The Web Site to Remember National Semiconductor's Series 32000 Family

National Semiconductor

National Semiconductor build many boards for the various members of the Series 32000 family. The customers could use this boards either for learning about the Series 32000 or to design very fast their own products. The boards came with a lot of printed documentation (not to be found on a CD - I'm writing about the 1980's ...). The boards presented here are:

All photos with the label are copyright of The John Gabriel Byrne Computer Science Collection. See Systems/Trinity College for further details.


Wow - this board must be one of the very first boards using an NS16000 chip set. The date on the board is 1981! At this time all the parts of NS16000 were in its infancy (buggy) or even not taped out (ready for production). Thanks to The John Gabriel Byrne Computer Science Collection for saving this rare item!

The documentation of the collection says that this board is the DB16000. DB means Demonstrator Board. Unusual the name is not found on the board like it is for the DB32000.

Fig. 1. The 16K DEMONSTRATION board from National Semiconductor

An excellent high resolution version of the photo in Figure 1 is available here. The board used the Multibus standard for an easy integration in existing equipment. 128 kBytes of DRAM memory (U36 - U51) and 4 sockets for EPROMs (AM2732) and/or SRAMs (HM6116LP-2) were provided. It is possible that the blue connectors could be used for memory expanding. I hope that the board is stored in a dark environment to protect the content of the EPROMs because there is no UV filter on top of them ...

Fig. 2. The CPU "cluster" of the 16K DEMONSTRATION board

The CPU cluster contains only the NS16032 CPU and the NS16201 TCU. The NS16081 MMU is replaced by an emulator board with 4 resistors. Later such an emulator board was not required if the MMU was missing. The NS16201 TCU used a plastic package. This is unusual because the TCU became hot in operation. Maybe the first versions were not so power hungry. It is possible that the design was changed later to support higher speed and more load. But this improvements come always with increased power consumption. It is interesting to note that both devices are not marked as engineering samples.

Fig. 3. The "ICU" of the 16K DEMONSTRATION board

At that time even the simple interrupt controller NS16202 ICU was not available. But a DM74LS92, which is a TTL device containing a divide by 12 counter, could do the job. Ok, joke - the ICU is not this simple.


The DB32000 is an impressive board simply for its size. Its only purpose was the evaluation of the Series 32000 family. All parts of the first generation could be used. For some unknown reason the board in Figure 4 has no CPU in its sockets, neither the NS32016 CPU nor the NS32032 CPU. It looks like the board is prepared to hold a second NS32032 CPU with its own MMU. But no information can be found about this option in the data books describing the DB32000. Maybe this board was prepared to use two of the mythical NS32132 CPUs.

The board has no standard bus connectors but features a large wire-wrap area for user circuitry. Up to 1 MBytes of DRAM memory is available. Eight 28-pin sockets for ROM/EPROM allow 256 kBytes of permanent memory. The DB32000 can operate stand-alone with no assistance from a host computer system. Optionally, the board can be operated in conjunction with a host, taking advantage of more powerful software development tools.

The collection owns three DB32000 boards. Two of them have the NS32032 CPU installed. They saved in addition a lot of printed documentation and some software on reel-to-reel tapes. I hope that someone will read the tapes one day and present the content in the web.

Fig. 4. The DB32000 board of National Semiconductor

The two black 25-pin connectors were used for serial RS232 communication links. It was possible to connect the board to a terminal and a host simultaneously. Please note that the board shown was manufactured in Japan.

The photo in Figure 4 is already very good. But if you want even more resolution look here. More photos can be found in a document at the SCSS Treasures Catalog. Use the link TCD-SCSS-T.20141120.007.pdf to open the document.

DB32016 (DB16000A)

In November 2017 I got a parcel from Finland. The main attraction was a DB32016 board. Beside this it contained two Evaluation Kits and a lot of documents. All together was laying around at an university for a very long time.

Fig. 5. The parcel from Finland was quite heavy and full of stuff.

The DB32016 board was first named DB16000A. It is very similar to the DB16000. One difference is the use of PALs.

Fig. 6. The DB32016 board from National Semiconductor. 128 kBytes of DRAM are located at the right edge

The system uses 5 V and +/-12 V for RS232. Current consumption of my board is 3.7 A at 5 V.

Fig. 7. The cage for the DB32016 board was designed for the Multibus and manufactered by Intel. It is very robust.

Fig. 8. The backside of the DB32016 board. The cage is ready to hold three more boards.

