PDIF spec doesn't forbid a receiver tolerating voltages higher than the mandated 0.6 V, it makes sense to use parts capable of accepting the full AES/EBU range as this simplifies interoperation and avoids accidental damage.

Besides the electrical specs, the S/PDIF and AES/EBU variants differ in their interpretation of the auxiliary data carried alongside the audio samples. Receiver ICs generally don't care and simply output the bitstream for consumption by a microcontroller. Transmitters usually support both formats.
 

Bombadil

Senior Member
Forum Donor
Joined
Jan 28, 2017
Messages
415
Likes
458
Location
Davis CA
I'd say the TOSLINK output is totally fine but is fed with a broken signal to begin with. The signal is already degraded (truncated to 16bits) before it even enters the device.
In the context of this discussion (how much influence does the physical connection has) this is a big difference of definitions.
I appreciate this clarification. The "non DAC" pathway through streamers seems a black box to me. If you could describe the pathway from wireless reception of the signal to the coax output that would really help. Would you agree that the Node 2i and current iteration are likely outputting a clean signal through the coax output that directly reflects the incoming information, i.e. if a 24/192 FLAC file is coming in is that what comes out? Does this vary significantly among different products? I can't help but wonder why one would spend almost 3K for a device like the Auralic Aries G1 wireless streaming transporter; it doesn't even include a DAC. Can we say that the sonic performance of a raspberry pi is identical to such expensive devices?
 

JonP

Member
Joined
Jul 7, 2020
Messages
14
Likes
11
I appreciate this clarification. The "non DAC" pathway through streamers seems a black box to me. If you could describe the pathway from wireless reception of the signal to the coax output that would really help.
(Teh Interwebs)->(SW player)->(stream o' digital data in standard formats)->(HW translator to AES/SPDIF)->(output to coax/optical physical media)

Something like that. Piece of software sets up delivery of some streamed music, sends that data to hardware chip translating it to the AES / S/PDIF digital formats and outputting it at the levels of those media types.. and you plug something in to the other end of the cable.

Path for a DAC would be, downstream of the software player the data output simply goes into the DAC chip
 

JonP

Member
Joined
Jul 7, 2020
Messages
14
Likes
11
And, I'll chip in my comments on the impedance concerns. While Sam Adam's thinking about the RCA to 75 ohm coax is correct from a RF theory point of view, I'd expect in this case the magnitude of the effect would be small enough to not to cause problems, especially w modern gear.

Not to say there could be crappy stuff out there, overly sensitive to things, but most of the "circuitry" is inside a chip. Gets hard to design badly in those cases.

Was going to request Amirm do some testing on older gear, to see how bad things could be, but from comments on here it sounds like he'd have to go pretty far back in the vintages.
 

JonP

Member
Joined
Jul 7, 2020
Messages
14
Likes
11
Is it possible to get TOSLINK tested? That’s the one I always wonder most about due to shortages of Toshiba quality connectors about a year ago.
I always heard that "TOSLINK has more jitter, avoid if you can" as audiophile wisdom, wonder how true it was/is. Might be interesting to apply a bit of myth busting to that. If it's easy for Amirm to do, and he thinks it's worth his limited time, of course. (Easy for us to come up w things to load on his table!)

Another note on TOSLINK, way back when I built a Twisted Pear Buffalo 1 DAC (DIY kit, one of the first ESS chip implementations) there was a type of TOSLINK adapter that would do 192khz audio they would try to supply, most were only able to sync to 96khz. The faster ones were going out of production?

Point being, your TOSLINK may vary, both in capabilities and quality. Also all the circuitry is inside the socket, photodetector to translating chip are molded into the connector body, at least in this style. No variations for product designers to mess with.
 

Tks

Major Contributor
Joined
Apr 1, 2019
Messages
2,905
Likes
4,577
I always heard that "TOSLINK has more jitter, avoid if you can" as audiophile wisdom, wonder how true it was/is. Might be interesting to apply a bit of myth busting to that. If it's easy for Amirm to do, and he thinks it's worth his limited time, of course. (Easy for us to come up w things to load on his table!)

