Twisted USB

[wraper]

The minimum USB speed is 1.5Mbits/s

https://en.wikipedia.org/wiki/USB_communications

So USB 2.0 can readjust from 480Mbps to 1.5 Mbps if the wiring is not good. Why did you guys say the speed is fixed?

Because you just wrote absolute BS despite being told multiple times how it is actually.

[wraper]

If the cable is very dodgy, it may look like HS is not supported and the link will work at FS.
If the cable is slightly dodgy, HS may be enabled and then fail to actually move data.

Only is USB sink device is specially designed to be able to reset interface on communication fail at HS and then restart without negotiating HS (not likely you'd find something like this).

[magic]

No, it can happen with any device. It may take tens of tries to get it right, but you can make any HS device work at FS by inserting the plug very slowly or at weird angle to make the contact poor enough. (It helps to watch dmesg in Linux in real time).

The reason it works is because a HS device first appears as FS and immediately sends a "high speed chirp" signal which old hosts/hubs ignore, but new ones recognize and respond to. If this signal doesn't reach the hub, the device will be enumerated as FS and every device which wants to support legacy hosts is prepared to deal with this.

In theory a device could refuse to work at FS at all, but I don't recall ever seeing such device. Even my USB 3.0 camera can fall back to FS with 320x240 resolution.

[wraper]

No, it can happen with any device. It may take tens of tries to get it right, but you can make any HS device work at FS by inserting the plug very slowly or at weird angle to make the contact poor enough. (It helps to watch dmesg in Linux in real time).

The reason it works is because a HS device first appears as FS and immediately sends a "high speed chirp" signal which old hosts/hubs ignore, but new ones recognize and respond to. If this signal doesn't reach the hub, the device will be enumerated as FS and every device which wants to support legacy hosts is prepared to deal with this.

In theory a device could refuse to work at FS at all, but I don't recall ever seeing such device. Even my USB 3.0 camera can fall back to FS with 320x240 resolution.

Well if you interrupt sending chirp signal, then yes. However I really doubt chirp won't go through just because of crappy cable.

[magic]

I think you're right, the chirps are supposedly 50μs in duration so it seems impossible to suppress them passively without breaking FS signaling as well.

So my dodgy insertion trick must be a matter of intermittency rather than resistance or anything like that.

[Ogitek]

I cut the USB 2.0 cables I own to see what is really inside in case there is a way to save $400 usb-lemo cable. But I found out none of them is even twisted!  So what you make of it? Fake 480Mbps and just FS of 12 Mbps?   







What branded USB cable do you know that is really twisted? have you open yours too?

In audio devices and ADCs. Do they really need 480Mbps, or do most run in 12Mbps? Remember this is already much faster than serial port of 115,200 baud (or 0.1152 Mbps)

[Ogitek]

I cut all my USB cables, even the flat one, and found out none is twisted. Just when I thought USB 2.0 was very slow and wanted USB 3.0 and just knowing now I haven't even tasted real 480Mbps. I have a Western Digital model that says USB 3.0. What software can I use to find out if it is running Superspeed of 5Gbps  and not USB 2.0 High Speed or even Full Speed?

[Ogitek]

I'm googling a certain information I can't find.

If the data lines are not twisted but only very short length like less than one meter. Can it still somehow run USB 2.0 High Speed of 480Mbps?

Maybe only when the cable is beyond certain length that twisting is necessary for High Speed? What is the magic length?

[Ogitek]

No, it can happen with any device. It may take tens of tries to get it right, but you can make any HS device work at FS by inserting the plug very slowly or at weird angle to make the contact poor enough. (It helps to watch dmesg in Linux in real time).

The reason it works is because a HS device first appears as FS and immediately sends a "high speed chirp" signal which old hosts/hubs ignore, but new ones recognize and respond to. If this signal doesn't reach the hub, the device will be enumerated as FS and every device which wants to support legacy hosts is prepared to deal with this.

In theory a device could refuse to work at FS at all, but I don't recall ever seeing such device. Even my USB 3.0 camera can fall back to FS with 320x240 resolution.

Can the manufacturer disable this on purpose, not able to negotiate at FS at all but only HS? Supposed you are manufacturing very expensive audiophile amplifiers and equipments, and you only want the customers to buy your $400 usb-lemo cable, and you didn't want any cable clone to duplicate it. Is it possible to only enable HS, or is the ability to negotiate to FS in every firmware that can't be bypassed or vetoed?
Remember Audiophile equiptment can cost more than $10000 and people fight over them at ebay.

[Andy Chee]

IBut I found out none of them is even twisted!  So what you make of it?

Whilst ethernet cable wire pairs have 5 twists per 1 inch, USB cables go as low as 1 twist per 5 inches.  So you will need to remove a whole lot more plastic sheath to reveal the twists.