The great question was: is the board still functional? I was very curious. I modified an old power supply and luckily found an edge card connector in my boxes for the serial interface (see it in Figure 8).

The start was promising. A message appeared on the screen, telling me that the board passes the selftest. But then the message was repeated and the process didn't stop. There was no reaction to any keyboard input. Not the result to open a bottle of champagne...

My hope was that some bits in the EPROMs had flipped over the last 30 years. I exchanged them with some new 27128 devices. But the result was the same. Therefore I have to dig deeper to find the root cause. My candidate now is the CPU. But I will first check what the program is doing.

In-depth Analysis

I started with a simulation of the program on my M32632. I had to add a simple Intel 8251 UART model. In addition I disassembled the code. Now I got a good overview about what was happening during the selftest.

I found the instructions which sent the mentioned message. I also found a signal which is only active during selftest. It is called "DIAG_MODE" and it can be used as a trigger signal for an oscilloscop or logic analyzer. Between "DIAG_MODE" becoming inactive and active again there must be the problem. I decided to use a logic analyzer. The measurement setup is shown in Figure 9.

Fig. 9. The measurement setup used an USB logic analyzer with 16 channels from Digilent. The red arrow points to the jumper which caused the loop behavior.

Most of the channels took address signals. It was easy to connect them to the empty EPROM socket. The signal trace in Figure 10 shows the moment when the error happened. Between -25 Ás and -15 Ás the wait loop at address x'102 is executed. After the loop a CXP (Call External Procedure) instruction should follow. In my opinion this instruction is not executed because the three writes for saving the program counter and the mod register are missing.

For me it was obvious that the CPU must be broken. I replaced the device - and got the same behavior. Uuuh! After some thinking about other possibilities and studying the schematics and the documentation I identified a wrong jumper. There are a lot of jumpers on the board and this one selects the 2732 EPROM. If 27128 are used the jumper has to be removed...

Fig. 10. The signal trace and the program code at the moment when the error occured.

Once again the start was promising. I got the desired TDS message together with the prompt for input. Success! A first dump showed some memory locations. But the disappointment quickly follows. The command "in" to fill the buffer of the editor was answered by the system with a "?". This is the sign for an input error!

Fig. 11. A screen shot after power up shows one successful command and one failing command.

A look at the command table of TDS showed that the "in" command is not contained in the table. Some other commands are also missing, for example the assembler. This is definitly different from what the documentation says. Therefore the board is useless in the moment.

I have no idea what is going wrong with this version of TDS. Hopefully there are other versions available somewhere.

Thanks to Gilbert!

At the moment I know only one person who has a running system using TDS. It is Gilbert who owns a DB32000. I described him my problem and he sent me a logfile of the communication between his system and the terminal. I saw immediately the cause of my problem: TDS needs after power up an initialization command. Obviously I have not read the manual carefully enough - a bad attitude.

The next screen shot in Figure 12 shows a successful command sequence. The DB32016 is connected to my NS32532 based TITAN3 which I use as the host. The green text on black background is the output of the DB32016. It shows the result of a floating point test program. Please note the unusual display of FP numbers.

Fig. 12. At the end DB32016 is doing what the user wants. That's the purpose of all computers.

TDS offers read and write commands for using a tape drive as a backup memory. I think it is a better idea to use an SD card for this purpose. It should be no big deal - just an interface to the Intel 8255 PIO and a voltage regulator. The software will be put in two EPROMs which use the free sockets. Then DB32016 is well prepared for being presented at vintage computer exhibitions.

I'm now very happy with the board. Since nearly 20 years I was not able to run programs on an NS32016 CPU. And it is my first running vintage computer.

Is it a surprise that I will be looking for a second DB32016?


National Semiconductor designed the ICM-3216 for OEMs who needed an embedded computer for their products. An example will be shown here in the near future.

The "Integrated Computer Module" ICM-3216 is a complete computer system contained on two 11.02 in. (300 mm) x 9.18 in. (233 mm) printed circuit boards. One board contains the CPU cluster, PROM sockets, serial interfaces, address mapping logic, SCSI interface, parallel port, memory interface and the MiniBus interface, see Figure 13. The other board contains the DRAM memory. The CPU cluster is based on the NS32016 CPU running at 10 MHz and used all support chips.

The SCSI interface was used to attach hard disk drives and tape units. It used an LSI SCSI device and a Z80B microprocessor. This solution took away the details of the data transfer from the NS32016.

Fig. 13. The ICM-3216 from National Semiconductor

There are a number of documents available describing the ICM3216 and the MiniBus.

