TI-82 Early Hardware and ROM Versions

[critor]

So all versions 3. to 12.0 seem to use the same "9TIETMB-30A" PCB.
Version 18.0 is using a new PCB, "9TIETMB-31".

[CVSoft]

So all versions 3. to 12.0 seem to use the same "9TIETMB-30A" PCB.
Version 18.0 is using a new PCB, "9TIETMB-31".

Indeed; hopefully I'll find out in which hardware revision that was changed. The only difference I could immediately notice between the two PCBs is the addition of resistor R21; whether this has to do with the change from OTP-ROM to Mask-ROM I am not certain, but it is likely. The added resistor is in the proper area for it to affect the ROM chip, yet I lack a complete enough schematic needed to confirm its purpose. A partial (and rather low quality) trace diagram I made can be found here.

I do not have pictures of the front side of the mainboard, and my ROM 18.0 TI-82 is rather damaged on that side; my repairs obscure many traces there. Spilling the keys is always fun to clean up .

[CVSoft]

After poking around in the assembly shell Ash, I ended up finding the ROM call/jump table for routines not in ROM page 0. Interestingly, it gradually increased in length as newer OS versions were made, and the insertions were made at various points in the middle of the table.

ROM version	Table entries	Table beginning
1.0.	476	$2E63
3.	478	$2E63
4*	478	$2E63
7*	478	$2E71
8.0	478	$2E71
9.0	480	$2E71
10.0	480	$2E72
11.0	481	$2E72
12.0	481	$2E76
15.0	483	$2E86
16.0	483	$2E86
17.0	483	$2E86
18.0	483	$2E86
19.0	483	$2EA0

While this information is more useful for assembly programmers, it does show a noticeable gap: it is extremely likely that there is a ROM version 9.0 whose table contains 480 entries. Furthermore, ROM 9.0 lies between the pre-'A' and 'E' hardware revisions; it could very likely exist either as June 1993 pre-'A' or as one (or more) of the hardware revisions between 'A' and 'D'. We have been unable to find the hardware revisions 'A' through 'D' as of this posting, so we have no evidence saying that these all use ROM 10.0. 9.0 has been discovered and dumped; it uses the same offset as ROM 8.0, but has a length the same as 10.0's.

[EDIT] Now that ROM 15.0 has been dumped, we now know it has 484 entries and begins at $2E86, the same as ROM 16.0. I have tested assembly shells Ash and CrASH, and both are compatible. The ROM call tables are minimally different, so programs should run just fine.

[CVSoft]

When I first assembled the TI-82 database, I noticed something odd: HW-'F' TI-82s always appeared after HW-'H'. As it turns out, HW-'F' is surprisingly unusual as it contains a mask ROM instead of the OTP-ROMs used at the time! These HW-'F' TI-82s use ROM 10.0 on a Sharp LH531 mask ROM IC; compare this with HW-'E', which uses ROM 10.0 on a Toshiba TC541001AF OTP-ROM. As it takes longer to produce a mask ROM compared to a OTP-ROM, this caused HW-'F' to appear later in production. Compare three hardware versions here, tabulated alongside the datestamp on the PCB (of year 1993):

HW-'E'	HW-'F'	HW-'H'
17 May - 23 May	28 June - 4 July	31 May - 6 June

Furthermore, there is no evidence of any post-assembly modification of components, unlike all OTP-ROM TI-82s I have observed. The ROM version, unlike HW-'E', is not written on the PCB.

HW-'F' was superceded by HW-'K', which uses ROM 16.0 on a Sharp LH531 mask ROM. These appeared in August 1993, only a little more than a month after HW-F calculators first appeared. HW-'H', with ROM 12.0, appears within approximately two weeks of HW-'E', leaving a narrow window where HW-'F' was prepared for production. This brings a new question: why would ROM 10.0 need to be replaced so quickly, given the cost of creating a new mask ROM?

[CVSoft]

News has been a bit slow, as the North American market has been depleted of interesting TI-82s. I have been focusing on finding TI-83 ROM 1.0 or 1.0100, and so far I have been unsuccessful; the task involves finding the TI-83's HW pre-'A', which is apparently very rare (if it exists). Other work of mine has involved running assembly programs on TI-82 ROM 10.0 with a modified version of Ash and programs reassembled to use ROM 10.0's different ROM call locations. After I solve some issues involving lockups upon exiting Ash, I will publish results; due to copyright, I cannot redistribute Ash, but my ROM call correction script can correct .82B files in addition to .82P.

Between HW-H and HW-Q (and probably changed beginning with HW-M), a new mainboard design was made. I have noted these differences between the HW-H and HW-Q TI-82s:

• On the display board, J9 is present and J10 is absent on HW-Q. On HW-H, J9 is absent and J10 is present.

I believe this change had an effect on the screen's contrast when many rows are dark. The MODE screen shows this change quite well.

	pre-'A'	'Q'
Self-Test Screen Low screen utilization
MODE Screen High screen utilization

While I cannot say for sure as I do not have HW-'J' and HW-'K' available for PCB images, pictures of 'J' and 'K' show this pattern. While this corresponds with what is believed to be the change from OTP-ROM to mask ROM, it could also indicate this change in the display board circuitry. The increased darkness towards the bottom of the screen on the ROM 3. calculator results from my repositioning of the foam strip below the bottom of the bottom of the visible screen area, such that the foam is no longer in the viewable area. The original position, however, presses down on the LCD, causing the darkening.

[critor]

Under french and european laws, I think you may be able to attach the modified version of Ash here, as it is for compatibility/interoperability.

Indeed, owners of TI-82 with ROM 12.0 and lower won't have another way of running released asm programs.


You may also ask the authors for their permission.

[CVSoft]

I have acquired a TI-82, 0873670 I-0893 K, and taken more circuit board images. Unsurprisingly, it uses ROM 16.0 on a Sharp LH531 mask ROM. It does not incorporate the display board configuration change seen in later TI-82s.

TI-82 'K'	Back	Front
Visible light
UV Fluorescence

In the UV Fluorescence Back picture, there appear to be fingerprints visible on the LCD board, near R28 and the LCD-mainboard ribbon cable.

Components:

• CPU: Toshiba T84C00AM-8, 93/28 (4 Jul - 10 Jul 1993)
• ROM: Sharp LH531, 128 kB, 93/33 (8 Aug - 14 Aug 1993)
• RAM: Epson SRM2A256LLM10, 32 kB
• ASIC: Toshiba TC14L010AF, 93/26 (20 Jun - 26 Jun 1993)
• PCB: 9TIETMB-30A, 93/30 (18 Jul - 24 Jul 1993)
• LCD: Toshiba T6A04

[CVSoft]

After a bit of Google, I found instructions about triggering a TI-82 OS bug here that states that ROMs 16.0, 17.0, and 18.0 are affected. Basically, create a program that contains one line of code:

Code: Select all

:Shade(1,2,6

with no closing parenthesis. Put any equation in Y1, and run your program. While the graph is drawing, press [ON] to break the program while the graph is drawing. Instead of the error screen, you'll be at the homescreen with a number in small text on the first line. Type in a character, press [ENTER], and the calc will crash in a manner that depends on which character is entered.

After a bit of testing, I found that the bug is present in ROM versions before 15.0, but the calculator merely fails to redraw the homescreen; it does not crash. The instability was introduced in ROM 15.0, and is present in versions up to and including 18.0.

ROM 19.0 properly handles the ERR:BREAK.

[critor]

Thank you very much for sharing.

This is out of topic, but any news on TI-81 dumping by the way ?

[CVSoft]

Thank you very much for sharing.

This is out of topic, but any news on TI-81 dumping by the way ?

I still don't know anything about the OCR used to dump the other ROM versions. I have (or at least, should have) the videos for both 1.1 and V2.00. The dumping program ran about twice as fast on the ROM 1.1 TI-81, and hopefully the LCD response time is fast enough. I can try to transcribe 1.1 by hand, but it'll be a laborious process.