Pi Pico[W] Peripheral Expansion Box Side Port Device

[JasonACT]

On 4/16/2024 at 5:57 AM, wiloson said:

my two computers are silver and black 

LTA2383 and my original system LTA4582.

They don't seem far from what others have as far as manufacturing dates.

One of yours was built the week before mine.  Maybe it's something else other than that date, like how much usage the TI got or how it was stored over the years.  It'll be interesting when I get my ebay machine, that's for sure.  I'm expecting to have to repair it since it's coming "untested".

[wiloson]

13 hours ago, JasonACT said:

One of yours was built the week before mine.  Maybe it's something else other than that date, like how much usage the TI got or how it was stored over the years.  It'll be interesting when I get my ebay machine, that's for sure.  I'm expecting to have to repair it since it's coming "untested".

I'm starting to suspect that the power supplies on both of my computers are a little too noisy to run this board.   I looked at the schematic of the TI power supply and I was surprised that the +5Volts supply is fed off of the 7812 regulator. 

I was thinking of replacing the 7812 and 7905 regulators with switching regulators like the N7812-1PH , as well as replacing some of the caps, but it may be better to replace the power supply with something newer.

What power supply did you replace your original supply with?

 

 

[Stuart]

1 hour ago, wiloson said:

I looked at the schematic of the TI power supply and I was surprised that the +5Volts supply is fed off of the 7812 regulator.

The +5V regulator is powered from the +12V supply, but the main +5V supply current path (through the TIP31 transistor) is not from the 12V supply.

[JasonACT]

On 2/22/2024 at 6:33 PM, JasonACT said:

I added a large (almost empty) perfboard with just a switch for the 12v input, after I tested this, but this is serving me well - +5v 2-3A, -5v 2-3A powered with a 12V 2A regulated wall wart.  I'm seeing +4.9v out of the side port, the other 2 voltages are 100% at spec.  I wouldn't call it a coffee warmer, but it does get a little warm...

 

I don't really have any details on the board itself, just that it does +5V & -5V and I'm using a 12V-2A power supply from an old 1TB WD external harddisk.

 

But this arrived today 

 

So I am now able to use the original power supply board and see what happens - I'm just a bit worried that the +5V is reading as 5.25V when it's not plugged in to anything.  But I'll see what it reads when it's back in the console, they say the Pico digital pins won't like 5.5V but is happy with 5.0V or below, but I think the 74LS chips I've got connected directly to the Pico will only go to around 4V for logic high anyway.

[JasonACT]

+5.23V  but nevermind, I've tested 2 of my boards with it (my rock-solid, didn't need an R4 at all one, along with the 16MB stacked chip one - both running at 266MHz) and it passed the SAMS test on both up to 1MB where I switched off the machine.  I also ran some other SAMS tests and they all passed too.  This isn't unexpected though, there's a few things about this machine which makes me think it's only ever had limited use - so is practically in brand new condition...  These are the things I know about it:

 

1. It came from Texas and doesn't have the V2.2 GROMs.

2. The felt strip in the GROM port is totally clean, I even left it in, which BITD I never would have done.

3. There was a hand-written 'extra' page in the manuals, clearly written by a child (pencil & large letters) with notes about TI BASIC.

4. The power board switch was broken, which I've replaced, the small metal conductors were not inside the machine, I checked before 1st power-up.

5. I've measured the master clock speed and I get repeating values of 3.030 - 3.030 - 2.941 MHz - so it's running at almost exactly 3.0MHz average.

 

I'd say the early 80's kid had broken the switch early on (bad design in the beige model, using the switch itself as a stopper) and it's been in storage ever since.

[RickyDean]

I have 3 of the first run Pi Pico[W] Peripheral Expansion Box Side Port Device boards for sale 6.00 each plus shipping to anywhere in the US fisrt class mail, and I will ship outside the US for the extra postage it will cost. PM me if interested. These are blue in color as you can see on a couple of the pages of this thread. There is plenty of information on the resistor mod on this board to make it function with the SPI SRAM chip in this thread, page 6 I believe.

