TI99/4A Motherboard Replacement (In Progress)

[rkrenicki]

I did find a number of inconsistencies on the original schematics, but nothing too crazy.  Mostly reference problems.   I found more issues on the HackMac KiCAD schematics that I "started with".. there were a large number of problems with that.   It really would have taken me less time to lay the whole schematic myself than to hunt and fix all of the errors.

On my motherboard, C100 and C101 matched the schematic.  They "look like" resistors, but they are absolutely capacitors.  The light green is sort of dead giveaway.   

 

C100 is Blue, Green, Black, Gold, Black which works out to 56pF
C101 is Brown, Red, Black, Gold, Black which works out to 12pF

These values match the schematics as well.

 

[HOME AUTOMATION]

Gotta hand it to you, in all the years... I've never come across this before.

 

In my defense; I am color-blind, they all look beige to me. I never try to determine their values, from those cute little bands.

[rkrenicki]

Yea, it was something far more common in the 70s/80s.  Pretty much everything now will just use a 3 digit code for a non-polarized capacitor.   There are a good number of these "green apple" colored caps scattered across the board.  Pretty much anything that is not a 0.1uF cap or polarized is in this form factor.   All of the 0.001uF keyboard/joystick bypass caps below the VRAM are like these, intermixed with the pull up resistors.  What really gets confusing are the "Crayola flesh colored" axial parts are caps too.

The only thing TI was consistent on, was their inconsistency.

 

[GDMike]

On 3/23/2021 at 1:30 PM, rkrenicki said:

am perfectly fine with a potential VGA, but DVI is okay too

Absolutely

[rkrenicki]

Due to a recent death in the family, progress on first revision testing has been slow.   I do have most of the passives and sockets installed.  I did find a couple of errors in some parts values on the HackMac schematic during assembly.   Nothing too crazy, just a couple of transposed numbers.    I also did find a couple of errors on the solder mask as well, which were artifacts from the Gerber import from sprint.   I thought I found them all before sending off to fabrication, but it looks like I missed 3 or 4 of them.

 

[rkrenicki]

Surprise second update in one day.    I finished installing the rest of the parts today, powered the board and checked voltages.   Everything looked good.  I then stuffed the board with all of the ICs.   I used new parts wherever I could (except for the 244/245s, since I can't find my tubes of those parts at the moment).

 

First boot:  I get a video sync, and the random tone from the SN94624, but it fails to actually start. 

 

I probed around the Address and Data lines, and I see activity with nothing obviously wrong.   The bigger problem I think is that my donor machine was of questionable origin.. and I never powered it on prior to stripping the board.   I honestly cannot say for sure that the custom chips actually work, so I may be dealing with one (or more) bad chips here.   My known working TI99 is a QI unit, so that is of limited use here.

 

I did buy a few of those "TMS9900 Chip Kits" on eBay from China, so I will try swapping parts around to see if I can get it to boot.. but I do not have extra 4116s, GROMs, or ROMs.   In the meantime, I am going to find another non-QI TI99 so I have a "known working reference" board to compare against.

 

 

[Stuart]

You're using a 48MHz crystal with the TIM9904? (As opposed to the 12MHz crystal (and associated timing components) for a TIM9904A?)

 

I see you're using quite a few 74HCT ICs. You might want to try 74LS. Although the HCTs are supposed to be compatible, I've found circuits where an HCT doesn't work but an LS does.

 

[fabrice montupet]

For a next motherboard assembly, I encourage you to install Tulip IC sockets. Dual-leaf ones are quite fragile, they don't like multiple insertion/removing of components and may have poor contact with used IC (de-soldered) so their pins have to be properly cleaned ( soldering residue removed), and perfectly flattened. And sometimes, even if they are new, chinese low-cost dual-leaf sockets are defective.
Yes, Tulip sockets cost a little more but they prevent any contact problems.  Personally, I only use this kind of socket.

[rkrenicki]

1 hour ago, Stuart said:

You're using a 48MHz crystal with the TIM9904? (As opposed to the 12MHz crystal (and associated timing components) for a TIM9904A?)

Yes to the 48MHz crystal.  I apparently forgot to order a new one, so the crystal is the original one from the donor machine, same as the TIM9904.

 

1 hour ago, Stuart said:

I see you're using quite a few 74HCT ICs. You might want to try 74LS. Although the HCTs are supposed to be compatible, I've found circuits where an HCT doesn't work but an LS does.

I did have this thought as well.  I largely have HCT and AHCT chips in my stockpile here, so I figured I would give them a try.  I do have the original LS chips from the donor, so I will be trying to swap those in one at a time.

