|
Post by tonycamphd on Sept 28, 2015 14:57:54 GMT -6
Anyone have any data, or an opinions on which performs better? BNC is unbalanced over a 75 ohm cable with an isolated clock signal, and AES is balanced and extracted over a 110 ohm cable as explained by lynx, my runs will be extremely short so length will not be an issue, but I am external master clocking 6 slaved converter boxes together, 2-8x8 channel ADDAC, 2-8 channel DAC, a stereo channel ADC, and a stereo channel DAC for final monitoring, i'm a little shaky on this stuff, so i'd like to learn as much as possible before committing to either.
thanx for any info
|
|
ericn
Temp
Balance Engineer
Posts: 15,012
|
Post by ericn on Sept 28, 2015 15:01:59 GMT -6
BNC comes to us from the video world, I have never seen a isolated clock on anything but BNC or RCA. I do believe That having the word clock isolated is the thing !
|
|
|
Post by Bob Olhsson on Sept 28, 2015 17:21:52 GMT -6
Word clock was an advantage with the first generation of incompetently designed digital transceivers but not with recent designs.
|
|
|
Post by LesC on Sept 28, 2015 17:23:09 GMT -6
I believe any setup of that complexity should be clocked using BNC. Tony, does this mean you received all the Ross Martin stuff? I hope so, and that it all works great for you!
|
|
|
Post by tonycamphd on Sept 28, 2015 19:23:51 GMT -6
I believe any setup of that complexity should be clocked using BNC. Tony, does this mean you received all the Ross Martin stuff? I hope so, and that it all works great for you! thanx pal, not yet, but i'm positive it's getting closer, i've had contact with him regularly, i will be sending it all right over to JW for preventative maintenance when i get them haha 8)
|
|
ericn
Temp
Balance Engineer
Posts: 15,012
|
Post by ericn on Sept 28, 2015 20:32:28 GMT -6
I believe any setup of that complexity should be clocked using BNC. Tony, does this mean you received all the Ross Martin stuff? I hope so, and that it all works great for you! thanx pal, not yet, but i'm positive it's getting closer, i've had contact with him regularly, i will be sending it all right over to JW for preventative maintenance when i get them haha 8) Since he's going to have them open, and knowing Jim he's going to anylize the clock I'd ask him!
|
|
|
Post by svart on Sept 28, 2015 20:45:12 GMT -6
Yep, I have some input.
BNC is 75ohm TTL(typically 5V) level with "wordclock" being the primary signal used for clocking. Wordclock operates at the frequency of sampling (I.E., 44.1K, etc), which is considered low frequency. The high amplitude and low frequency make it relatively immune to low amplitude noise from the outside world. Coax cables are used for all kinds of signals up into the terahertz range, so coax itself is not a problem at all. Where problems occur is the multiple connections (daisychaining) of the BNC through multiple devices. Each device can add noise in the form of ingress (noise that couples to the conductor through EMI/RFI) or through noise figure (NF is noise added to the signal from amplifiers and active components) or through other means. Termination with 75R is a requirement on the last piece of gear to "sink" signal currents and keep reflections low. Just as with a wave of water in the pool bouncing back from the side, electrical signals do the same thing without a termination resistor sucking the energy out of the reflection.
For wordclock, the jitter on the original signal is the baseline for the jitter for the system.
AES3 is typically differential 110R, up to 7V (at transmitter output) with the clock signal embedded into the datastream. The encoded clock is a "biphase" clock typically at 128x-256x the sample rate. The clock must be digitally extracted from the signal and divided down to a wordclock (Dividing clocks also divide the jitter). Since the signal can be in the low MHz range, the differential data is more for allowing long runs of signal without rounded edge transition problems, than for noise immunity. The large amplitude of the signal is rather immune to most noise sources, much like coax.
AES3 can also be single-ended as well as 75R.
For AES/SPDIF, some receiver ICs have internal PLLs which will lock to the extracted clock on the incoming signal and reproduce a "cleaner" version. Some refer to this as "jitter cleaning". Some designers believe this is just wishful thinking, and some claim that it helps. What does help is the division of the incoming clock signal, which divides the jitter down by the same amount the signal is divided.
"Superclock" was at one time the same idea behind the extracted clock of AES, but over a wordclock-like setup. It would have been a 128X or 256x clock signal (I.E., 11.289600MHz for 44.1Khz x 256 sample rate) that was over BNC, so that it could be divided down by each unit internally, for superior jitter performance. Unfortunately it never really caught on and the increase in costs caused customers to balk at upgrading.
