Misplaced Pages

This is an old revision of this page, as edited by 137.112.137.189 (talk) at 00:40, 24 June 2005 (→Can someone please explain usb cable types). The present address (URL) is a permanent link to this revision, which may differ significantly from the current revision.

Use of SCSI Command Set by USB

Q: I think it would be useful to mention that USB uses the SCSI command set, despite adopting a different physical architecture. This is actually highlighted in the SCSI article. As the current article stands SCSI and USB seem like completely unrelated technologies. Nick 08:50, 20 Aug 2004 (UTC)

A: See USB Mass Storage Rationale of SCSI over USB.

Please specify USB version

"USB has a Full Speed rate of 12 Mbit per second." Is this version 1.0 or 1.1?

no one knows. - Omegatron 14:36, Jan 15, 2005 (UTC)

• ? whaddya mean nobody knows? VERSION NUMBERS MEAN NOTHING TO THE SPEED. DO NOT PAY ATTENTION TO VERSION NUMBERS. Full speed is 12Mb. Always, has been, since .9 at least. SchmuckyTheCat 19:25, 15 Jan 2005 (UTC)

• update to myself, Full Speed has always been in the spec, with bandwidth testing of multiple FS devices being included as reference at 0.9 SchmuckyTheCat 17:11, 8 Mar 2005 (UTC)

Can someone please explain usb cable types

I know there are at least two and maybe more types of cables for USB - Could someone differentiate them? Is one type of plug/cable for USB 1 and another for USB 2? Or does the cable/plug type not matter? I know they are compatible, does one cable/plug limit the bandwidth?

I have tried to find this on the internet but my searches only turn up vendors trying to sell stuff - it is overwhelming.

Could someone please research this and maybe even put pictures of the USB plug types up on the page?

Oh, and I know that a USB 1 hub would have to be replaced to handle the bandwidth of USB 2, but I do not know if the cables make a difference.

While Richard's answer (below) is true it is slightly unclear. These are some difficult questions. The facts in my post are based on information made available by the USB Implementers Forum, including the USB 2.0 spec.

First you should know that captive cable means a cable where the device end is either permanently attached to the device, or is detachable, but uses a connector other than the ones specified by the usb standard.

Also note that usb cables cannot contain usb ports. (USB extention cables are not permitted by the USB Spec).

The legal cables are:

• A to B
• A to mini-b
• A to captive
• mini-A to mini-b
• mini-A to captive

There is also one legal adapter which consists of a USB A connector and a usb mini-A socket. (This is the only legal use of a mini-a socket.)

The usb 1.0 cables will probably work with high-speed usb 2.0 devices, but it is possible that problems may occur. If a cable claims USB 2.0 support it should work for all devices.

The link to the USB.ORG site includes access to the specs for cables which include pictures (drawings). The A end hooks to the host and the B end hooks to the device. As noted above, in the extension called On-the-go, there is also a hermaphroditic socket which will accept either the A or the B end of the mini-version of the standard cables.

There is no plug/socket change for High speed (480Mpbs) but the cable spec was tightened in the 2.0 version to allow for the higher transfer rates. A 1.1 Spec. hub will work on a 2.0 system but will limit the maximum speed of any down-stream devices to 12Mbps regardless of whether the downstream devices are High-speed capable or not. - richard

USB A-A cables

Can someone write in detail on the topic of USB A-A cables? From what I know, there are 2 types of these:

• Plain A-A cablle
• A-A with electronics

I don't know what the first one is used for, it costs ~$5 (nominal cost - cable and 2 terminals). The latter can be used to connect 2 computers and includes some kind of device, so that files can be copied (~$50). I have some info that files cannot be copied from computer to computer on the plain ones because USB is an asynchronous protocol, but I don't understrand the technical details. Any info on this? Any software workarounds for file transfers using the cheap cables possible?

