 |
| Home | The Simulator | Telnet Clients | PDP-11 History | Links of Interest | Code Archives |
|
Early Computing at Wofford Original Witch Witch Reincarnation Access the WITCH Wofford Witch Documents Image Gallery PDP-11 Collection (w/documentation) Projects Message Board Videos |
 |
THE DEVELOPMENT OF THE DH11 MULTIPLEXER John E. McNamara, former DEC employee and the designer of the DH11 multiplexer, tells the story of the design of the DH11 This story is reproduced here with John E. McNamara's permission, from an email correspondence with W. Ashley Carder dated March 12, 2005 Hi Ashley, Great website! Back in about 1972, the PDP11 communications group undertook two simultaneous 16-line multiplexor projects, the DJ11 and the DH11. We referred to the DJ11 as the "simpleplexor" and the DH11 as the "sexyplexor." Both products used two identical line cards utilizing eight of the recently-developed UART chips. Both products also used a "silo buffer card" that stored received characters in first-in first-out buffer chips, also recently developed. The PDP11 was notorious for long interrupt entry and exit times, so the idea was that software would respond to an interrupt and suck as many characters (along with their associated line numbers) from the silo/fifo as possible on each interrupt service. The neat thing was that the heavier the traffic was, the more characters the software serviced per interrupt, so the efficiency actually improved with load! As interrupt service times improved in later processors, this idea fell into disfavor due to the context switching involved by having to sort out the traffic from sixteen different lines. There were four major difference between the DJ and the DH . The first was that the line speeds and formats in a DJ were configured by hard-wired jumpers, while they were fully programmable in the DH. The second was that the DJ was EIA-only or 20ma-only, depending upon the "distribution panel" used. In contrast, the DH (at least the -AA and -AC models) had a multi-slot distribution panel that could accommodate a mix of EIA, 20 ma, modems, etc. The third difference was that the DH had the option of full modem control, using a set of modules called the DM-11-BB, a remnant of a previous multiplexor called the DM-11. The fourth and final difference was that the DJ was program-interrupt on both receive (the silo/fifo) and on transmit. While the DH11 was program interrupt on receiver (the silo/fifo), the transmission was DMA. One of the consequences of the DH's many features was that it took a 9-slot "double system unit," whereas the DJ took only a single 4-slot "system unit." It also took lots of power and generated lots of heat. Another problem turned out to be cost. While the marketing folks suggested that I should design a version with 8 speeds rather than 16, this would have meant using 8-to-1 mux chips rather than 16-to-1 mux chips, saving maybe $5 in manufactured costs. A little research determined that the really big expense was the external distribution panel, which had its own power supply and could accommodate any mix of level converter boards. By redesigning some of the boards in the main product to include EIA level converters on board and thus use the el-cheapo DJ distribution panel, we could cut the manufactured cost by 40% and still offer all of the features of the DH except the mix-and-match level converters. The new products were the DH11-AD and -AE (with and without modem control, respectively). All-in-all, the DH was a very successful product, and over 100,000 were produced. It was on the price list for ten years, as it was used in PDP10's and some DECSystem20's in addition to PDP11's. Within the next year or two, progress in semiconductors made the DZ11 possible. As you know, this was a single board containing everything needed for 8 lines with modem control. I do not remember whether the DZ11 has programmable speeds and formats or not. I think it does. Looking at the pictures of your 11/40, I'm not sure that you would want a DH, as it would take a lot of space and power, especially the DH11-AA or -AC. Emulex and Able made newer technology, program-compatible versions, however. Able's product, the DH/DM is a single hex board with limited modem control, but claims to have all the other features. I've not seen one. As for my other projects, the major one was the DV11, a 16-line synchronous multiplexor. I also did some work on card readers. I look forward to hearing of your future adventures! I (Ashley) told John McNamara that I would try to get a DH11 up and running on one of my systems since it was part of the original Wofford Witch, and he said, "I would love to know that one of my "children" was alive someplace ;-)" Now I must find a DH11 and get it up and running in a multi-terminal timesharing environment. |