VOGONS


First post, by feipoa

User metadata
Rank l33t++
Rank
l33t++

Recently, I created an interposer module to translate the 168 pins (486 socket) of a Texas Instruments TI486SXL2-G66 to that of a 132 pin socket (386). From what I could discern after reading the TI486SXL Reference Guide, found here: https://archive.org/details/bitsavers_tiTI486 … eGuide_19799789 , the TI486SXL2 PGA168 should be the same as the QFP144 and PGA132 variants with minor exceptions. QFP144 and PGA168 have the FLUSH circuits built-in (although you do not have to use them if you don't wire up MEMW#), they have a FLT#, which can be left floating as it contains a built-in pull-up resistor, they run at 3.6 V, and have a few more GND (Vss) and voltage (Vcc) pins.

So what would prevent someone from just wiring up the PGA168 pins to the location of the PGA132 pins on an interposer socket and using a voltage regulator module to run the CPU at 3.6 V? The TI486SXL2-66 chips in PGA168 format are readily obtainable new for $15. An upgrade company made such an interposer for the QFP144 TI486SXL2-66 and another user here has tested it. The TI486SXL2-66 (PGA-132)... & repair! and Re: The Systempro; Benchmarking (TI486SXL2-66)

Aside from the IBM BL3 and a Transcomputer module, the TI486SXL2-66 would have been the fastest upgrade for your aging 386. The SXL2 is still architecturally a 386, but with cache and 486 instructions added. Considering how much the SXL2-66 upgrade QFP144 sold for recently, I figured there might be interest in creating an interposer board for the more common TI486SXL2-66 PGA168. I planned to start with a proof of concept prototype, then possibly create a proper PCB if there was sufficient interest.

Here is the prototype:

SXL2_PGA168_to_PGA132.jpg
Filename
SXL2_PGA168_to_PGA132.jpg
File size
1.3 MiB
Views
7743 views
File license
Fair use/fair dealing exception

I built the prototype using a snap-able solder boards, a MIC29302WT (LDO = 0.37V @ 3A), 10 uF cermaic input cap, 10 uF tantalum output cap, a resistor, a trimmer, and solid core 28 & 30 AWG wire. I tested several linear and switching voltage regulators before deciding on the MIC29302WT. I finished the prototype after about 10 days of casual soldering.

Unfortunately, I could not get the prototype to work in any of my 386 motherboards. In each motherboard, the TI486SXL2-50 PGA132 CPU worked fine though.

SXL2-50.jpg
Filename
SXL2-50.jpg
File size
149.38 KiB
Views
8448 views
File license
Fair use/fair dealing exception

I basically soldered the PGA168 socket directly to the pins of the PGA132 socket using the template found in the TI486SXL reference manual.

Terminal_Assignments.jpg
Filename
Terminal_Assignments.jpg
File size
396.01 KiB
Views
8448 views
File license
Fair use/fair dealing exception

The exception is that the Vcc pins on the PGA168 side get wired to the output of the voltage regulator instead of the 5 V line from the PGA132 side (the motherboard). I have the trimmer set such that the voltage to the CPU is 3.60 V. I confirmed this after the prototype was powered on in the motherboard. I have checked each solder connection three times now and also confirmed that none of the wires sorted to any of the other pins, e.g. while soldering some pin, the cladding on an adjacent wire melts and unintentionally caused bare wire to touch. I checked every wire using the beep function on the multimeter to ensure there was no accidental connection somewhere. Perhaps there is still some cross-talk?

I noticed on the QFP144 upgrade module that there are some extra components, such as additional decoupling capacitors, a NOR gate, and a PAL. Here is what it looks like,

TI486SXL2-G66-HBN.jpg
Filename
TI486SXL2-G66-HBN.jpg
File size
310.95 KiB
Views
7743 views
File license
Fair use/fair dealing exception

I have come across many 486 upgrade interposer boards which also contain programmeable logic device, such as PALs or GALs, and gate circuits. On many motherboards, these were not required. I believe that they were usually added to increase compatibility with various systems, mostly archaic or non-standard motherboards. Considering that all my 386 motherboards work fine with my 5V PGA132 SXL2 chips, I do not think any special logic would be required on my prototype. Could there be some other difference between the PGA168 and PGA132 SXL2 chips that has eluded me? My gut feeling is that there is something very fundamental I have overlooked. Any ideas? Help!

