F18A MK2

[matthew180]

10 hours ago, Papalapa said:

At this point will be not possible to build a new F18A VGA batch to sell?

The FPGA (Spartan-3E 250K) used on the original F18A is not in stock anywhere and I think Xilinx is either no longer making it on a regular basis, or at all.  I can't find an official status from Xilinx about the Spartan-3E, but I'm pretty sure it is discontinued at this point.  This was one of the main reasons I started working on the MK2, which uses a newer Spartan-6 FPGA, which is also now cheaper than the Spartan-3E.

 

I realize people are wanting the MK2, and some have been waiting for an F18A since 2017.  I apologize for the situation, I really am putting as much time into this as I can.  The whole H D M I problem stalled me for almost a year, so I'm playing catch up now.

 

Edit: So it seems there are some of the Spartan-3E FPGAs available, I messed up my previous search.  They are more expensive now, but I could possibly look into making an intermediate run.  It will take time away from the MK2 though.

Edited October 30, 2019 by matthew180
Corrected stock status

[ja2who]

1 hour ago, matthew180 said:

Edit: So it seems there are some of the Spartan-3E FPGAs available, I messed up my previous search.  They are more expensive now, but I could possibly look into making an intermediate run.  It will take time away from the MK2 though.

Honestly I'd be more inclined to wait for you to finish your work in your own time than to ask for an expedited run of the old solution. I'm just eager like everyone else here. Do you think the MK2 might surface before 2020 or are we more likely looking at next year?

[wolhess]

Hi Matthew180,

 

please take the time you need for the development of the mk2. 

I like to wait until the mk2 version is ready. Thank you very much for your effort.

 

Wolfgang

[GDMike]

I agree, plz, don't kill yourself over it for sure..it's a hobby #1, and it's your hours. Enjoy life, I can't wait to get a MK2, but not if you're heart wasn't in it too 

 

[matthew180]

23 hours ago, ja2who said:

Do you think the MK2 might surface before 2020 or are we more likely looking at next year?

Hard to say.  If you had asked me in June of 2018 I would have said it would be out by Aug 2018, and indeed that is the only reason I announced the project (see my first post about that).  I work on the MK2 as much as I can carve out the time.  But, for example, if there is a time when I planned on MK2 work, yet my son asks me if I can take him to the archery range, guess what I will be spending my time doing (my family will always get the priority).

 

The MK2 is my top hobby priority.  That's about the best I can say right now.  As things get done I have been posting the progress here.  That's the best I can do for now.  Also, I'm not adverse to help, so if anyone is good with VHDL or Verilog (*cough*) and wants to help out, let me know.  There is also prototype building and testing, so if anyone is good at SMD assembly and microsoldering (0402 and 0.8mm pitch BGA work required) and wants to help with that, also let me know.

[Papalapa]

On 10/31/2019 at 8:24 PM, matthew180 said:

Hard to say.  If you had asked me in June of 2018 I would have said it would be out by Aug 2018, and indeed that is the only reason I announced the project (see my first post about that).  I work on the MK2 as much as I can carve out the time.  But, for example, if there is a time when I planned on MK2 work, yet my son asks me if I can take him to the archery range, guess what I will be spending my time doing (my family will always get the priority).

 

In fact this is not a problem, I'm sure that we can organise here a collect to rent the services of Brady Ellison for example to train your son in archery.

(I'm joking of course, I'm 55 and I well know that family is the first priority ?)

 

[Papalapa]

Well, trying to solve the colors problem on my TI-99/4A this morning I was excited when I was about to test my second hand E74 IBM TRC monitor. This was the first non-TFT monitor tested with my TI-99/4A but unfortunatelly the result was the same. ?

 

 

A really good quality image with retrocomputers (better than the TFT monitors of course) but colors are not correctly shown. I already tried three different video converters (the monitor has only VGA input) unsucessfully.

 

I'll continue investigating and testing....

 

Edited November 2, 2019 by Papalapa
Double picture ¿?

[fabrice montupet]

Oh! I just discover your pictures. I have never had a so poor RGB color rendering using my PHA2037 adapters, as well with my CRT monitors as my TFT ones. I think that your problem come rather from your RGB converter that handle the signal differently depending the home computer you attach onto.

 

EDIT: Do you have tried to adjust the RGB signal using the potentiometer inside the PHA2037, maybe yours is not calibrated?

Edited November 2, 2019 by fabrice montupet

[Papalapa]

14 hours ago, fabrice montupet said:

Oh! I just discover your pictures. I have never had a so poor RGB color rendering using my PHA2037 adapters, as well with my CRT monitors as my TFT ones. I think that your problem come rather from your RGB converter that handle the signal differently depending the home computer you attach onto.

 

EDIT: Do you have tried to adjust the RGB signal using the potentiometer inside the PHA2037, maybe yours is not calibrated?

That's the problem, I do not have the PHA2036 nor PHA2037 adapters, only the TI-99/4A.

 

On the other hand I have connected to the VGA converter a ZX81, ZX Spectrum, Amstrad CPC464, VIC-20 and Commodore 64 and all of them show good quality images.

 

[fabrice montupet]

9 minutes ago, Papalapa said:

That's the problem, I do not have the PHA2036 nor PHA2037 adapters, only the TI-99/4A.

On the other hand I have connected to the VGA converter a ZX81, ZX Spectrum, Amstrad CPC464, VIC-20 and Commodore 64 and all of them show good quality images.

 

Can you tell me which VGA converter you use?

At a time, I tried the GBS 8220 and I have been disappointed by the picture quality. After some days of testing, I have forgotten it in its box in my basement. But with a such cheap converter,  I didn't hope for a miracle.

[GDMike]

Let's get Matt help! someone!! Wonder what timeframe you need their help for?

I definitely can solder, just not at that level, unfortunately or I'd be there. I suppose that's an obstacle.

I can assemble larger components well.

Hopefully someone can give to the mission.

 

 

[Papalapa]

10 hours ago, fabrice montupet said:

Can you tell me which VGA converter you use?

At a time, I tried the GBS 8220 and I have been disappointed by the picture quality. After some days of testing, I have forgotten it in its box in my basement. But with a such cheap converter,  I didn't hope for a miracle.

After testing some converters now I have the "NPG Real Game & Video Box HD".

 

It is not expensive because is very simple and quality is acceptable for me...

 

[+dhe]

Hope you would consider making a peb version of the mk2 as sort of a digit/snug video card replacement.

I always liked to have my video card in the peb.

Dan

 

[+OLD CS1]

4 hours ago, dhe said:

Hope you would consider making a peb version of the mk2 as sort of a digit/snug video card replacement.

I always liked to have my video card in the peb.

Is there documentation on how an external video card works on the 4A?  I assume the OS does not recognize the external video chip and it has to be activated by software.  Is it just a simple video pass-through like an 80-column card on the Commodore 64 which switches between the VIC-II and the on-board CRT VDP?

[+arcadeshopper]

Is there documentation on how an external video card works on the 4A?  I assume the OS does not recognize the external video chip and it has to be activated by software.  Is it just a simple video pass-through like an 80-column card on the Commodore 64 which switches between the VIC-II and the on-board CRT VDP?

Pretty sure it required a cable to the console motherboard to replace the 9918 and route the a signals to the card

Sent from my LM-G820 using Tapatalk

[+dhe]

On the cyc, there is a 70 page manual for the digit card, but the relevant console mods that had to be done was in these two para:

 

3.4.
Modifying the console
3.4.1.
Cutting the trace
The TI-99/4A was designed in such a manner that the video control data generated by the CPU was notmultiplexed onto the main system bus but went directly to the VDP. In order to control the AVPC in thePEB, it is necessary that the VDP and the CPU have free access to data via the main bus. To do this themultiplex control gate must be disconnected from the video block decoder. There is no way of doing thiswithout cutting a lead or trace on the printed circuit board.
Take the computer printed circuit board with component side up and the I/O port to the right. Refer todiagram "Trace Cut Location" which is a simplified drawing of the lower right hand corner of the printedcircuit board. Locate the trace pattern between IC 74LS03 and IC 74LS32. Note that it consists of a 4 dothole pattern. It is necessary to cut only the lead from the topmost dot. The cut should be made with anX-ACTO knife or single edged razor blade. Carefully cut a piece the width of the trace, out of the trace.Inspect the area with a magnifying glass and verify that only this trace is cut and that the trace iscompletely severed.

 

3.4.2.
Video Processor jumper
Congratulations, the messy work is done. Now we shall install the Jumper on the Video Display Processor.Note that it is one of the few socketed integrated circuits on the board. Also note the index on the chippoints toward the center of the printed circuit board. The index is the "U" shaped indentation on the topof the IC that guides automatic insertion machines involved in mass production.
Refer to the diagram "VDP Jumper Detail". Pry the IC gently out of its socket with a flat blade orscrewdriver by alternately inserting it under each end. The pin to the left of the index is pin #1 and pinsare numbered counter-clockwise (CCW) from this point. Count down to pin #14 and, gripping it firmlywith a pair of needle-nosed pliers or tweezers, bend the pin out 90 degrees. Reinsert the VDP into itssocket by carefully engaging one row of pins in its respective side of the socket. Press gently and rotatethe IC until the other row of pins engages. Be sure all pins are engaged and that the index is pointingtoward the center of the board. Then press the IC uniformly and firmly into the socket until it is seated.
Now remove the Jumper from its plastic bag and note that it has three legs: a Red hook-clip, a Blackhook-clip and a leg with a tiny one-pin socket. Slide the socket over pin #14 of the VDP. Clip the Blackhook onto pin #15 of the VDP making sure that the shank of the clip is in the pin #14 space and the hookis away from pin #14. Clip the Red hook on the capacitor lead on the opposite side of the VDP as shown.The use of clip hooks eliminates the need for soldering and is quite reliable. Dress the Jumper wires sothat they do not interfere with the replacement of the computer shields and secure them with a littlescotch tape or masking tape to prevent them from moving around when you replace the shielding.

 

[MrPix]

How about a nice 2020 status update, Matthew?

[+FarmerPotato]

On 11/21/2019 at 11:32 AM, dhe said:

On the cyc, there is a 70 page manual for the digit card, but the relevant console mods that had to be done was in these two para:

 

3.4.
Modifying the console
3.4.1.
Cutting the trace
The TI-99/4A was designed in such a manner that the video control data generated by the CPU was notmultiplexed onto the main system bus but went directly to the VDP. In order to control the AVPC in thePEB, it is necessary that the VDP and the CPU have free access to data via the main bus. To do this themultiplex control gate must be disconnected from the video block decoder. There is no way of doing thiswithout cutting a lead or trace on the printed circuit board.
Take the computer printed circuit board with component side up and the I/O port to the right. Refer todiagram "Trace Cut Location" which is a simplified drawing of the lower right hand corner of the printedcircuit board. Locate the trace pattern between IC 74LS03 and IC 74LS32. Note that it consists of a 4 dothole pattern. It is necessary to cut only the lead from the topmost dot. The cut should be made with anX-ACTO knife or single edged razor blade. Carefully cut a piece the width of the trace, out of the trace.Inspect the area with a magnifying glass and verify that only this trace is cut and that the trace iscompletely severed.

