22

At the time when the USB standard was being developed, there were a number of widely-available connectors that they could have used instead of making a new custom connector. Why did they not use mini-DIN or DE-9 or some other common connector of the time, instead of inventing entirely new physical connectors, which would then need to be manufactured with entirely new (expensive) machinery?

I can think of a few possible partial causes (e.g. something new would be less confusing to consumers, or maybe the new connectors are more mechanically sound than anything else available at the time) but I'd like to see a more developed analysis of what was available at the time and why it wasn't suitable--and hey, maybe someone who actually worked on the spec can give the actual definitive reasons, if we're lucky.

Hearth
  • 27,177
  • 3
  • 51
  • 115
  • 40
    Common connectors equal confusion. If you have two DE-9 connectors one for RS323 and one for USB somebody will certainly plug in the wrong interface and things are bound to go BANG! – Oldfart Dec 07 '19 at 17:55
  • 7
    Main reason is that the bus requires a particular physical topology - upstream/downstream ports through hubs. This is why there are type A and B connectors that are physically different (i.e. don't just use polarity keys or labels). Also making a custom connector wasn't ground breaking. Firewire had already done this. – Jon Dec 07 '19 at 18:05
  • @Jon Not groundbreaking, but also not cheap. – Hearth Dec 07 '19 at 18:06
  • 13
    Mini-DIN is an abomination and a disaster in practice when used frequently and by people not committed to keeping the connector intact. | Relatively the original USB connectors were mechanically works of relative genius. || DE9 is not too bad but physically far too large and already the host to substantial interconnect confusion. || Mini USB was an interim step with a major failing that the sockets and not the cables were the weak points. ||... – Russell McMahon Dec 08 '19 at 08:04
  • 6
    Micro USB was driven in large part by a then signifcant manufacturer's "need" for a lower profile socket BUT had the advantage of making the cable the weaker link. ||... || If the world is going to use connectors in their zillions then cost of tooling up is a non issue and IN-compatability with existing connectors a bonus. – Russell McMahon Dec 08 '19 at 08:05
  • A new physical design can be copyrighted, and licensed out for an income stream. Got to pay for it all somehow. – Criggie Dec 08 '19 at 10:50
  • USB is not the only interconnect that uses a new (specified by the standard and not compatible with any existing connector) connector; Infiniband certainly does. Other interfaces are the subject of multi-source agreements (such as SFP, SFFP, POP4, SNAP 12 and more) that are all compatible with each other physically and electrically but *only* support a specific interface. – Peter Smith Dec 08 '19 at 14:13
  • 13
    The common connectors causing confusion point can't be understated. Take a look at the fiasco that's arising from everything being loaded into using USB-C and almost nothing saying clearly what it actually supports using USB-C for. – Austin Hemmelgarn Dec 08 '19 at 18:53
  • It was an action of the connector manufacturers/standards associations complex, to increase profits. – Hot Licks Dec 08 '19 at 23:38
  • 2
    @AustinHemmelgarn The whole slew of different USB-C flavors definitely contribute to the problem. There are clearly different "modes" that USB-C can be in but there is no clear way to tell whether the cables supports it, the hardware supports it, the hub, or whatever else. It's a giant FUBAR. – Nelson Dec 09 '19 at 03:09
  • @Nelson, Type-C is FUBAR... It is complicated. What do you suggest instead? Micro DP, mini-DP, DP, Micro-HDMI. mini-HDMI, HDMI, USB-A,B,C? TB1-2-3? USB4 2x2? I believe the time will come when you just plug a Type-C-C cable, and all will work just fine, display will power up, touchscreen activated, etc, without customer having any clue what he/she just did... but it will take some time. – Ale..chenski Dec 09 '19 at 06:15
  • @Ale..chenski Sorry, but your monitor only supports DisplayPort over USB-C while your phone only supports Reverse UVC (not sure if that actually exists). No connection for you! – user253751 Dec 09 '19 at 10:45
  • @AustinHemmelgarn [USB-OTG ](https://en.wikipedia.org/wiki/USB_On-The-Go) provided a preview of the confusion to come, and they still made it more confusing with USB-C – Chris H Dec 09 '19 at 11:25
  • Could have been worse, when the consumer AV crowd tried a standard connector they came up with possibly the worst abomination since the phono plug, the SCART connector... Unreliable, fragile and actually not quite enough pins to do what they needed it to. USB (ANY version) is a wonder of mechanical design for mass production in comparison. – Dan Mills Dec 09 '19 at 14:09
  • @DanMills I wish we'd had a good unified A/V connector before HDMI, really. Whenever I have to work with component video (not that often anymore fortunately) I even start feeling envious of the territories where they had SCART. – Hearth Dec 09 '19 at 14:57
  • 1
    @Ale..chenski The issue isn't that there are a bunch of standards using the same connector, it's that almost everything is optional and essentially nothing documents what it actually supports over USB-C. The net result is that you can't be certain that two USB-C devices connected together will actually do _anything_ (and even then, the cable might be at fault, not the devices). – Austin Hemmelgarn Dec 09 '19 at 16:29
  • There was already confusion over DE-9 which was used for voltage-incompatible RS232 and RS422 as well as other non-standard interfaces. Adding yet more confusion to this would result in yet more equipement destruction. – houninym Dec 10 '19 at 11:43
  • Speaking of which, I understand USB-C is a disaster in what is or is not supported by the counterpart. But why it took folks like 20 years to design a USB connector that can be plugged upside-down? – Oleg Lobachev Dec 10 '19 at 21:48