Concerning the increased number of decoupling capacitors found on the QFP144 upgrade unit; while they may help, I do not believe their absence is causing the no signs of life symptoms. For example, in my experience with these IBM 5x86c-100 chips, the dozen+ decoupling capacitors did not change the stability of the system, not even at 133 MHz. I have two such interposers; on one I removed all the caps, and on the other, I left them on . Both demonstrated equivalent stability in the system I tested them on. While I could add one or two extra decoupling capacitors to the TI486SXL2 prototype, I don't really think it is going to help with the absence of life. The prototype, when installed in the motherboards, does not show any signs of life - that is, the screen never displays anything.

IBM-5x86c-Thinkpad_TP365E_mod_for_PGA.jpg
Filename
IBM-5x86c-Thinkpad_TP365E_mod_for_PGA.jpg
File size
274.03 KiB
Views
8448 views
File license
Fair use/fair dealing exception

With any luck, I might be able to borrow one of these QFP144 SXL2-66 upgrade kits from a CPU collector and try to determine what they did differently. Obviously the manufactured QFP144 unit is interrupting some pins and taking them to the NOR gate and to the PAL. Unfortunately, I won't know what the PAL is doing, but at least I'd know which pins are interrupted. Perhaps the PGA168 and QFP144 units need some trigger to wake-up, and that is what the PAL is performing? Perhaps the mystery is with FLT#

I have three of these TI486SXL2-66 CPUs and I tried them all, but the result was the same. I also tried setting FLT# to ground at startup and also to set FLT# to GND after power-up, but the result was the same. FLT# is for floating the bus signal. FLT# and MEMW# are the only two pins the QFP144 and PGA168 chips which are not on the PGA132. I do not really understand what FLT# does, but I will copy/paste here what the reference manual says about it.

FLT#
Float (active low). This input forces all bidirectional and output signals to a 3-state condition. Floating the signals allows the microprocessor signals to be driven externally without physically removing the device from the circuit. The microprocessor must be reset following assertion or deassertion of FLT#. This signal may be used in conjunction with an upgrade socket. FLT# is internally connected to a pullup resistor to prevent it from floating active when left unconnected.

MEMW#
Memory Write (active low). This input is used in the cache interface logic which flushes the cache in systems that hold the CPU during DMA and MASTER cycles.

MEMW# is the pin which leads to the hardware flush circuit (NAND gate and an inverter) inside the PGA168 and QFP144 chips. You connect MEMW# from the motherboard's ISA bus to MEMW# on the CPU. This is why the prototype has a single header. Anyway, the motherboard I'm testing already has this circuit built in, so this pin is not needed. The motherboard's circuit drives FLUSH# directly, which performs the same function. Connecting or not connecting this should not prevent the motherboard from powering up.

I tested the prototype using 50, 66, and an 80 MHz oscillators. If there is cross-talk issues, I figured running the CPU at 25 MHz would have at least showed signs of life; it did not. I tested the motherboard with MR BIOS and the original AMI BIOS. I tested with the FPU installed, removed, and with the co-processor jumper set to off. Still no signs of life.

The next logical step would be to test the TI486SXL2-66 in a 486 motherboard. Problem is that I haven't seen and do not have any boards which have jumper settings for this CPU. Although the TI486SXL2-66 has a nearly identical pinout to the Intel 486SX, a few modifications must be made to a non-supporting 486 board to use this CPU. The TI486SXL reference manual goes through the few design considerations in Appendix D for motherboards wanting support for the TI486SXL. There is about a four jumper difference, which would be fairly easy to implement on your 486 board if one doesn't mind modifying their boards. This would just verify that the TI486SXL2-66 chips I have are indeed functional. It does seem unlikely that three NOS chips would be bad though.

I also tried using the Transcomputer 486HPi upgrade module with the TI486SXL2-66 installed. I set the Transcomputer module up for 486SX pinout, but I did not see any signs of life when powering on the system. An Intel 486SX CPU, on the otherhand, worked just fine on the same 386 motherboard. This confirms that there are differences which need to be considered with getting a 486SXL2-66 working on a 486 motherboard. I tried the TI486SXL2-66 in the Transcomputer at 5 V and at 3.6 V (by using a variable voltage VRM on top of the Transcomputer).

EDIT: Here are two URL's to the TI486SXL and TI486SXLC reference book
https://ia803003.us.archive.org/35/items/bits … rence_Guide.pdf
http://www.bitsavers.org/components/ti/TI486/ … rence_Guide.pdf

Last edited by feipoa on 2022-12-16, 01:14. Edited 5 times in total.

Plan your life wisely, you'll be dead before you know it.

Reply 1 of 1156, by feipoa

User metadata
Rank l33t++
Rank
l33t++

I also tried using the Transcomputer 486HPi upgrade module with the TI486SXL2-66 installed. I set the Transcomputer module up for 486SX pinout, but I did not see any signs of life when powering on the system. I can even use the Transcomputer with an AM5x86-133 chip and still have the 386 motherboard work fine.

Transcomputer_486HPi_Upgrade_Module.jpg
Filename
Transcomputer_486HPi_Upgrade_Module.jpg
File size
936.07 KiB
Views
8446 views
File license
Fair use/fair dealing exception

An Intel 486SX CPU also worked fine on my 386 motherboard. To me, this confirms that there are differences which must to be considered when installing a 486SXL2-66 onto a 486 motherboard. However, a 386 motherboard is already wired up for a TI486SXL, so the prototype should work! I tried the TI486SXL2-66 in the Transcomputer at 5 V and at 3.6 V (by using a variable voltage VRM on top of the Transcomputer). Actually, I tried three different VRM modules on the Transcomputer, but still did not see any signs of life.

VRMs.jpg
Filename
VRMs.jpg
File size
1.18 MiB
Views
8446 views
File license
Fair use/fair dealing exception

This is everything I have tried to date and am out of ideas. Any thoughts on how to get this prototype working would be much appreciated.

Plan your life wisely, you'll be dead before you know it.

Reply 3 of 1156, by feipoa

User metadata
Rank l33t++
Rank
l33t++

I remember back when I was creating a native PS/2 mouse implementation. Native PS/2 mouse implementation for 386/486 boards using the keyboard controller

I thought I had everything perfect, then someone commented to ensure that I connected the IRQ12 line. That was all that was needed. My gut feeling is that the issue now with the PGA168-to-132 prototype is something similar. Hoping someone stumbles upon this thread with the magic password.

oerk wrote:

Since I can't offer any constructive advice, I'll just remark that you are a much more patient man than I am! 😀

I can't even imagine soldering that many wires in such a tiny space by hand.

It wasn't nearly as bad as I thought it would be. Just listen to some podcast and get going. There was no deadline. If you can spare 1 hour per day, I think anyone could do it.

Plan your life wisely, you'll be dead before you know it.

Reply 4 of 1156, by Anonymous Coward

User metadata
Rank l33t
Rank
l33t

Have you tried using a logic analyzer to see if there are any signs of life at all? I don't think at this point you're going to get very far without one.

"Will the highways on the internets become more few?" -Gee Dubya
V'Ger XT|Upgraded AT|Ultimate 386|Super VL/EISA 486|SMP VL/EISA Pentium

Reply 5 of 1156, by luckybob

User metadata
Rank l33t
Rank
l33t

I have a 486 > 386 upgrade interposer in my ibm model 60. fits a standard 486 into a 386 socket. Would pics help?

There is QUITE a bit of glue logic going on. Even on my 386 > 286 upgrades. The chip you are wanting to use might have the internal glue logic physically disabled. Thats my first line of thought.

It is a mistake to think you can solve any major problems just with potatoes.

Reply 6 of 1156, by Anonymous Coward

User metadata
Rank l33t
Rank
l33t

Err....wasn't the model 60 a 10MHz 286?

"Will the highways on the internets become more few?" -Gee Dubya
V'Ger XT|Upgraded AT|Ultimate 386|Super VL/EISA 486|SMP VL/EISA Pentium

Reply 7 of 1156, by feipoa

User metadata
Rank l33t++
Rank
l33t++

I have one of those ISA logic analyser boards. I didn't think to try it because they have never really helped me in the past. Perhaps I don't really know how to use it. My understanding is that the final code displayed on the LCD is where the error lies? I'll plug in the board and see what code it stops on.

If the issue were rooted in glue logic as the root cause of the problem, I would imagine that the TI486SXL manual would cover what it takes to "turn them on". Interesting, though, is that the reference manual does not mention the 66 Mhz QFP144 or the 66 MHz PGA132; it does, however, mention the 50 MHz QFP144 and 50 MHz PGA168. As I noted in my original dialogue, I might need to modify one of my 486 boards to accept the SXL2 pinout to confirm that nothing extraordinary is required. I think I have a buggy PCCHIPS M919 I can try, also have an extra PCCHIPS VLB board. However, if I go through this effort and the 486 board does not turn on, it doesn't really help me any.

Plan your life wisely, you'll be dead before you know it.

Reply 8 of 1156, by feipoa

User metadata
Rank l33t++
Rank
l33t++
luckybob wrote:

I have a 486 > 386 upgrade interposer in my ibm model 60. fits a standard 486 into a 386 socket. Would pics help?

Pics of 386 to 486 interposer boards are always welcome.

Plan your life wisely, you'll be dead before you know it.

Reply 9 of 1156, by luckybob

User metadata
Rank l33t
Rank
l33t
Anonymous Coward wrote:

Err....wasn't the model 60 a 10MHz 286?

yes. sorry. Its actually a 80-A21. Looks identical.

@feipoa

I'll get the system out of the pile and crack her open. Hopefully tomorrow. I did take some pics of the 286 > 386 upgrades I have: http://imgur.com/a/M5YzG

It is a mistake to think you can solve any major problems just with potatoes.

Reply 11 of 1156, by feipoa

User metadata
Rank l33t++
Rank
l33t++

Jesus, that Kingston thing is loaded with glue logic chips, far more than what the Transcomputer 486HPi has on it. Does it work with a DX4, Cyrix 5x86, or DX5? And is it specific to a particular system? Do you have any of the Transcomputer modules?

Plan your life wisely, you'll be dead before you know it.

Reply 12 of 1156, by feipoa

User metadata
Rank l33t++
Rank
l33t++

When I install the PC analyser card, the code which is displayed is 01. I beleive this is the first sequence of tests. The booklet which came with the PC analyser card says that 01 is:

Processor Test 1, Processor status (1FLAGS) verification. Test the following processor status flags: carry, zero, sign, overflow. The BIOS sets each flag, verifies they are set, then turns each flag off and verifies it is off.

Does anyone know how this information can help in my diagnosis?

The good news is that when I turn on the motherboard without the prototype installed, I do not even get an error code. The LCD display reads --.

Plan your life wisely, you'll be dead before you know it.

Reply 13 of 1156, by luckybob

User metadata
Rank l33t
Rank
l33t

dont know. its possible, but i have little reason to try. 386-25 to 486-25 is plenty fast enough.

I have been thinking.

a lot of these adapters have the math copro onboard. 486 is in-chip, but maybe the glue logic catches certain copro op codes and make sure its seamless to the mobo.

also, maybe there is a wiring mistake on your adapter. one crossed address line... i dont doubt your diligence, I'm just looking for the most likely issues.

It is a mistake to think you can solve any major problems just with potatoes.

Reply 14 of 1156, by BloodyCactus

User metadata
Rank Oldbie
Rank
Oldbie
feipoa wrote:
When I install the PC analyser card, the code which is displayed is 01. I beleive this is the first sequence of tests. The boo […]
Show full quote

When I install the PC analyser card, the code which is displayed is 01. I beleive this is the first sequence of tests. The booklet which came with the PC analyser card says that 01 is:

Processor Test 1, Processor status (1FLAGS) verification. Test the following processor status flags: carry, zero, sign, overflow. The BIOS sets each flag, verifies they are set, then turns each flag off and verifies it is off.

Does anyone know how this information can help in my diagnosis?

The good news is that when I turn on the motherboard without the prototype installed, I do not even get an error code. The LCD display reads --.

an isa diag card wont help with an interposer board. they rely on a working cpu, and they only do certain specific tests which wont really help you.

you need a real logic analyser (not a usb toy one either). I have a HP 1670G ist a beast. with all the exposed pins on the interpose board it should be a cakewalk to connect and probe pins.

--/\-[ Stu : Bloody Cactus :: [ https://bloodycactus.com :: http://kråketær.com ]-/\--

Reply 15 of 1156, by feipoa

User metadata
Rank l33t++
Rank
l33t++

The only way the issue could be co-pro related is if the 486SXL2 was a a 486DX. Remember, the 486SXL2 PGA168 is the same as the 486SXL2 PGA132 chip. So the motherboard is already wired up to work with a seperate co-pro. It is only when you insert the 486SXL2 PGA168 into a proper 486 mothreboard that changes must be made concerning how the co-processor is wired.

The most likely issue is wiring. But I checked in 3 times - once while soldering, once when everything was finished, then again after it didn't work following my first test run. After I had finished, I did find two wires (out of 168) that I messed up the pin location. When you've been soldering too long and are always turning the chip over, you can sometimes confuse the top and bottom.

I really don't know what to check now. I did more reading in the manual concerning FLT#, and it was recommend that this pin be used for testing only. So it sounds like it is not where the issue lies.

Plan your life wisely, you'll be dead before you know it.

Reply 16 of 1156, by feipoa

User metadata
Rank l33t++
Rank
l33t++
BloodyCactus wrote:

you need a real logic analyser (not a usb toy one either). I have a HP 1670G ist a beast. with all the exposed pins on the interpose board it should be a cakewalk to connect and probe pins.

136 channels? that thing is a beast. Are you offering to help? It would take forever to hookup the 136 cables. I assume you have thing enough clips to reach the centre and inner pins of the PGA132 while it is inserted? I do not - only the outer pins, or inner pins of the unit is not plugged into the motherboard.

Assume you have all channels hooked up, wouldn't you need to know the expected waveform on each pin and compare it with what you are receiving? How do you know what is expected and at what point doing the boot process? I haven't had the fortune to work in industry in this area of expertise and have only had a few courses at university dealing with expected waveforms from logic chips, but that was more than a decade ago now. At university, we pretty much did everything in simulation as well, so...

Best I have is a 2-channel budget 100 MHz oscilloscope. But what to probe and what to look for? I don't really have the time, energy or a quiet environment to read up on archaic CPU-motherboard interaction.

Plan your life wisely, you'll be dead before you know it.

Reply 17 of 1156, by luckybob

User metadata
Rank l33t
Rank
l33t

You need to find one of these: http://www.ebay.com/itm/250687130392

but obviously for the socket you are using. like this: http://www.keysight.com/en/pd-1000001563%3Aep … or?cc=US&lc=eng

*edit like these: http://www.logic-analyzer-adapter.com/pga_monitor.htm

wow those things are expensive AF.

It is a mistake to think you can solve any major problems just with potatoes.

Reply 19 of 1156, by BloodyCactus

User metadata
Rank Oldbie
Rank
Oldbie
feipoa wrote:

Assume you have all channels hooked up, wouldn't you need to know the expected waveform on each pin and compare it with what you are receiving? How do you know what is expected and at what point doing the boot process? I haven't had the fortune to work in industry in this area of expertise and have only had a few courses at university dealing with expected waveforms from logic chips, but that was more than a decade ago now. At university, we pretty much did everything in simulation as well, so...

well you dont need to hook up EVERY pin, and the waveform timings are in the intel manuals etc.

start small.

test clock pin, is the CLK signal coming through the interposer?

is the cpu sending out its mem fetch signal on the bus? test address line 0, it should pretty well oscillate as the lowest address for fetching data..

work up, then you can monitor the entire data + address bus or something.

I dont have enough time to really help out much and its waaaaay to heavy to post (you can get 1670G's off ebay pretty cheap).

--/\-[ Stu : Bloody Cactus :: [ https://bloodycactus.com :: http://kråketær.com ]-/\--