Helix84 13:25, 3 Nov 2004 (UTC)

There shouldn't be any topic of USB A-A cables. The ones that have electronics in the middle are just another device. The electronics just expose a bulk interface on each side, one to each host controller. These usually require some special software that sends files or whatever over the bulk interface to the same software running on the other machine. These aren't deserving of their own topic, they're just another device.

Those that don't have electronics in the middle are invalid according to the spec. Since the first machine you plug one into is sending 5v and the next machine you plug it into EXPECTS to send 5v, you're just likely to blow the motherboard of one or both machines. Just bad, bad, bad. Of course you occasionally run into some no-name stupid device that uses an A-A as a device cable, avoid them like the plague. If they couldn't even get the cable spec right it's unlikely the device will work well AT ALL. -- SchmuckyTheCat 8 Dec 2004

At some point I was going to add a section on those little USB PC-to-PC networky things (but I confess I've never used one, never seen one, and I'm really not sure what they're officially called). That section would have a one-liner explaining why an A-A cable wouldn't work. I suspect that in addition to the "you cooked my motherboard" phenomenon, there's the unavoidable fact that the low-layer protocol handlers (built into the relevant host controllers) aren't built this way (unlike those in USB-to-go) and anyway the resistors that balance the datapair will be doubled (which isn't good). - John Fader 03:31, 9 Dec 2004 (UTC)

we bought a bunch of one brand and reverse engineered our own driver for a few purposes: throughput, loopback, and bit verification testing. yeah, it might be worth a few sentences of "dispelling common misperception" in the discussion of cable types.

- SchmuckyTheCat 22:42, 8 Dec 2004

USB 2 HS vs. FireWire

I would like to comment on USB2.0 being in direct competition with IEEE 1394. Here are a list of features that makes then resolutely different:

Application domain

• USB addresses needs for a wide range of devices (mouse, keyboard, modem, hard disk drives, scanner, printer...) that do not exist in IEEE 1394.
• IEEE 1394 addresses needs of audio-video devices such as videorecorder, digital camera that have no real equivalent on USB.

USB's isochronous mode does apply to streaming devices including videorecorders and audio devices such as speakers, microphones, etc. The High-speed (480Mbps) use of isochronous is directly comparable to use of IEEE 1394/Firewire for these devices - richard

Communication paradigm

• USB provides host to peripheral communication. A host computer is required in the system.
• IEEE 1394 provides non-centralized networking. You may build a network with simply a VCR and a TV tuner.

Sylvwild

Yes, the system requres a host - but the USB On-the-go extension allows for the possibility of a unit that is usually used as a device to become a host for the purposes of a point to point conversation as you describe. The initial setup is determined by which end of the cable is plugged in to which unit but is switchable under software control so that the initial device may assume host and vice versa. The USB OTG sockets are hermaphroditic - can accept either the A or B end of the cable. - richard

"Throughput"

"USB 2.0 boasts 480Mbps throughput"

I don't think so. USB 2.0 does signaling on the wire at 480 megacycles per second. The physical layer transports up to 480 million bits in one second. But that doesn't mean it has 480 Mbps throughput. In fact, it pretty much guarantees the throughput is less than that.

Looking at the spec, the fastest way to transfer data seems to be with a high-speed bulk transaction with a data payload of 512. This gets you (see p.55) 53248000 bytes/second of bandwidth, or just under 426 Mbps throughput. And that's assuming you can saturate the line with 100% high-speed bulk transactions; I think you have to have other transactions going on to request all that data, which is why even the theoretical throughput is lower than 426 Mbps.

Am I missing something?

I haven't seen the Firewire spec, but it probably has something similar going on. I doubt it has 400 Mbps of throughput. But it does seem to perform better than USB 2.0: this should be a good indicator that USB doesn't really "boast" more "throughput".

USB 1.0, 1.1, 2.0

"Confusingly, the USB Forum has renamed USB 1.1 to USB 2.0 Full Speed; and USB 2.0 to USB 2.0 High Speed."

