R/E/P Community

Please login or register.

Login with username, password and session length
Advanced search  

Pages: 1 2 [3] 4  All   Go Down

Author Topic: a £400 digital interconnect  (Read 22997 times)

Jon Hodgson

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1854
Re: a £400 digital interconnect
« Reply #30 on: September 03, 2006, 07:49:37 AM »

Bubblepuppy wrote on Sun, 03 September 2006 12:22


"AES and MADI, they have been designed such that what you put in is what you get out, with no further error correction, in fact there is no provision for error correction in them, only detection.

Is simply not true, what you put in uses redundancy and error-correction to rebuild what was originally sent in spite of the errors. They are always there.

No I am not an EE, just deal with them on a daily basis.
Cheers!




Firstly I am an EE, and a Software Engineer also, and I've known about Reed Solomon for nearly 20 years, before I even went to university for my degree, incidentally I've known about the reason that resistance of a conductor increases with temperature for a similar amount of time, just because it's new to you doesn't mean that engineers haven't known about it and designing with it for a long time.

Secondly, do you have the MADI spec to hand? The AES spec?

I do, and I can understand what they say.

Neither of those systems transmits redundant bits for error correction at the receiver.

Actually many transmission channels don't, the connections on a chip, connections between chips, rs232, midi, none of them has any kind of error correction.


Logged

Bubblepuppy

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 72
Re: a £400 digital interconnect
« Reply #31 on: September 03, 2006, 08:57:55 AM »

Firstly I am an EE, and a Software Engineer

I wont hold that against you I promise.

just because it's new to you doesn't mean that engineers haven't known about it and designing with it for a long time.

Its not new to me, its just enginners think the lab is the real world.

Finally, your position is MADI & AES digital communications never have errors?

I give up, if its written in the spec then it must be true....
I think I will sell my new MAC dual core intel go find a 386 and live with the blazing speed of 100Mbits that never has any errors and more than eneough speed to meet all my audio needs.

PLEASE.

Logged
"Trust Your Ears Not The Gear"

Jon Hodgson

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1854
Re: a £400 digital interconnect
« Reply #32 on: September 03, 2006, 09:15:30 AM »

Bubblepuppy wrote on Sun, 03 September 2006 13:57

Firstly I am an EE, and a Software Engineer

I wont hold that against you I promise.

just because it's new to you doesn't mean that engineers haven't known about it and designing with it for a long time.

Its not new to me, its just enginners think the lab is the real world.

Finally, your position is MADI & AES digital communications never have errors?

I give up, if its written in the spec then it must be true....
I think I will sell my new MAC dual core intel go find a 386 and live with the blazing speed of 100Mbits that never has any errors and more than eneough speed to meet all my audio needs.

PLEASE.


No I didn't say they never have errors

I said that if everything is in spec, including the environnent, then there are no errors. I am also saying that the format does not have any contingency for error correction, only detection. If it gets corrupted, it's lost, end of story.

You claimed that all communication channels use error correction systems involving the transmission of redundant data, you were wrong, PERIOD.

And what does the speed of the processor have to do with the speed of the link? Nothing. Just because I maintain that MADI is perfectly adequate to do what it is supposed to do, transmit 32 channels of 96kHz digital audio over distances up to 50m, does not mean that I don't think that there are situations where more channels, video, or some other feature is desirable. In that case I will look at an interface that handles that, and the cabling required.

You try to argue your position by quoting numbers and specifications you don't acually understand, then when you're called on them you accuse engineers of being the ones tied to specs?


Logged

Bubblepuppy

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 72
Re: a £400 digital interconnect
« Reply #33 on: September 03, 2006, 09:42:29 AM »

Isnt detection a contengcy for dealing with errors?

Smile
Logged
"Trust Your Ears Not The Gear"

Bubblepuppy

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 72
Re: a £400 digital interconnect
« Reply #34 on: September 03, 2006, 09:47:37 AM »

Have you ever tried to send 100mbits with a 286?
Logged
"Trust Your Ears Not The Gear"

Jon Hodgson

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1854
Re: a £400 digital interconnect
« Reply #35 on: September 03, 2006, 09:54:17 AM »

Bubblepuppy wrote on Sun, 03 September 2006 14:42

Isnt detection a contengcy for dealing with errors?