So for me, I would choose AES simply because of the clock dividing. The physical connection doesn't make a terrible difference unless going more than 25ft or so.
|
|
|
Post by LesC on Sept 28, 2015 21:13:36 GMT -6
It's interesting that Dan Lavry has (or had) the opposite take on it. He believes that a 96 Hz clock is very easy to generate and detect compared to a clock extracted from AES. Here's a thread that basically asks the same question, with answers from Lavry and some others: repforums.prosoundweb.com/index.php?topic=12452.0This thread is from 2006, so it's quite possible that in today's world Lavry would agree with Svart.
|
|
|
Post by tonycamphd on Sept 28, 2015 22:41:55 GMT -6
really great stuff guys, thanx! i'm interested to hear from svart what he thinks of that link? and what is the deal with "AES black"? i've never heard of that until reading that thread....hmmm.
|
|
|
Post by Ward on Sept 29, 2015 6:44:28 GMT -6
AES3 might be better, but I've never had an issue with BNC . . . . but DO remember to terminate!!
|
|
|
Post by svart on Sept 29, 2015 7:03:59 GMT -6
really great stuff guys, thanx! i'm interested to hear from svart what he thinks of that link? and what is the deal with "AES black"? i've never heard of that until reading that thread....hmmm. It's an interesting take. I'm sure he's done testing and is aware that the preambles are for syncronization of the decoder, not necessarily for clocking. In fact, the preamble is not BMC (biphase mark coded), which means that the clock is not encoded into this portion of the signal.. In layman's terms, the preamble does not contribute to the clocking of the signal. Only the audio data is BMC coded, so in more layman's terms, the audio data is the only portion that contains clocking information. For him to suggest that the preamble affects clocking directly is questionable. However, if he meant that it could affect the receiver and have the potential to cause a glitch, then he's correct in theory. Most AES/SPDIF chips are now mature devices and it's highly unlikely there is any form of glitch or clocking issues. In fact, there are only 3 variations of preamble, X,Y and Z as specified by AES3 standard (taken from wikipedia so I don't have to type it all): X (or M) : 111000102 if previous time slot was 0, 000111012 if it was 1. (Equivalently, 100100112 NRZI encoded.) Marks a word for channel A (left), other than at the start of an audio block. Y (or W) : 111001002 if previous time slot was 0, 000110112 if it was 1. (Equivalently, 100101102 NRZI encoded.) Marks a word for channel B (right). Z (or B) : 111010002 if previous time slot was 0, 000101112 if it was 1. (Equivalently, 100111002 NRZI encoded.) Marks a word for channel A (left) at the start of an audio block. So his claim that preambles vary depending on audio data is making me scratch my head a bit.. Anyway, for some anecdotal evidence, I placed an oscilloscope on the recovered clock of a modern Wolfson digital receiver IC and it was as clean and glitch free for 5+ minutes as you would expect any clock to be. I suspect that Dan was mainly speaking of older technology on all this. I mean, I gotta give him the benefit of the doubt.
|
|
|
Post by Bob Olhsson on Sept 30, 2015 11:26:23 GMT -6
There was an AES paper about data corrupting clocks back in the '90s that undoubtedly led Dan to making some measurements.
Where the rubber met the road was when the same clowns were hired to create digital video interfaces and their incompetence at clock recovery was revealed in picture glitches they couldn't argue required ABX testing. What was learned has trickled down to new audio designs over the past decade.
The thing that sucks is that Bell Labs figured out how to do all of this properly in the 1950s but no audio chip manufacturer had done their homework. We no longer have the best and brightest designing audio gear which had been the case prior to the 1960s.
|
|
ericn
Temp
Balance Engineer
Posts: 15,012
|
Post by ericn on Sept 30, 2015 12:00:57 GMT -6
There was an AES paper about data corrupting clocks back in the '90s that undoubtedly led Dan to making some measurements. Where the rubber met the road was when the same clowns were hired to create digital video interfaces and their incompetence at clock recovery was revealed in picture glitches they couldn't argue required ABX testing. What was learned has trickled down to new audio designs over the past decade. The thing that sucks is that Bell Labs figured out how to do all of this properly in the 1950s but no audio chip manufacturer had done their homework. We no longer have the best and brightest designing audio gear which had been the case prior to the 1960s. No money in the software even less in the hardware! Besides most care more about how many songs can it hold, remember when it was about how reel it sounded?
|
|
|
Post by Ward on Oct 1, 2015 7:07:40 GMT -6
The thing that sucks is that Bell Labs figured out how to do all of this properly in the 1950s but no audio chip manufacturer had done their homework. Please tell me more!!
|
|
|
Post by Bob Olhsson on Oct 1, 2015 7:39:17 GMT -6
I wish I knew more but my knowledge comes out of conversations I've had with people in a position to know at AES conventions. I'm a nerd who attended technical papers even when they were way over my head because I wanted to understand digital audio as well as I understood analog from hanging out in studio and radio station shops. For example dither was/is an integral part of our digital telephone network that was developed in the 1950s. Today we have people ignorantly claiming that it should be optional while it should just be handled with nobody needing to know or think about it.
|
|
|
Post by svart on Oct 1, 2015 8:56:39 GMT -6
As a side note, one of the reasons I chose the specific digital receiver I use is that I found an old whitepaper comparing these types of chips, and found one that reduced transport-induced jitter by buffering and reclocking the output data to a DPLL locked to a precision clock source, but in a novel way. A few open folks around the world have also posted a lot of testing info that seems to support the claims of vastly superior jitter performace compared to other manufacturers, including the one used in another very well known product on this forum.
