There are many documents (e.g. this) describing the twist on two-drive floppy disk cable on IBM PC compatibles. While this is not the most proud example of clean hardware design, how much of this was originally planned in advance, or was it just improvised on the go?
Having the desired pins aligned on the pin out nicely enough to make this happen suggests, that this was intentional at least to some degree.
This was a good piece of production engineering.
IBM expected to sell lots of twin-floppy PCs, even after the launch of the XT (personal experience in a UK reseller in late 1984).
To assemble a twin-floppy PC and have the drives respond correctly to the drive select wires in the cable, there are four choices:
There is no question that option 4 will give the best results (cost, time and quality) in the IBM assembly plant. If the cable supplier can be persuaded to make the twist at a reasonable price and with good quality, it's an excellent solution which also saves time and effort in field service.
There seems to be some confusion about how this worked clouding the categorization of what to call it.
Let's compare the interfaces between how Shugart designated signals at the drive (and how other makers used them) vs how IBM re-designated them at the controller in order to support the twist idea. (I've used A and B on the IBM version to reduce confusion)
pin Shugart IBM before flip IBM after flip "2nd drive" interprets
10 /DS1 /MOTOR_A /MOTOR_B ignored
12 /DS2 /SELECT_B /SELECT_A /DS
14 /DS3 /SELECT_A /SELECT_B ignored
16 /MOTOR /MOTOR_B /MOTOR_A /MOTOR
IBM then jumpered all drives to use /DS2 on pin 12. And being standard Shugart drives (or equivalent) they expected their motor enable on pin 16 just as Shugart intended. So you have drives that, regardless of their actual role, expect one signal on pin 12 and one on pin 16, and don't pay any attention to what is going on at pins 10 or 14.
As such, the pre-twist drive connector becomes DOS drive B, presenting the original Shugart /DS2 select signal on pin 12, and the motor enable on pin 16, all in the very ordinary vanilla way (except that the BIOS should only set the bit to drive the traditional motor signal when using drive B)
But after the twist, the signal on the motor enable is what Shugart would call /DS1, but IBM is instead cleverly using as a "new" motor enable A. And the select signal arriving on pin 14, would be that which originated on the controller end pin Shugart would call /DS3, but which IBM is using as drive enable A.
Given that this scheme works with standard unmodified Shugart-interface drives, I'd categorize it as a clever "hack" or bit of manufacturing engineering. IBM changed only their controller card (which already had to be customized to match their bus design) and cable, and just bought the drives off the shelf.
You can see the manual for Shugart's longer-predating-the-IBM-PC original 1976 single-sided SA400 drive which originated this interface (as an evolution of the earlier 8-inch drive's interface) here. When Shugart extended the scheme to double sided drives such as the SA450 (still before the IBM-PC), they put four more places on the far end of the connector adding a side select and a "spare" and corresponding grounds, but didn't change the part of the interface used for drive selection and motor control. Regardless if IBM's drives had their logo molded into the front plastic, the technical reality is that they were and remain interchangeable with standard Shugart interface drives like the SA450. Peel the four wires that Shugart had previously added to create their 34-pin double-sided interface off the cable, and even getting an old 30-pin 1976 SA400 to work would have been merely a matter of software support.
Having the desired pins aligned on the pin out nicely enough to make this happen suggests, that this was intentional at least to some degree.
Since the pinout is an IBM one, made especially for the IBM PC (*1), and this twisted cable was used from the very first PC model, it's pretty clear that it was done on purpose.
Interlude:
hack or intended design?
While 'hack' is a rather fuzzy term, used for a variety of assignments - including great ideas one would not have thought of - it is exclusive tied to modifying something after market to enable a new usage, not planned by it's original manufacturer.
Keeping that in mind it's clear that the twisted cable is part of the PC design, as it is based on the floppy connector pinout which the PC engineers did device for theit machine. The Connector used is not a standard one used before, but on purpose made to enable this 'twist'. Standard (Shugart) drives could not be used with this connector. And a standard bus connector does not allow this 'trick' as it holds the signals in different order.
*1 - The original Shugart bus had only one motor signal.
Definitely a hack. A systems with floppy disk drives had separate selection wires in the connection cable, where one of them was to choose between on of 2 or 4 drives connected in parallel. There was a jumper on the drive to activate one of these wires resulting in an address.
Setting up this jumper was somewhat error prone and also extra work. The trend was to make more and more components of the PC self configuring. With the twist all drives could be jumpered on select 1, and a different wire would connect to that.
Other examples of self configuration are the PCI bus for addresses and interrupts, and USB for peripherals.
It's intended. The outbound cable at the motherboard is for a slave drive B. But it has settings for the master drive A as well. So to make all the drives alike, you have a twist in the cable. They all come like that.
Hard drives sort the issue with a jumper at the back.
There's a similar device with the PS/2 plug, which uses five of the six pins. If one switches the mouse into the unused chanel, you can plug both keyboard and mouse into a single socket, as is done on Laptops. I have recently acquired a keyboard with two tails (mouse + keyboard), and it can daisychain the real mouse as well.
Edit
The hardware arrangement supposes at the front of the box, that drive A: is above drive B:. A cable without any kinks in it would go A: to B: to the motherboard, providing a 3½ and 5¼ connector before and after the twist.
So a twist is needed between A: and B: if these are to be identical kinds of drive. There is no twist needed between B: and the motherboard.
As time goes by, you can drop various connectors, if you don't plan to support that kind of drive. This is done simply by not providing headers on the cable.
Because a direct cable remains between B: and the motherboard, dropping B: means that the connectors below the twist are dropped, and the one above the twist (aka A:) is set.
It is possible to swap the drives in software, but the balance of things is that one gets better assembly and running if one reduces the amount of user fiddling: ie use cables that set the identical floppies to A: and B:. A is above the twist, B is below.
Likewise, swapped cables allow one to use things like Null modems (which are a serial or parellel cable, with two 'computer' ends, and a twist in the middle.)