8 Answers8

47

Most likely many reasons, at least the following.

It was a requirement for the connector to support hot-plugging. The connectors you mentioned do not support hot-plugging.

Using an existing connector also means it is possible for someone to plug in two incompatible devices together just because they use the same connector.

The connector also needs to support enough mating cycles to be usable.

And the connectors you mention are old, as technology to make connectors advances, it makes sense to use that technology to make a modern connector that is smaller and more durable and otherwise better, and most importantly, a connector that is suitable for the specific application.

Justme
  • 127,425
  • 3
  • 97
  • 261
  • 12
    +1 for your plugging. The outer contacts are longer so that power connects first. – Transistor Dec 08 '19 at 09:32
  • 3
    @Transistor, yep, so that power and ground connect first and disconnect last. – TonyM Dec 08 '19 at 17:41
  • 6
    And connector shield is the first to make contact which should equalize potentials and ESD charges even before supply ground connects. – Justme Dec 08 '19 at 19:37
  • @Transistor well, you could still specify that (unequal pin lengths) for an otherwise identical existing receptacle or plug. – Peter - Reinstate Monica Dec 10 '19 at 12:41
  • SATA likewise supports hot-plugging (and for the curious, the connector fingers are more visible). Some people used to hot-plug ATA ribbon-cable drives, and sooner or later found out the hard way that it's a bad idea. – nigel222 Dec 10 '19 at 16:23
  • @Peter: identical looking plugs with different pin lengths? Nooooooo! – Transistor Dec 10 '19 at 16:31
  • @Peter-ReinstateMonica So in other words... a new connector design, absolutely incompatible with the one it looks like, but that looks identical to the old one to most users? That doesn't sound like it'll cause lots of devices to fry as they were trying to connect power but ground wasn't present (because they're not different lengths because I used an identical-looking plug) – Delioth Dec 10 '19 at 18:07
  • A D connector carrying RS-232 (for instance) supports hot plugging, simply by ensuring that the inputs tolerate any voltage overload from having an unconnected ground or whatever. – Rich Dec 11 '19 at 01:01
  • @Rich There is no guarantee which pins of the D connector mates first and last depending on the angle of insertion. The key point was that the D connector itself has no features to support hot plugging, so to support hot plugging with a D connector, time, effort and protective components must be put into designing an interface that has enough protection for it to work, and it costs money to do that. – Justme Dec 11 '19 at 06:33
33

Compatibility with existing connectors is an anti-feature

Let's imagine they used the then-common DB-9 DE-9 for USB. What would happen? Lots of mice used a DE-9, and it's used in a lot of other serial ports. People would be plugging a serial mouse into the USB port, and become frustrated when it didn't work.

Availability of USB connectors is a problem that would solve itself

Of course, the hardware person wants to see pre-existing connectors so they could just buy them at Digi-Key. However, this problem would solve itself fairly soon, as manufacturers started offering the new design.

manassehkatz-Moving 2 Codidact
  • 2,522
  • 11
  • 19
Harper - Reinstate Monica
  • 21,639
  • 27
  • 68
  • 1
    Is there any precedent for USB-like connectors though? It seems like they'd need completely new tooling to manufacture them. Of course that works out in time too. – Hearth Dec 07 '19 at 18:04
  • 7
    @Hearth I think their viewpoint was they'd be making these by the billions, so they'd achieve economies of scale on their own, and there wasn't any savings to borrowing tooling. – Harper - Reinstate Monica Dec 07 '19 at 18:20
  • 3
    This is basically what happened with USB C. USB C cable/socket can either be USB 2.x or USB 3.x and can support any number of extensions (Thunderbolt 3, DisplayPort, USB PD etc.) and there is no way for user to verify what given cable/device supports by just looking at it. – Matej Drobnič Dec 08 '19 at 19:44
  • 3
    @Hearth USB-A is similar to a PCB edge connector surrounded by a (mechanical & electrical) shield, in fact some devices actually implement a USB-A on a PCB. Edge connectors, including cable-mounted ones, have been around essentially forever. USB-B is folded up but conceptually similar. These connectors are nice because as well as fixing the connection sequence of the pins, they stay in without screws, and they wipe the contacts quite well on insertion, while holding less dust in them than say DE-9, handy for things you put in a pocket. Not perfect, especially smaller/recent versions, but OK – Chris H Dec 09 '19 at 11:32
  • @ChrisH I never thought of it like an edge connector, that's a good point. – Hearth Dec 09 '19 at 11:49
  • Tooling requirements are really not that out of the realm of contemporary connectors. Sure a production line would need to be set up (or programmed for a run), but the tooling stamping and crimping the shape plus molding the plastic or printing on plastic (or setting a PCB) was and is not that unusual for this domain, perhaps more exotic than a simple screw machine churning out gazillions of round pins, but still available. – crasic Dec 10 '19 at 07:09
  • Indeed. Many years ago I worked with a device that used a DB-25 connector, but had 48v on one of the pins, frying serial interface cards if you plugged it into one. One customer managed to work their way through a selection of 20 spare serial cards before calling us. – Rich Dec 11 '19 at 01:03
19

