First-Hand:Experiences with the Membrain 7700
Submitted by Tanj Bennett
I worked on the Membrain 7700 that was effectively a minicomputer version of the MU5. This was implemented in bitslice ECL, the Intel 3000 series, starting work in 1975 and the first working versions were in the labs at Membrain by late 1976. I recognize the instruction set of the MU5 from my memory of the instructions in the MB-7700 (which I documented, as well as writing a lot of the device drivers). The architecture was slimmed down for a 16 bit system, but we still had segmented memory and even could swap memory to hard drives by 1977. The OS was a port of MUSS, and the assembly language was MUPL. Several of the key engineers were graduates of Manchester.
The target market was Automated Test Equipment. The MB-7700 was physically the size of a desk with the CPU and storage (both 8" floppies and a hard disk were supported) about the size of a set of drawers on the right side. The surface of the desk had a CRT and KB to the right, and the rest of it was given over to the interface electronics for testing circuit boards. Those boards generally loaded onto a bed of nails fixture customized to match the board, and the computer had analog and digital I/O that could drive signals and measure responses. A fully configured machine had 64kB of memory, which was enough to run the OS and various applications. The applications included the full development suite for the tests, such as editor and digital circuit simulation. I do not know how many of them were built, but it remained in production until about 1980 and probably about 100 would have been sold. The company Membrain was sold to Schlumberger around 1978/79 and the founder (Tony Davies) eventually left Schlumberger and bought control of CTL some time in 1979.
The Transputer also had a passing resemblance to the MU5 instruction set.
I was told the port to Intel bit slice ECL was done mainly by one engineer in less than a year. He left some of the microcode unfinished (I cleaned up some corner cases) but it was amazing how small a team could build a new machine back then.
The management was apparently upset it took us 15 months to get from design to market instead of planned 12. That included new CPU, ported OS, new compiler, and new application suite. The whole engineering team would have been less than 40 people. There were never more than 20 of them in software, I never met more than 10 on hardware (but we probably had some others for areas like chassis, power, etc.). It was a simpler time, and amazing what you could get done when nobody told you it was impossible.
Looking through some old email I found that Chris Hedges, then at Intel, did the microcode for the Intl 3000 version of the instruction set as well as consulting on board design. Here are some other items I can share from that email:
Membrain Ltd was founded in 1972 by Anthony Davies OBE, it manufactured Automatic Test Equipment for electronic systems, evolving through memory (whole board) testers, wire-wrap backplane testers and multiple generations of functional circuit-board testers. Membrain was acquired in 1977 by Schlumberger.
It appears that Membrain bought MUSS around 1976 and ported it to their MB7700 , which was a machine that was embedded in test equipment. It is an obscure machine for which there seems to be very little information. The machine was functional by the end of 1976, and it was in production equipment by mid-1977.
It seems that the Membrain MB7700 was based on the Intel 3000 series. The Intel 3001 was the program control unit which executed the microcode and drove a set of 8 x Intel 3002 bit-slice ALUs. There might have been a third chip. The microcode included multiply and divide, and synchronous IO to floppy and HDD heads. The minimum configuration was 32KB, but most units in practice had 64KB of DRAM. The board was designed by a contractor in California.
One former Membrain employee who worked for them in Ferndown near Bournemouth from 1978 to 1981 reports a product generation that ran an operating system based on MUSS, this product generation was already established in 1978.
Another former Membrain employee says it was shaped like a desk and you could sit at it and interact with it in a way which was unusual then – a side effect of designing it to support test jigs on top of the desk with IO cabling to the back for the measurement instruments. This employee wrote an interactive screen editor for it in early 1978. He says they also added some of the first floppy disks available and had an electrostatic printer which for the time was amazingly small.
The information in this section comes from Rob Jarratt, Prof. Gregory Egan, Doug Duke and Tanj Bennett.
- H. Barringer, P.C. Capon & R. Philips, "The Portable Compiling System of MUSS", Software - Practice and Experience, Vol. 9, pp645-655, 1979
- Derrick Morris & Roland N. Ibbett, "The MU5 Computer System", Macmillan, 1979