Comment by vel0city
2 days ago
I'm just meaning people here are looking for a "does it work", a binary yes/no. Not a precise measurement of exactly how perfectly the digital signal is making it through the wire. They don't care exactly what the SNR is or whatever, they just want to know when they plug in the cable for a given need will this cable work?
You don't need a six figure oscilloscope or logic analyzer to know "does it work".
They can determine that today without extra parts or test equipment at all: They plug the cable in to the equipment they already have, and it either works in the way they care about or it does not.
If that's the extent of the thing that they want to do, then there is nothing further to discuss: They already have everything they need to begin that kind of evaluation immediately.
It doesn't give them any viable forecast as to whether the cable will work or not with some other bits of gear, though, or on some other day. It just evaluates it as a binary pass/fail for right now.
"Digital" cables are often dismissed as simple things, wherein: A string of bits goes in one end of a random collection of copper wires or metal coat hangers or whatevers, and it always comes out the other end the same way. After all, it's digital.
But reality isn't that way at all, and there's a lot that goes into a reliable cable that connects things together. Characteristic impedance is only one of the things[1].
---
Interrogating the emarker doesn't provide further practical information, either. The information that can come from there is similar to what wording on the package on the shelf at the store is: Both things merely document what someone else once said about what the cable should be capable of doing.
Like a printed URL vs a QR code that each lead to the same web site: It's the same information, differing only in presentation. One of these presentations is human readable and other is not.
And neither thing neither thing tells the user whether that website is still reachable or whether it fell off the face of the earth, or whether it is friendly or evil.
It's all just a fixed string of information and that information can be wrong. It can even be deliberately wrong. The words on the box, the bits burned into the emarker chip, the URL, and the QR code: In the best case, they may or may not lead to a useful outcome.
In the worst case, they can be lies.
emarker chips set a hard not-to-exceed performance limit, but they don't provide a promise of performance below that limit.
---
[1]: Like the RG-6 cable connecting a cable modem to the wall: It functions with a very specific characteristic impedance of 75 Ohms. Or the SATA cable in the desktop computer of someone with some Linux ISOs: That's twinaxial cable, with its center pair of straight, parallel wires and the shielding all interacting with eachother at a controlled impedance. USB 4, meanwhile, uses cables with a characteristic impedance of 90 Ohms.
Characteristic impedance can be complex, and even dynamic. Like the ye olde POTS phone lines, for voice audio: That was designed and built very specifically in order to behave itself within those seemingly-minor constraints. 600 Ohms at 1 KHz was the target, and that was good enough for a very long time. But when frequencies increase into the MHz range with things like DSL, that characteristic impedance tends to drop to somewhere in the realm of 100 Ohms.
100 Ohms is, not coincidentally, around the same characteristic impedance that our Ethernet cables seek to provide at the frequencies we care about in that space. :)
Whatever that characteristic impedance is (at whatever frequencies are being used), when it matches what the termination impedance of the two endpoints are then we get an ideal transmission line. (That's always the target; an ideal transmission line is a glorious thing, indeed.)
But bad impedance match results in not just seemingly-obvious things like attenuation; it also introduces things like reflection -- signals bouncing back and forth, causing temporal distortions that obscure the desired signal. Bad matches also result in standing wave patterns where the length of the wire makes a difference, but shorter isn't better: The pattern of these standing waves within a wire are a function of the wavelength of the signal, not the overall length of the cable.
It's easy to make any of those things work within their designed uses and designated systems, where all of the effort to get this stuff right happens behind the scenes.
It's much harder to replace with bits of copper wire from the junk drawer and get reliably-good results.
For 1KHz in modern-ish POTS lines, that wavelength is around 40km. That's easy-enough to hand-wave away for lots of uses. The impedance matching can all be completely screwed and it really doesn't matter much for short runs.
But at 12.8GHz (the frequency maximum of USB 4) in a USB cable, the wavelength is only around 16mm. It's kind of a big deal to seek to have the little parts exactly-right in these seemingly-innocuous, hard-to-see ways. :)
And the impedance changes in-use, too. A kink in the wire produces an impedance aberration at that point that can be measured, and that can also be bad enough to produce practical issues that affect real-world performance.
Bad enough to fail the binary test you described today, even though the test was completely successful yesterday.
I'm a ham and took a number of EE/electromagnetism classes in college, so I totally understand what you're talking about with wire characteristics. I agree there's a ton of hidden complexity on these high bandwidth cables we can buy in the discount bin at electronics store.
> They plug the cable in to the equipment they already have, and it either works in the way they care about or it does not.
