R/E/P > Dan Lavry
Why longer is generally better for an S/PDIF Digital Cable
Andy Peters:
lfeagan wrote on Wed, 19 July 2006 18:52
Dan, thanks for reading the message. Though, by no means did I intend for you to take your time to reply. It was only meant to dispel the incorrect belief that ATA did not have a minimum length specification.
I suppose I spoke off-the-cuff regarding ATA cable length specs, as I had not read the spec. The correction and the link to the spec are appreciated.
-a
Andy Peters:
lfeagan wrote on Wed, 19 July 2006 01:32
Not that I agree with audioengr as Dan et al. are quite correct about the length. However, there is a minimum length on ATA cables. For information on this, I suggest reading the http://www.t13.org specs. The T13 Technical Committee are the people who define ATA. In their minutes someone asked about this. Their response:
"The minimum length was set to prevent problems associated with reflection. There is also a maximum distance between two drives at the end of the cable. This requirement is present to solve impedance and capacitance issues."
http://www.t13.org/docs2002/e02144r0.pdf
The older ATA-3 standard working draft has a lengthy discussion of the issues with reflection they faced in the development of the standard. http://t13.org/project/d2008r7b-ATA-3.pdf
OK, I've had a chance to read the spec and the various discussion. First and foremost, the draft spec referenced above (from 1997) actually does specify cable length -- on page 7, in section 3.2, "I/O Cable," where we are told that "Cable total length shall not exceed 0.46 m (18 in)" and that "Cable capacitance shall not exceed 35 pf." Nowhere in the (draft) spec is there something that says, "cable length shall be minimum x m (y in)."
Yes, there is considerable discussion in Appendix C about signal integrity, most having to do with how they're kinda stuck with an interface original implemented using old-skool LS TTL, a technology whose (by modern standards) slow rise time and wimpy drive meant that you had no problem driving an 18" cable at 12 MHz or whatever. Move to faster edge rates, and you've got problems; most of the discussion involved using an 18" cable. One of the goals was to maintain backward-compatibility, so suggestions like using Thevenin termination at both ends, a-la SCSI, were rejected.
The meeting minutes in T13/e02144r0, from 2002, reference a discussion about minimum cable lengths, and in fact the Apple rep wanted to know where a minimum-cable requirement originated.
A later draft (ATAPI-5) tells us, in section D.3.3, that we shouldn't exceed the 18" length, but also doesn't specify a minimum cable length. I think that any discussion of minimum cable lengths, in the case of ATAPI, is in the context of workarounds to ensure that very old disks function with brand-new controllers (and the converse).
-a
danlavry:
Andy Peters wrote on Fri, 21 July 2006 08:58
lfeagan wrote on Wed, 19 July 2006 01:32
Not that I agree with audioengr as Dan et al. are quite correct about the length. However, there is a minimum length on ATA cables. For information on this, I suggest reading the http://www.t13.org specs. The T13 Technical Committee are the people who define ATA. In their minutes someone asked about this. Their response:
"The minimum length was set to prevent problems associated with reflection. There is also a maximum distance between two drives at the end of the cable. This requirement is present to solve impedance and capacitance issues."
http://www.t13.org/docs2002/e02144r0.pdf
The older ATA-3 standard working draft has a lengthy discussion of the issues with reflection they faced in the development of the standard. http://t13.org/project/d2008r7b-ATA-3.pdf
OK, I've had a chance to read the spec and the various discussion. First and foremost, the draft spec referenced above (from 1997) actually does specify cable length -- on page 7, in section 3.2, "I/O Cable," where we are told that "Cable total length shall not exceed 0.46 m (18 in)" and that "Cable capacitance shall not exceed 35 pf." Nowhere in the (draft) spec is there something that says, "cable length shall be minimum x m (y in)."
Yes, there is considerable discussion in Appendix C about signal integrity, most having to do with how they're kinda stuck with an interface original implemented using old-skool LS TTL, a technology whose (by modern standards) slow rise time and wimpy drive meant that you had no problem driving an 18" cable at 12 MHz or whatever. Move to faster edge rates, and you've got problems; most of the discussion involved using an 18" cable. One of the goals was to maintain backward-compatibility, so suggestions like using Thevenin termination at both ends, a-la SCSI, were rejected.
The meeting minutes in T13/e02144r0, from 2002, reference a discussion about minimum cable lengths, and in fact the Apple rep wanted to know where a minimum-cable requirement originated.
A later draft (ATAPI-5) tells us, in section D.3.3, that we shouldn't exceed the 18" length, but also doesn't specify a minimum cable length. I think that any discussion of minimum cable lengths, in the case of ATAPI, is in the context of workarounds to ensure that very old disks function with brand-new controllers (and the converse).
-a
Thank you for the report.
My point is – one can do a good clean job when one is familiar with the principles of electronics. The issue of cable length in context to cable reflection should first and foremost take into account the ratio between the signal rise time and cable length (cable travel time, we call it cable delay). Reflection issues need to be taken care of when the rise time is fast in relation to the cable length (cable delay). Reflections become a non issue when the rise time is very slow compared to cable delay. In other words, a shorter cable is always good from reflection stand point.
We do not terminate audio cables (for analog signals to say 100KHz), because the rise time of audio is very slow compared to even hundreds of cable feet. On the other hand, we are careful to terminate even a few feet of say AES or SPDIF signals, where the rise time can be, say around 10-20 nanoseconds.
As a rule, shorter is better. One needs to have a specific reason to go longer, a reason that has nothing to do with simple transfer of signals from point A to point B. It is “lucky” for us that we do not need to terminate audio analog cables. Imagine what a mess it would be: An AES signal (34.72V peak to peak) into say 110 Ohm cable would require about 0.3A current drive. Without termination, say the load is 10KOhm, and then the peak current is only 3.472mA.
Question: What about a 10KOhm cable impedance? Will that not be the solution?
Answer: First, there is no such a thing as 10KOhm cable. There is no 1KOhm, nor 200 Ohm… Practical cables impedance is in the range of say 30Ohm to 180Ohm (and that is already as large a range as possible). The impedance is a log function, so a 300 Ohms impedance will take 2 very thin wires separated by a distance greater then earth to the moon!
Question: What about driving all audio signals with a power amplifier (capable of 34.72V peaks into say 100 Ohms)?
Answer: Yes that would work, if the objective were to drive 100Ohms. In fact, we do use power amp to drive low impedance, such as 8 Ohm speakers.
BUT: We do not need to terminate audio cables, so we don’t. We do not need to terminate audio signals because they are slow relative to the cable length we use for audio signals. Take the SAME cable and run a much faster signal on it, and you may need to terminate. Alternatively, instead of terminating, make the cable very short, and the reflection problem goes away.
Those basic principles will not go away just because someone is trying to sell some expansive 1.5 meter as better then 1 meter…
Regards
Dan Lavry
Navigation
[0] Message Index
[*] Previous page
Go to full version