Are those definitely absolutely the same? Please provide a reference. I find conflicting info online. - Omegatron 22:39, Nov 6, 2004 (UTC)

subsequent edits deleted that info, it was wrong.

Maximum cable length

I think about adding a mention of maximum cable length, for people (as I was) concerned about the distance between two equipments. There's an answer on the USB.org USB Info: Frequently Asked Questions. Any comment ? - --Olivier Debre 09:02, 18 Jan 2005 (UTC)

Certainly this should be included. Explain why cable length matters and explain the changes in recommendations by the USB.org. (Originally it was a length, but they changed it to a signal integrity spec instead.) - Omegatron 21:49, Jan 26, 2005 (UTC)

interrupts and synchronization

Which way are frames initiated? Does the host/hub send out a frame to each device every ms, or do the devices send out frames every ms? Since the clocks of the devices would vary slightly I would imagine the host does. How does the USB host interrupt the CPU? Where can I find more details about these, since they may be too technical for this article? - Omegatron 21:47, Jan 26, 2005 (UTC)

• the USB host sits on an internal bus in the PC. HCD --> bus --> CPU I don't know of any implementation of USB that isn't PCI, but maybe Mac's or other non-x86 architectures use some other internal bus, I've never had to care. On Windows the USB stack transfers URB messages between driver components, URBs are equivalent to IRPs in the rest of the Windows kernel messagaging interface. SchmuckyTheCat 03:58, 31 Jan 2005 (UTC)

:-) Well, at least I know of some acronyms to look up now... - Omegatron 04:06, Jan 31, 2005 (UTC)

Host controller device (USB host), Interrupt return pointer, USB Request Block - Omegatron 04:11, Jan 31, 2005 (UTC)

• also, i gave you an incomplete answer. the bus has 10% of the traffic reserved for control purposes - so devices DO keep in synch. it also uses control to make sure everybody gets a chance to talk, but who talks is dependent on the type of traffic they want to send (see the transfer types under technical details). transfer types have priority: control, interrupt, isoch, bulk. interrupt devices can interrupt but they don't send large amounts of data, so this shouldn't interfere. isoch devices reserve bandwidth in an OS dependent fashion (on windows it's first-come first-served and no hoarding) so they can talk as much as they want in their reserved pipe. bulk devices talk when the HCD tells them too and they get whatever bandwidth is left over. they have to scale back when something else is going on. simple enough. but then: see also how the virtual 1.1 HCDs in EHCI create virtual busses on the same wire, and how the TT virtualizes 1.1 traffic in a 2.0 hub. i think i'll do a re-write of portions of the article for this, the article is currently too technical for simple concepts to come out of it.

I don't really understand all of that. One thing I was asking was, since the host and devices each have their own clocks, which clock is used to synchronize transfers between them? I assume the host clock, and the devices only send data when they are spoken to? - Omegatron 18:42, Jun 6, 2005 (UTC)

Types of USB connectors

Would anybody who knows this topic well care to add pictures and information about the various types of USB connectors, such as Type A, Type B, and mini USB? — Brim 08:28, Feb 5, 2005 (UTC)

Which information should be hidden or removed?

I simply wanted to add information that some people might want to know: USB 1.1 introduced Full Speed, or 12.5 Mbit/s transfer speed. In my first edit, the wording was poor but not factually incorrect when I said 1.1 had Full Speed. Then an unnamed editor reverted it, stating that version 2.0 also has support for Full Speed, not just 1.1. Not a valid argument at all for a revert, but a true statement nonetheless, I thought, so I simply added more information to this for clarity and said 1.1 introduced Full Speed. This was yet again reverted by the same editor. It seems like this person has an agenda of removing information from the main article. Strange, since one would think an Encyclopedia would be a place where you add, as opposed to remove, information (especially the ones people would likely to be interested in). As a consumer, I would like to know as a basic fact what speeds are associated with the version number of a USB port. SchmuckyTheCat thinks otherwise, and believes that this information must be hidden from plain view. At his/her request, I have created a section where s/he could respond, hopefully without the caps-lock on. --69.214.224.74 20:19, 8 Mar 2005 (UTC)

