classic digital inputs / ouputs : Aes vs Bnc vs Rca vs Toslink. Does it matter?

[laurent]

Apart from Firewire or USB Dac, many people use an interface (Lynx Aes 16 like Chris or Weiss Afi1 for example) to a "standard" Dac. This may be the old way but it still works.

I would like to know which ouput should be preferred over another and why?

 

I have made a few experiences and I am a little confused.

For instance, I have used a Dac2 as a digital converter with Aes to a Dac7 and the results were good compared to my current set up with a usb card and glass toslink. On the contrary, I have used the dac2 with Rca on a A1008, and the sound was really not as good as the same usb card and optical glass connection to the A1008.

I also don't understand why a glass toslink is so much better than a plastic one. I tried this with the Dac7. It should not have anything to do with jitter or electrical noise?

 

I think I have read on this forum Mr Rankin stating that if he were to go for a usb interface (a path he has abandonned due to his asynchronous design) he would use BNC to the Dac. Why so? I think it is also the way chosen by Belk canto for their usb link.

 

Sorry if I seem to crush open doors, but help me be a little bit less confused (and ignorant).

 

Thanks

 

laurent

 

Hereunder an informative note on the differences between the standards :

 

http://www.rane.com/note149.html

 

[BobH]

Hi Laurent,

 

You should prefer the one that sounds best to you. It really is as simple as that. There may well be valid technical reasons explaining why one sort of connection would be preferred over another but these reasons are technical ones and may not be relevant to your system components and how they react with each other.

 

Various manufacturers will favour various connection types for all sorts of various reasons - IOW there is no hard and fast rule! But if a manufacturer recommended the spdif connection on their piece if kit, then that is where I would start looking for the best possible sound!

 

Grab a bunch of different cables, (they don't have to be expensive ones), listen to each for a week or so and go with the one you like best. If plastic Toslink sounds best to you, in your system, then the fact that it is generally regarded as 'rubbish' is irrelevant. Once you're happy with the type of connection that sounds best in your system, you can start obsessing over how to much spend on a really good one and whether or not it should be solid core or multi stranded copper with, or without, silver plating!!

 

And just to get off the fence a little, as far as Dac's are concerned, I've not yet heard a good implementation of USB (cue flame-war!) and I have always preferred AES to spdif - but that's just me! I still use whatever sounds best to me and that depends on what is being connected. There is no answer other than to 'suck it and see'.

 

Have fun!

 

[Andrew S.]

I agree with Bob but....

 

"And just to get off the fence a little, as far as Dac's are concerned, I've not yet heard a good implementation of USB (cue flame-war!) and I have always preferred AES to spdif"

 

I would argue that that all depends...on your platform, operating system and audio engine (by which I mean the sw making the tunes..).

 

Certainly nothing wrong with AES/BNC - they have been pro standard for years. Firewire is now the pro standard and usb is in it's infancy. Give it 12 months and I am sure we swill see more asynch come onto the scene.

 

But generally the devil is in the detail and implementation is everything. No one is going to say a Alpha Dac with AES sounds bad. Or an Accustic Arts with spdif... On the other hand many folk (me included) will tell you M Audio (ummm just about anything...) sounds distinctly average through AES. Or anything else.

Wadia use glass very very effectively.

 

Sorry no hard and fast rules.

 

And just because the implemenation is excellent you may not like the sound of the dac...eg , for me, the Benchmark's. Or the Lavry DA10. Both have a signature I don't care for but I would never say they weren't excellent dacs well executed. Great design great build a whole host of features but....not for me.

 

So my advice get out there and listen to a whole bunch of stuff...or save yourself the agony and just buy a Weiss Dac2. Great place to start...

 

 

 

[Audio_ELF]

BNC, generally, should be marginally better than a phono connector as the BNC is rated at 75ohm, same as the SPDIF spec. However there are many more factors that will affect this. I'm not sure if anyone has managed to compare the BNC and Phono inputs on a Bryston DBA-1 for example?

 

Optical connections can have the advantage of being electrically isolated input to output, however the extra level of conversion electrical - optical - electrical can add it's own problems.

 

AES, where provided, can be the best if implemented correctly. The cables are balanced so any external interference is minimised, this allows longer cables from computer to DAC which can eliminate the troubles of noise from a MacPro or tower PC required to accommodate an AES16 or similar card.

 

I would agree with Bob though, you need to try each interface and see if one is better than the others. There are so many factors that will affect SQ that it's impossible to say interface A is better than interface B. Some manufacturers will concentrate everything on the AES interface, adding the optical and co-ax SPDIF connections at the last minute. Some DACs will have separate receivers for AES and SPDIF, others will just utilise one receiver and carry out conversion for the other input type. Of course any FireWire or USB interface is also going to have a "preference" of AES over Co-ax, or Co-ax over Optical, etc.

