old/new ELF II BASIC/Math board

Most recent revision of this page Apr 23 2025(c) Herb Johnson. In March 2025, Ed Kilan offered a new build of the Elf II BASIC/MATH board. I obtained one of the new-layout cards. I've worked on obtaining and disassembly of the BASIC software and obtained docs for it and its National MM57109 math processor. This page provides those documents. Because the National MM57109 is very different from other 8-bit microprocessor products, I've also found and described other applications of the National processor from the era. This page is work in progress as of date given above, so check this page over time, and please inform me of errors and corrections.

Another Web page introduces the Elf II I obtained from Will Dotson in 2024. Work in 2024 was to figure out his modifications. This Elf II is not operational as of this date.

Some of these links are to a cosmacelf groups.io discussion group or their file archive. You may need to be a member of the group to get file access. Check the home page of the cosmacelf groups.io for membership information. - Herb

Newly produced BASIC/Math board, ROM recovery and operation

Andrew Wasson's AVI Elf II github site has an image and new schematic for the BASIC/Math board layed out by Josh Bensadon in Feb 2025. Here's a PDF of the new schematic by Josh. There's a few changes from the Netronics schematic. (Also the new PC board includes the jumper wire that was added to the original board as per the assemble instructions.) The Gerber files are in another section of the site.

The AVI Elf II github project site is maintained by Andrew Wasson. The project is a reproduction of the Netronics Elf II (rev D) as redesigned by the late Ed Keefe, with work by several developers as listed on-site. This Web page is focused on the original and Bensadon/Kilian's reproduction BASIC/Math board of the Elf II, the BASIC ROM software, and the National MM57109. The Gerber files for this BASIC math board, and the other redeveloped Elf II boards, are on the AVI Elf II github site.

Here's Ed Keefe's BASIC/ROM board less processor. Ed Keefe restored an ELF II, including the BASIC board. His Web pages describe his process. (Ed's ROM dump, BASIC board image and other content here are courtesy of the Ed Keefe estate.)

Here's the board produced by Alan Kilan and the solder side. It happened to be produced with ENIG gold (gold over nickel) over the entire board, in order to have some kind of gold over the edge connectors. Some other producer might provide ENIG edge-connectors, or hard-gold, or some other treatment. The Web will inform you about PC board production features.

The layout for the reproduction board was created in response to a request from Alan KIlan in mid-Dec 2024, posting "WTB: Basic board" on the cosmacelf groups.io discussion list. Alan wanted a replica to add to his replica Elf II, while providing his original Elf II to a museum. By Feb 2025 Josh Bensadon made a CAD layout of the board. Alan had boards produced and distributed his extras in March 2025. Discussion continued about who had original ROMS, where the ROM images were archived, and various progress reports on use of the original or replica boards.

I checked the ROM images available in Feb 2025 and found they were faulty. When I checked around to find original ROMs or their dumps, Josh Bensason was able to obtain Ed Keefe's ROM dumps. Also Josh dumped the ROMs from a BASIC board he had from the estate. The recovery from two sources marched. Disassembly of these ROM images (by I and others) confirmed they were likely good. These images are on the cosmacelf groups.io files area and are also on this Web page.

build results

On Mar 26 2025, Andrew Wasson reports on cosmacelf groups.io, that he got the new BASIC/MATH board running with the ROM images loaded into memory (via a WAV file loaded into RAM) and the BASIC appears to run and execute immediate PRINT commands. He's in the process of finding and programming 2716 ROMs for the board. (Apparently he made his own boards from the Gerber files on github.) Alan Kilian reports on Mar 25 that he has his board running but also needs to find and program PROMs.

backward diode, Giant board fixes: ".. and if you're building a new board, remember the polarity of Zener diode D1 on the BASIC/Math/ROM board. It shows backwards in the original Netronics schematic [but not in Josh's new schematic and as indicated on the new board]." Also don't forget, "There's a [Giant Board diodes] modification to the EF2 signal because it's used (now) by both the Netronics Full Basic/Math/ROM board and the cassette output function on the Giant board." - Alan Kilian, April 2025. Alan got his new board and BASIC running.