 

52 minutes ago, fabrice montupet said:

For a next motherboard assembly, I encourage you to install Tulip IC sockets.... Yes, Tulip sockets cost a little more but they prevent any contact problems.  Personally, I only use this kind of socket.

I assume by "Tulip" you are referring to Turned Pin sockets?     This has been a long debated topic about which is better, but I personally prefer Dual-wipe sockets, and I know how to properly prep pulled ICs.   I do stick with name brand sockets though.

[Kchula-Rrit]

That is so cool!

 

Are you going to offer these for sale?  That would be so neat to get one!  I have a motherboard, that died from a previous attempt at modification, that I would like to resuscitate.  32K on the 16-bit bus is something I would love to have.  I assume the video output is composite.  Is the output improved over an original?

 

K-R.

 

Edited April 1, 2021 by Kchula-Rrit

[rkrenicki]

8 hours ago, Kchula-Rrit said:

Are you going to offer these for sale?  That would be so neat to get one!  I have a motherboard, that died from a previous attempt at modification, that I would like to resuscitate.  32K on the 16-bit bus is something I would love to have.  I assume the video output is composite.  Is the output improved over an original?

The design will be freely available, but I can have some fabricated for sale once all of the dust settles.   At this stage, the video output type is not changing.  I will try to implement any improvements once I get to that stage.

Otherwise, I did try swapping in all of the original LS logic with no change.   I also swapped the CPU with a spare, also with no change.  I am going to open up my QI console and socket the SRAM, ROM, and GROM chips so that I can test these.

[fabrice montupet]

17 hours ago, rkrenicki said:

I assume by "Tulip" you are referring to Turned Pin sockets?     This has been a long debated topic about which is better, but I personally prefer Dual-wipe sockets, and I know how to properly prep pulled ICs.   I do stick with name brand sockets though.

Anyway, the result is really fine, good work ?

[fabrice montupet]

I remember the very first TIny-99/4A (v1) I have assembled a few years ago. When I started up it for the first time, I had no video signal but a continuous tone. I tested all the IC (all are socketed), resistors, diodes, transistors and caps. All these components were OK and I always had the problem. Finally I replaced the L602 inductor by an other one and the computer started.
Maybe a direction for your troubleshooting.  

[+dhe]

 A few random thoughts...

  1) That's a very sharp looking board!

  2) I hope you can release a BOM, I'd like to put as many NEW parts in as possible.

  3) I can see this starting a whole thread of - what to test next! ?

        People might want to pitch in a make something like the Horizon assembly manual.

  4) I hope you consider gold fingers on the I/O port, it might improve reliability for just a bit more cost.

  5) I've always wanted a TI motherboard with machine screw (tulip) sockets, it makes the dread of working on something less.

  6) TI Seemed pretty adamit that the TMS9900 need bus bars to work properly, I follow with interest that suggestion.

 

Thanks for the work you have put in to this project.

[rkrenicki]

32 minutes ago, dhe said:

  2) I hope you can release a BOM, I'd like to put as many NEW parts in as possible.

  4) I hope you consider gold fingers on the I/O port, it might improve reliability for just a bit more cost.

  5) I've always wanted a TI motherboard with machine screw (tulip) sockets, it makes the dread of working on something less.

The entire design will be available freely, including the gerbers and the BOM.   I am not really planning on selling any significant quantity of these, but anyone would be able to have their own made up with whatever options they want (gold fingers, ENIG, fancy colored soldermask, etc)

 

32 minutes ago, dhe said:

6) TI Seemed pretty adamit that the TMS9900 need bus bars to work properly, I follow with interest that suggestion.

In order to distribute 3 different power rails and ground on a 2 layer board, then bus bars or jumpers would be needed.   The 4-layer gets around that need.

 

 

1 hour ago, fabrice montupet said:

I remember the very first TIny-99/4A (v1) I have assembled a few years ago. When I started up it for the first time, I had no video signal but a continuous tone. I tested all the IC (all are socketed), resistors, diodes, transistors and caps. All these components were OK and I always had the problem. Finally I replaced the L602 inductor by an other one and the computer started.
Maybe a direction for your troubleshooting.  

That is a good point, I was unsure on the value for that location.  I see 33uH on the TI schematics, but the HackMac schematics say 0.33uH.  The QI version of the schematics say 3.2uH.   What value are you using for that inductor?   In the meantime, I will use the one that I harvested off the original and see if that changes anything.

 

[fabrice montupet]

The L602 inductor is 3.3uH (orange, orange, gold, silver).  This is the value of the inductor soldered on the TI-99/4A motherboard. And as you have seen, it is not the value written on the TI-99/4A schematics (".33 uH").