First off, please don't make a technical discussion into a personal conflict by making comments about other users and what you imagine their motives to be. Let's just stick to the article, okay. Now, it's clear that there's disagreement both about what speeds were introduced when, and how we should present this. I think we should have a "feature table" showing the different versions of USB, their features, and significant changes. Now, as to which version did what: sources need to be cited, particularly so now that there is controversy. A source in this case is ideally the version, page, and line number (and ideally a quote) from the official USB spec. When this (surely an unimpeachable) source is available, bickering without citing exact references is a waste of everyone's time. -- John Fader (talk · contribs) 20:50, 8 Mar 2005 (UTC)

This info was wrong in the past and someone fixed it. See above. Please don't change it unless you know for certain you are right. Cite your sources to prove it. - Omegatron 22:38, Mar 8, 2005 (UTC)

• Full speed has always been part of the spec, probably since the version number was just an idea in a few engineers head. It was never introduced, it has always been. I'm actually interested to see why you think this is true. Is there another website or product literature saying this? I understand the user confusion, but the version number of a port doesn't correspond to speed. Like all technical interfaces, the new interface is "faster, better, cheaper" than the last but backwards compatible. You never, for instance, talk about RAM as being DDR v2.0. It's DDR266 or PC2100, and of course the faster stuff slows down in a slower slot. Unfortunately, some USB device marketers started advertising the version number as a speed indicator, which is enitrely misleading. A USB 2.0 device might only do Full Speed (12mb). What sells more, a 12mb device that says "12mb" or a 12mb device that says "USB 2.0"? A device might even do 480Mb, but the manufacturer fails to get it certified (maybe it fails other parts of the spec). They can't put the Hi-Speed logo on the box, but they can say 2.0. "Hi-Speed" is trademarked. "2.0" is not. A smart consumer should look for certification logos and declared speeds, not version numbers. SchmuckyTheCat 15:16, 9 Mar 2005 (UTC)
• Also, see previous discussions on this on this talk page under "Please specify version number" and "USB 1.0, 1.1, 2.0" SchmuckyTheCat 18:07, 9 Mar 2005 (UTC)

What is the difference between USB 1.0 and 1.1? If they supported the same device speeds, someone had better go tell list of device bandwidths, which may or may not be confusing Low-Speed/Full-Speed with 1.0/1.1. (It lists 1.1 as being faster than 1.0; can someone who knows the difference go look?) What benefits did moving to 1.1 offer for devices that made up for breaking compatibility with 1.0 host controllers? grendel|khan 19:05, 2005 Mar 9 (UTC)

• krikey, what a mess that list is. I will go correct. If someone writes a USB history section the difference between 1.0 and 1.1 would be: pretty much nothing. SchmuckyTheCat 19:44, 9 Mar 2005 (UTC)

impedance

cable has a characteristic impedance of 90 ohms. each driver should be 45 ohms ± 10%. for high-speed, each receiver shuold also be 45 ohms, standard nonreflective impedance matching. not sure if the 45 ohms are required for low/full speed devices though. - Omegatron 20:22, Jun 6, 2005 (UTC)

Actually it's more complicated than that:

"When the full-speed driver is not part of a high-speed capable transceiver, the impedance of each of the drivers (ZDRV) must be between 28 Ω and 44 Ω, i.e., within the gray area in Figure 7-4. When the full-speed driver is part of a high-speed capable transceiver, the impedance of each of the drivers (ZHSDRV) must be between 40.5 Ω and 49.5 Ω, i.e., within the gray area in Figure 7-5." - Omegatron 20:36, Jun 6, 2005 (UTC)