Apr 22-23 2025 Costas Skordis, negative supply: Costas reports on cosmacelf groups.io of building the Math board for his Elf II rev D (new version), getting BASIC to operate, but no operation of the Math chip. BASIC hung up whenever a math operation was performed.

Diagnostics as suggested, showed the math chip was being I/O addressed by BASIC (at the logic-signal level, results of I/O operations 62 and 6A, and EF2). Also, that the local oscillator was running (but at 233KHz and a .022uf cap). I asked about negative voltage for the MM57109 chip. Costas said there *seemed to be* -4V available at the Vss pin of the MM57109 chip. However, Costas didn't get the chip to respond until 1) he replaced the D1 Zener diode in the -4V supply circuit and 2) he fiddled with the Elf II rev D jumpers related to the negative 12-volt converter on the (new Rev D) Elf II board. But after this work, his math chip responded and BASIC performed accordingly.

Andrew Wasson, reported in reply, he had similar Math board problems until he also "played with the 8-volt, minus 8-volt jumpers" on his Rev D Elf II.

I was confused by this negative supply business. there's no negative power supply on the Elf II original board and none noted in the original documents. Then Costas told me he was running a Rev D Elf II, not an original Elf II.

Documents for the orignal Elf II and Giant board, showed that in the era a -12V supply was an optional add-on, not part of the Elf II supply or wired normally onto the Elf II bus. A negative supply was used by the Giant board to generate RS-232 (negative voltage) signals. Both the Giant board and the original (and new) Math board specify that a negative supply be wired to pin 15 of the bus as a user upgrade. (I do not know if the Elf II Rev D has pin-15 connected across all bus boards.) Thus, it's important to source a negative supply for use of the Math board, connect it to the Math board bus pin 15, and verify that voltage.

Regarding Math chip oscillators: Kilian previously reported his new-build Math board has an oscillator that runs at 250Khz; he changed the resistor value to increase it. I checked various R/C logic-buffer oscillator circuits on the Web, on how frequency depends on the R and C values that produce a clocking analog voltage. These circuits vary, but a typical calculation is F = 1 / (1.39 × R × C). 18K ohms and 100pf seems to calculate out to 399.6 kHz. David Madole suggests that a non-50% duty cycle may cause problems at higher frequency; but notes the National docs state details and a max frequency of 400Khz. I suggested a larger-than expected cap would slow the frequency down; Costas reports his cap is .022uF (22000 pF?).

Supporting Netronics BASIC/ROM docs, ROMs, codes

Ed Keefe restored an Elf II some years ago. Here's his Web page about that work. including efforts on the BASIC/ROM card. Web search will find several sites about the Netronics ELF II.

From Charles Morford as uploaded to the cosmacelf groups.io files archive, are Netronics docs as follows. chapters 1 and chapters 2 and 3 of "Full BASIC Level III" for the BASIC board. The BASIC is contained in three 2716 ROMs (6K) as supported by the National math processor chip. The original Netronic build and schematic for the board, are in this document "Full BASIC Math/ROM Board - Assembly Instructions".

Here's a dump of the three ROMs on Full BASIC board. The first ROM dump available was incorrect, the ROM for 0000H was all EDh and FDh bytes. Those were replaced by Josh Bensadon March 8 2025 with a good set of ROM images, from Josh's original BASIC board, and a ROM-dump file set from Ed Keefe's estate. They were identical.

I and others are disassembling the ROMs. I've performed a rough disassembly of the three BASIC ROMs. I'll update it and/or provide other disassemblies as they become available. The Mar 19 2025 version, has been edited then assembled, the binary compared to the original ROM dump, with no differences in the binaries. Further disassemblies will identify 1802 coding methods and call out ASCII and National-format data; this will take time. As I'll describe, the features of the BASIC and processor are unique even for the early 1980's era of 8-bit microprocessors.