 

Anyway ... thats my thoughts ...

Eloise

 

[cfmsp]

 

lots of good advice already.

 

judging by the amount of money people spend on coax S/PDIF cables, I'd want to go with AES/EBU, I think, all other things being equal. But they seldom are, of course. BNC would also seem to have a theoretical advantage over RCA coax.

 

 

happy listening,

 

clay

 

 

 

[Wavelength]

Gang,

 

I prefer BNC 75 ohm cables and setups. AES is crap as what XLR connector can be a true 110? RCA the same... basic physics... diameter over diameter. If the center section to outside section of a BNC gives you 75 ohms what does that make the RCA??? it's easily... well let's measure..

 

BNC center 0.035", RCA center 0.125 or more than 3.5 times larger.

 

Guys don't tell me that all these connector companies guarantee 75 ohms for RCA's... it's marketing.. should have them tell some of you guys to jump off a bridge and see if you will do it.

 

Toslink... other than isolation which is what should be done anyways... it's a worthless cheap interface.

 

Thanks

Gordon

 

[laurent]

To Gordon,

 

Many thanks for your straightforward and rational answer.

I am a little surprised about xlr cables. I thought that their being symetrical would make the Xlr interfaces less cable "sensitive" if I may say so. When I tried modulation cables, xlr always outperformed rca

and the differences between xlr cables were minor.

Concerning the cable "sensitivity" of FW or USB interfaces, it seems that the quality of the cable also matters. I have heard it with a ultraviolet cable usb cable. We can also find usb cables costing hunderds of dollars.

In your experience, do you think that usb cables play a significant role ? And if it is the case, couldn't we imagine a "usb key" real Dac without any usb cable just a male usb plug. Not so inconvenient for lap top with usb on the side.

Sorry for changing the subject in my own thread and posting a certainly ridiculous suggestion.

 

Laurent

 

 

[Wavelength]

Laurent,

 

Again RCA can't be 75 and that is why it's easy to modulate it. Try a really good BNC to BNC cable like the Nirvana T2 and try and modulate that.

 

Most of the best BNC outputs are differential so that should not be a issue. The big deal is that AES/EBU has a lot higher output and BNC is only 0.5vac which means you also have take the percentage of the modulation.

 

Thanks

Gordon

 

[syganymede]

As I understand it.... the music industry is against letting you send anything above CD 16bit/44.1kHz signals over any non-encrypted link. Check out DVD-Audio and Blu-Ray audio to see what I mean. If you want to send out 96kHz or higher then the digital stream has to be encoded and S/PDIF (TOSINK or Coax) and AES are not allowed. You have to use Firewire of HDMI with encryption. Both ends of the link have to be "qualiyfied" equipment i.e. with built in encryption keys and systems.

 

 

 

[hifitubes]

I'd like to hear more about the non-existence of 110ohm XLR connectors.

 

I'm considering moving from a BNC system to AES/EBU.

 

thanks

hifi

 

[ar-t]

What do you want to know about 110 ohm XLRs? They do measure 110 ohms, at low frequencies, but go all wrong at higher frequencies.

 

[manisandher]

How important is ar-t's above post if you have alongside an AES cable a separate 75 Ohm wordclock cable terminated with BNCs? I'm assuming impedance matching is of most importance when considering the transmission of the wordclock, or is this not totally correct?

 

Another assumption I've always made is that using balanced XLR is helpful in preventing noise entering the ADC/DAC. Again, not totally correct?

 

Mani.

 

[hifitubes]

So could you qualify "high frequencies" .....18kHz, 100kHz, 200mHz?

 

After asking various designers, seems like AES

is fine for auido use.

 

[ar-t]

Some engineers think AES only has to pass 15 MHz to work. Yeah, it works, but works better (they all do) if you have 50 MHz or greater bandwidth.

 

The AES/EBU format was thought up by a bunch of audio engineers. That is what the AES membership is. However, almost none are RF engineers. I had along talk with the guy who headed up that group, but not until after it was finalized. Pointed out all the flaws in it. They corrected some of them. But the XLR part was cast in stone.

 

Most designers aren't RF engineers, either. Which is why 99% of all DACs have lousy input stages.

 

[ar-t]

Missed a post.........

 

Terminating the word clock is not as important. First, it is at word clock frequency (right?), so 44.1 kHz will have almost no energy @ 100 MHz. Any mismatch will have less noticeable impact. Second, you are not extracting the clock from an encoded signal, which is where the problem really lies.