They Learned from Parallel & Serial Ports

Until the IBM-PC came along, the "typical" parallel port was the 36-pin "Centronics" connector. Then IBM decided to use a DB-25F connector for the parallel port on the IBM PC. Oops, that was already very commonly used for serial ports on computers and terminals. So they switched the serial port to a DB-25M connector, which meant new cables or gender changers were needed if you put an IBM-PC in place of a terminal, even if the pinout was the same.

So with USB, they created something new, and that was a very good idea.

manassehkatz-Moving 2 Codidact
  • 2,522
  • 11
  • 19
  • 6
    Not to mention 25-pin SCSI ports, and sundry 25-pin pinouts on industrial-control cards.... – rackandboneman Dec 09 '19 at 00:40
  • 3
    @rackandboneman Yeah, SCSI was a mess of connectors. I don't fault the industrial/etc. stuff - but IBM should have known better. Then again, to them a terminal was coax or twinax. – manassehkatz-Moving 2 Codidact Dec 09 '19 at 00:43
  • 1
    My college professor tried to make us understand his bewilderment about the design decision of a male outward facing connector. "Why would anybody do that? It's a plug facing out: Exposed blank pins so that any touch could short them?" – Peter - Reinstate Monica Dec 10 '19 at 12:48
14

At the time when the USB standard was being developed, there were a number of widely-available connectors that they could have used instead of making a new custom connector. Why did they not use mini-DIN or DE-9 or some other common connector of the time, instead of inventing entirely new physical connectors, which would then need to be manufactured with entirely new (expensive) machinery?

In practice Mini-DIN, when used frequently and by people not committed to keeping the connector intact, is an abomination and a disaster. Bent and broken pins "just happen".

Relatively the original USB connectors were mechanically works of genius.

Db9 is not too bad mechanically, but physically is far too large, and was already the host to substantial interconnect confusion.

Mini USB was an interim step, with a major failing that the sockets and not the cables were the weak points.

Micro USB was driven in large part by a then significant manufacturer's "need" for a lower profile socket BUT had the great advantage of making the cable the weaker link. After many insertions the mini-USB in-device socket tends to fail. With micro-USb the design means that the plug connector is the weaker point.

If the world is going to use connectors in their zillions then cost of tooling up is a non issue and IN-compatability with existing connectors is a bonus.

Russell McMahon
  • 147,325
  • 18
  • 210
  • 386
  • Yep - the dimensions of the regular USB-A plug allowed for nice flat casings compared to the size of DIN / mini-Din or DE-9 connectors ... – eagle275 Dec 09 '19 at 12:44
  • 4
    Also, while we're at bashing mini-DIN: people already are annoyed by USB-A's need to plug in in the correct orientation out of... **2**. With all those round DIN connectors, it's more like one out of 360 possible orientations that'll work... – leftaroundabout Dec 10 '19 at 09:23
  • @leftaroundabout I'm not so much bashing mini-DIN as seeing the pins bent and broken off by people who have no idea of how to treat a connector. | According to one HP publication, USB-A has 3 orientations -> The way you try 1st. Then 180 degrees flipped. Then 180 degrees flipped again to get what actually the right way after all. | I have a USB A in line tester (V, I, mAh, time, ...) with an omni (out of 2) orientation connector. The joy!!! :-). – Russell McMahon Dec 10 '19 at 11:22
  • @RussellMcMahon I have one of those too, but since the two-way connector means the contacts have to flex no matter which way you plug it in, I'm doubtful of its longevity. Fortunately it was cheap. – Hearth Dec 10 '19 at 23:44
  • @Hearth - yes, you could build one badly. But with "proper engineering" [tm] you should be able to move the contact plate without any life degradation. Whether they do so is moot, but one can guess :-). You could make a housing with a spring loaded shell that moved as required with the contact plate fixed relative to the electronics - whci may be what they do - must look. – Russell McMahon Dec 11 '19 at 01:16
  • Mini-DIN invited trying to twist the connector into orientation by feel if it had to be inserted out of sight (eg reaching behind equipment). Perfect way to destroy the pins. – rackandboneman Dec 11 '19 at 22:31
3

There might or might not have been foresight regarding USB2.0 and its signal integrity requirements.