Features of the board, Netronic BASIC and National math processor

The BASIC from Netronics, uses "Polish" or RPN (Reverse Polish Notation, also called postfix notation) to represent math expressions. RPN was in use on many HP brand scientific calculators of the late 1970's. Hand calculators almost always operate in decimal - digits 0 thru 9. The byte-I/O-based National processor, derived from National calculator designs, likewise used BCD and RPN processing . All data in and out is in decimalized digits, commands are byte values. Thus the Netronics BASIC is RPN and in BCD, because the National math processor is RPN and in BCD. That is relatively unique for that era.

The BASIC/ROM board consists of the National MM57109 math processor, two 8-bit data latches, and four ROM sockets. The National processor operates from 1802 INP OUT instructions. OUT 5 from the 1802 loads a byte into the processor, INP 2 on the 1802 reads a BCD value from the processor. There's a "ready" from the National processor that is read by /EF2 to see if the processor is ready for another operation. Also data ready (bit 6 INP2) and error (bit 5 INP2).

The four ROM sockets on the board, support 2716 ROMs, 2K bytes wide and 5-volt powered ROMS. Only three ROMS support the Netronics BASIC, which run in the lowest part of memory (0000H-17FFH). RAM in the highest part of memory is required - BASIC looks through memory to find available RAM. There's other INP and OUT ports assumed by BASIC for input and outputs from BASIC, their hardware is elsewhere on the ELF II.

"The MM57109 processor operates on a 9V power supply. In order to make it TTL compatible, it can be operated from supplies of 5V and -4V." That's from the National data sheet. In the Netronics schematic, pin 21 Vdd is connected to +5V, and pin 15 Vss is connected through a 270 ohm resistor to a -8V supply. The -8V supply also is a negative reference for some of the MM75019 lines through 10K pull-down resistors.

In the original BASIC/ROM docs, a wire and PC layout change is called out, including a connection to edge connector pin 14. That change is part of the Bensadon layout and need not be made. The original schematic shows the results after the change (and as represented in the Bensadon layout).

Caution: do not attempt to use TI brand "TMS2716" ROMs on this board. They are not compatible with other brands of 2716 ROMS. The TI brand TMS2716 uses three DC supply voltages - like a 2708 ROM. Web research will explain why this situation occurred.

Read the Netronics documents and National documents as referenced on this Web page for more details. Discussions of RPN and early HP calculators can be found on the Web. Because of the unique features of the National math processor, I've gone to some efforts to find and describe the few uses of the MM57109 processor in the period. My hope is, more information will providemore opportunities to understand it. The fear is, "more information creates more confusion". My regrets if some people are overwhemled or uninterested. - Herb Johnson

COSMAC results

Ipso Facto and ELF II math test codes

The Ipso Facto small-press magazine covered Netronics and other COSMAC products. Ipso Facto issue 28, Mar 1982 published an article "Full Basic Number Crunch Sans Basic" by Tony Setaro. He produced some simple 1802 codes to access the processor and perform some simple math. These test the National processor without the BASIC ROMs. The codes were hand assembled, hex bytes with comments. I edited the codes into a binary file and disassembled them, to regenerate 1802 assembly language source for the codes. Thanks to Marcel van Tongeren, who did the same and verified our codes match.

Here's a ZIP file of the Ipso Facto article and the disassembled codes. The code as given needs further testing and refinements, it's work in progress. So contact me with any results and improvements. I'll update it accordingly.

EMMA 02 emulation

An EMMA 02 emulation of the BASIC as part of ELF II emulation is in progress as of late March 2025. Check the EMMA 02 Web site for status and details of his emulation of the board and math processor.

Questar Engineering NOM, a variation of the Netronics MATH board

Thanks to Andrew Wasson, who found the following articles in Radio Electronics. Apparently Questar Engineering produced their own MM57109 math board and provided a kit. It's built on a 4 X 6 inch board with 44 pin connector. It appears at a glance, to have the same math processor features as the Netronics board, but no on-board ROMS (or Netronics BASIC). Here's an image of the Questar Engr. schematic for your comparison to the Netronics product.

