Return LossReturn Loss can be calculated. Here’s the formula:
-20 log * difference/sum
That gets you the value in dB. Here’s what this means in terms of loss:
10 dB = 0.50 dB additional loss
13 dB = 0.22 dB additional loss
15 dB = 0.14 dB additional loss
18 dB = 0.08 dB additional loss
21 dB = 0.04 dB additional loss
23 dB = 0.02 dB additional loss
26 dB = 0.01 dB additional loss
Now half a dB may not sound like a lot, but consider that a 48-kHz AES signal has a bandwidth of 6.144 MHz. Category 5 has an attenuation of 5 dB/100m at 6 MHz. So a half a dB is an additional 10-percent loss in signal strength, or a 10-percent reduction in cable distance.Every so often, someone sends me an e-mail or corners me at a trade show to tell me their weird story about wire. Sometimes, a certain kind of wire worked in a situation where it shouldn’t.
For instance, I once gave a talk about using Category 5 (computer) cable for running RS-422 or RS-485 control applications. After it was over, a number of engineers took me aside, one by one. Each one whispered, with almost a guilty smile, that they had been using Category 5 for these applications for years. And they didn’t know why or how it worked, but it did.
At least they didn’t know why until my talk. They were all proud of their discovery, now that the truth had been revealed.
In the same way, I was surprised to read an article by Eric Hoehn in the Aug. 2, 2000, issue of this newspaper titled, "WETA: All-Synchronous Digital," describing the design of an all-AES digital radio facility. In this case, the entire system was based on the AES-3id standard that runs digital audio down coax cable.
Now, there are dozens of TV stations wired up with digital audio on coax. But for them it makes sense, because they already use coax to run video. In fact, they can use the same cable to run both digital audio and video (although the quality of cable used for video is overkill for digital audio).
Then we have the recording studio, which is unquestionably the "high end" of audio. There are a few recording studios wired up for digital audio on coax, most notably Hollywood Digital in (where else?) Hollywood, Calif.
But the RW article was the first time I have read of an all-coax digital audio install for radio. And that is probably because radio has been a twisted-pair audio world since the beginning.
So it’s natural that most radio engineers would think twisted pairs for digital too. And why not? The original AES/EBU standard was twisted pairs, although quite a different animal from the analog twisted pairs. With a characteristic impedance of 110 ohms and very low capacitance, these cables are more data cables than audio cables. But then, the digital audio signal is more data than audio.
The advantages to using coax are very long runs and very small connectors (BNC) compared to XLRs or other connector options.
In fact, you can order some digital audio consoles with all BNCs. You can imagine the space savings on that back panel, not to mention savings in weight. And BNCs can be connected to coax cable a lot faster than XLRs can. Their performance, even into the gigahertz, is common knowledge.
The disadvantage to coax and BNCs is the loss of a balanced line with its common-mode noise rejection. On the other hand, digital signals are inherently noise "resistant," because noise often can be filtered out, leaving the data untouched.
So should you go with coax for audio? Let’s just say it’s an option you might consider when you get to that point.
In the same article, Hoehn mentions using Category 5 cable for digital audio. I’ve mentioned this in previous columns; let’s look at this in greater detail.
The AES/EBU spec for balanced line cables requires a characteristic impedance of 110 ohms +/- 20 percent. That means a cable between 88 and 132 ohms should work fine. Category 5 cable is specified as 100 ohms +/- 15 ohms, or a range of 85 to 115 ohms. You will note that, unless the Category cable is at the very low end of its allowed tolerance, it will fit well into the required spec for digital audio. And there are bonded-pair Category cables with much tighter tolerance that fit easily into the digital audio requirement.
So why not use Cat 5 for digital audio? Hoehn mentions one reason: impedance mismatch. This mismatch leads to "return loss." Transmitter engineers will recognize return loss as VSWR. At high frequencies, this is an impedance discontinuity that reflects the signal back to the source (that’s the return), and looks like attenuation at the other end (that’s the loss).
So is there "return loss" when you use a 100-ohm cable (Cat 5) on a 110-ohm device (AES/EBU digital audio)? Sure. But calculations show the return loss to be a miniscule —26 dB loss. (See table.) Not a whole lot to worry about. And this assumes that the source and destination devices, and cable itself, have exactly the stated impedance, which they often do not.
How about the worst case, putting 85-ohm twisted pairs on a 132-ohm digital audio device? Then you have a serious return loss of 13.3 dB.
More likely, the digital audio devices probably are closer to 110 ohm outputs, especially if they are active balanced inputs or outputs. In that case, return loss of a worst-case Category 5 (85 ohms) would be 17.8 dB.
For these digital applications, because you have no way of knowing just where your Cat 5 cable would be in impedance, your best bet would be to use Category cables with tighter tolerances than generic versions. Tight-tolerance bonded-pair Category 5e cables have typical specs of 100 ohms +/- 7, a lot better than generic Category 5. You can see their advantage. Their worst-case return loss (110 ohms vs. 93 ohms) would then be a return loss of —21.5 dB.
Have you ever used a cable in a weird application, perhaps one that all your engineering friends insisted wouldn’t work? I’d love to hear about it. In my next column, I will regale you with some of the stranger stories of wire and cable I’ve experienced. Some of these never have been explained, even by "experts." Maybe you can solve the mystery!
Tune in next time.
Return LossReturn Loss can be calculated. Here’s the formula: