F18A DIY?

[matthew180]

The SPI flash needs to be 1MB (8Mb) and 100MHz or better, since assumptions are made in the firmware about access speed and size.  It also needs to match the footprint.  Other than that, it can be larger in capacity.

 

As for the shorts between FPGA pins, there are a lot of design changes I would make if I reworked that board.  One of them being that I learned a lot about routing, and I will never connect adjacent pins on the FPGA directly.

 

I have reused a single FPGA multiple times, but rework is easily 2x harder than soldering, and heat definitely kills components and damages the PCB.

 

It is easy to get too much solder on the FPGA pads and have solder bridges behind the legs.  Even harder to clean the flux back there so you can see.  The only way to clean it all seems to be the ultrasonic cleaner.

 

On 8/10/2021 at 5:55 PM, WicoKid said:

If anyone has a board out can you verify continuity or not between positive C18 and negative C8 after FPGA placed.

The resistance between C8 and C18, on the FPGA-side, is about 400-ohms through the FPGA.  None of the caps should be shorted to ground on the FPGA-side (the other side *is* ground).

[WicoKid]

Thank you for all of the input/advice. After a bit of sub-millimeter bending of FPGA pins and then re-bending the C8/C18 short is gone; resistance thru FPGA from nongrounded C8/C18 is, well, inconsistent (695 once then 1195). At this point will wash the board, continue to work thru traces along other edges of FPGA and solder the bus transceiver when it arrives. Will likely be a week or so but will update.

Edited August 12, 2021 by WicoKid

[matthew180]

The bus transceiver is a PITA to solder.  Worse than the FPGA IMO.  Just so you have something to look forward to. ?

 

The resistance through the FPGA will vary since it is a massive black box with lots of devices inside.  The main point being, there is resistance in the few hundred ohm range, i.e. nothing shorted to ground.

[WicoKid]

Well...got a package from UTSource and Ya Da Ya Da Ya Da......

 

 

Thank you Matthew for putting the effort into this and releasing it into the wild. Appreciate this thread and all the information/pictures from the other posts as well. Next step is case modification.

 

[WicoKid]

Spoke too soon. Modded case, reassembled, sat down to try Parsec and booted to this. Since then have retried 8223 and it works. With FA18 however have reflowed pins, reseated multiple times....now rarely getting the pictured screen. usually blank screen or no signal, sometimes LED bright then flickering. I took the FA18 out and applied 5V and ground only....steady LED no VGA signal or output at all. May have fixed it till broke it. 

Edited September 2, 2021 by WicoKid

[Hans23]

I would try swapping the socket first.  In my experience, a bad connection between the motherboard and the F18A causes this behavior, and your description of intermittent problems also point in that directions.  You could try, to verify the theory, to put a second socket into the VDP socket and then putting the F18A into that second socket.  Avoid machined sockets and use the dual-wiped variant.

 

-Hans

[WicoKid]

Now not getting above screen all....just no VGA signal. I tired a different machine. Same behavior.  If I just plug in 5 volts (F18 not installed at all), should the VGA put out the screen above?

Edited September 2, 2021 by WicoKid

[Hans23]

Yes, the F18A internal screen is what you get before the host initializes the F18A.  Moat of the problems that I had were caused by solder bridges that I could only find under the microscope. Did you check the regulator output voltages yet?

[+dhe]

Thanks for the debugging tutorial Hans. I didn't know that was the screen you got, until the host was initialized. It's always easy when others do the work, but it is still good information to know.

 

[Tursi]

I can't go too deep, but there are a few things I do know. If you are getting the screen, most of the F18A is working fine (it has loaded its bitstream, VRAM is working, VGA out is working, TI Graphics Mode is working) and the likely probably is the interface to the computer.

 

If you aren't getting that screen, then it's power, the FPGA isn't loading correctly or the video output section has problems. (Kind of a duh statement, I guess, in other words everything...)

 

That you had the screen and lost it suggests to me that you probably introduced a solder bridge (or, if you were a little light on the solder, a bad connection).

 

I'm not sure that just 5V gives you the screen - I tried it once myself and did not succeed. I think you may need the reset pulse, but on the other hand, my results surprised Matt, but I didn't dig any deeper.

 

[WicoKid]

I left it for a few hours and re- tried....the F18A boot screen was back, so its still alive, it was strong for 10-15 seconds then some color changes, flickering and black again. The LED initially strong then goes dim and flickers soon thereafter. I will check the voltages but truthfully am not real sure what I'm checking but will look closer at schematic.

 

I do not have any  extra sockets to try and piggyback.

Edited September 2, 2021 by WicoKid

[Jimhearne]

Sounds like it could be a voltage regulator shutting down, either over current or over temperature, see if any are getting hot.

Or it could just be a faulty regulator , i've not had good experience of UTSource parts though it was a few years since i last used them.

 

Jim

 

[matthew180]

Let's see, where to start.  I'm sure I will repeat some things already discussed, and Tursi's summary is pretty much right on.