Smile


Contingent yes, sufficient NO

Try learning about Reed Solomon and similar codes before quoting them at people
Logged

Bubblepuppy

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 72
Re: a £400 digital interconnect
« Reply #36 on: September 03, 2006, 10:00:28 AM »

"I am also saying that the format does not have any contingency for error correction, only detection"

"Contingent yes, sufficient NO"

I think we need error correction here.
Logged
"Trust Your Ears Not The Gear"

Jon Hodgson

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1854
Re: a £400 digital interconnect
« Reply #37 on: September 03, 2006, 10:17:47 AM »

Bubblepuppy wrote on Sun, 03 September 2006 15:00

"I am also saying that the format does not have any contingency for error correction, only detection"

"Contingent yes, sufficient NO"

I think we need error correction here.



You're playing semantics despite having been wrong.

AES allows for detection of some errors, it does not allow for correction of those errors. Is that clear enough for you?
Logged

Bubblepuppy

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 72
Re: a £400 digital interconnect
« Reply #38 on: September 03, 2006, 03:48:56 PM »

You claimed that all communication channels use error correction systems involving the transmission of redundant data, you were wrong, PERIOD.

Question,
Then explain to me
Cycle Redundancy check sum (Byte 23)
Logged
"Trust Your Ears Not The Gear"

Jon Hodgson

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1854
Re: a £400 digital interconnect
« Reply #39 on: September 03, 2006, 04:06:45 PM »

Bubblepuppy wrote on Sun, 03 September 2006 20:48

You claimed that all communication channels use error correction systems involving the transmission of redundant data, you were wrong, PERIOD.

Question,
Then explain to me
Cycle Redundancy check sum (Byte 23)



Quoting from John Watkinson's "The Art of Digital Audio"

"The final byte in the message is a CRCC which converts the entire channel-status block into a codeword. The channel-status message takes 4ms at 48kHz and in this time the router could have switched to another signal source. This would damage the transmission and also result in a CRCC failure, so the corrupt block is not used. Error correction is not necessary, as the channel-status data are either stationary, i.e. they stay the same, or change at a predictable rate, e.g. timecode. Stationary data will only change at the receiver if a good CRCC is obtained".

Note 2 things
1) Error DETECTION, not correction
2) The CRCC applies only to the status block, not the audio data.

There is one other error detection mechanism in the transmission, that is a parity bit on each word transmitted. This is only capable of detecting a single bit error, which indicates that errors are only expected under very extreme circumstances.
Logged

Jon Hodgson

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1854
Re: a £400 digital interconnect
« Reply #40 on: September 03, 2006, 04:58:09 PM »

Yannick Willox wrote on Sun, 03 September 2006 12:02

As an end user, I did not invest in MADI (although I really need a multichannel digital carrier that is robust enough for live use), just because of these reservations.

The cables that are sold are not OK. I was on the point of buying two MADI/AES converters, but was not able to source non-plastic cable from audio dealers. Contacting the constructor did not help. I had to contact a network company (eg Black Box) to have the glass cables constructed and test myself (!). This was for distances above 50m.

I would say either it is under engineered, or the real world performance is neglected.

Of course 50m with plastic would work, but then you start using 32 ch at 96K, what happens ? On a bad day - in a live situation ? I do not want to be bothered by this, and I have the impression manufacturers do not like to be bothered by this either. They think safe/studio/fixed/short installations.

Also, does MADI not carry the clock ? If I understand correctly, the interface gets near its theoretical maximum when using all channels - so what happens to the CLOCK in this case ?
Did anybody do some tests on a MADI interface with all channels sending/receiving data ?

Yes you can still use wordclock, but doesn't Bubblepuppy have a point, if it had been overengineered instead of just enough bandwith engineered, there would be no potential problem.


In today's world where gigabit ethernet comes built in to a Max at no extra cost it is very easy to look at AES/EBU and MADI and say they were underengineered. You have to look at the situation when they were devised.


AES was devised over 20 years ago, to replace a number of similar but incompatible standards, it also had to work along cables that people had already laid in their studios, and be affordable.

MADI is almost as old, when it was brought in even 10 MBit networks were rare, a 100MBit connection ove one cable really was cutting edge. Sure they could have devised something that could transmit 1 gig, do 300 channels of audio or include video etc, but it would have been way too expensive for anyone to implement, and been stillborn.

