ICL 1900 Series Computers

Non-ICL 1900 Systems

Apart from ICL themselves, the 1900 architecture was implemented by the Polish company ELWRO ELWRO on wikipedia

ELWRO obtained specifications for the 1905, and presumably later, for the 1905E and 1902A/3A. They did not obtain technical drawings and instead re-implemented the machines using technology available to them. This meant that their ODRA 1305, whilst it would run executives built for the 1905E was built in TTL logic and not using discrete transistors. ICL only started using SSI TTL with the 'A' series machines.

ELWRO extended the concept of the ICL 'standard interface' allowing not only ICL 1900 'Standard Interface' peripherals to be connected but also IBM 360 style I/O channels were provided permitting IBM peripherals to be connected. ELWRO also manufactured IBM 360 'work alikes' and had peripherals for them available.

ODRA 1300-Series machines were used in a wide set of areas, including academia, power-station control, railways, as well as military uses.

They created several machines in the ODRA 1300 series:

System Year #Made Brief Details
ODRA 1304 1970-1973 90 Based on the ICL 1905 specification, these machines were the first 1900 clone made by ELWRO, using discrete transistors (like the 1905) they ran E4BM executive. 32KW maximum store
ODRA 1305 1973-1983 362 Odra 1305 'Technical Console' (Engineer's Console)
The buttons at the bottom right are set to present a value of 4. This is the micro-program address at which the machine's bootstrap is entered. The bootstrap address is on the switches at the bottom centre (switches 19, 20 and 22 are on = 11010 = 26).


Based on the ICL 1905E specification, these machines were apparently widely used within Poland. One ODRA 1305 was in use until 30th April 2010 at a railway facility in Wroclaw Brochów. ELWRO implemented all of the 1905E specification including the multi-processor and realtime facilities, and extended it to include some elements of 'high-availability' systems. They implemented the floating-point hardware as standard and also offered the multi-processor and possibly even the paging options. Following their 1905E parentage, they supported store sizes from 32KW to 256KW. This is the ODRA model that we have by far the most information on.

The machine was similar to the 1905E in that it employed a stored microprogram approach, with a 48 bit 'fixed store' to hold the microprogram. However the layout of the two machines' micro-intstruction words are radically different, as was the entire internal architecture. The system had facilities in the microprogram to perform some diagnostics and to detect failure of other processors in a multi-processing cluster, allowing the remaining CPU(s) to take over the tasks from the failed CPU where practical.

They used an interleaved address scheme for the store modules, allowing a higher degree of overlap between machine cycles and store cycles. Accesses to consecutive logical locations were distributed around the 32K store blocks.

The 1305 used a different assignment of slow, fast and multi-channel peripherals to I/O channel numbers. ICL machines used channels 4 through 21 as 'slow hesitation' channels and channels >= 24 were the fast/highspeed channels. The 1305 also had the channels from 0 to 21 as character/slow hesitation channels and the higher numbered channels were 'autonomous' (PAC).They only supported multi-channel devices such as multiplexers and scanners on channel 21. ICL added extra channels in blocks of 6, the 1305 used blocks of 4. In multiple CPU systems, failure of a CPU would raise an interrupt to the remaining nodes by raising the interrupt for channel 32, possibly defining the upper limit on the channel numbers at 31.

Just above 16 bottom centre is a neon between EXM (Exec mode (lit)) and ESM (22AM) called PM (Priority Mode) showing that the Realtime / Priority mode was implemented, just like they were on the 1904E upwards as well as the 1902A series machines.
ODRA1325 1973-1979 151 The available information on this machine is very sparse. It is most likely derived from the 1902A/3A specification, but it could be based on the earlier 1902/3 Stevenage specification. The available information is inconsistent, with some sources stating a store capacity of up to 128KW, others 16KW or 32KW. The available documentation (in Polish) states executives EX2M and EX2P as being employed on these machines, which supports the unsuffixed view. These machines were capable of using 'realtime' or priority mode programming - EX2P was a variant of the EX2M executive with the 'realtime/priority mode' package included.
RODAN 10 1974 - 1986 135 A 'ruggedised' variant of the ODRA 1325
RODAN 10/79 ? ? RODAN 10 with 2 'cassette' memories. What were they? - My guess is removable store cartridges that allowed all operational program and data to be removed from a machine so that it may be maintained without risking unauthorized access to any data in the store.
RODAN 15 ? 34 A 'ruggedised' variant of the ODRA 1305. http://pl.wikipedia.org/wiki/Rodan_(komputer)
Data for the above from http://www.aresluna.org/attached/computerhistory/articles/odra?language=pl (in Polish)

ELWRO took the 1900 and extended it in ways that ICL didn't. ICL were focussing on the later 'A' series, whereas ELWRO only had the 1305, the 1304 had done the same startup job as the 1905 and allowed software to be developed. This allowed them to use realtime facilities for industrial process control. - Realtime hardware was built-in to the 1904E, 1904A and 1902A/3A machines, but no ICL documentation is currently known to exist about this. Naturally we would like to hear from anyone with information on these facilities.

Links:
There are several articles in the publication "Informatyka" (Computer Science) available at delibra.bg.polsl.pl which provide a good insight into the Polish use of their excellent 1900 'clones' in academic and scientific institutions.

Thanks:
We must give thanks to ELWRO (taken over in 1993 by Siemens and closed down shortly afterwards) and the Polish enthusiasts who are preserving the remaining excellent ODRA machines in Poland. Their open preservation of these documents has allowed us to understand the 1900 series much better. We currently have little or no documentation regarding detailed instruction specifications from ICT/ICL sources, the Polish documents have provided a level of detail previously unseen especially regarding options such as multi-processing, although I am certain that it must have existed within ICL.