Jump to content
IGNORED

Pi Pico[W] Peripheral Expansion Box Side Port Device


JasonACT

Recommended Posts

@RickyDean

 

I'm not sure why, but Polpo on GitHub has again changed his pin settings (3 days ago, still no fast slew & back to 2mA strength instead of 4mA) which I already tested months ago and wouldn't work for me at even 250MHz.  Retested that anyway with the ESP chip, but no-go, errors galore.  Maybe he's seeing issues with different chips and is trying to come to some sort of "works for everything" setting, I don't know?

 

Then the idea hit me, move from my settings of fast slew & 4mA but up to 8mA.  I was able to pass the 1MB SAMS test at 266MHz (133MHz PSRAM) with the ESP chip (which I again verified still wasn't working at that speed just before I tried 8mA).  I've now lowered the speed to 258MHz (129MHz PSRAM) which is one notch above the working minimum speed, and am running one last pass through the memory tester now.  Half way through, and no issues, touch wood...

PPEB2.ino.uf2.zip

  • Like 1
Link to comment
Share on other sites

13 minutes ago, JasonACT said:

@RickyDean

 

I'm not sure why, but Polpo on GitHub has again changed his pin settings (3 days ago, still no fast slew & back to 2mA strength instead of 4mA) which I already tested months ago and wouldn't work for me at even 250MHz.  Retested that anyway with the ESP chip, but no-go, errors galore.  Maybe he's seeing issues with different chips and is trying to come to some sort of "works for everything" setting, I don't know?

 

Then the idea hit me, move from my settings of fast slew & 4mA but up to 8mA.  I was able to pass the 1MB SAMS test at 266MHz (133MHz PSRAM) with the ESP chip (which I again verified still wasn't working at that speed just before I tried 8mA).  I've now lowered the speed to 258MHz (129MHz PSRAM) which is one notch above the working minimum speed, and am running one last pass through the memory tester now.  Half way through, and no issues, touch wood...

PPEB2.ino.uf2.zip 379.1 kB · 0 downloads

Well here is the picture of my results after programming this. I didn't do anything on this today as I did other things. I'll try changing this psram out tomorrow and very carefully solder one pin at a time, leaving some cooling time between to see if I'm killing it somehow.  Have a good evening, it's 1am my time.

20240118_010023.jpg

  • Like 1
Link to comment
Share on other sites

My best chip, the APS6404L on my first V2 board I built, that ran @276MHz/138MHz, is now running at 282MHz/141MHz with the full strength value (passes the 1MB SAMS test successfully at that higher speed).

 

@RickyDean I don't think you're killing them when just soldering pin-by-pin, not sure about the heat-gun though, but we can see in almost all the errors you and I get, a single bit is being merged into the next position along...  From my tests tonight, I'm fairly sure it's the shape of the clock signal making a difference for me (we're almost certainly only getting read errors, because read "2:" almost always works properly).  The shape of the clock is what changes the timing of when the Pico can see the PSRAM's correct single-output signal being used here to read.  A curved clock signal will delay the correctly read value (which I think is happening for you at a lowish speed, and at an overclocked speed for me) but the sharper clock signal with the higher strength brings it forward for me and I get better results.

 

While it might be your particular Pico, and might be worth building one other board to test, the other big difference we have is you pulled off the 3rd party installed pins.  Before I'd go out and buy a "no header" Pico to test that, which would be my next move, I would be heating up the bottom of the V2 board with the heat-gun on the corner of the Pico near the PSRAM chip to see if I can produce some smoke (to burn off anything hidden there between the board and the Pico).  If there's some contaminate in that area of the pins being used for the PSRAM, it may be having an effect on the shape of the clock's signal even at lower speeds.

 

Not 100% sure, but I think that's all I've got for suggestions.

Link to comment
Share on other sites

5 hours ago, RickyDean said:

20240118_010023.jpg

5 hours ago, JasonACT said:

@RickyDean How come your last bank goes from E000 -> 9E9E ?

 

Is the memory test loaded into my device or a FinalGROM?

 

This has been on my mind for hours now, my initial two thoughts on this were:

 

1/ if it was running from a FinalGROM then the cartridge or TI this is being tested on might not be 100%

2/ if it was running from my device, then OK, the PSRAM being bad might cause that...

 

For option 2 though, that would require the PSRAM being "used" for the 8KB ROM this memory test program uses, but I never use PSRAM for the first 8KB of any loaded cartridge - Even when enabling the multi-base GROM Library function, I ensure the 1st 8KB ROM is in real (fast) Pico RAM (like the Disk DSR ROMs which are working).

 

There appears to be something wrong here, maybe unrelated, that I just can't explain.  I think I feel a migraine coming on!

Link to comment
Share on other sites

15 hours ago, JasonACT said:

@RickyDean How come your last bank goes from E000 -> 9E9E ?

 

Is the memory test loaded into my device or a FinalGROM?

The test is loaded from FinalGrom, as I've only been concentrating on the memory issue, before I learn any more of the system.

 

I tried this last program on both these units, but didn'tt work. These 4 picture are the failures for the primary unit (the one without risers) that we have been testing and changing psram chips on. I had the Idea to change around the 74lvc245's, the differences you see in the photos are from that testing. I don't know if they mean anything or not, I'll let you determine that.

20240118_161138.jpg

20240118_160948.jpg

20240118_160740.jpg

20240118_160544.jpg

Edited by RickyDean
Link to comment
Share on other sites

I've went ahead and replaced the mem chip. I'm soldering 1 leg, then letting cool, then the next. There will be no reflowing with the blower on this one at all. I'm not disposing of the pulled chips, if there is a way to test them out of circuit later, I'll test them before disposal. I have two more unused Pico's, but they also came with the headers soldered on, so they have to have those removed or I find another socket/riser arrangement to mount them to the board.

Edited by RickyDean
spelling
Link to comment
Share on other sites

13 hours ago, RickyDean said:

I tried this last program on both these units, but didn'tt work. These 4 picture are the failures for the primary unit (the one without risers) that we have been testing and changing psram chips on. I had the Idea to change around the 74lvc245's, the differences you see in the photos are from that testing. I don't know if they mean anything or not, I'll let you determine that.

I don't think it means anything, strange that it changes, but if you can get into the menu and swap pages with the spacebar, then all of those chips are working.

 

5 hours ago, RickyDean said:

I've went ahead and replaced the mem chip. I'm soldering 1 leg, then letting cool, then the next. There will be no reflowing with the blower on this one at all. I'm not disposing of the pulled chips, if there is a way to test them out of circuit later, I'll test them before disposal.

I really would have thought we've tried enough already, and instead, there's a signal issue (I think it's the clock, because the Pico PIO state machine running all of this just toggles the clock and reads the data pin, but getting the right data value depends on how long the clock takes to transition states).

 

5 hours ago, RickyDean said:

I have two more unused Pico's, but they also came with the headers soldered on, so they have to have those removed or I find another socket/riser arrangement to mount them to the board.

Yeah, I would have thought that was fine (using sockets) given mine worked.  You also cleaned up the desoldered one, so that should have been OK too, but here we are still trying to get it to work.

 

I'll look into writing a stand alone memory tester .uf2 which flashes the LED that you can try on the bench (not plugged into the TI) over the coming weekend.  That will rule out all the other chips having any issues.  You could still try a different SD card though, or run the TI memtest with it ejected after power-up.

  • Like 2
Link to comment
Share on other sites

@RickyDean The memory tester attached, as mentioned above, is for stand-alone usage on the bench.  Once loaded, it will run at a low frequency (250MHz) and try to pass the read/write test, then move onto the next higher speed.  It will blink a few times a second while running.  When a test eventually fails, it will blink slowly a certain number of times, then remain off for 5 seconds and then repeat that slower-blink number.  Count them and report back, do this 2 times, once with the SD card inserted and again with no SD card.

 

I can report, my ESP chip tops out at the speed mentioned previously with an SD card inserted, but with no SD card inserted the test finishes with no errors and stops blinking altogether (no blinks = pass at the highest speed, 141MHz).

 

Running this test with my "do not use" SD card from the junk-box, it gets errors a little bit earlier at a slower speed, so different SD cards do make a difference.

memtest.ino.uf2.zip

  • Like 2
Link to comment
Share on other sites

2 hours ago, RickyDean said:

Report unit with riser: w/SD card= 9 blinks, wo/SD card= 0 blinks

      unit without riser: w/SD card= 1 blink, wo/SD card=14 blinks

The unit with riser w/SD gets an error @ 266MHz (133MHz PSRAM) ... No error detected wo/SD all the way up to 282MHz (141MHz PSRAM).

The unit without riser w/SD gets an error @ 250MHz (125MHz PSRAM) ... and an error wo/SD @ 280MHz (140MHz PSRAM).

