roffe skrev:
This must be a flawed logic. If you give the reflections "enough of a round trip", you may well be influencing the timing of the next bit instead. The best solution in my book is either perfect impedance matching to avoid reflections or good buffering and reclocking circuitry in the DAC, which would pretty much make the DAC immune to jitter variations. My money is on the latter, which would also allow for simple and inexpensive cable solutions.
But I suppose you can create a jitter-minimizing cable, even if it is not necessarily based on the technology an logiq presented here. The question is how much you get for your money compared to the inexpensive cable solutions and a good reclocking DAC.
I would also like to see your rationale for claiming a bandwidth requirement up to 100MHz for spdif.
No flawed logic at all!
Let me explain something about reflections; what they are and how they work.
Let us assume that we have an ideal 75 ohm source, that is 75 ohms from "DC to light". And we connect up an ideal 75 ohm coax. (Can't really say "DC to light", as talking about characteristic impedance of a coax below several hundred kHz is pointless.)
When the signal reaches the RX end, it will most likely not meet the same "DC to light" qualities. Most of the signal will be absorbed, but a small part will not. That part will be reflected back to the source. The amount that is reflected is predictable.
Let us say that the RX is 100 ohms. (A very popular D/A box did this. I asked the designer why, and his response was "Hey, 100 ohms is what we had that day. Most guys leave it unterminated. Give me a break; we are a step in the right direction.")
Ok, the reflection coefficient, called
rho is:
(100-75)/(100+75) = 0.143
That translates into -16.9 dB. Not that great. Keep in mind that there will be some stray reactance to take into account, so in practice it will be worse.
Anyway, we now have around 14% of the original signal bouncing back to the source. Well, if we have a perfect 75 ohm cable and source, all of the reflection will be absorbed, and there will be no further reflection.
Of course, that is not reality. So, let us say that it also has the same rho of 0.143. 14% of 14% will then be reflected back to the RX. Delayed in time by the propagation time of
twice the cable length.
If the cable is very short, you could end up 2% of the original signal arriving a few nSec later.
Ok, 2% is obviously not enough to affect whether it is a 1 or a 0.
But it is enough to affect the timing of the decision point. And this is what matters.
So, what do you do about it???
1.) Design a reclocking circuit to get rid of that problem. Guess how many do that. Hint: not many.
2.) Make a perfect impedance match. Yes, that helps. But if you ever measure the impedance of any DAC or transport, you will realise that is also a pipe dream. Almost no one does that. They treat SPDIF just as they treat inputs and outputs in a preamp. Mystery wire, steps to eliminate EMI (necessary in most commercially produced units), and just plain bad engineering is what you will find. If you lived in Texas, I would gladly measure your gear. You would not be pleased with the results. Maybe you should be thankful that you are far away from our lab.
3.) In lieu of those steps, since you are going to need to buy some sort of cable, why not buy one that costs the same (or even less!) than the dreck already on the market? There are cables on the market, labeled as being 75 ohms, but are really 50 ohms! (Yes, they exist. Yes, I have measured them. No, I am not naming them.)
OK, you asked about BW. Let us say that you agree that a very sharp rise time is needed to prevent those nasty reflections from mucking things up. (Faster rise time, sharper slope, less chance for mucking up timing.)
If you have a 7 nS rise time, the -3 dB point (assuming a single-pole network), you get 0.35/(7^-9) = 50 MHz. Cut the rise time in half, and the BW is now 100 MHz. Yes, it will generate tons of EMI. But, there are units around that are very fast. So, 100 MHz is not that outlandish of a claim.
OK, there is a lot less energy in that part of the signal. Still, there is some. Since it is much harder to get a decent rho (at 100 MHz) in a circuit that some only feel needs to be "somewhat resistive" to only 10 MHz or so.......well, take my word for it........try making one sometime. It is not as easy, especially if you stick a transformer in the mix. (Read my posts on AC to see how successful some of my attempts are. I know that few units on the market come close.)
The problem is further complicated by the fact that damn near every unit out there uses a Crystal receiver. Guess what.........try measuring the rho of their stock circuit. The one damn near every copies right out to the data sheet. Guess what......measure one, and if your DAC does not have some fancy reclocker, you will think our cable is the best gadget to come along in quite a while.
Briefly on "peer-reviewed" papers, such as one will find in the JAES.
JAES has an agenda. It protects its sacred cow members, primarily in industry. (This trend started under Ray Dolby's watch.) Briefly, how many here have equipment that has at its heart design principles based on the works of John Curl, Walt Jung, Marshall Leach, and some guy named Matti Otala.
Anyone care to guess how many JAES papers have been written that claim those 4 designers are all wrong? All papers peer-reviewed, and published in the JAES.
Sorry, but the fact that they have been "peer-reviewed" does not hold water with me. If all of those papers are correct, many of you will have to throw out a lot of your gear.
The same holds true of the papers that SC submits. Nothing but advertising, artfully hidden amidst a lot of measurements. I know for a fact that I can rig a test to show that most any transformer is better than all the rest. So much for "peer-reviewed". You may know the old saying: "To catch a crook, you need to hire a crook."
Well, I am not President Nixon, but I know how to concoct a SPDIF test to yield most any result that you would like. Again, look at my AC posts.
Ok, my fingers need a break, and I am sure that you want a break from me. Until "tomorrow"..........ok, tomorrow morning in Texas.
AR-T