 

3.4.2.
Video Processor jumper
Congratulations, the messy work is done. Now we shall install the Jumper on the Video Display Processor.Note that it is one of the few socketed integrated circuits on the board. Also note the index on the chippoints toward the center of the printed circuit board. The index is the "U" shaped indentation on the topof the IC that guides automatic insertion machines involved in mass production.
Refer to the diagram "VDP Jumper Detail". Pry the IC gently out of its socket with a flat blade orscrewdriver by alternately inserting it under each end. The pin to the left of the index is pin #1 and pinsare numbered counter-clockwise (CCW) from this point. Count down to pin #14 and, gripping it firmlywith a pair of needle-nosed pliers or tweezers, bend the pin out 90 degrees. Reinsert the VDP into itssocket by carefully engaging one row of pins in its respective side of the socket. Press gently and rotatethe IC until the other row of pins engages. Be sure all pins are engaged and that the index is pointingtoward the center of the board. Then press the IC uniformly and firmly into the socket until it is seated.
Now remove the Jumper from its plastic bag and note that it has three legs: a Red hook-clip, a Blackhook-clip and a leg with a tiny one-pin socket. Slide the socket over pin #14 of the VDP. Clip the Blackhook onto pin #15 of the VDP making sure that the shank of the clip is in the pin #14 space and the hookis away from pin #14. Clip the Red hook on the capacitor lead on the opposite side of the VDP as shown.The use of clip hooks eliminates the need for soldering and is quite reliable. Dress the Jumper wires sothat they do not interfere with the replacement of the computer shields and secure them with a littlescotch tape or masking tape to prevent them from moving around when you replace the shielding.

 

 

What is the effect of these changes?

 

Pin 14, CSW*, has been disconnected from the motherboard.
Pin 15, CSR*, is still connected but shared with the AVPC.

 

The theory I see this working under is:

•   The AVPC decodes A12,A13,A14,A15,DBIN,WE* to derive CSW*, CSR*, MODE0, MODE1
•   Cutting the trace means that the side port can respond to a read from >8800 (VDPRD) or >8802 (VDPSTA) instead of contending with the 9918's driver.
•   The 9918 outputs data during CSR* but the motherboard doesn't gate it onto the 16 bit bus.

 

 

Also:

 

• The AVPC can allow motherboard writes to the 9918  within the first 16K?
• Interrupts are no longer synchronized with VSYNC if the 9918 generates them.

[RedBaron]

I just wanted to say thank you Matt for all your work on the F18A.  I do agree that family comes first.  Once the Video module becomes available I would be interested in buying a few.  Thank You Ricki

[matthew180]

Apologizes for yet more update delays.  I know people are waiting anxiously for the MK2, so I always feel guilty that my progress is never as much as people would like.

 

So I finally got time to build and test Rev J... (yes, there have been revisions A, B, C, D, E, F ,G, H, I, J).  I can tell you right now there will have to be a Rev K, which I hope will be the final PCB change.  I have been spending a lot of time on this over the last few months, and I have made some further decisions that should help get the MK2 into physical production.

 

The biggest PCB problem has been the digital video (TMDS) output, especially the connector.  The fix was to use a 20-pin FPC connector and provide a header-board with a DVI connector.  Since I'm only officially supporting DVI, I felt the need to also provide an audio output jack on the header board.  The problem with that is, 19 of the 20 pins are already used for the TMDS signals, and the only available pin for audio was right next to the one of the TMDS pairs.  Needless to say, a lot of noise was coupled onto the audio and made it completely unacceptable.  So I will be changing to a 22-pin FPC connector to give the audio an isolated pin that does not interfere with the 19 predetermined pins, which means yet another small header board to convert the 22-pin FPC to 20-pin FPC to support people who will use 3rd party headers... sigh.

 

The other problem I have not yet solved is the back-feed from the monitor keeping the FPGA powered after the computer power to removed.  This prevents the MK2 from powering back up cleanly and the solution usually requires unplugging / re-plugging the monitor.  This is not acceptable either, and apparently this is a problem other projects are having as well.  I think I have a solution and I'm just waiting on the parts to arrive so I can test.

 

Since I have to make at least one more PCB revision to fix the power problem, I have also decided to try and support VGA output as well.  This is only made possible by going to the 22-pin FPC connector, and it will require a 3rd header with the resistor DAC, VGA connector, and audio connector.  It will also probably require a different firmware to be loaded in the FPGA, or at least a way to select a firmware and reset...  I have not figured out the details of this yet.