That's the thing though, I don't always have these things handy or ready to just plug something in quick and test. I'd like some kind of small device which can actually do some quick tests and see if it passes or fails at whatever power and bitrate more than just what the little identifier says it can do. Because, I agree, you can't trust the tag, there's a lot going on. And a cable that tested fine a year ago might be too messed up to work today.
In the end, it's also sometimes messy testing bandwidth given two random devices. Are they just being buggy with their handshake? Are we getting strange driver issues? Something else going on? A more dedicated tester tool can just be sure these two things definitely, without question should work just fine at whatever bandwidth they're trying to test with. And with a button press it does it and clearly gives the results. If it's not too expensive and you end up handling lots of questionable cables, sounds kind of nice!
> And the impedance changes in-use, too. A kink in the wire produces an impedance aberration at that point that can be measured, and that can also be bad enough to produce practical issues that affect real-world performance.
This is still true of a six figure oscilloscope testing the wire. If you're not manipulating the wire in realistic ways your super expensive tool isn't really going to answer that any more as the cheaper pass/fail. What might look just fine on the scope may look really bad when you bend it wrong or it gets a little internal crack or too much outside interference or something. You'll potentially be clued in to it being closer to being a messed up signal, but not necessarily.
> I'm a ham and took a number of EE/electromagnetism classes in college
Excellent! I had to learn all of this background the hard way by working with things in the field and asking myself why things behaved in certain ways. Finding answers that fit functionally into a larger mental model was not always very fast.
I should get a ham license one of these days. On a whim with no preparation, I recently managed to pass a technician-class practice test well-enough that a real test would not have been a problem. But I got tripped up badly on some of the questions about packet radio because many of the most-correct answers (in absolutes) were considered wrong (in the perspective of the older-timers who wrote that part of the test).
I mean, I know of plenty about TCP/IP. As my own brand of old-timer, I even know of KA9Q's work in packet radio and I think I even used his software ~35 years ago when running PPP under MS-DOS for dial-up internet, since that was the style at the time. I found that part that part pretty frustrating.
The rest of the practice test all made sense. :)
> I don't always have these things handy or ready to just plug something in quick and test.
Me, neither. I don't have any of this higher-speed stuff to play with at all. If I scored a great deal on some future-proof-ish all-singing, all-dancing USB cables, I'd have no way to validate their performance today. They'd be placed in a Rumsfeld bucket of known unknowns until something else changed, like the appearance of any kind of equipment to test with.
> I'd like some kind of small device which can actually do some quick tests and see if it passes or fails at whatever power and bitrate more than just what the little identifier says it can do.
Agreed. Perhaps it can be built. It's conceptually simple to say that the high-speed receiver and transmitter widgets exist as general-purpose COTS chips that can just get used in a circuit with some kind of supervisory intermediary to task them with Doing Verifiable Stuff Real Fast and provide performance information. Like UARTs, say.
But reality doesn't seem to match that very well at all. They instead seem to be very special-purpose ASICs that work in strictly-defined ways. This part is a motherboard/CPU glue chip (whooo boy), while this other part connects to a display (and only usefully-operates in display-world). None of them seem to be general purpose when I've looked into it.
And, of course: If it were easy then everyone would already be doing it. We'd just pick up some proper ludicrous-speed cable testers on AliExpress for fifty bucks or something. :)
(I guess we could tape out our own ICs, but I'm not prepared to visit that rabbit hole.)
> You'll potentially be clued in to it being closer to being a messed up signal, but not necessarily.
The point I was driving at with more-abstracted analyzers is that they can conceivably go beyond a pass/fail test.
Like: An edge-case cable that already presents a kind of smeary mess on the hypothetical Lambo-scope might actually work perfectly well, right now, in real use -- and I really do mean perfect. But is more likely to fail in noticeable ways after its performance degrades a bit and/or with less-tolerant electronics at either end. When every-day use requires it to work at 10/10ths, there's no room for things to get worse.
Meanwhile: A superior cable that presents a much clearer picture on the Lambo-scope might only be operating at, say, 7/10ths in every-day use. It has some room for its performance to degrade, or for less-tolerant electronics to be used, and still behave absolutely perfectly in the real world. This cable will tend to have better longevity and compatibility.
So above-and-beyond performance matters, I think. If I could make an informed choice between the edge-case cable or the superior cable for the thing I toss into my bag before a trip, I'm definitely not choosing the one that I know to be an edge case.
But without the moral equivalent of a Lambo-scope, I don't know how to characterize this or otherwise discriminate between these two cables' characteristics to that level of detail.
We don't have the technology to inexpensively do analytical work at these frequencies.
And this may ultimately be unsolvable: If we did have the tech, then some clever person would adopt it immediately to make an even faster data bus for the world to use and we'd be right back where we started, where the bus is faster than our ability to analyze it. :)
1 reply →