128MB SDRAM Troubleshoot

[flipflop]

I assembled a 128 mb module using the instructions on the wiki.

The first time i plugged it in it showed a lot of errors in memtest. So I reflowed the U3 inverter and some pins on the memory chips that looked like they might not have the best contact.

Now it seems to be completely dead to memtest though

When I load cores that use it they seem to work though I wonder if it will work on the more intensive ones.

I read somewhere that the U3 is problematic but how can i test if thats the problem ?

I reflowed some more and now it doesn't work at all. Not in memtest and not in the arcade cores that worked before.

The silkscreen came off a bit in the corner after messing too much with a bridge there. I measured that even though the bridge seems to be removed there is a short between the last 3 legs of the chip.

Any ideas ?

How come the capacitors are all 1uF on the pdf in the hardware repo ?

https://photos.app.goo.gl/19cbGmXp9ogzbdP17
https://photos.app.goo.gl/iRUUTo3cuu4ZDhSZ6

[Nat]

If you can include some images of the memory module you have made it will be easier to help with trouble shooting the problem, however if you have created a permanent bridge by removing the solder mask, it might not easily be fixable.

These are the capacitor values I use:

MiSTer-FPGA-128MB-SDRAM-Capacitor-Values.png (102.91 KiB) Viewed 5072 times

[flipflop]

I fixed the link to the image..

I didn't think the removed solder mask would cause a bridge in itself

Added a whole board image

[yellperil]

If you’ve only visually inspected the pins, try mechanically testing them as well.
I have found a very light push on each pin with fine tweezers will quickly identify which ones are not attached.

The tolerance for the connector under the memory and the memory is very small on this board, so the chips may not have been sitting completely flat. Which lifts the pins up. Even under a microscope it’s hard to just see.

Bridges are easy to fix, use flux and solder wick.

[Nat]

The silkscreen came off a bit in the corner after messing too much with a bridge there. I measured that even though the bridge seems to be removed there is a short between the last 3 legs of the chip.

Looking at the two images, I suspect this is the issue.

In my opinion your best bet is use a hot air rework station and salvage the two SDRAM chips, capacitors and inverter, and have another try on a new PCB.

[yellperil]

Ignore the solder mask. It’s not the issue.

Get your multimeter out and test those pins.

They are 52, 53 and 54

52 and 54 are meant to be connected to each other.

53 is not meant to be connected to 52 or 54.

If you have a spare pcb you should be able to verify the
Ground pins. Also look at the data sheet.

[flipflop]

I have confirmed the short between pins 52,53,54 using a multimeter. I just don't see where the problem could be hiding. I have ordered a new pcb and will try to move the components.

If I solder wick the pins and then use hot air will the chips then be clean enough to lay flat on a new pcb ?

[yellperil]

Some ideas, I’ve used both of these to clear shorts before.

If you have a scope with a bright light try to look between the
Pins

It’s possible the short is behind the pins of the chip along the exposed traces. You could try adding more solder to those pins ( deliberately bridging them ) then wick it away. The intention being for it to flow a little under the chip then suck out the short.

The wick won’t be effective if it can’t draw out the blobs from underneath. That’s why you need to add more solder first.
Use flux!

Also use quite thin wick 3mm’ish. Otherwise you need too much heat.

If you used paste it may not have drawn up so use some pcb cleaner get right under the chip and gently scrape a sharp thin sharp tool between the legs and under the chip if you can see any residue to break it up then flush it again.

[Nat]

If I solder wick the pins and then use hot air will the chips then be clean enough to lay flat on a new pcb ?

Yes they should be, my suggestion is use a large nozzle and get even heat over the chip, remove each chip individually, the PCB being 4 layers does soak up a fair bit of heat.

While heating keep gently nudging the chip with metal tweezers as soon as the solder melts completely lift the chip off and allow to cool, base your hot air temperature on the type of solder you used.

[legacypixels]

lots and lots of flux

[flipflop]

I gave it another rework shot after learning that it can work as a 64 mb module with only the U1 ram chip. I removed U2 with hot air and found the bridge which was under the chip in the corner where the silkscreen has come off.

After getting that removed the module worked at 150 MHz as 64 mb

Then I cleaned up the U2 area with flux, solderwick and IPA, made the solder points for the connector even flatter and installed the U2 chip.

After what I thought was a careful drag soldering it did not work.

I found the loose pins under my cheap banggood microscope by nudging them with a tweezer and seeing if they moved. This was a tip from someone I can't remember where I read it but it was a great tip.

Ran without errors for 7 hours at 140 while i slept

Now I just regret having ordered an assembled board from misterfpga.co.uk but maybe I can find someone who wants one.