 

That said, be careful: there are numerous errors that this official document is containing. Clearly, TI engineers have not controlled the conformity of these schematics destined to the external publication "TI-99/4A Console and Peripheral Expansion System Technical Data" . It is a real lack of seriousness, no acceptable for a big brand like Texas Instruments. A second point of deception: the Technical Service Data for the TI-99/4A released by SAMS also contains errors, giving the impression that SAM has taken technical information from the TI Schematics without verification (surprisingly, same errors are found in these two publications). This tarnish the Howard W.Sams company reputation.  Now, when I read  SAMS schematics (for all computers) I doubt about its contain, it's a pity.

 

So, in my TIny-99/4A development, my first step was to control each component, each value and each trace to find all the errors made by TI, otherwise it would have been very difficult to finalize the project.

 

Edited June 27, 2021 by fabrice montupet

[rkrenicki]

My original part is actually an open wound inductor.  I had no color bands to go by.   L602 and L207 are the only parts that I was unable to reconcile a value from the actual donor board, but yes.. there are a number of inconsistencies which I corrected along the way.

 

[Stuart]

I think the schematic is actually correct ... it specifies the clock IC as a 74LS362, also known as the TIM9904. This uses a 48 MHz crystal, and the values of L602 and C603 are calculated using an equation to relate to a frequency of 48 MHz. The values in the schematic of 0.33uH and 22pF satisfy that equation (the capacitance of the PCB also has to be taken into account). TI then switched to a TIM9904A, which uses a 12 MHz crystal. The values of L602 and C603 are calculated to relate to a frequency of 12 MHz. Values of 3.3uH and 50pF (or thereabouts) satisfy that equation.

 

So the L602 and C603 will be different values depending on whether you have an older motherboard with the TIM9904 or a later board with the TIM9904A. So you'll see different values in the original schematics and the QI schematics for example.

 

@rkrenicki - if you use the clock IC, crystal, L602 and C603 from your donor board, then they should be matched. You're using a 4-layer board though with a ground layer which will affect the board capacitance, and I don't know if that will be significant enough to affect things.

[rkrenicki]

Okay, that is good news.   I have corrected the schematic (and therefore the silkscreen) to reflect 0.33uH.  I have the correct 22pF cap (already installed) and a 0.33uH inductor on hand.  Since neither of these parts are tied to ground or 12v, I cannot imagine that the board capacitance would play too much of a role here, but I can try doing a keep-out zone underneath them on the next revision.  I do also have 0.33uH new inductors to try, which I have not done yet. 

 

In related news, I also tried some other TMS9900s, TMS9901s, and TIM9904s that I had on hand.   The only ICs that I have not yet tested in another machine are the ROMs and RAM/SRAMs.

[fabrice montupet]

53 minutes ago, Stuart said:

I think the schematic is actually correct .

At the beginning, my motherboard was equipped with a 74LS362, a 48 MHz crystal and 0.33uH inductor because I didn't want to move away from what I saw in the TI schematics. And it doesn't worked.   It worked when I replaced the inductor. We can consider the capacitance of the motherboard has played a role*... Anyway, that not prevent that  TI schematics has errors.

 

*My motherboard is really different from the TI one.

Edited April 2, 2021 by fabrice montupet

[+9640News]

How many gigawatts are you applying to the flux capacitor?  If you don't have enough gigawatts, it may not be enough for a 'timing' sync. ?

[fabrice montupet]

Not sure to understand. I'm sorry, I do my best to read and write in english but I do not have the command of English as you.

[+mizapf]

[x] You want to watch the movie "Back to the Future". ?

 

[fabrice montupet]

I didn't made the connection with this (great) movie because some replicas have been differently translated in French!  The "Flux Capacitor" as been replaced by "Convecteur temporel" ("Time Convertor" en English).

[rkrenicki]

33 minutes ago, 9640News said:

How many gigawatts are you applying to the flux capacitor?  If you don't have enough gigawatts, it may not be enough for a 'timing' sync. ?

Why, 1.21 Jigga-watts of course.

I did put a copper pour keep-out area around these parts for future revisions.  I also removed the values, so I can make a little grid of values nearby detailing the 9904/9904A use.

 

1 hour ago, fabrice montupet said:

At the beginning, my motherboard was equipped with a 74LS362, a 48 MHz crystal and 0.33uH inductor because I didn't want to move away from what I saw in the TI schematics. And it doesn't worked.   It worked when I replaced the inductor.

So, just to be 100% clear:   You are using a TIM9904/74LS362 with a 48MHz Crystal, 3.3uH Inductor and a 22pF capacitor, and that works for you?

 

I am trying to find a datasheet for the non-A version of the TIM9904, but I have not yet found one...