I am going to do a modifcation to the second run board to accommodate SOP and USON Spi Sram as I have 10 USON ic's looking for a home someday. 

Edited April 20 by RickyDean
subtracting content

[JasonACT]

It arrived, and I've tested it, LTA3183, 1981 ROM non-QI...  Note the curly 3 key, they say they are Alps keyboards when beige...  No, a rare Futaba:

 

Works, all tests pass (to 1MB, @266MHz) and I may have picked up a couple of extra items in the process 

Edited April 22 by JasonACT

[JasonACT]

On 4/22/2024 at 6:31 PM, JasonACT said:

I may have picked up a couple of extra items in the process 

LTA3083, 1981 ROM non-QI...  Same keyboard as my NTSC console, the "higher spacebar" keyboard 

 

 

Works really well too.  Like my NTSC console, the TI COMPUTER 99/4A metal sticker isn't peeling up on this one 

 

I also received a PHA2036 and a PHA2030 PAL Encoders - these tests were done with the UHF Channel 36 adaptor - it gives a nice picture.  The '2030 VHF Channel 4 unit was awful, but I had prepared some RCA receptacles, BC548 transistor & resistors/caps to do the composite output mod on one of these.  So I picked that one for the job...

UHF-36 adaptor:

 

 

Converted VHF-4 adaptor:

 

 

The still picture here doesn't look bad, and it's got the thing working well enough and is sharper, but there's a few things wrong - given how bad it was, this is still an excellent mod, but I'm wondering if I've done something wrong?  Maybe @ti99iuc can advise?

 

There's constantly moving colour shadowing (interference maybe) you can see it when zoomed in on the photo, but it's pretty noticeable due to the constant change.

The colours are slightly different, seems a bit lighter, red has lost its vibrance.

The left margin is eating into one or two pixels - mostly noticeable on text screens that use all columns.

 

But yeah, pretty cool all these things still work after so long and I'm happy with the composite mod in any case.  The unmodified '2030 was really the only thing which has any issue 

 

Edited April 23 by JasonACT
Add pic

[JasonACT]

.

Edited April 23 by JasonACT
Dup

[JasonACT]

.

Edited April 23 by JasonACT
Dup

[JasonACT]

1 hour ago, JasonACT said:

There's constantly moving colour shadowing (interference maybe) you can see it when zoomed in on the photo, but it's pretty noticeable due to the constant change.

Interesting, the longer the console is powered up, the faster the shadows move and they become a fair bit less noticeable.  This was my number one concern, so it's a win.  Might be heat related, just needs to warm up a bit?

[JasonACT]

Switched over my PAL consoles, from the warm one to the cold one.  Initial power up has the shadows moving fast, but after some minutes you can see them slow down, that's where I took this photo - at the point they completely stopped.  Then they start going in the reverse direction as the console heats up, until it becomes fast enough to be less distracting:

 

The yellow bar to green bar shows the strange shape of the colour change, at the point everything is completely steady.

[ti99iuc]

@JasonACT 

Unfortunately, these kinds of problems are quite normal with the mod. I never understood, with absolute certainty, what it could depend on; it could be various things because being old equipment connected to modern monitors anyway already this could be a factor. 

For example, I had noticed that using CRT monitors worked better with the mod, but the result changed also only after changing various brands of monitors/LCD TVs. 

A whole host of factors can create signal changes and I could not figure out why (the beauty of analog signals  ). Often some components deteriorate over time and could give different problems, either from modulator parts or Monitors/TVs. But it can also be just more or less compatible with the LCD monitor/TV, depending on brand-

Another example, on the same TV monitor I had improvements by using the RGB SCART port of the TV interposing an RCAtoSCART adapter on the cable instead of using the direct RCA connectors. 

However, I see you are doing a lot of trial and error and maybe you will be able to optimize the signals better.

Edited April 23 by ti99iuc

[Gary from OPA]

12 minutes ago, ti99iuc said:

@JasonACT 

Unfortunately, these kinds of problems are quite normal with the mod. I never understood, with absolute certainty, what it could depend on; it could be various things because being old equipment connected to modern monitors anyway already this could be a factor. 

