4 ms·

Why on earth is so much more expensive than an arbitrary pair of devices that could max out a cable?

ChrisMarshallNY 2mo ago
Think about it. You need to generate a precise, calibrated, 40GHz signal, transmit it, then precisely measure the returned signal for phase, amplitude, and waveform integrity -at that speed. Oscilloscopes and logic analyzers that run at that speed, are damn expensive.
- CamperBob2 2mo ago
  Someone could do an equivalent-time sampling instrument for not much money, if the market were there. It's probably not, though.
  - ChrisMarshallNY 2mo ago
    Time sampling isn’t enough. You need to at least look at waveform integrity (which can be approximated, by doing things like looking for harmonics). For example, you may send in a 40 GHz square wave, and get back a wave full of ringing, or a sine or sawtooth wave. It’s damn difficult to maintain clean square waves, at that frequency. The cable could also double as a microwave antenna. Since it’s digital information, some waveform alterations are OK, but not too much. You need fairly clear state transitions. Ringing, or waveform distortion, can introduce extra (or fewer) transitions; thus, completely changing the data.
    - CamperBob2 2mo ago
      That's not really what I mean. You don't have to diagnose the signal, just the medium, which can tell you what will happen to the real signal. It requires a fast edge (or a very broadband sweep), but nothing much beyond that. E.g., here's what a bad connector at the far end of an SFF-8654 cable looks like on a TDR: https://i.imgur.com/AiLnSuC.jpeg https://i.imgur.com/AiLnSuC.jpeg . This type of plot can be interpreted as impedance versus distance along the line. It could be made with a small handheld instrument, but currently a human is needed to read the tea leaves. While I was making that measurement, I couldn't help thinking that it would be nice if it didn't require either a 50 GHz VNA or equally-expensive training to read. Equivalent-time sampling gets you out of needing an expensive network analyzer, but it doesn't help with the diagnosis itself. And of course when checking a USB cable you are most likely just after a go/no-go result, not the gory details. That further calls the market economics of such a gadget into question. Still, if somebody offered one at a reasonable price, they'd sell at least one to me.
      - ChrisMarshallNY 2mo ago
        Fair 'nuff. As long as we don't get false positives. They are worse than false negatives, and is what happens, when you just read the markers from the chip. As we know well, lots of dodgy C.M.O.T. Dibbler-types will deliberately program the chips to emit bogus markers.
- roaringrocky 2mo ago
  But for bandwidth can't you just have two simple computers at either end that can send a 100GB file and if it arrives in 20 seconds, you know you have 40gbps of bandwidth?
  - traceroute66 2mo ago
    > But for bandwidth can't you just have two simple computers at either end that can send a 100GB file and if it arrives in 20 seconds, you know you have 40gbps of bandwidth? Because that is not reproducible and introduces all sorts of other random variables to your test. Expensive equipment is the only answer here. Its a bit like when you ask an electrician to install network cable (yuk !). They consider their job done after a continuity test (if they bother to do a continuity test at all). Meanwhile, if you put a Fluke on it, you will pick up all sorts of subtle faults that the electrician's continuity tester will never pick up, but will cause subtle and hard-to-diagnose problems when you try to push data over the cable. Its the same thing with all those cheap USB testers. They might have all sorts of fancy displays, but in the end they are no substitute for the real thing and will end up lulling you into a false sense of security.
    - criddell 2mo ago
      If you buy decent cables from reputable companies (CableMatters, LTT, Apple, etc...) you could end up with a drawer full of high quality mystery cables. A cheap tester is probably all you really need to sort out that mess.
    - chainingsolid 2mo ago
      You mention a Fluke, which I'm assuming is in relation to an ethernet cable tester (made by Fluke). I've seen/used continuity testers, but not something better (I've been using iperf3) Could you link/include the sku/model number of what you are refering too?
      - fragmede 2mo ago
        https://www.homedepot.com/pep/Fluke-Networks-LIQ-100-LinkIQ-Cable-Network-Tester-LIQ-100/335131563 https://www.homedepot.com/pep/Fluke-Networks-LIQ-100-LinkIQ-... Only $2,500!
- LorenPechtel 2mo ago
  But do you actually need that at the consumer level? Make some properly formatted garbage, send it down the wire to the other end of the tester and see if you can read the result. It won't tell you if the cable is marginal, but how often is that actually an issue? Network cables are connected to hardware that retries, masking such errors.
LiamPowell 2mo ago
Testing in that way will only tell you if the two devices work with the given cable. It won't tell you what sort of margins you have. Not all devices are made equal. A device isn't even necessarily better if it works with one cable when another doesn't, it might just happen to have a slightly lower impedance on its internal traces that happens to better match an out of spec cable, or any other one of a number of parameters.

