Talk:UNIBUS parity
From Computer History Wiki
18-bit mode
Would it be appropriate to mention 18-bit data on the UNIBUS here? I understand the KS10 and some of its devices could use the two parity bits for data. Maybe some of the 18-bit PDPs too? Larsbrinkhoff (talk) 12:40, 23 January 2019 (CET)
- Yeah, good idea. This hack is mentioned in a number of places (e.g. MX15-B Memory Multiplexer, RH11 MASSBUS controller, etc). Once I update here, I'll link them in.
- I wonder if this possibility was thought of up front, or if this hack was only developed later, when someone had a brainwave? Jnc (talk) 14:28, 23 January 2019 (CET)
- OK, done - see what you think.
- Also, I wonder about the RH11 in the KL10 front end - could it do 18-bit transfers over the front end's UNIBUS? ISTR that the RP's on that were double ported, and the RH10 could get too them too, so maybe no need for the 18-bit mode on the front end (and probably not high enough performance anyway). Jnc (talk) 21:43, 23 January 2019 (CET)
18-bit MASSBUS tape controllers
Circling back to this, I checked the TM03 User Guide and on page 2-16 it says it can do 18 and 36 bits. Page 1-8 talk about the bit fiddler, but doesn't specify how it works with >16 bits. TM03 magtape controller says it can have a "M8915 for use with a PDP-10". Larsbrinkhoff (talk) 16:28, 6 February 2025 (CET)
- I am totally stunned (absolutely literally - I had to lie down and sleep to recover).
- The TM03 does seem (in some configurations) to be able to send 18-bit words over the MASSBUS (when in 'PDP-I0 Core Dump Mode'). Tables 2-8 and 2-12 give the details: 4&½ 8-bit characters (on the tape; the other ½ character on the tape is unused, and contains '0' bits) hold a 36-bit word, which can be sent over the MASSBUS in two 18-bit cycles. (I wonder if it can read in that mode, or just write?)
- It would apparently usually have been attached to an RH10 for operation with a PDP-10. So, operation with an RH11 running in 18-bit mode is un-charted territory (as the expression in English goes).
- The TM03 comes in two configurations; with an M8906 or M8915 'bit fiddler' board. 'PDP-10 Core Dump Mode' is only supported with an M8915. (There are two variants of the M8915's: M8915 and M8915-YA. The latter is only needed for operation at 125 ips - such as with a TU77 transport.)
- The KS10 is apparently often configured with a TU45, which does use a TM03 (packaged in the first drive); some documentation says the TU45 can also be used with a TM02.
- ITS machine configurations claims that AI-KS had a TM03 (with a TU77). So now the questions are:
- Was that TM03 configured for a PDP-11 (i.e. with an M8906) - because it was on a UNIBUS - or for a PDP-10?
- If the latter, was it utilized in 36-bit mode? (I suppose you could look at the code and see.)
- Now I need to go back to sleep again! Jnc (talk) 05:37, 7 February 2025 (CET)
- It would also be good to know if KS10's were usually sold with the PDP-10 version of the TM03.
- EK-KS10-TM-PRE (and probably other places) says that "'DEC FS' will not maintain systems that do not have TU45 magtape" (pg. 9 of the PDF); that could likely have been because they needed the capability to load 36-bit data (for diagnostics), which would have required the PDP-10 version of the TM03.
- That manual does say (pg. 13) that the "TM03/TU45" supports "18-bit (NPR) data transfers over Unibus". So I guess that answers that.
- Interestingly, that manual does show (pg. 27) a "Typical KS10 System Configuration", with a TU45 attached through a "TM02/TM03"! The "TM02" is printed; the "TM03" is written in by hand! An apparent contradiction between i) no 18-bit support on the TM02, and ii) a need for 18-bit support, perhaps to load diagnostics? (One can imagine all sorts of ways that came to be. A simple error in an early version of the documentation? Someone in FS realized that they needed a way to load 36-bit data, which meant it had to have a TM03?)
- I give up! Jnc (talk) 13:34, 7 February 2025 (CET)