For example, I had noticed that using CRT monitors worked better with the mod, but the result changed also only after changing various brands of monitors/LCD TVs. 

A whole host of factors can create signal changes and I could not figure out why (the beauty of analog signals  ). Often some components deteriorate over time and could give different problems, either from modulator parts or Monitors/TVs. But it can also be just more or less compatible with the LCD monitor/TV, depending on brand-

Another example, on the same TV monitor I had improvements by using the RGB SCART port of the TV interposing an RCAtoSCART adapter on the cable instead of using the direct RCA connectors. 

However, I see you are doing a lot of trial and error and maybe you will be able to optimize the signals better.

I had similar issues with my Toshiba 32 flat screen even tho it accepts yellow, white, red rca directly l, I bought a cheap RCA to HDMI converter and used the HDMI port on the TV and it was much better. Shame as the whole reason I bought that TV to began with as it still had RCA connections on it which is rare these days.

[JasonACT]

12 hours ago, ti99iuc said:

However, I see you are doing a lot of trial and error and maybe you will be able to optimize the signals better.

I had a bit of a go, but I think you're right, it's just the nature of the beast.  My USB2 capture device has no issue with the lost 2 pixels on the left, they are there.  You can change resistors and get colour red to be more vibrant, but at cost to the other colours that may get washed out.  I tried different caps but that made little difference until the unit failed altogether, but I've ended up picking a single cap that works OK, rather than two connected ones from the original circuit.

 

You just need to let the video chip warm up, I think.  This is what I've got now, a 1K pot to get the colours right in various games, but it's much the same as the original circuit.  Thanks for making it available 

 

I snipped the tin side panel so the cable can go through, rather than poke it through the bottom-panel hole - as the unit bulged otherwise, when screwed together.

[Gary from OPA]

On 2/6/2024 at 9:42 AM, wiloson said:

Does anyone in Canada want to build this project with me?   I can order the PCB and assemble & test but it's cheaper to build if we order a batch of 5 or more.

I'm located in Toronto.

 

Just noticed your message about this, as I been reviewing all the pages as recently @RickyDean mailed me one of his extra boards.

 

I been making up a list of parts I need to order from mouser canada and amazon, and I was wondering if you built yours yet or did you grab a extra board from Ricky as well?

 

Let me know, as I located in Scarberia that place to East of Toronto, maybe we can get together as well for coffee, since I have not been in contact much with local fellow TI99'ers.

 

Looking forward to hearing from you. Send me a PM if you wish. Take Care, and enjoy the rest of your day here in GTA at least the sun is out today, but still on the chilly side of the weather!

[RickyDean]

I've been tinkering with @JasonACT's toy. I am adding two uson pads (circled) for APS6404L-3SQR-ZR uson memory ic's as I accidently bought 10 of these on my first try of ordering the BOM for this device instead of APS6404L-3SQR-SN sop. It was a time trying to figure out where to put the first one that corresponds to U2, the one in U7's footprint was much easier. I hope these don't violate design rules to much and are producable. And as @JasonACT states below the 74LVS245 is 74LVC245 now on my drawing. This is just for me to use up my stock of uson memory.

 

 

Edited April 26 by RickyDean
added content

[JasonACT]

@RickyDean I hope they work OK 

 

The '245 chips seem to be mislabelled though, "LVS" should be "LVC".

[RickyDean]

@JasonACT, I figure if I can't fit them on, I can always use the bigger memory, no foul.

[Gary from OPA]

30 minutes ago, RickyDean said:

@JasonACT, I figure if I can't fit them on, I can always use the bigger memory, no foul.

I wonder if anyone ever made a tiny PCboard adapter to convert a uson to sop.

 

I only seen uson to dip adapters and sop to dip ones.

[RickyDean]

32 minutes ago, Gary from OPA said:

I wonder if anyone ever made a tiny PCboard adapter to convert a uson to sop.

 

I only seen uson to dip adapters and sop to dip ones.

Yea, I searched a few days ago without finding anything, that is why I am going this route.