Can the manufacturer disable this on purpose, not able to negotiate at FS at all but only HS?

If they did, the product would no longer be "USB compliant", completely defeating the term Universal Serial Bus.

[Ogitek]

IBut I found out none of them is even twisted!  So what you make of it?

Whilst ethernet cable wire pairs have 5 twists per 1 inch, USB cables go as low as 1 twist per 5 inches.  So you will need to remove a whole lot more plastic sheath to reveal the twists.

Can the manufacturer disable this on purpose, not able to negotiate at FS at all but only HS?

If they did, the product would no longer be "USB compliant", completely defeating the term Universal Serial Bus.

Are you saying all USB 2.0 cables have the data lines twisted? How about this. It's a flat all the way, there is no way to twist any wire:






So they are the only USB 2.0 cables that don't have a twist?  And it's a sure way to force any HS device to renegotiate at FS by suppressing the chirp?

Also any reference for the chirp signal sent that can heard over the other end?

[tooki]

IBut I found out none of them is even twisted!  So what you make of it?

Whilst ethernet cable wire pairs have 5 twists per 1 inch, USB cables go as low as 1 twist per 5 inches.  So you will need to remove a whole lot more plastic sheath to reveal the twists.

FYI, that’s a gross oversimplification. Neither the Ethernet nor cable standards actually specify the number of twists per inch, only performance targets (like bandwidth and crosstalk), with the manufacturer free to meet them however it wants. But the most important thing to remember is that each pair uses a different number of twists per inch, to reduce crosstalk by avoiding running in parallel regularly.

The number of twists per inch goes up with higher category cables, also.

[Andy Chee]

IBut I found out none of them is even twisted!  So what you make of it?

Whilst ethernet cable wire pairs have 5 twists per 1 inch, USB cables go as low as 1 twist per 5 inches.  So you will need to remove a whole lot more plastic sheath to reveal the twists.

Are you saying all USB 2.0 cables have the data lines twisted? How about this. It's a flat all the way, there is no way to twist any wire:

Those flat cables are extremely unreliable in a high electrical interference environment.

[Andy Chee]

IBut I found out none of them is even twisted!  So what you make of it?

Whilst ethernet cable wire pairs have 5 twists per 1 inch, USB cables go as low as 1 twist per 5 inches.  So you will need to remove a whole lot more plastic sheath to reveal the twists.

FYI, that’s a gross oversimplification. Neither the Ethernet nor cable standards actually specify the number of twists per inch, only performance targets (like bandwidth and crosstalk), with the manufacturer free to meet them however it wants.

USB standard also does not specify twists per inch, only performance targets.

[tooki]

IBut I found out none of them is even twisted!  So what you make of it?

Whilst ethernet cable wire pairs have 5 twists per 1 inch, USB cables go as low as 1 twist per 5 inches.  So you will need to remove a whole lot more plastic sheath to reveal the twists.

Can the manufacturer disable this on purpose, not able to negotiate at FS at all but only HS?

If they did, the product would no longer be "USB compliant", completely defeating the term Universal Serial Bus.

Are you saying all USB 2.0 cables have the data lines twisted? How about this. It's a flat all the way, there is no way to twist any wire:


So they are the only USB 2.0 cables that don't have a twist?  And it's a sure way to force any HS device to renegotiate at FS by suppressing the chirp?

Also any reference for the chirp signal sent that can heard over the other end?

You can’t see the twisting by stripping just a few mm. 

If you were to slit the cable down its length and peel off the jacket, you should see twisting of the data pair — assuming the cable is actually compliant, which is not a given, thanks to the glut of shitty cables out there.

Attached are photos of what proper, compliant cable should look like (including some USB 3 just for illustration). A ton of USB cables out there are not compliant, and work by luck and/or using up all the margins for error in USB’s design.

[Ogitek]

Here I cut 2 more inches (uncrop so you can see every details). There are no twists in the green and white or any wire, or is it possible to have one turn every say 5 inches? Whatever, what brand is guaranteed to have twists and proven to be USB 2.0 compliant. Or the most Milspec USB 2.0 availab.e  I'd like to hold a genuine one to feel a real year 2000 technology, which I missed.

[tooki]

Here I cut 2 more inches (uncrop so you can see every details). There are no twists in the green and white or any wire, or is it possible to have one turn every say 5 inches? Whatever, what brand is guaranteed to have twists and proven to be USB 2.0 compliant. Or the most Milspec USB 2.0 availab.e  I'd like to hold a genuine one to feel a real year 2000 technology, which I missed.

I guess you didn’t read all the earlier replies, especially this one (bold added by me):

IBut I found out none of them is even twisted!  So what you make of it?