The fact is that 20 years after they were devised both these stamdards still do such a good job, in so many situations (want to transfer 32 channels of 96kHz audio from your recorder to your mixer along one cable? No problem, MADI does that fine. Want to connect your converters to your recorder? it can do that.), to me that's a pretty well designed interface.

Could we do a better job if we were starting today? Sure, but these guys were working 20 years ago.
Logged

Andy Peters

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1124
Re: a £400 digital interconnect
« Reply #41 on: September 04, 2006, 08:35:21 PM »

Bubblepuppy wrote on Sun, 03 September 2006 06:47

Have you ever tried to send 100mbits with a 286?


So you're blaming lack of processing horsepower as the reason why you can't keep a 100 mbps pipe full?

If you compared apples with another fruit, your arguments would hold more water.  As it is, you compare apples to Buicks.

-a
Logged
"On the Internet, nobody can hear you mix a band."

Andy Peters

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1124
Re: a £400 digital interconnect
« Reply #42 on: September 04, 2006, 08:54:44 PM »

Bubblepuppy wrote on Sun, 03 September 2006 04:22

Every digital device has errors, Period.


Actually, that's not true.  Error-free digital electronics is in use every day, all the time.

Quote:

The use of Reed-solomon code.

(snip)

When audio data is (sic) transmitted or stored, it is coded and accompanied by redundancy.


See, now you really miss the important point here: "when audio are transmitted or stored" is too simplistic.  CDs, as an example, use RS coding because it is expected that the discs can suffer physical damage simply because they are mechanical items and are handled by users who won't take care of them.  Hard disk data are similarly encoded because as a mechanical medium it is susceptible to physical wear and tear (although no amount of RS coding will help when the bearings die).

If you transmit data over a medium that's subject to change (wireless transmission, transmission over long distances, transmission over a medium that a user can disconnect/change) then error detection and correction makes sense.

On the other hand, if you're interested in getting data from processor A to memory B over a well-defined electrical backplane where operation is in accordance with all relevent specs and design parameters, then you can dispense with the overhead of ECC.  For example, it's always been possible to buy a computer that uses ECC memory.  In practice, most users don't bother, and it's not a problem.  Why?  Because as long as you use the memory specified by your motherboard manufacturer and you don't do something stupid like overclock, then you're operating correctly and you won't have problems.

People who DO have memory problems either use the wrong memory, or they've installed it incorrectly.  In some rare instances, some memory locations may go bad because of a static zap.  When this happens, the whole system tends to lock up; ECC doesn't do much good when the results of a memory read are all 0xFF for all bytes and parity bits.

Quote:

No I am not an EE, just deal with them on a daily basis.


You seem to hold them (errr, us, as both Jon and I, and Dan, and others, are EEs) in contempt.

What's your background?

-a
Logged
"On the Internet, nobody can hear you mix a band."

danlavry

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 997
Re: a £400 digital interconnect
« Reply #43 on: September 05, 2006, 01:38:38 PM »

Bubblepuppy wrote on Sun, 03 September 2006 13:57

Firstly I am an EE, and a Software Engineer

I wont hold that against you I promise.

just because it's new to you doesn't mean that engineers haven't known about it and designing with it for a long time.

Its not new to me, its just enginners think the lab is the real world.

Finally, your position is MADI & AES digital communications never have errors?

I give up, if its written in the spec then it must be true....
I think I will sell my new MAC dual core intel go find a 386 and live with the blazing speed of 100Mbits that never has any errors and more than eneough speed to meet all my audio needs.

PLEASE.




I am back from a long weekend, and I see that the subject has changed completely. Instead of dealing with the comment about the effect or lack of effect of temperature on attenuation, instead of acknowledgment that the load DOES matter, instead of agreement that for audio frequency applications (under say 50KHz) the load resistance is higher then 600 Ohms (often 10KOhms) and that for 25MHz signals, which is most of digital audio interconnect, you chose to make a sharp turn, and talk about future formats that raise the frequencies to 1GHz.

Well, most readers that are not technical, would need to have a CLEAR DISTINCTION that will enable them to know that as long as they are dealing with analog audio or AES or SPDIF the temperature DOES NOT MATTER to cable attenuation. I am doing my part - I keep stating the parameters (less then 100 feet and under 25MHz) and that is a lot of margin!