High speed digital signals are effectively HF to UHF signals, with all the problems they bring.

Generic Sub-D or DIN connectors are NOT designed to be viable VHF/UHF connectors. Certain EXACT models of Sub-D parts might be viable (as in, you have the part number and data sheet and have measured and tested that exact part).

A DIN or Sub-D connector will be electrically suboptimal already for a 12Mbps signal (though they have been used for protocols in that speed range, eg AUI for ethernet transceivers). Using it for a 400Mbps signal with a requirement of 500ps risetime could be considered ludicrous - if you ever toyed with experimental setups dealing with digital signals with even sub-10ns risetimes you will find that every inch of undisciplined wire or connector pin (as in, it is not part of a coax, or near a groundplane, or part of a twisted pair line) has a good chance of introducing gremlins into your system - or turning something innocent into an antenna.

Granted, VGA uses a sub-D connector for (analog) high speed signals that can be in VHF range - and problems (signal degradation, ghosting....) are not that uncommon, even though there has been ample time for manufacturers to optimize HD15 connectors (which are probably mainly VGA use today) for their most common usage. Some professional computer systems (SGI or IBM workstations ...) chose to use BNC connections or 3W3/13W3 connectors (a sub-D with coaxial inserts) for a reason for their analog monitor connections.

So, USB2.0 would likely have been impossible to keep physically compatible - and at the same time reliable - if a legacy connector like that would have been used.

rackandboneman
  • 3,085
  • 9
  • 13
  • Reminiscing :-) - 10base2 (?) internet with RG-58 coax at 10 mbps used BNC connectors. Workstations T'd off the main cable with short extensions from a BNC-T. Replacing even one of these short runs (say 1 or 2 metres long) with visually near identical 75 Ohm coax would bring the whole LAN down. Ask me how I know :-(. – Russell McMahon Dec 11 '19 at 01:23
1

In addition to the other fine answers, I'll add that mouse and keyboard connectors during the birth of USB were DIN connectors. The switch to USB mice and keyboards was not immediate, and picture holding a device with a DIN connector on the end of a cable, and trying to figure out which DIN socket to plug it in to.

Personally, I miss the days of stacking 25 to 9 pin adapters, gender changers, and null modems, and then tweaking baud rates until something worked.

Scott Seidman
  • 29,274
  • 4
  • 44
  • 109
1

Connector technology is an advancing field.

D-style connectors were developed around manual soldering, megahertz-range signals, commercial/industrial use cycles (a system would typically be setup and left for months on end), etc.

By the time USB emerged, the state of the art in (cheap) connectors and assembly allowed for denser connector pitch, smaller pins, better duty cycles, etc. (There was also an aspect in the A connector of being able to use a PCB directly as a (fairly unreliable) male connector).

With USB-C, these things have moved on even further.

(But why do current printers and musical devices use the ancient full size B connector?)

Rich
  • 231
  • 1
  • 3
  • 1
    Maybe because a full size B connector, and especially the jack for it, is far less likely to be destroyed beyond repair if you accidentally push a printer against a wall with the cable in between... or if you get all kind of stage dreck into the connector of eg an effects pedal and/or step on the cable where it is plugged in.... Also, people will recognize a B connector if they need to replace the cable, the smaller connectors often lead to guesswork (ok, is that mini or micro or micro-usb-c or kodak special or....) – rackandboneman Dec 11 '19 at 22:20
0

Second answer, since it is about very different aspects:

At the time the IBM PC was introduced (and introduced certain connector standards with it), desktop computers (be they IBM compatible or not) were considered industrial/professional/enthusiast equipment, expensive enough to expect users to be knowledgeable and careful enough to not require foolproof design. Also, a lot of hardware and wiring needed to be custom-built or custom-adapted, so using generic connectors available from any electronics parts distributor catalog, which could be hand soldered with some skill (soldering Sub-D and DIN connectors is not a real beginner task!), was a real feature.

At the time USB was introduced, PC compatibles were already becoming a common household item, where this kind of standard could no longer be applied. Also, custom hardware was already less common (and continued to use the legacy ports oftentimes), and was much more difficult to design to interface with USB anyway: You can make some piece of circuitry behave as a centronics printer with a 7407, and 74221 and not much more... RS232 just needs some cheap and common level converter chips... To turn something into a USB device, the parts needed are much more specialized.

rackandboneman
  • 3,085
  • 9
  • 13