 

BTW: The list of frequencies is in the doco, except I've added 282MHz as one more on the end, for this tester.

 

This is perhaps what I'm seeing also, your one "with riser" probably still has the AP6406L, which for me is only slightly affected by my SD card, the ESP chip seems to be far more sensitive for us both...  My tester isn't as thorough as the TI, so it might pick up errors at a lower speed, but it does look to me like it might pass on both your boards if you boot the latest 12mA software with the SD inserted (you need the DSRs loaded to configure the SAMS) but eject the SD before starting the TI memory test.

 

You might like to re-run the bench tester using a handful of other SD cards you have lying around (they don't need to be formatted any particular way, just have them in the socket) to see if you can find one that affects the ESP chip less.

  • Like 1
Link to comment
Share on other sites

4 hours ago, RickyDean said:

unit without riser: w/SD card= 1 blink

This actually looks like something else is wrong, I've put every SD card I own through the test now with these results:

 

ESP + Generic 8  = Error @ 11 / 135MHz
ESP + Toshiba 32 = Error @ 12 / 136MHz
ESP + Lexar 16   = Error @ 12 / 136MHz
ESP + Lexar 32(P)= Error @ 11 / 135MHz
ESP + Lexar 64   = Error @ 11 / 135MHz
ESP + SanDisk 32 = Error @ 12 / 136MHz
ESP + SanDisk 64 = Error @ 11 / 135MHz
ESP + SanDisk256 = Error @ 11 / 135MHz
ESP + Lifelong 8 = Error @ 11 / 135MHz

 

And the ESP chip on mine has no issues passing the rated spec. of 133MHz with any SD card attached.

 

2 hours ago, JasonACT said:

My tester isn't as thorough as the TI

Also, this isn't as accurate as I should have written...  I only test the first 64KB (not the full 1MB SAMS range) because I'm not looking for bad bytes, but rather issues with signals at different speeds.  The TI memory tester does 12 runs of 4 tests = 48 tests for 32KB on the first set of pages where most of your errors show up (now that I look closely at the code) but mine is doing 125 runs of 2 tests = 250 tests over 64KB.  Actually, I thought to modify mine to add all 4 tests, but it makes no difference, mine all fail at the same place no matter which 4 test types are being used.

 

In a way, I think I'm being more thorough than the TI memory tester.

  • Like 1
Link to comment
Share on other sites

Tonight, I've been soldering up a set of short Dupont wires, I actually took a picture, but I've decided not to post it - it's a rats nest - this is just the bare components and no circuit board.  But I am able to get an ESP PSRAM chip and SD card socket (with or without an SD card inserted) to fail at 250MHz.  If I leave every wire plugged in except the clock to the SD socket - then the ESP PSRAM works all the way up to 141MHz...  This shows wires just hanging off the PSRAM chip and going nowhere affects it.

 

Truth be told, I don't know why yours @RickyDean is behaving differently to the 3 I've built.  We have cut out all components except the Pico, SD, PSRAM and circuit board (they sometimes have issues too) without success and I don't want you wasting any more money on this.

 

Find attached a version of the .uf2 with SAMS disabled (reverting to standard 32KB expansion memory) with the PCode card also disabled.  I had to disable the PCode to get back some memory for the other functions to continue working.  You will probably need to continue using a FinalGROM, while I still have all the code enabled for loading cartridges into PSRAM, it may or may not work with whatever issue is showing errors in the memory test.

 

It would be interesting for you to view the Bad Apple Demo though - all the code in that runs from the scratchpad - so any data errors reading from PSRAM shouldn't crash the TI and you might see some sparkles on-screen or in audio.

 

I'm sorry it didn't work out for you mate.

PPEB2.ino.uf2.zip

  • Like 1
Link to comment
Share on other sites

3 hours ago, JasonACT said:

Tonight, I've been soldering up a set of short Dupont wires, I actually took a picture, but I've decided not to post it - it's a rats nest - this is just the bare components and no circuit board.  But I am able to get an ESP PSRAM chip and SD card socket (with or without an SD card inserted) to fail at 250MHz.  If I leave every wire plugged in except the clock to the SD socket - then the ESP PSRAM works all the way up to 141MHz...  This shows wires just hanging off the PSRAM chip and going nowhere affects it.

