TI-74 TI-95 8x32K banked RAM/FLASH cart soon, ROM images needed

[Jeff_Birt]

(10-18-2024 06:22 PM)prino Wrote:  I think there was a later revision of the TI-74 that could see 32K. If I remember correctly, it was supposed to have a three pixel-wide separator in Calc mode.

I wonder what serial # range this started in. One TI-74 I have is #0000383 and the other two are in the #0005xxx range.

[dmh]

Could you please provide more details on exactly what to look for regarding the "three pixel-wide separator". We can compare and provide serial numbers and see if we can work out if/when the full 32K visibility became supported.

(10-18-2024 06:22 PM)prino Wrote:  

(10-18-2024 06:16 PM)toml_12953 Wrote:  When I had the TI 32K developer's cartridge, the 74 could see all 32K as BASIC program memory and added it to the 8K built in for a total of 40K. I didn't have the 74S (or at least it wasn't marked that way anywhere on the case).

I think there was a later revision of the TI-74 that could see 32K. If I remember correctly, it was supposed to have a three pixel-wide separator in Calc mode.

[Jeff_Birt]

I did get a hold of all three TI produced TI-95 ROM carts and borrowed the missing TI-74 carts. My friend adapted a device we had used with PC-1500 carts to dump them. The results were surprising. The ROMs seem to be much slower than their data sheets would indicate. I seem to recall seeing this mentioned somewhere but can't find the forum post now.

The ROMs are 32K with the upper and lower 16K banks swapped. We are doing a direct dump from the cart and did not expect this, but the banks have to be swapped to work on the FLASH cart. Again, this 'swapping' of banks has been mentioned in forums but with no explanation. After trying to suss it out for hours I think the ROM must be inverting A14.

A few of the carts still don't sump properly. This may be further timing issues and coincidently revolve around A14. Some more experimenting this evening to try and figure it out. Thanks to my co-conspirator for helping figure this out and sending me new code to test (even when he is on holiday.)

[badacktor]

I wonder to what extent the hackiness of those cartridges is related to the trouble TI had during the doomed Project X and TI-88 development. I don't know if it has ever been confirmed if there's much of anything in the architecture of the chipsets of the 88 and the 95/74 that share any development or lineage. I should probably go re-read the Datamath articles on those topics. I think I just always assumed they did because it seemed narratively obvious.

[Jeff_Birt]

We did finally get all the data from the cart by dumping the whole 64K address range. Which resulted in two copies of each 16K bank in this order: B2 B2 B1 B1. Cutting out the extra copy and swapping banks to: B1 B2 provided a working image that could be saved to the new FLASH cart.

My friend who made the dumping apparatus we are using confirmed the board was laid out correctly. Last evening I was able to manually set each address line (there are CLI commands for that) and confirmed the correct signals make it all the way to the cart.

It is a very odd problem. Perhaps there is some weirdness with my Teensy 4.1? When my coconspirator returns from aboard, he will repeat the experiments with his system.

[Jeff_Birt]

We found the ROM datasheet and it turns out that it is an exceptionally slow part, i.e. measured in uS not nS. datasheet

Will try some tweaking on the read timing of the cart dumper and see if this oddness makes more sense.

[brouhaha]

(11-14-2024 08:03 AM)badacktor Wrote:  I don't know if it has ever been confirmed if there's much of anything in the architecture of the chipsets of the 88 and the 95/74 that share any development or lineage.

The chips show little or no sign of common lineage. The microcontrollers in the TI-88 are an evolution of the TMS0100 and TMS1000 series of 4-bit micros, and those of the TI-95 and TI-74 are TMS7000 series 8-bitters, with no significant similarities. TI had a separate line of word register (vs TMS1000 digit register) calcualtor chips, typified by the TMC0500 series used in the SR-50, -51, -52, -56, -60, TI-58, and -59, but the TI-88 chips came from the digit register family.

The TMS7000 series was developed as general-purpose competition with the 8048, MK3870, and similar low-end 8-bit microcontrollers.

Oddly enough, Intel, who had pioneered both 4-bit and 8-bit microprocessors with the 4004 and 8008, did not make any single-chip 4-bit microcontrollers, and was not first to market with a single-chip 8-bit microcontroller. (That may have been Mostek with the MK5065 introduced in 1974, which had been designed by Motorola.)

Aside from DSP processors, which TI has been very successful with, for much of its history after the introduction of the TMS0100 and TMS1000 familiies, TI has been an "also-ran" regarding general-purpose microcontrollers and microprocessors.

[Jeff_Birt]

(11-15-2024 02:41 AM)brouhaha Wrote:  Aside from DSP processors, which TI has been very successful with, for much of its history after the introduction of the TMS0100 and TMS1000 familiies, TI has been an "also-ran" regarding general-purpose microcontrollers and microprocessors.

TI sure seems to have 'half-A$$ed' a lot of projects. I can't help but suspect this was a management issue. They had the engineering chops but lack of a coherent vision as to what the heck they were trying to do w.r.t. computers. All the way back to the TI-99 fiasco (or before).

[AnnoyedOne]

(11-15-2024 02:41 AM)brouhaha Wrote:  ...TI has been an "also-ran" regarding general-purpose microcontrollers and microprocessors.

I like the TI MSP430 line. If low-power is your thing they're the solution. At a conference I once saw one design powered by a lemon.

As for the TMS7000 a company I worked for in the late 80's used a lot of them in a large system. I didn't have to develop and hardware/firmware for them specifically although a project I did work on required some knowledge of their design. Even back then it seemed like an odd choice... Not mine.

As for the ROM cart--weird.

A1

[Jeff_Birt]

OK, we tracked down the weirdness with carts. There were two issue that combined caused a lot of confusion.

First problem was that the TI-74 technical manual had the pin designations of A14 and A15 swapped. Taking apart an original ROM cart and comparing the wiring with the ROM datasheet (which was a bit hard to read) indicated this was most likely. Taking apart a vintage 3rd party EPROM cart confirmed this.

Second problem was the exceptionally slow access time of the TI ROM (~4.5us read cycle). Once the correct datasheet was found the dumping code was modified with the correct timing and to swap A14 and A15 we were able to dump the carts correctly.

Fortunately, it is an easy process to fix the ~12 carts we currently have made up. I modded one tonight and now things work and make sense.