On average, in modern devices, USB has overtaken the jitter crown. Unless you're RME for example that implements specific mitigations against jitter with their 'femtosecond clocking', you usually see COAX and TOSLINK pretty much on par (depending on the device, one is a hair worse than the other with no real pattern). Keep in mind, when I say "worse" than USB, it's virtually inconsequential. None of the differences are within realm of audibility as far as we know. The worst you'll see these days is sometimes tones very close to the fundamentals being somewhat high. Aside from that, all three inputs are pretty good. I just like seeing the TOSLINK results since thats the one I find most use for and is the one I feel will go quite wrong if any of the inputs are themselves going to be botched (I don't even know what people honestly use COAX for these days in modern devices, and if you're going to use an input that doesn't require drivers, might as well use TOSLINK with it's isolation inherent property).
 

pseudoid

Addicted to Fun and Learning
Forum Donor
Joined
Mar 23, 2021
Messages
545
Likes
270
Location
SoCal
TOSLINK has become a necessary evil of sorts, but nothing beats a ground loop like it.
The connectorization has always been 'problematic', at best.
202110_ToslinkMetalBody.jpg

I can only hope that @amirm' toslink gets runover on the way back from his milk-run and he replaces it with this type of metal connector housing version.
 

audio2design

Major Contributor
Joined
Nov 29, 2020
Messages
1,350
Likes
1,259
Coax Output Measurements
My Audio Precision analyzer has a clever mode where on digital input it can extract clock jitter and then analyze it using the rest of its tools. So let's use that to measure its own S/PDIF output to input, compared to output from DAC-SQ3:
Does the AP give you a jitter spectrum, or only the potential impact on the resulting audio and pk-pk jitter? (On a side note, has AP fixed their jitter generator to allow more control over the jitter spectrum?)

But such is not the case because the clock extraction logic in the DAC receiver acts like a flywheel to filter out high frequency jitter (think of heavy platter of a turntable for you analog heads).
Jitter however is frequency sensitive. The higher it is, the more it is visible. Our dashboard is run at just 1 kHz. So let's up the stakes by running the standard J-test where the primary tone is at 12 kHz (so 12 times magnification of jitter):

w.r.t. Jitter is frequency sensitive, you are stating that higher analog frequencies are more susceptible to jitter I assume?

w.r.t. the flywheel, there is also the aspect of the jitter w.r.t. frequency. That "flywheel" (the ones in audio usually) are really good at getting rid of high frequencies, >10-20Khz, but not so great at low frequencies.

AKM's switch-cap architecture really solved the jitter issue. The architecture is inherently pretty jitter immune. They have moved away from that on some of their highest end DACs, I expect a combination of process limitations and "absolute" performance for the marketing people.

Good article.
 

audio2design

Major Contributor
Joined
Nov 29, 2020
Messages
1,350
Likes
1,259
If I look at the review of the DAC-SQ3, the jitter is nothing to worry about. Are there not far worse devices and more jitter-prone dac's, especially cheap China-tech on Amazon or AliExpress that would make a far more interesting test?

I think you will find that the cheap "China-tech", as long as they are getting their levels right on SPDIF and not totally screwing up the power supply are not the issue. They will be using AKM/ESS DACs, with a generic I2S receiver. The I2S receiver will get rid of the jitter >10-20KHz, and modern AKM/ESS DACs are pretty effective (arguably better in some cheaper/older chips) at getting rid of low frequency jitter.

I would be far more worried about a boutique roll-your-own DAC that costs thousands.
 

garbz

Member
Joined
Sep 17, 2021
Messages
55
Likes
77
TOSLINK has become a necessary evil of sorts, but nothing beats a ground loop like it.
It is nothing of the sort. Most semi-competently designed DACs will have transformer isolated coax inputs which far outperform crappy Toslink, or they simply galvanically isolate digital and analogue signals between the receiver and the DAC chips. And by outperform I mean in terms of a reasonable looking waveform at the receiver and also the liklihood of longer cable runs or cheap junk cables actually simply working.

The DAC-SQ3 which was used for this test is no exception.
 

pseudoid

Addicted to Fun and Learning
Forum Donor
Joined
Mar 23, 2021
Messages
545
Likes
270
Location
SoCal
It is nothing of the sort.
May sound like a smoke screen until you witness one when trying to interconnect three or four different types/mixes of interfaces, including analog audio.
If you tell me again that ground loops don't exist (and/or "nothing of the sorts"), I shall ignore your comments. Cheers!
 

garbz

Member
Joined
Sep 17, 2021
Messages
55
Likes
77
If you tell me again that ground loops don't exist (and/or "nothing of the sorts"), I shall ignore your comments. Cheers!
As I will yours if you don't read my comments properly. The idea that I said "ground loops don't exist" is a fantasy you made up. I exclusively said toslink isn't a necessary evil, and it's not. Not only are most coax inputs isolated but you can put passive isolators in the line as well if you are dealing with devices which are (to put it nicely) designed by an idiot and not fit for service.
 
Top Bottom