Though that also increases the power dissipated by 10%.
Can you just try the RAM tester on your current prototype? Doesn't matter if the PSRAM section is a dud, even just getting a non-corrupted video out from the tester is a feat.
Platform A2 board is doing MisoYume just fine with no fan, with Vdd supplied by the 2.0 A, 2.0 V supply of Platform A4 board.
So, think the plan is to go back and start from A2, replace the Vdd supply with the 2A one, but keep the 3.65 VIO supply to more efficiently power 3.3V LDOs.
Now thinking that 500 mA is enough for that. Especially since the 2A supply won't give me 3.65 V for some reason that can't figure out...
I don't want to take any chances with the A2 PSRAM layout. Seems to work perfectly, according to tester at 350 MHz.
Built first of Rev. A5. It's also good to 400 MHz with Vdd=2.0.
But, there's a solder jumper one can undo and get Vdd back down to 1.8 V (but not me).
One, kind of mistake, is that the Fan#1 power port is a little too close the two Eval headers.
The Parallax adapter seem to just clear, but my 96 MB board does not .
Accidentally ordered these in HASL instead of ENIG.
But, fortunately there are a few tweaks to make, so not a complete waste...
@Rayman said:
Built first of Rev. A5. It's also good to 400 MHz with Vdd=2.0.
But, there's a solder jumper one can undo and get Vdd back down to 1.8 V (but not me).
One, kind of mistake, is that the Fan#1 power port is a little too close the two Eval headers.
The Parallax adapter seem to just clear, but my 96 MB board does not .
Pretty tidy looking board.
You can try right angle/horizontal JST connectors if that helps and gives more clearance.
Memory tests OK, but at least slightly worse than before for some unknown reason. Think the layers should be the same...
It's good at 340 MHz.
Fails instantly at 350 MHz (this used to work).
8-bit at 350 MHz works (this is good).
slow mode at 350 MHz works (guess that's a good fallback, if needed).
@Rayman said:
Memory tests OK, but at least slightly worse than before for some unknown reason. Think the layers should be the same...
Unless they were the exact same chips used for both tests you may not be able to conclude much given that the APS6404 PSRAM chips are probably only rated to operate up to 133MHz (266MHz P2), or perhaps you found some different devices with higher specs...?
Comments
@Wuerfel_21 think vdd=2 is going to solve a lot of the problem.
Though that also increases the power dissipated by 10%.
Can you just try the RAM tester on your current prototype? Doesn't matter if the PSRAM section is a dud, even just getting a non-corrupted video out from the tester is a feat.
Grab it here: https://forums.parallax.com/discussion/comment/1563561/#Comment_1563561
Need to put in your config.spin2 from NeoYume that sets up the RAM/video/audio (doesn't use USB or SD).
Will do
If pushing up into 1.9-2.0V territory it may pay to carefully check the overshoot and droop with dynamic load changes ?
Platform A2 board is doing MisoYume just fine with no fan, with Vdd supplied by the 2.0 A, 2.0 V supply of Platform A4 board.
So, think the plan is to go back and start from A2, replace the Vdd supply with the 2A one, but keep the 3.65 VIO supply to more efficiently power 3.3V LDOs.
Now thinking that 500 mA is enough for that. Especially since the 2A supply won't give me 3.65 V for some reason that can't figure out...
I don't want to take any chances with the A2 PSRAM layout. Seems to work perfectly, according to tester at 350 MHz.
A2 with Vdd supplied by A4 at 2.0 V can do vga up to 375 it seems. Vga output works at 380 but serial is messed up.
Cranked up Vdd on A4 board to 2.24 VDC. Can do VGA at 400 MHz. Off rails at 410 MHz.
Letting it run at 400 MHz for a while now, with fan. Been good for ~30 minutes so far...
You know, at that voltage, a load stress test will probably need upwards of 4 Amps on Vdd. And 10 Watts of cooling.
Vdd is stable at 2.24 VDC with 2-A supply.
But, only 2 cogs running...
Restarted 400MHz VGA and still 2.24 VDC.
Just using wearable fan on it now. Kind wear around neck. Battery powered, so doubt even 1 W, but don't know...
The failure mode above 400 MHz is interesting...
VGA output is stable, just not right.
That'd make fine abstract artwork for a wall.
Wow really interesting failure mode
I think any fan / air movement at all really useful. Doesn't need much airflow, just something so it doesn't warm its own bubble of air
It'll need serious cooling if ever doing any hard work.
Built first of Rev. A5. It's also good to 400 MHz with Vdd=2.0.
But, there's a solder jumper one can undo and get Vdd back down to 1.8 V (but not me).
One, kind of mistake, is that the Fan#1 power port is a little too close the two Eval headers.
.
The Parallax adapter seem to just clear, but my 96 MB board does not
Accidentally ordered these in HASL instead of ENIG.
But, fortunately there are a few tweaks to make, so not a complete waste...
Pretty tidy looking board.
You can try right angle/horizontal JST connectors if that helps and gives more clearance.
@rogloh Just moved it, not a problem. Want to get new boards with ENIG anyway.
Memory tests OK, but at least slightly worse than before for some unknown reason. Think the layers should be the same...
It's good at 340 MHz.
Fails instantly at 350 MHz (this used to work).
8-bit at 350 MHz works (this is good).
slow mode at 350 MHz works (guess that's a good fallback, if needed).
Unless they were the exact same chips used for both tests you may not be able to conclude much given that the APS6404 PSRAM chips are probably only rated to operate up to 133MHz (266MHz P2), or perhaps you found some different devices with higher specs...?
Ok, built 3 boards now and they can all pass memory test at 344 MHz and run MisoYume.
Guess declare victory with Vdd=2.0 saving the day.
Only 1 of 3 can pass memory test at 350 MHz in fast mode. They all pass in slow mode though.
What happened to the 400 MHz at 2V?
Investigating that…
So far one is ok at 400 but flickering at 405
Another seems to die at 380 or so.
But need to give it a fan and try again
Ok, it's interesting... Got 3 new boards with Vdd=2.0 V.
One can do 400 MHz and palette flickers at 405 MHz (labelled C).
One can do 405 MHz and about 10 pixels flicker at 410 MHz (labelled B ).
The other can do 380 Mhz when started that way (labelled A).
But, in this freq stepping test code it dies when stepping to 375 MHz.