 

Truth be told, I don't know why yours @RickyDean is behaving differently to the 3 I've built.  We have cut out all components except the Pico, SD, PSRAM and circuit board (they sometimes have issues too) without success and I don't want you wasting any more money on this.

 

Find attached a version of the .uf2 with SAMS disabled (reverting to standard 32KB expansion memory) with the PCode card also disabled.  I had to disable the PCode to get back some memory for the other functions to continue working.  You will probably need to continue using a FinalGROM, while I still have all the code enabled for loading cartridges into PSRAM, it may or may not work with whatever issue is showing errors in the memory test.

 

It would be interesting for you to view the Bad Apple Demo though - all the code in that runs from the scratchpad - so any data errors reading from PSRAM shouldn't crash the TI and you might see some sparkles on-screen or in audio.

 

I'm sorry it didn't work out for you mate.

PPEB2.ino.uf2.zip 377.79 kB · 0 downloads

Okay, I'll try these soon, I've been dealing with a bad PEB box and had been testing my Geneve(s) with a clonded Gal video jedec, but accidently killed the 5 volt line in my PEB box. So that has been taking my time yesterday, and still will be.

Link to comment
Share on other sites

On 1/21/2024 at 2:44 AM, RickyDean said:

I've been dealing with a bad PEB box

I'm sorry to hear that, I hope you get it all working again soon.

 

I've been trying different things out in the meantime, just to know the limitations of these chips (they seem to be the weakest link)...  I made the memory tester more aggressive too in that it runs as fast as is allowed by the specs. (the doco states you must raise *CE for a certain period after accesses for the refresh circuit to work and I was previously being very generous, like a TI would be).  That lowered my numbers on the ESP chip a little (the AP chips don't seem to be affected as much).  Then I thought to build a minimum test board, and not clean it up like I normally would, in fact, I made it using more solder than needed just to get as much flux on the board as I could.  I'm not saying your board isn't totally clean, but this board does the same thing as yours, with no SD card inserted it passes tests all the way up to 141MHz.  With my good SD card inserted, it only passes 127MHz.  With my bad SD card inserted it only barely passes 125MHz - in fact, because it's on risers, I needed to wiggle the Pico for good contact to make it pass at all - sometimes it fails at 125MHz.  Maybe I've got a bad chip and it's not the flux?  I kind of want to leave it this way in case I think of another test I can do with it.  Oh, one last thought, since the AP memory chips seem better, maybe one of those will pass for you with the 12mA build of the software?

 

image.thumb.jpeg.7100a44bf751f2055f2e44a96a89b905.jpeg

IMG_20240123_172942_HDR.thumb.jpg.cd41ed60ffcbde673bb767d60d1d2253.jpg

Edited by JasonACT
All MHz values are for PSRAM
  • Like 1
Link to comment
Share on other sites

Yeh, I haven't had a lot of free time, been dealing with doctors and PT for my shoulder surgery and then soldering up some thngs like the second Tang and other things. Been checking my PEB boxes to see if I have another good one, but at this time it seems I don't. Will have to rebuild one. I'm going to use an ATX power supply and get my Rave 99 board running, waiting on this to show up. 24/20-Pin ATX DC Power Supply Diagnostic/Breakout Board Module from Amazon. I'll try tp pay some attention to this today or this evening and give an update.

Link to comment
Share on other sites

@JasonACT

I have programmed your last Pico file onto the board that has the Pico soldered to the board and the 32k passes. So that makes this useable as is.  My pico on riser board is not detecting the ram, so I'll have to work it. 

 

Edit: Just changed back to one of the old APS6404L-3SQR-SN on the riser board and it is now working as a 32k memory, at least it is now usable too!!

Edited by RickyDean
  • Like 2
Link to comment
Share on other sites

21 hours ago, JasonACT said:

I'm sorry to hear that, I hope you get it all working again soon.

 

I've been trying different things out in the meantime, just to know the limitations of these chips (they seem to be the weakest link)...  I made the memory tester more aggressive too in that it runs as fast as is allowed by the specs. (the doco states you must raise *CE for a certain period after accesses for the refresh circuit to work and I was previously being very generous, like a TI would be).  That lowered my numbers on the ESP chip a little (the AP chips don't seem to be affected as much).  Then I thought to build a minimum test board, and not clean it up like I normally would, in fact, I made it using more solder than needed just to get as much flux on the board as I could.  I'm not saying your board isn't totally clean, but this board does the same thing as yours, with no SD card inserted it passes tests all the way up to 141MHz.  With my good SD card inserted, it only passes 127MHz.  With my bad SD card inserted it only barely passes 125MHz - in fact, because it's on risers, I needed to wiggle the Pico for good contact to make it pass at all - sometimes it fails at 125MHz.  Maybe I've got a bad chip and it's not the flux?  I kind of want to leave it this way in case I think of another test I can do with it.  Oh, one last thought, since the AP memory chips seem better, maybe one of those will pass for you with the 12mA build of the software?