ICM-3216 Advertisement

ICM-3216 Preliminary Datasheet

ICM-3216 Final Datasheet

ICM-3216 CPU Board Specification

ICM-3216 Memory Board Specification

MiniBus Specification

MiniBus Interface Chip Specification


The VME532 was the most powerful board level product of Series 32000. It is based on the NS32532 CPU. The VME532 consists of two boards which are closely connected by a local bus: a CPU board and a memory board. The boards are made according to the VME bus standard and can be mounted in a VME rack. The size of the pcbs is 233 x 160 mm. Figure 14 shows the frontside of the VME532 pair.

Fig. 14. The front side of VME532. The selector for the CPU ID is placed on the memory board.

I got the photos from the only person I know who owns a VME532. It is astonishing for me how rare this boards are. Maybe an explanation is that the owner aren't willing to sell this nice board at ebay...

The VME532 CPU board in Figure 15 shows the NS32381 FPU sitting in a larger socket. The board has the option to use the NS32580 FPC instead of the NS32381. For further information see the application brief AB-40. Using the FPC requires also the use of the Weitek WTL3164 FPDP for which the now empty 144-pin socket XU62 is provided.

The owner of the board has long tried to get the NS32580/WTL3164 pair. But he was not successful. It seems that this parts are even rarer...

Fig. 15. The CPU board of VME532. This photo is available in higher resolution here: VME532 CPU.

For high performance the CPU board has a 64 kBytes cache. The data memory of the cache is made of 8 fast SRAMs from Performance Semiconductor. The SRAM device P4C188-25PC is organized as 16k x 4 bits and has an access time of 25 ns. They are located at the left edge. The tag memory of the cache is made of four special SRAMs from SGS-Thomson. The tag memory chip MK41H80N-25 contains a 4k x 4 bits memory array and a 4-bit comparator. 12 address bits and one valid bit are used for the tag. The result is a 256 MBytes cachable address space. The tag memories are located at the right edge. Both types of memory use a 22-pin DIP package.

The VME532 was designed for multiprocessor systems. Up to 16 of these boards could be used in one system. A thumbwheel switch enables user to assign unique CPU identification numbers, ranging from 0 to 15. The local memory could be accessed by other VME masters including other CPUs.

Fig. 16. The back side of the VME532 CPU board. Some wire fixes were necessary.

The backside of the CPU board reveals an interesting detail: some smaller ICs are placed below the bigger ones. For example the device U59 is placed below the 512-kbit EPROM XU60. The EPROM is the device with the label "SYSMON VME532 8/22/88" on it. Two ICs could be placed below each 40-pin DIL device which are the NS32202 ICU and the MC2681 UART.

Fig. 17. The memory board of VME532. This photo is available in higher resolution here: VME532 Memory.

The memory board in Figure 17 contains 4 MBytes of parity protected DRAM. 36 1-Mbit devices in ZIP package are used. The DRAMs are soldered to the board. Therefore an easy upgrade to 16 MBytes is not possible although the 1-Mbit and 4-Mbit DRAMs are pin-compatible.

The 1988 data book mentions a 12 Mbytes Local Memory Expansion. Whether this is an additional board plugged in the connectors on the upper side of the CPU board or an equipment option between 1-Mbit or 4-Mbit DRAM device is not clear to me.

The devices in 124-pin PGA packages with the marking SCX6244 are custom chips. The VME532 development team designed them to implement the DRAM controller and the delayed write buffer. The biggest available CMOS gate array at National Semiconductor of this time providing 2800 gates was used. The design was done manually because of lack of synthesis tools. Nevertheless the custom chip was fully functional at the first shot. A datasheet of SCX6244 can still be found in the internet.

In my view the memory board contains the interface to the VME bus. Quite a lot of SN74AS652 bus transceivers and SN74AS374 register chips are placed near the VME bus connectors at the edge of the memory board.

Fig. 18. The back side of the VME532 memory board. Again some wire fixes were done.

Fig. 19. Another VME532 memory board.

The VME532 memory board in Figure 19 is a noteable one. According to the serial number "3" at the left edge it must be one of the very first boards. Another hint is the marking "PROTOTYPE" on the gate arrays. Very unusual is that no labels are placed on the PALs. How they are identified? A lot of bipolar PALs from MMI and Texas Instruments were used. One electrically erasable PAL from Lattice was already used too. In the photo the GAL16V8 can be found in the middle of the lower edge. In 1987 GALs were very new.

This chapter was last modified on 11 February 2018. Next chapter: Opus