 

Transmitting a word clock, separate by itself, will greatly reduce jitter. But, it will not eliminate it. Most modern DAC chips use the bit clock, as the crucial clock. The word clock merely tells it when all the bits are loaded. (Look at the PCM1704, as an example. The conversion takes place 2 bit clocks pulses after the word clock changes state.) It would seem to me that there has to be some sort of PLL, to derive the bit clock from the word clock. Obviously, this is a much easier matter than getting the bit clock from the SPDIF (or AES/EBU signal), but there will be a measurable amount of jitter.

 

Audible? Perhaps not. It will be Gaussian in nature, and the threshold of audibility is higher, than with data-correlated jitter.

 

Once again, throwing around jitter numbers without specifying whether it is word clock or bit clock, the amplitude, the frequency content, and whether or not it is data-correlated is all meaningless numbers. You have to take all of those into account, to get a proper perspective of its importance.

 

[hifitubes]

Good stuff, thanks for the insight. The new (?) Weiss DAC seems to have BNC word clock i/o. I wonder why many pro and hifi units always seem have BNC as word clock and AES/EBU via XLR; I haven't seen any that use BNC for both outputs.

 

My system currently has an ADC as a source and I'm clocking my transport to the ADC as well. I'm planning on moving to AES/EBU via XLR to move the clock to a dedicated BNC output from a Mytek ADC192.

 

[ar-t]

It wouldn't be AES/EBU if it didn't have XLR. The only differences between it and SPDIF are:

 

Impedance: 110 ohm, balanced vs 75 ohm, unbalanced.

 

Level: 5 V vs. 0.5 V

 

Originally, they had some silly scheme where it could have 3 transmitters and 7 receivers. (Those numbers may be off one or two.) Didn't take long to figure it didn't work right. That is part of the reason for the higher level. The other is greater noise immunity, as it was designed to operate in a noisy commercial setting.

 

When I pointed out the flaws to the multi-unit scheme, they laughed! Guess who had the last laugh. (Insert mental image of your favorite smiley icon here.)

 

[hifitubes]

Yes, just that on an EMU card, for example, with no true XLR output, you can choose to output AES/EBU, but you would need an adapter. Sorry for any confusion.

 

http://images.asia.creative.com/images/inline/products/E-MU0404PCI/feature_1.jpg

 

Some comments by Lavry:

 

http://www.lavryengineering.com/lavry_forum/viewtopic.php?f=1&t=102&view=previous

 

[ar-t]

Interesting post, you have linked to. Some points I would take issue with:

 

"The connector alone has nothing to do with jitter. The cable alone has nothing to do with jitter. The issues making jitter are, as I mentioned already, the lower and upper cutoff of the transmission link, the termination tolerance, the cable impedance tolerance and more. The biggest potential issue is the skin effect (non issue at very short distances), which has to do with the dimensions of the cable, who much copper, how many wire strands and so on."

 

Uh, not quite. "Cable impedance tolerance" has to include the "impedance tolerance" of the connector. Any deviation from the design impedance with create reflections. That is simply the laws of physics. Where some folks fall into the trap is thinking the BW required is not that high. If you subscribe to the low BW school of thought, then the impedance problems of the connector are negligible. (Say 10 MHz or so.) But, if you believe that a higher BW is required (50-100 MHz), the connector is a problem.

 

I don't suppose the guy who wrote that post has ever done any work correlating impedance mismatches to recovered clock jitter. I have, so I know the problem exists. How much? Depends on how picky you are.

 

"The statement: "Those optical TA&T is an idealistic interface as its transmitter & receiver are the root cause of jitter (20 times of BNC)" can be highly misleading as well."

 

The problem with the ATT link is that it was designed to work at lengths > 1 km. Will it work at shorter distances? Sure. But it will be plagued with pulse dispersion problems. If you want to know the physics behind that, you can contact me. If enough folks want to know, I can bloviate on why that is so.

 

"The statement: "75? is the perfect among three interfaces since its jitter is the smallest, BNC connector is better than RCA" is wrong."

 

Well, the person who wrote the first response was on the right track, they are a bit mixed up. A true 75 ohm BNC will have the lowest reflection coefficient, of any of the connectors being discussed. An XLR has horrible discontinuities. A RCA is nowhere near the 75 ohm, called for in the SPDIF spec.

 

"In short, ONE OF THE MANY mechanisms that may cause jitter has to do with the transmission of the digital audio. The mechanism is based on BOTH signal rise time and signal droop, which is in fact the upper frequency and the lower frequency of the transmission bandwidth."