 

image.thumb.jpeg.7100a44bf751f2055f2e44a96a89b905.jpeg

IMG_20240123_172942_HDR.thumb.jpg.cd41ed60ffcbde673bb767d60d1d2253.jpg

I'll try the files you have posted this whole thread again, in the next day or two, on both units. Will give an update. I eventually will continue to build two more as I have two pico's left, but my desoldering gun glass tube broke while cleaning the other day, so I have to wait for a replacement to get here, coming from Spain.

  • Like 1
Link to comment
Share on other sites

On 1/18/2024 at 5:06 PM, RickyDean said:

Well here is the picture of my results after programming this. I didn't do anything on this today as I did other things. I'll try changing this psram out tomorrow and very carefully solder one pin at a time, leaving some cooling time between to see if I'm killing it somehow.  Have a good evening, it's 1am my time.

20240118_010023.jpg

I've tested my flux covered board a lot further today, I updated the bench tester to report via the serial port in the same format as the TI memory tester.  I'm getting errors that look very much like this picture of what you were getting.  This is at a speed too low to work (252MHz / 126MHz PSRAM) if I complete the board and used it on my TI.  I guess it's time to clean the board up to see if it's just a bad chip...  Long weekend here starts after work tomorrow, so I'll sleep on it for one more night.

 

I'm glad you had a little bit of success today :)

Link to comment
Share on other sites

On 1/23/2024 at 6:02 PM, JasonACT said:

I'm not saying your board isn't totally clean, but this board does the same thing as yours, with no SD card inserted it passes tests all the way up to 141MHz.  With my good SD card inserted, it only passes 127MHz.  With my bad SD card inserted it only barely passes 125MHz - in fact, because it's on risers, I needed to wiggle the Pico for good contact to make it pass at all - sometimes it fails at 125MHz.  Maybe I've got a bad chip and it's not the flux?

@RickyDean My board is totally clean now, and this chip is still failing.  It now passes 128MHz but fails at 129MHz after the clean with my good SD card.  With the bad SD card, it fails at the same spot.  Strangely, with no SD card inserted it fails at 140MHz now it's clean, but that's still close to what was passing before.

 

I guess they are "working" to spec. and I'm just expecting too much for them to co-exist with an SD card on the same Pico pins.  So my advice is to avoid the ESP chips for this build.

 

I'm still interested in what happens when you try the 12mA software with the AP chips.

Link to comment
Share on other sites

15 minutes ago, JasonACT said:

@RickyDean My board is totally clean now, and this chip is still failing.  It now passes 128MHz but fails at 129MHz after the clean with my good SD card.  With the bad SD card, it fails at the same spot.  Strangely, with no SD card inserted it fails at 140MHz now it's clean, but that's still close to what was passing before.

 

I guess they are "working" to spec. and I'm just expecting too much for them to co-exist with an SD card on the same Pico pins.  So my advice is to avoid the ESP chips for this build.

 

I'm still interested in what happens when you try the 12mA software with the AP chips.

I  was busy today with PT and such. I will try today after sleep. to rerun some of those files, I should say, din't realize it was already 1am.

Edited by RickyDean
added content
  • Like 1
Link to comment
Share on other sites

On 1/23/2024 at 6:02 PM, JasonACT said:

IMG_20240123_172942_HDR.thumb.jpg.cd41ed60ffcbde673bb767d60d1d2253.jpg

@RickyDean As you can see from my picture here, this chip (plus 3 others in the 5 I bought from Adafruit) has the surface sanded off.  Only one of mine was a real ESP chip with the company logo still showing - and that's the one that worked.  I've now tried another one and it's also failing, though it gets a little bit further, the fail point is 131MHz.

 

Can you check what they sent you and let me know if you can see the ESP logo on any of the chips?

 

EDIT: I've written an email to my supplier here asking what's going on in my case, we'll see what they say.

Edited by JasonACT
Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Loading...
  • Recently Browsing   0 members

    • No registered users viewing this page.
×
×
  • Create New...