The build articles are in Dec 1978 Radio Electronics mag and Jan 1979 Radio Electronics mag Two part article is titled "Build this NOMCard For The 1802" by L. Steven Cheairs. The article starts, "Add-on math board for an 1802-based microcomputer. Based on a number-crunching IC, this board speeds execution time, reduces software overhead and saves memory". Those are claims similar to those in the National documentation. NOM means "Number Oriented Microprocessor".

The article series includes a schematic, PC layout, parts list and how-to-order info for Questar Engineering (Mesa AZ). Questar also offers an "ELFII to SB-44 converter board", and a PROM with 1802 monitor and code to operate the NOM (no BASIC). Apparently Questar Engr. intended to offer a similar board for the S-100 bus. and other 44-pin boards consistent with their "SB-44 bus". Cheairs may be affiliated with (or actually is) Questar Engr. as other R-E articles he authors are based on kits available from Questar Engr. (This Questar is not the same company as the Questar in California which produces optics and telescopes.)

Non-COSMAC applications and docs of the National MM57109

National documentation

A copy of the National datasheet for the MM57109. My local copy.

My version of the MM57109 instruction set and data format.

A National Semi 10 page AN-186 app note on the processor. And here's my local copy.

The US patent on the processor.

Z80 application - PSI Comp 80

A UK Z80 computer "PSI Comp 80" used the processor. BURP was the ROM BASIC which supported the processor. There's many other Web pages about the PSI Comp 80.

Another PSI Comp 80 Web page, which links to the several Wireless World magazine construction articles about it. One of the articles includes a hex dump of the 2708 ROMs for it, and some patches. Again, this is Z80 code that implements a BASIC that is supported by the National processor.

SWTPC application - MP-N

The SWTPC 6800 had an optional I/O board which supports the MM57109 called MP-N. The board amounts to a 6820 PIO that operates the processor. There does not seem to be any SWTPC software products that were supported by this board. A "calc-1" program was listed and described in the MP-N docs to demonstrate the board. The MP-N seems not to have sold well, and they are uncommon to find in the 21st century.

Here's the calc-1 source and here's a symbol table for it. The MP-N manual has this description of the calc-1 program. There's a correction for some of the manual and code, by Daniel Long as I explain below.

A Univ of North Texas student project in 1979 by Daniel Paul Long, produced a document about the MP-N board for 6800 math calculation in assembly for a subsequent project. His report includes the MP-N SWTPC and National document exerpts, comments and corrections on calc-1, and some results of use. He chose *not* to use the processor - it was too slow - but instead used 6800 assembly for his trig calculations. Thanks to Neil Cherry for finding this document.

Here's the original document but this is my on-site copy and the OCR text of his report with some tables he produced or modified. The bulk of his paper copies the SWTPC MP-N document, as available in the Deramp.com SWTPC archive two paragraphs above.

Another paper by Long, is about his use of a SWTPC 6800 in assembly code (without the MP-N) to perform operant conditioning reinformcement experiements on rats. Statistical processing of results is in PL/I-like code on some mainframe computer. Code is listed in the document. The 6800 portion of the code is mostly about operating the test apparatus on a timed interrupt, a terminal interface, and a bit of seconds-counting and multiplication math. Again, Long chose not to use the MP-N card in this code.

Other vintage math coprocessors

In the same era and class as the National MM57109 were the AMD 9511 and 9512. A licensed version of these were the Intel 9511 and 9512. (They used one of two floating-point representation schemes but were pin and logic compatable otherwise.) They found use on a number of vintage computers of the late 1970's. Otherwise, like the COSMAC 1802, 8-bit vintage computers implemented floating-point software for mathmatic processing. Later processors either used floating-point chips, or incorporated whole floating-point ALUs into their CPUs as is done with today's consumer-class processors. Here's a Web page I maintain on vintage floating point primarily in the Intel 8080 - Zilog Z80 world. - Herb

Copyright © 2025 Herb Johnson