 

First, the "F18A Title Screen" was added to help with troubleshooting.  The 9918A powers on with everything disabled (all VDP registers reset to 0) and the VRAM in some random state, but that is not very useful when things are going wrong.  So, like Tursi said, if you see the F18A Title Screen you know the FPGA has loaded its bit-stream from the flash, you know the 3.3V and 1.2V voltages are ok, and you know the 50MHz oscillator is working.  The VGA "output" section is just a resistor DAC and is rarely the problem.  Although, a cold solder joint on one of the 100-ohm resistors for the Horz or Vert signals will prevent your monitor from syncing.

 

The LED on the F18A is not technically a "power" LED, it is driven by logic from the FPGA and only comes on when the FPGA has successfully configured (which takes about 200ms).  If it is flickering then something is *really* wrong, probably with the power section.

 

The only thing the F18A needs to power up and display the F18A Title Screen are: 5V, Ground, and pull the reset pin high.  A 5V power source and 3 wires on a breadboard are about all you need, although I would recommend pulling reset high with a pull-up resistor.  The reset pin is a "soft reset", i.e. is only affects the loaded circuit and will *not* cause the FPGA itself to reset; only a power cycle will trigger the FPGA to reload its bit stream from flash.

 

Host communication relies on the level shifter and the one discrete AND-gate IC (next to the level shifter).  The level shifter requires 3.3V, but the AND-gate is powered from 5V.  If the F18A powers up and you see the F18A Title Screen, it is possible the level  shifter or the AND-gate have failed.  If the level shifter is kind-of failing, you would expect to see gabled output as the host tries to communicate.  But a solid F18A Title Screen would indicate the host is either not trying at all, or all 8-bits of the data bus are blocked.  The AND-gate controls the direction of the level shifter for the data bus and is controlled by the host #CSR and #CSW signals.

 

The regulators are VR1 1.2V @ 150mA, and VR2 is a dual 3.3V @ 150mA and 2.5V @ 150mA.  The 2.5V is the JTAG reference.  All caps are rated for 16VDC (at least the ones I used), but the Absolute Max input rating for VR1 and VR2 are 6V.  Any more than that for too long and they will blow.  IIRC they have over current and output short-circuit protection.

 

The two jumpers that control the GROMCLK and CPUCLK generation, and their respective output pin are critical for some systems.  The 9918A, and thus the F18A, produces the GROMCLK which is required in the 99/4A for GROMs and sound to work, which is critical for normal operation of the console.  I had a problem just the other night with an F18A I was repairing that was showing the F18A Title Screen.  I had just replaced every IC on the board (the console power supply failed and put 12V on the 5V rail), so I was sure it was not an IC failure, all the voltages were good, and I had video output.

 

The problem turned out to be a cold solder joint on the input of the jumper pins to the FPGA, so the F18A was not producing the GROMCLK, which meant the 99/4A was not operating, which meant the F18A was just sitting and waiting for the 99/4A to talk to it.

 

The jumpers pull the inputs low, and the FPGA has weak internal pull-ups on those inputs, so if you don't have a jumper, the selection for that jumper position will be a "1".  The reason for this design decision was, for anyone who was so inclined, to allow switches-to-ground to be used on the jumper pins, i.e. it makes using switches easy in place of jumpers in case you want to make the default-sprite-selection an external switch, etc.

 

Just realize the problem is not always straight forward and you need to understand what is supposed to be happening, and make sure the jumpers are set correctly for your target computer.  For example, if you pull all the jumpers and put the F18A in a 99/4A, the computer will not work.  But put it in a CV with no jumpers and it will work fine.

 

[WicoKid]

Beautiful and unexpected. Thank you for posting this. This evening, will reinspect for cold joints and test voltages. If looks good will then go for the breadboard as its back to no vga output again. When I had the original bridging on the FPGA I was pretty aggressive on removing the solder.

 

 I am an amateur learner  following instructions and get most of this, but thinking very concretely about the reset and resistor. So if I go to the socket diagram I can locate the reset pin and I put a 100 k resistor between it and ground the remove it to put reset high?

 

 

 

[matthew180]

Any resistor between 1K and 10K is fine for a pull-up.  No need to move it from ground to 5V, the F18A does not need the reset, but it will honor it.

 

Also, depending on where you got the VGA pig-tail, I have seen those fail too, at the soldered wires on the back of the HD15 connector.

 

When doing SMD, or any soldering for that matter, keeping things clean and using copious amounts of flux is critical.  Use a quality no-clean flux (for example Kester brand), and a tacky-flux can help keep the parts in place while you work.  Use 99.9% alcohol for cleanup, NOT the stuff you get at the store (look up Miller Chemicals 824).

 

Heat control is very important for making good solder joints.  There are a lot of videos and such related to soldering, and you have to watch a bunch to find the real ones.  The right amount of heat and flux properly applied to the parts will make all the difference.  I use a 1mm chisel tip and keep the temperature set at 640 F (leaded solder).