Whilst ethernet cable wire pairs have 5 twists per 1 inch, USB cables go as low as 1 twist per 5 inches.  So you will need to remove a whole lot more plastic sheath to reveal the twists.

And you’ll need to do it carefully to ensure you’re not straightening out the twist as you strip it. Slit the insulation down a foot of the cable and peel it off like you would a banana peel. Don’t strip little by little by cutting around the circumference and sliding down the wires.

[Ogitek]

Yes, there appears to be some kind of twists but the Red and Black wires are also twisted along with the green and white. But then, couldn't it be during manufacturing, the twisting of the wires are random?

Anyway. I'm trying to find references about the numbers of twists per x inch in USB wires. Where can one find it? Because I saw the following that there must be 4 twists per inch to avoid crosstalk. Also where are you supposed to measure the impedance of 90 ohms? in between green and white? but with respect with what resistance (since it has distance between them and not shorted)?

[Andy Chee]

Because I saw the following that there must be 4 twists per inch to avoid crosstalk.

Crosstalk in ethernet cable (and bundled telephone copper pairs in general) doesn't exist with USB, because USB only has one channel i.e. there's no other channel to crosstalk!

That said, if you bundled several USB cables in parallel, then maybe you would get crosstalk.  But long parallel USB cables could be avoided by using a single cable with a USB hub on the end.

[Ogitek]

Ok. So if the Vention is really USB 2.0 that can negotiate at 480Bps well and you just replace one of the head with a Lemo. why wont it work? What made a cable "dodgy"? does dodgy refer to the manufacturer, so if one choose a good brand like Vention and add a Lemo plug. It wont be dodgy and can work?

[Andy Chee]

If you just replace the plug at the very end of the cable, that should be ok.

But if you try to replace a large 24 inch section of cable with untwisted wire, then that is a very different scenario which may impact signal integrity.

[tooki]

Yes, there appears to be some kind of twists but the Red and Black wires are also twisted along with the green and white. But then, couldn't it be during manufacturing, the twisting of the wires are random?

No, that’s not how cable is manufactured, the twist has to be added on purpose.

Anyway. I'm trying to find references about the numbers of twists per x inch in USB wires. Where can one find it? Because I saw the following that there must be 4 twists per inch to avoid crosstalk.

As you’ve already been told, the USB specification doesn’t specify a number of twists, only that it be twisted, and then performance specs that must be met.
There are clearly huge differences in the number of twists used, and as I’ve said, many cheap cables would fail compliance testing if they were actually ever subjected to it. They just happen to be barely good enough to work most of the time. But they might fail in electrically noisy environments, for example.

Twisting not only reduces crosstalk, it also reduces sensitivity to electromagnetic noise, because the twisting causes the noise to sort of cancel out.

Also where are you supposed to measure the impedance of 90 ohms? in between green and white? but with respect with what resistance (since it has distance between them and not shorted)?

You can’t measure this yourself. (You would need a very expensive piece of equipment called a Vector Network Analyzer.) The 90 ohms isn’t the resistance, it’s what’s called “characteristic impedance”, which is a function of its behavior with high frequency signals.

[tooki]

Because I saw the following that there must be 4 twists per inch to avoid crosstalk.

Crosstalk in ethernet cable (and bundled telephone copper pairs in general) doesn't exist with USB, because USB only has one channel i.e. there's no other channel to crosstalk!

That said, if you bundled several USB cables in parallel, then maybe you would get crosstalk.  But long parallel USB cables could be avoided by using a single cable with a USB hub on the end.

Well, with USB HS, FS, and LS. With USB 3.0, there are multiple pairs.

And I suppose that theoretically, noise on the power rails could be coupled to the data pair, or vice versa.

But I suppose the primary reason for twisting in USB is to reduce common-mode noise.

[Ogitek]

If you just replace the plug at the very end of the cable, that should be ok.

But if you try to replace a large 24 inch section of cable with untwisted wire, then that is a very different scenario which may impact signal integrity.

What if I cut the original Venton USB in half and put Lemo at one end of one of the pairs? In other words, does the manufacturer have to design the twisting of each cable with certain numbers and design each complete cable or can they just cut their roll of wiring for every one meter and put termination into it?

[Andy Chee]

If you just replace the plug at the very end of the cable, that should be ok.

But if you try to replace a large 24 inch section of cable with untwisted wire, then that is a very different scenario which may impact signal integrity.

What if I cut the original Venton USB in half and put Lemo at one end of one of the pairs? In other words, does the manufacturer have to design the twisting of each cable with certain numbers and design each complete cable or can they just cut their roll of wiring for every one meter and put termination into it?

You can cut to any length and replace the plug.  The twists do NOT have to be "integer" lengths, so to speak.