 

Yes. But as he states, there are many mechanisms. And reflections are a big part of that. For some reason, I get the impression he minimizes that part of the problem.

 

Rise time is important, which is why many of us feel you need to have a fast driver. Most any cable will not have any effect on the rise time, even if it is ridiculously long.

 

Now, here is where he is on the right track, but may (or may not) realize that the "recommended" approach is all wrong.

 

Warning: Insane rant alert, and possible inflammatory remarks follow.

 

For some obtuse reason, almost the entire industry has been bamboozled into using a brand of transformer that is a piece of crap. For those of you who have read some of my other posts, you may recall a link to Audio Circle, where I show measurements of said POC. It has more LF droop, than other transformers. Bad for performance! And, it has less HF BW than other brands. (And it has horrible reflection problems, if you hook it up the way the app note shows.)

 

Now, if this designer actually uses those parts (I have no idea if he does, and have no interest one way or the other if he does), but if he does, he is doing the opposite of what he should do, based on what he has posted!

 

As he states:

 

"My advice: don't believe everything you read."

 

Agreed. Even if it is by me. Take some time to research and learn about this stuff, before you propagate disinformation. Unfortunately, a lot of these concepts are tricky to comprehend. Even for "professionals". The 'Net is full of well-intentioned, yet confused, "amateurs" taking these concepts, misinterpreting them, and spreading more confusion. Hell, even the "professionals " at the AES have shown they don't fully comprehend them.

 

What can you expect from "audio" engineers. Well, in my experience, a complete lack of understanding of RF.

 

Explains why there are so many problems with SPDIF and AES/EBU, which allows for endless futzing by equipment owners. If there were much tighter adherence to the proper impedance, both on the TX and RX end, then cables would all sound pretty much all alike. But, as most of us have found out, you can change cables all day long, and they all sound different. It isn't BW: the cables have little effect on that. It is all impedance problems, and how they interact with everything else in the chain. It all has to be considered as an entire chain. One part, taken aside and characterized, will not predict its performance in a system.

 

 

 

 

 

[hifitubes]

It's very strange how the de facto standard is to offer up RCA digital i/o on high end gear (Weiss 202, Mytek ADCs, REM Fireface UC). I can't get over it. Just be done with it already, especially on "high end" gear; people can buy a BNC>RCA adapter or cable.

 

I have heard Steve mention you can't just have these devices modded to replace the RCA jacks w/ BNC because you need to have traces on the boards changed/modified as well....I'm not an expert, that's just what he said (on Head-Fi?).

 

Does the fact the XLR is perfect and RCA is junk actually have any bearing in the real world? To quote a friend (hope you don't mind RT!):

 

Now none of that may actually matter in the real world - the fact that wire hangars function as reasonably acceptable AES cables attests to that. These things affect signal quality in various ways - jitter is only one of them - but jitter does happen to be one of the few distortions involved which won't completely corrupt the signal. From a pure engineering standpoint it would be swell to use nothing but coax of some sort for such digital transmissions, but from an upgrade perspective it's still worth evaluating this on a purely psychoacoustic basis for the potential room for improvement.

 

[ar-t]

XLR is perfect? Who told you that one? It is far from perfect. I can not think of a poorer RF connector.

 

OK...........maybe USB. Haven't put any on the TDR, as I see no use to fret over a interface that is unterminated, with a pull-up resistor on one leg. (Who is the guy who thought that up? Oy vey.)

 

Back when I had a real job, we spent a lot of time seeing how lousy of a cable pair you could push a T-1 stream through, and still make the Bell System mask. You would be surprised.

 

No, maybe not. Big difference between "It works. Somehow." and "WOW, this is fantastic!" Audiophiles are willing to spend $$$ to go from one to the other. Even when that step is a very small one.

 

As for modding...........yeah, the traces do play a part. There are ways around that. But, the mere fact it has RCAs is not the end of the problem. It goes beyond that. Stupid stuff like a LPF on the input. (One of the biggest names in the bidnis did that one.) Diode clamps, Schmitt trigger inputs, crappy transformers, other obtuse stuff. You will have to trust me on these points.

 

[hifitubes]

Sorry for the typo. I meant NOT perfect. Thought the gist of my comment would have that shine through. Sorry.

 

[manisandher]

rom661

"The only thing I can advise with confidence is that SPDIF is superior to toslink, at least when a CD transport is involved. I know there is some debate among some people but I have made comparisons for fifteen years. Toslink is cheap. In terms of traditional transports, in ascending order I consider it to be toslink, SPDIF(coax)and AES."