Instead of conceding in a clear way that what I said is correct under the conditions (or that as a non EE you are in no position to know it), you keep arguing a different point, so in some indirect sense, you are "dragging" the argument to a different area where you may be correct. In GHz applications the skin effect is important, and so are other factors.

There have been some systems designed to distribute a lot of audio over high speed links, from Cat5 to optical... Some have been better then others. The engineering challenges of high speed communications will not be solved by the audio industry, and any audio engineer that want to send audio or anything else over the Internet will most likely use cables that meet the standards of Internet communications.

But coming here to an audio forum with a general claim that temperature rise will effect cable attenuation, INITIALLY WITHOUT A STATEMENT ABOUT THE VERY HIGH FREQUENCIES is way off. It is only AFTER I wrote my post that the issue does not exists at lower frequencies, that you started talking about GHz frequencies.

Again, in the cases that are of Daly interest to the audio people, such as cables for analog line levels and frequencies, or mic cables, or AES cables or SPDIF cables, the cable;e temperature is NOT a factor effecting attenuation!!!

When an audio person will need to send GHZ of data accross a network, they will resort to cables that are appropriate for a network. You can argue about skin effect, temperature and more for such applications. But a general statement on an audio forum that a cable temperature will cause attenuation creates the false impression that audio people should worry or consider the temperature of their cables. There are some important cable characteristics that are important factors for audio, for the sound itself. Temperature is NOT one of them.

Dan Lavry
http://www.lavryengineering.com

   

Logged

Jon Hodgson

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1854
Re: a £400 digital interconnect
« Reply #44 on: September 05, 2006, 01:57:59 PM »

Yannick Willox wrote on Sun, 03 September 2006 12:02

As an end user, I did not invest in MADI (although I really need a multichannel digital carrier that is robust enough for live use), just because of these reservations.

The cables that are sold are not OK. I was on the point of buying two MADI/AES converters, but was not able to source non-plastic cable from audio dealers. Contacting the constructor did not help. I had to contact a network company (eg Black Box) to have the glass cables constructed and test myself (!). This was for distances above 50m.

I would say either it is under engineered, or the real world performance is neglected.
.

What is your reasoning for this? Do you have any figures that show that when used with cables of the required spec, be they plastic or glass, that it doesn't do what it's supposed to?

Quote:

Of course 50m with plastic would work, but then you start using 32 ch at 96K, what happens ? On a bad day - in a live situation ?

Whether you are transmitting one channel at 32kHz or 32 at 96kHz the situation is the same. MADI is a fixed data rate transmission, what changes is how much of that data is filled in the transmitter and used at the receiver.

Quote:

I do not want to be bothered by this, and I have the impression manufacturers do not like to be bothered by this either. They think safe/studio/fixed/short installations.


Actually they think of a balance between different performance objectives and cost, with presently available technology (MADI was specced around 89). For the same cost they could possibly have guaranteed transmission over longer distances... but with fewer channels, or they could possibly have given it more channels, but with shorter distances. They might possibly have been able to improve either without reducing the other, but then the cost would have gone up, and bear in mind that when MADI was introduced it was already really expensive, it took many years before it became viable as a project studio format.

Quote:

Also, does MADI not carry the clock ? If I understand correctly, the interface gets near its theoretical maximum when using all channels - so what happens to the CLOCK in this case ?

MADI doesn't carry the clock, you need a seperate wordclock or AES connection for that.

Quote:

Did anybody do some tests on a MADI interface with all channels sending/receiving data ?


Yes, everyone who's ever used it. As I said MADI always transmits the same amount of data at the same rate, what changes is the content of that data.

Quote:

Yes you can still use wordclock, but doesn't Bubblepuppy have a point, if it had been overengineered instead of just enough bandwith engineered, there would be no potential problem.


If it was overengineered it would never have been viable at the time it was introduced, they needed a solution that could be implemented then, not one that they would have to wait years before it became financially and technically viable.
Logged
Pages: 1 2 [3] 4  All   Go Up
 

Site Hosted By Ashdown Technologies, Inc.

Page created in 0.09 seconds with 18 queries.