 

In trying to get the MK2 physically built and into people's hands sooner, it will not be released with any of the big enhancements, i.e. the 512K RAM will not be accessible, no new GPU features, modes, VDP register changes, etc.  It will pretty much be the V1.9 firmware adapted to the MK2 board and put out TMDS (i.e. DVI) video, or VGA.  This allows me to focus on making sure the hardware design is correct, and I can introduce the new features in a future firmware release.

 

However, the MK2 will have the audio enhancement since that is part of the hardware; so start preparing now if you can't solder.  You will need to get a wire from the computer's internal audio to pin-1 of the MK2 board.  On the 99/4A a good place to pick up the audio is L201, on the CV it is C88.  The MK2 has a 12-bit ADC that converts the audio into digital, and then spits it back out via a 12-bit DAC to the header's audio line-out.  Since the audio is brought into the FPGA, that means the GPU can mix, modify, replace, or do whatever with the audio data on its way out to the DAC.  This basically gives the MK2 12-bit audio capability on top of the system's existing audio.  The GPU audio is not yet designed, so I'm considering any suggestions at this time.

 

I have tested the MK2 in the 99/4A and CV (see photos) and it works as expected.  The main TODO list for the hardware:

 

* Test the fix to the power-problem.  Add parts to PCB and reroute as necessary to make room.

* Finalize audio subsystem coupling components, adjust PCB routing as necessary.

* Decide if the ADC and DAC really need precision 0.2% voltage regulators, and somehow fit them on the PCB if the answer is yes.

* Change 20-pin FPC to 22-pin FPC, reroute affected pins.

* Find and route six more signals from the FPGA to the 22-pin FPC connector to support the VGA header.

* Make the schematic and PCB for the VGA header.

* Make the 20-pin to 22-pin FPC adapter.

* Order Rev K prototype, build, test.

* Get some beta testers to test on more hardware, monitors, etc.

 

Believe me when I say I'm dumping many many hours into this, to the detriment of other things (like doing taxes... ugh, or planning the family vacation).

 

I am also very nervous about the release of the MK2.  The demand is pretty high right now, and the first batch will most likely be 500 or more boards.  This is actually freaking me out since it is a lot of money, and if a problem is found with the board, it would be really bad.  That's another reason why I'm spending the time to really test this and trying to get it right.

 

So, the photos (in no particular order):

 

* Testing in the CV with a 3rd party header.  Finally no tall-pins or smashing the VRAM capacitors!  So much nicer.

* Testing the MK2 audio on the breadboard.

* The MK2 DVI header, works as expected.

* Have you ever *seen* the Tombstone City's title song Hell in Texas?  The yellow trace is the original audio input to the MK2, the purple trace is the output from the MK2.

 

[+arcadeshopper]

Awesome!  

[+Ksarul]

Thanks for the update--and I will patiently wait for as long as it takes to get it ready to release into the wild.

[+jmazzy]

Thanks for the update.

 

I'm also happy to wait patiently.  As much fun as it is to see new upgrades like this one, it's a hobby after all so there's no need to rush.

 

Good luck with the ongoing development, and thanks, too, for all your hard work on the project.

[moulinaie]

Wonderful work! Take as much time as necessary.

[ndp630]

11 hours ago, matthew180 said:

Apologizes for yet more update delays.  I know people are waiting anxiously for the MK2, so I always feel guilty that my progress is never as much as people would like.

 

So I finally got time to build and test Rev J... (yes, there have been revisions A, B, C, D, E, F ,G, H, I, J).  I can tell you right now there will have to be a Rev K, which I hope will be the final PCB change.  I have been spending a lot of time on this over the last few months, and I have made some further decisions that should help get the MK2 into physical production.

I'm sure a lot of people would gladly pre-order and pre-pay for the units... especially here on AtariAge, to defray some of the upfront cost. I definitely would.