Think about it. You need to generate a precise, calibrated, 40GHz signal, transmit it, then precisely measure the returned signal for phase, amplitude, and waveform integrity -at that speed. Oscilloscopes and logic analyzers that run at that speed, are damn expensive.

CamperBob2 2mo ago
Someone could do an equivalent-time sampling instrument for not much money, if the market were there. It's probably not, though.
- ChrisMarshallNY 2mo ago
  Time sampling isn’t enough. You need to at least look at waveform integrity (which can be approximated, by doing things like looking for harmonics). For example, you may send in a 40 GHz square wave, and get back a wave full of ringing, or a sine or sawtooth wave. It’s damn difficult to maintain clean square waves, at that frequency. The cable could also double as a microwave antenna. Since it’s digital information, some waveform alterations are OK, but not too much. You need fairly clear state transitions. Ringing, or waveform distortion, can introduce extra (or fewer) transitions; thus, completely changing the data.
  - CamperBob2 2mo ago
    That's not really what I mean. You don't have to diagnose the signal, just the medium, which can tell you what will happen to the real signal. It requires a fast edge (or a very broadband sweep), but nothing much beyond that. E.g., here's what a bad connector at the far end of an SFF-8654 cable looks like on a TDR: https://i.imgur.com/AiLnSuC.jpeg https://i.imgur.com/AiLnSuC.jpeg . This type of plot can be interpreted as impedance versus distance along the line. It could be made with a small handheld instrument, but currently a human is needed to read the tea leaves. While I was making that measurement, I couldn't help thinking that it would be nice if it didn't require either a 50 GHz VNA or equally-expensive training to read. Equivalent-time sampling gets you out of needing an expensive network analyzer, but it doesn't help with the diagnosis itself. And of course when checking a USB cable you are most likely just after a go/no-go result, not the gory details. That further calls the market economics of such a gadget into question. Still, if somebody offered one at a reasonable price, they'd sell at least one to me.
    - ChrisMarshallNY 2mo ago
      Fair 'nuff. As long as we don't get false positives. They are worse than false negatives, and is what happens, when you just read the markers from the chip. As we know well, lots of dodgy C.M.O.T. Dibbler-types will deliberately program the chips to emit bogus markers.
roaringrocky 2mo ago
But for bandwidth can't you just have two simple computers at either end that can send a 100GB file and if it arrives in 20 seconds, you know you have 40gbps of bandwidth?
- traceroute66 2mo ago
  > But for bandwidth can't you just have two simple computers at either end that can send a 100GB file and if it arrives in 20 seconds, you know you have 40gbps of bandwidth? Because that is not reproducible and introduces all sorts of other random variables to your test. Expensive equipment is the only answer here. Its a bit like when you ask an electrician to install network cable (yuk !). They consider their job done after a continuity test (if they bother to do a continuity test at all). Meanwhile, if you put a Fluke on it, you will pick up all sorts of subtle faults that the electrician's continuity tester will never pick up, but will cause subtle and hard-to-diagnose problems when you try to push data over the cable. Its the same thing with all those cheap USB testers. They might have all sorts of fancy displays, but in the end they are no substitute for the real thing and will end up lulling you into a false sense of security.
  - criddell 2mo ago
    If you buy decent cables from reputable companies (CableMatters, LTT, Apple, etc...) you could end up with a drawer full of high quality mystery cables. A cheap tester is probably all you really need to sort out that mess.
  - chainingsolid 2mo ago
    You mention a Fluke, which I'm assuming is in relation to an ethernet cable tester (made by Fluke). I've seen/used continuity testers, but not something better (I've been using iperf3) Could you link/include the sku/model number of what you are refering too?
    - fragmede 2mo ago
      https://www.homedepot.com/pep/Fluke-Networks-LIQ-100-LinkIQ-Cable-Network-Tester-LIQ-100/335131563 https://www.homedepot.com/pep/Fluke-Networks-LIQ-100-LinkIQ-... Only $2,500!
LorenPechtel 2mo ago
But do you actually need that at the consumer level? Make some properly formatted garbage, send it down the wire to the other end of the tester and see if you can read the result. It won't tell you if the cable is marginal, but how often is that actually an issue? Network cables are connected to hardware that retries, masking such errors.

LiamPowell 2mo ago

Testing in that way will only tell you if the two devices work with the given cable. It won't tell you what sort of margins you have. Not all devices are made equal. A device isn't even necessarily better if it works with one cable when another doesn't, it might just happen to have a slightly lower impedance on its internal traces that happens to better match an out of spec cable, or any other one of a number of parameters.