The extracted clock and the external clock are buffered and phase locked, then the data is re-aligned to the new clock, without the problems of jitter that is present on the incoming streams.
What you get is a signal that is nearly free of transport-induced jitter, but would still include some accumulated jitter and wander from the clock/DPLL, but these are orders of magnitude lower than the average transport jitter.
Through my own testing, the resulting phase noise is on par with good internal clock sources, so I have no reason to believe that AES3/SPDIF has any worse potential than any other scheme, but it is still rather sensitive to the regular problems of ADC clock jitter and encoding/justification jitter.
Modern parts have rendered most of the arguments invalid about digital transport streams being poor clocks invalid IMHO.
I also find it amusing that during my daily work, I deal with modulated signals in the GHz range, DACs with outputs in the hundreds of MHz, and with signal levels akin to mouse farts within a noisy restaurant.. So I'm used to dealing with clocking schemes that need jitter immunity 10-100x better than the "best" audio clocks, yet I continue to see people squabbling over jitter numbers that seem almost alien to me.
|
|
ericn
Temp
Balance Engineer
Posts: 15,012
|
Post by ericn on Oct 1, 2015 9:20:56 GMT -6
I wish I knew more but my knowledge comes out of conversations I've had with people in a position to know at AES conventions. I'm a nerd who attended technical papers even when they were way over my head because I wanted to understand digital audio as well as I understood analog from hanging out in studio and radio station shops. For example dither was/is an integral part of our digital telephone network that was developed in the 1950s. Today we have people ignorantly claiming that it should be optional while it should just be handled with nobody needing to know or think about it. Bell labs died with deregulation, My Uncle spent 3 years being trained for the first digital telco switch install, he like others were swept away when Siemans and others entered the market and offered insane money that the U.S. Telcos couldn't match. From the winters in Frozen Rice Lake WI barely making ends meet to Sunny Boca and another digit on your pay check! The old days of Bell Labs meant you could be working on something for NASA and your own pet project now it's just make it smaller and cheaper!
|
|
ericn
Temp
Balance Engineer
Posts: 15,012
|
Post by ericn on Oct 1, 2015 9:28:26 GMT -6
As a side note, one of the reasons I chose the specific digital receiver I use is that I found an old whitepaper comparing these types of chips, and found one that reduced transport-induced jitter by buffering and reclocking the output data to a DPLL locked to a precision clock source, but in a novel way. A few open folks around the world have also posted a lot of testing info that seems to support the claims of vastly superior jitter performace compared to other manufacturers, including the one used in another very well known product on this forum. The extracted clock and the external clock are buffered and phase locked, then the data is re-aligned to the new clock, without the problems of jitter that is present on the incoming streams. What you get is a signal that is nearly free of transport-induced jitter, but would still include some accumulated jitter and wander from the clock/DPLL, but these are orders of magnitude lower than the average transport jitter. Through my own testing, the resulting phase noise is on par with good internal clock sources, so I have no reason to believe that AES3/SPDIF has any worse potential than any other scheme, but it is still rather sensitive to the regular problems of ADC clock jitter and encoding/justification jitter. Modern parts have rendered most of the arguments invalid about digital transport streams being poor clocks invalid IMHO. I also find it amusing that during my daily work, I deal with modulated signals in the GHz range, DACs with outputs in the hundreds of MHz, and with signal levels akin to mouse farts within a noisy restaurant.. So I'm used to dealing with clocking schemes that need jitter immunity 10-100x better than the "best" audio clocks, yet I continue to see people squabbling over jitter numbers that seem almost alien to me. One again a great post, as to your last paragraph remember a lot of audio guys don't really understand clocks and jitter and all the "get " is the absolute of zero! As far as Dan Lavry, he is one of those absolute geniuses, like many just because he's brilliant in no way means all his theories even in his field are are valid interesting yes but prove able no !
|
|
|
Post by Bob Olhsson on Oct 1, 2015 10:26:34 GMT -6
I'd say Wall Street killed Bell Labs and not deregulation. In fact, it's death happened while Carly Fiorina was Lucent's president.
|
|