 

barrows

"... I find it very curious that you would rank an AES digital interface as superior to SPDIF? AES is certainly technically inferior to SPDIF in every way. As just one example, the XLR connector will never have proper impedance for a digital interface (it was never meant for high frequency signal transmission). Properly implemented SPDIF (bnc connectors, properly terminated 75 ohm, with a true 75 ohm cable) will outperform AES for signal integrity every time."

 

Mani

"... with a bit of motivation from Juergen (JR_Audio), I decided to see if I could improve the AES connection. What I did was to feed the ADC/DAC with an analog signal (from vinyl) and compare:

1) direct ADC -> DAC output

2) indirect ADC -> AES cable -> PCI interface -> AES cable -> DAC output

 

What I found was that I could tune the system so that there was absolutely no audible difference between 1) and 2), in double-blind listening and extended listening tests.

 

My conclusion: my AES cables and PCI-interface are transparent enough for my system.

 

barrows

"Of course, real world performance will vary, as many SPDIF interfaces are improperly implemented, and do not achieve 75 ohms (like the many that use RCA connectors, or that use BNC, but for some odd reason choose 50 ohm BNC connectors instead of 75).

I am just on a bit of a crusade here to try and hold manufacturers responsible to making the right design implementation choices when it comes to digital interfaces, as I feel it just plain silly to handicap a great DAC or transport with an interface that cannot achieve a proper impedance and allow for the lowest possible jitter in the interface.

 

Mani

"I'm no RF engineer, but I know that Pat (from ar-t) is/was, and that he doesn't think much of AES/EBU. However, there are other digital designers (JR_Audio) who believe that other things are important, as well as impedance-matching and reflections. Things such as ground loops and common-mode noise. Certainly, AES/EBU can help with the latter. As for impedance-mismatch/relections, well the effect of these can be reduced by using longer cables - something that AES/EBU works well with.

 

rom661

"All of the reference Levinson stuff I sold for years sounded better to me and everyone on my staff via AES. I am offering no explanations for it; it was a subjective evaluation done by experienced listeners."

 

barrows

"If an AES interface sounds better (than SPDIF) on a given piece of gear, it is likely that the SPDIF interface is not properly implemented (from a technical standpoint)-this imperfection in the interface could be at the transport, it could be the cable, it could be the DAC, or it could be a combination of all three."

 

"Common mode noise rejection for the cable lengths used in audiophile systems is a non-issue for digital signals, as the digital transmission is immune to noise (at the levels present) being a factor."

 

JR_Audio (quotes taken from another thread)

"I would highly recommend using mainly balanced digital audio interfaces"

 

"The best input schematic would be to have a digital input transformer with a static shield between the primary and secondary winding, and that this shield is only connected to pin 1 of the xlr jack. This is the way I have done a lot of AES/EBU inputs. So the static shield between the input transformer is connected only to the shield of the cable and on the sender side of the cable, connected to ground, By doing this, you have a very high common mode rejection of any common mode voltage and so the best isolation between the sender (transport) and the receiver (DAC). Only the differential signal of the AES signal is transferred to the DAC. This is the reason why I do like balanced interfaces for digital line so much. With SPDIF you can't have this 'features."

 

"For digital transformers, I personally suggest the best common mode rejection on every side, on the receiver and on the sender side. But keep in mind, this is my recommendation."

 

Mani

"In any event, my comparison was not 'AES vs. spdif', it was 'AES vs. absolutely no cable'... and a number of people couldn't hear any difference whatsoever in double-blind tests. And I haven't hear any difference in extended listening.

 

barrows

"For a home audio system, better to have a properly implemented SPDIF interface that maintains 75 ohm impedance throughout, as the S/N on the cable should not bother the SPDIF receiver one bit."

 

I think this gives the gist of the exchange on the 'AR DAC8' thread.

 

Mani.

 

 

 

[barrows]

thanks for moving this here, and I approve of the excerpts here as getting the gist of what we were saying.

I would like to add one small point. I realized I did not address the issue of noise coming in on the digital signal, and getting into the DAC's analog circuitry via ground contamination (as opposed to noise causing problems for the DIR). Yes, I would think running AES balanced can help here, but I do believe proper implementation of a good digital input transformer will take care of such noise on a SPDIF interface without resorting to a balanced solution.

If Pat (ar-t, hope I got it right this time) is still monitoring this thread, maybe he will add some additional comments, right now I am trying to decipher what (brand) is a good digital input pulse transformer, and so far the internet is not helping!

 

[hifitubes]

I use AES to feed my Zodiac+ from an TYS16-E, and have to float the grounds on the external switching AC adapters for both DAC and mini-ITX PC. The results in a very audible lessening of the noise floor.

 

At some point I would like to run these off linear or battery supplies.