 

Use leaded solder for hobby stuff!  Lead-free is so much harder to work with due to the higher temperatures required, you can really kill your parts trying to get them soldered.  Use a temperature controlled soldering iron, and a brass sponge (look it up), NOT a wet sponge.  Always apply a little solder to your tip before you put it in the holder, and wipe it clean on the brass sponge every time you pick it back up.

 

Hot-air workstations are useful too, but they require a different approach, so you should practice with them before trying them on something you care about.

 

Being able to see what you are doing make the difference between being successful or not.  For that reason I always work under magnification (and because getting older sucks).  I invested in a stereo boom microscope (AmScope, about $850 on Amazon).  With that I can work on stuff all the way down to 0102, but never if I can help it.  The MK2 uses 0402 caps and some 1mm x 2mm QFN and BGA parts.  It is amazing what you can solder when you can see what you are doing.

 

[WicoKid]

Disappointing, did get to see the screen at one point anyway, but I think this ones done. Tried new video cable. Voltages measured OK. Then found a bridge with multimeter (could not see under magnification but cleared with flux and heat)on the bus transceiver, made another one while clearing the first and removed. Then sparks between pins 21-23 area and smoke of bus transceiver. Hooked it up to breadboard with voltage, ground and reset (with 2 K resistor)...no output and drawing 0.5 amp; voltage regulators bus transceiver and FPGA all hot with noise on the oscillator. 

 

Going to give it a rest for a while. Thank you for all of the information, hopefully will help someone else. On the lookout for the F18A MK2. 

[matthew180]

The first test after assembling a board is to *always* use the multi-meter to check for a short from any power rail to ground.  You do this with the power off.  Use the diode beep test if your meter has that feature, otherwise the resistance reading should be at least some tens or hundreds of ohms (3.3V @ 50mA is 66-ohms, for example).

 

If you have less resistance than what your regulator can give (E=I*R, where E=voltage, I=current, R=resistance), then you know you have a problem somewhere.  The F18A regulators are 150mA for each 1.2V, 3.3V, and 2.5V.  I just measured a working F18A, the 2.5V and 3.3V rails were greater than 10K, and the 1.2V rail was about 230-ohms, so very easy to detect a short if you have one.

[GDMike]

5 hours ago, WicoKid said:

Disappointing, did get to see the screen at one point anyway, but I think this ones done. Tried new video cable. Voltages measured OK. Then found a bridge with multimeter (could not see under magnification but cleared with flux and heat)on the bus transceiver, made another one while clearing the first and removed. Then sparks between pins 21-23 area and smoke of bus transceiver. Hooked it up to breadboard with voltage, ground and reset (with 2 K resistor)...no output and drawing 0.5 amp; voltage regulators bus transceiver and FPGA all hot with noise on the oscillator. 

 

Going to give it a rest for a while. Thank you for all of the information, hopefully will help someone else. On the lookout for the F18A MK2. 

Yeah, I'm pretty good at making smoke. Waiting on mk2

[WicoKid]

Another F18A installed and working. Thank you Han. Ordered and sent right out. Originally, I was hesitant to order from overseas but it arrive within a week. While I learned a lot while attempting to build, in the end, it was a lot of time and expense and it is sure nice to have a working unit and to just use the machine. Thanks again. 

[J-Data]

On 4/12/2021 at 6:30 PM, J-Data said:

I still have a bare board up for grabs

Offer still stands, I have a spare F18A bare board for anyone who want's to try building their own.  I also have a few spare Micron EOL flash parts, which is just about the only part you can't get directly from DigiKey.  Just cost you a stamp. 

[broettger]

4 hours ago, J-Data said:

Offer still stands, I have a spare F18A bare board for anyone who want's to try building their own.  I also have a few spare Micron EOL flash parts, which is just about the only part you can't get directly from DigiKey.  Just cost you a stamp. 

I would gladly take you up on that offer.   

[J-Data]

11 hours ago, broettger said:

I would gladly take you up on that offer.   

Sorry, you were a little slow on the draw and two others IM'ed me first.  A set of 5 boards only cost me $7 from JLCPCB delivered, so that may be your best route.  All you have to do is drag and drop the gerber zip file from this thread on their quote page.  Takes about 10 days. 

[GDMike]

That's a fast reaction.

Hmm.we need more built .. but I can't pay xxx for a built one.

What's a good fair price for built product anyway?

 

 

[J-Data]

On 10/23/2021 at 10:46 AM, GDMike said:

What's a good fair price for built product anyway?

Hard to say.  The BOM cost is around $45, half of which is the FPGA.  Assembly isn't trivial due to the fine pitch components.  What's your time worth?

[Archimedes5000]

There are now many service providers who offer assembly in addition to the circuit board. Has anyone already had experience with it? For example aisler.net, PCBWAY, JLCPCB or OSH Park.
It must be possible to prepare the information in such a way that the circuit board including assembly can be ordered. Then everyone should be satisfied, because then it only matters that the IC is available.

Greetings Axel