Classic Case of Bad Advice

Page 2 - Seeking answers? Join the What HiFi community: the world's leading independent guide to buying and owning hi-fi and home entertainment products.
A

Anonymous

Guest
Hi Malcolm - fanatstic post, really interesting and educational for me!

Bloney
 
A

Anonymous

Guest
[quote user="MalcolmH"]A simple retransmission protocol might go: "data data data checksum". If checksum doesn't match some simple function of "data data data", then ask for a retransmission. However a more sophisticated protocol might ask for "data data data data data data checksum", where checksum would not only show if one of the data elements had been corrupted in transmission but you'd be able to calculate from the combination of the checksum byte and the data which byte had been corrupted and what the corrupted byte would have had to have been for the checksum to match. For a well-known example of this in the real world, that's what happens in RAID5 drive arrays, where the any one of (say) four hard drives can fail and yet the whole array can be rebuilt from what's left.[/quote]

But I think in video transmission there isn't time to retransmit. Once the packet's gone, it's gone. Therefore you would need some polynomial-based error-correction, either FEC or BCH, which HDMI actually uses. There is an interesting and accessible thread here on ArsTechnica: http://episteme.arstechnica.com/eve/forums/a/tpc/f/67909965/m/450005880831?r=363006290831

It's also worth noting that data coding is used to minimise errors at the physical level, as does Ethernet and many other data transmission protocols.

BCH (like FEC which I know a little about) can only rectify a certain number of errors and these errors should conform to a Gaussian distribution, i.e. not too many too close together. If I have understood correctly the data 'islands' to which error correction can be applied is fairly small, which in video terms is only a few pixels. Therefore I would expect that a small amount of data degradation would be visible in terms of miscoloured pixels, rather than gross blocking errors or a complete loss of signal.

Therefore, on that basis, I would expect the quality of cables to affect the visual quality to a noticeable degree in terms of 'noise' or visual distortion. I would not expect it affect 'global' quality parameters such as colour accuracy. If such parameters were affected I would say that something somewhere is badly implemented or faulty.

But then, unlike Claire and the gang, I don't spend hours comparing different HDMI cables so who am I to argue? It is probably beyond the scope of What Hi-Fi? but I'd like to see an in-depth comparison test with data integrity tests, just to see how different HDMI cables degrade the signal. I would not have thought that would be difficult to do - there must be such things as HDMI testers/analysers out there?
 
A

Anonymous

Guest
[quote user="Halloway"]But I think in video transmission there isn't time to retransmit. Once the packet's gone, it's gone.Therefore you would need some polynomial-based error-correction, either FEC or BCH, which HDMI actually uses. There is an interesting and accessible thread here on ArsTechnica: http://episteme.arstechnica.com/eve/forums/a/tpc/f/67909965/m/450005880831?r=363006290831

BCH (like FEC which I know a little about)

[/quote]

Brilliant, some real references!
emotion-1.gif


Admittedly, the HDMI spec itself is a too technical for me. I looked on Wikipedia for "Error Correction", and learnt that "FEC" is short for Forward Error Correction (what I was describing as an alterative to detection and retransmission), that BCH is a form of Forward Error Correction that is much studied within coding theory, and that CDs and DVD use another type of Forward Error Correction called Reed-Solomon (look here, Will!).

[quote user="Halloway"]BCH (like FEC which I know a little about) can only rectify a certain number of errors and these errors should conform to a Gaussian distribution, i.e. not too many too close together. If I have understood correctly the data 'islands' to which error correction can be applied is fairly small, which in video terms is only a few pixels. Therefore I would expect that a small amount of data degradation would be visible in terms of miscoloured pixels, rather than gross blocking errors or a complete loss of signal.[/quote]

That's not how I'd interpret the source you quoted. It says that data is sent over HDMI in data islands (blocks), and that it can correct a certain amount per block; anything more than that and it fails for that whole "data island". So I interpret this as meaning precisely the opposite, and what I said before, namely that a small amount of degradation is corrected to a perfect signal, but if it gets too bad for that you start losing big chunks of data.

Also, you refer to "miscoloured pixels". Does anyonne know what format the data is sent across the HDMI in? Is it a raster scan or similar (in which case it's fair to talk about pixels), or is it a higher level encoding such a MPEG2 or H.264? If the latter, no amount of corruption would affect "pixels" per se, because MPEG etc doesn't describe pixels, but rather broader "areas" covering multiple pixels, that have to be decoded down to a raster later in the decoding process.

[quote user="Halloway"]I'd like to see an in-depth comparison test with data integrity tests, just to see how different HDMI cables degrade the signal.[/quote]

How this test is done is critically important, if it's going to decide this debate. I certainly agree that you can stick an oscilloscope on the end of a cable and prove that one cable degrades the electrical/optical signal more than the other cable does. So what? What we're interested in is the result after error correction is applied, in other words, not does one cable make a difference that is measurable in a test lab, but does a cable make a difference to your TV. This means testing the cable as part of a system including the HDMI interfaces, not the just testing the cable as a physical item alone.

Unfortunately a fully objective system test is probably going to be very difficult to organise, and would certainly require specialist equipment.
 
A

Anonymous

Guest
louisv6 You asked for it, now you'll have a scientific method for evaluating HDMI cabling & accessories. It's called DPL(Digital Performance Level) Rating Program. Strictly based on performance not packaging, pricing, or brandname. All of the tests involved are focused on how the device performs over and above the HDMI minimums. Eye mask margin differentials, DDC integrity both with Eye Patterns and I2C dynamics just to name a few. For a lot more info you can contact them at www.dplrating.org. Good Luck
 

TRENDING THREADS