[JasonACT]

New PPEB firmwares to test, mainly for @wiloson but also @chue and @RickyDean too.

 

These are the 2MB SAMS and 8MB SAMS versions (8MB only for use on the v3 board, or piggybacked mod) but both are running at 250MHz and one NOP.

 

You can now change the settings for these in autoload.cfg to increase them using a line saying "MHZ=252" up to 280, and "NOPS=2" up to 9.

 

I've also cleaned up the speech PWM to use 32KHz refreshes instead of 16KHz (I'm far too old to hear 16KHz, but there was some 16KHz noise in test-recordings I had made in Audacity, and they are gone now - your pets and/or grandkids will be grateful).

PPEB2.zip

Edited April 26 by JasonACT
Edit: only use even MHz values.

[JasonACT]

I should explain what increasing MHZ and NOPS does, so here is a picture of the read timing from the technical data manual:

 

NOPS adjusts the #1 settling time, the time between detecting *MEMEN going low and sampling the address value

MHZ affects how fast the PSRAM can output a byte, which needs to be less than the #2 access time shown in the diagram

 

Ideally, you would set 9 NOPS and 280 MHz and it would all fit nicely in the diagram.  But as we have found, the PSRAM chips are dodgy more often than not, and our TIs are all different too.  So we have to search for our particular working values.  All of mine work at 250MHz and 1NOP (with this newly optimised firmware), but some TIs may need more (#1) NOPs and this will push-out (#2) access time - so with more NOPs you may need to increase the MHz to pull it back within spec.  But then you are fighting the dodgy PSRAMs, which starts to give bit errors on reads, when the speed is too fast for them.

 

Finally, here is a recording of a real speech synthesizer (first) vs my device (second) saying "HELLO" in extended basic.

 

 

 

I wrote somewhere that I was not happy with the signal noise my device introduced to the TI (through the sound chip) but I can say the real speech synth is about 2-3 times more noisy than my device (when you turn the volume right up to listen, when nothing is playing).  The second word (my device) in the recording does have a more metallic sound to it, but I think the volume is pretty close between the two - and I'm more happy with my device overall after doing this test. 

[+chue]

3 hours ago, JasonACT said:

I should explain what increasing MHZ and NOPS does

Didn't see the above explanation, but did test with values of 266 Mhz and 2 NOPs.  I picked those based on what I had used in the past.  The SAMS burn-in worked on all 3 of my TIs.

 

Will test 250 Mhz and 1 NOP next.

[JasonACT]

8 hours ago, chue said:

Didn't see the above explanation, but did test with values of 266 Mhz and 2 NOPs.  I picked those based on what I had used in the past.  The SAMS burn-in worked on all 3 of my TIs.

 

Will test 250 Mhz and 1 NOP next.

I'll be interested in what you find out, I recall one of your machines started working properly with my device when we increased the NOPs.

 

I've also done some calculations for what adding an extra NOP means to the base setting of 250MHz:

 

1 clock = 4nS @ 250MHz (basically 1000/250) which is the time it takes to execute a NOP in the Pico at this speed.

 

For the PSRAM to read 1 byte = 48 clocks...4*48 = 192nS

 

Now, adding in an extra NOP increases the start time by 4nS, so we have to shorten the PSRAM read time by as much to get the same timing as before...

 

1000/252=3.968253r....*48 = 190.476190r - no not enough
1000/254=3.93700......*48 = 188.9763779 - no not enough
1000/256=3.90625......*48 = 187.5 - yes, it's now more than 4nS shorter

 

So we need to increase the frequency by around 6MHz for each NOP that's added, but that's only a rough guide.  It really depends on the TI, for how many NOPs are needed and for when the sample takes place by the multiplexer - which both seem to be different amongst consoles.

 

EDIT: Actually, there's a lot more going on in calculating the address to read, so the best case scenario of 192nS actually blows out to about 560nS (almost 3 times as long) when tested with an analyser...  That would make each NOP worth about 2MHz in real terms.  It also makes me think there's more optimising that can be done, but I just can't quite see it at the moment.

Edited April 27 by JasonACT
Checked with the logic analyser