SDRAM board

[tomywong]

I want to make sdram card for myself. Can someone tell me what is the thickness of sdram card PCB ? Thank you.

[THaase]

'Normal' 1.6mm

[tomywong]

Thank you THaase

[hyiger]

I'm a bit late to game in this forum, just started with the MiSTer project but some notes with regard to soldering the SDRAM board. I've done a lot of soldering over the years. I also have a reflow over, hot air station and of course a hand iron. I've built 5 RAM boards (for myself and friends) and consistently get 160+ with Alliance RAM.

1. First I suggest using Alliance memory, getting cheap or NOS memory on eBay will most likely end if grief. Don't panic about not achieving 160+mHz in a memory test, 150mhz is sufficient. I suspect 99%+ of the cores can't go that high anyway. A typical fMax on DE10-Nano boards with the T80 core (used for most of the arcade cores that require a Z80) is probably around 150-155 and that's on a good day... Probably more like 130-135mhz is to be expected. In other words if you can pass at least a 150mhz test then you are fine. In fact you are probably OK with on 140mhz.

2. Any SMD 0805 package capacitor in the range of 0.1uf - 1.0uf will work. That is, you don't need to get the "exact" Kemet part.

3. No need to go overboard and use a stencil with solder paste and a reflow oven (unless you have one of course). However, I suspect if you have a reflow oven you don't need my advice . The chip can easily be hand soldered even if you are new to soldering.

4. I suggest soldering the RAM using a "drag solder" technique and plenty of liquid flux (there are probably plenty of Youtube videos on this). There is no need to solder pin by pin, besides it will take forever. Drag soldering takes seconds. For SMD hand soldering, I use Kester 959T Soldering Flux and Kester 32117 Solder (1.57mm) with a small chisel tip. I've tried many types of solder over the years and I've found that Kester is the best. Secure opposite corners of the ram with solder, apply lots of flux to the pins (but don't go overboard), apply solder to a clean tip then drag it across the pins. Repeat 1 or 2 times as needed. Apply a little bit more flux to clean up any bridges with iron. Should take only a couple of passes and it's quick and effective. Inspect with a 10x loupe or microscope. Note: High quality solder and flux is a must! Do not use cheap solder! If soldering by hand I suggest using 60/40 tin/lead solder (unless you are environmentally conscience, but lead-free soldering can be tricky)

5. Remove flux and debris afterward. Although there are differing views on this, RAM can be sensitive to stray capacitance and certain fluxes can cause problems in this area (I find I get better performance after cleaning the board). Note: you do not need to buy specific "Flux Remover". I use Isopropyl alcohol (ethanol or methanol work fine as well) in a high concentration (around 90%) and a soft-bristle toothbrush. This can be purchased easily in a pharmacy (chemist). I usually rinse the board immediately afterward with demineralized, distilled water but this is probably not necessary. Do not rinse the board with tap water!

5. Again, inspect the solder around the pins using 5-10x magnification to look for: not enough solder, too much solder and most importantly solder bridges. Also make sure there are no debris around the pins (especially fine copper wire if you used a solder braid to clean up excess solder).

[uigiflip]

I know some people are now testing 64MB sdram, is it true there could be possibility of plugging in a second 64MB sdram on unused pins for more memory for neogeo core (and possibly others core maybe minimig)?

[Sorgelig]

I know some people are now testing 64MB sdram, is it true there could be possibility of plugging in a second 64MB sdram on unused pins for more memory for neogeo core (and possibly others core maybe minimig)?

There will be an option later.
Cannot tell any specific details as it's still in development.

[PhantombrainM]

By the way...if we have ~700mbit with 64MB SDRAM Module for NeoGeo and the biggest games have 706mbits...
...can't we just copy the sound ROM's to DDR3?
I think this will free up the left 6mbit's and I don't think that soundfiles/music needs high data throughput?

[THaase]

It will be either 32MB (256MBit) OR 64MB (512MBit) - not both together. (How do you count ~700?)
1GBit Chips are only available as BGA - this will cause (at least) new board, 4Layer and big soldering problems for normal people.
And I'm also not sure (have not count) it we would have enought address lines at all (possible).

[quack75]

MiSTer now supports 128MB dual SDRAM ?

[THaase]

No - why do you think so?
At the moment Sorgelig is thinking about exchanging the 256MBit SDRam chip with a 512MBit one with the same footprint afaik.

[quack75]

No - why do you think so?
At the moment Sorgelig is thinking about exchanging the 256MBit SDRam chip with a 512MBit one with the same footprint afaik.

Cause SmokeMonster said so and he is generally well informed:
https://twitter.com/SmokeMonsterTWI/sta ... 4714981376

MiSTer now supports 128MB dual SDRAM and Sorge has been heavily editing @furrtek
's Neo Geo core. Although this doesn't necessarily mean that 128MB will be required--options are still being explored.

[THaase]

Yes - you're right - I can also see it in the commits from Sorgelig .... well let's wait

[NobleX13]

I just wanted to drop in and say that I built my first SDRAM boards today. Getting just the right amount of solder on the connector pads can be finicky, but after some practice it is pretty easy. I used some MG Chemicals solder paste for the actual memory module and the capacitors. Worked a treat!

I also quickly learned to use the DE10-nano to determine the spacing for the three-pin header. I had to desolder my first one and spend some time cleaning out the vias to get it fixed.

Both of the modules I have assembled so far have passed the memory test core at 167 MHz. Pretty pleased with this so far.

I have another eight boards to build. This project was the boost I needed to get into SMD soldering seriously and I cannot thank you enough. Obviously I only need one board for my setup, but I figured I would build all ten as practice.

I also went a little crazy and ordered parts for a batch of IO and USB boards.

Here is a completed pic. Please ignore the flux residue; the black solder mask makes it look worse than it is.

[Ashenshards]

So I'm testing my second SDRAM board and both so far are failing almost exclusively at 90MHz ( like the red number just races up on 90 MHz). I've got problems with lots of glitches in SNES games and sometimes freezing.

any idea why my SDRAM is passing all tests but not 90MHz ? I've ordered both times from someone that has lots of sells and 100% feedback. I'm going to try yet another ram module (this time from a different seller) when it arrives and see if it has the same problem.

IS 90 MHz achieved by applying specific voltage to ram module? Maybe I can check power pins for some strange fluctuations ?

Maybe it has something to do with my De 10 nano board itself?

Anyone else have a similar problem?

[Sorgelig]

Usually below 100MHz even badly soldered SDRAM works fine.
Looks like some anomaly..

[merlin4782]

Usually below 100MHz even badly soldered SDRAM works fine.
Looks like some anomaly..

Usually below 100MHz even badly soldered SDRAM works fine.
Looks like some anomaly..

Hi Sorgelig,

I am the seller that Ashenshards purchased from and I fully refunded him when I was able to replicate the issue at 90MHz. After further testing, I confirmed that 4 different SDRAM boards are able to pass for at least 1 hour at 167MHz. In fact, I tested all 4 boards at 80, 90, 100, and 167MHz, they all passed without errors at all speeds except for 90MHz. Ashenshards confirmed that his board also successfully passes the test at all speeds except for 90MHz on his DE-10 NANO. This was all tested with the latest MiSTer and core releases (full update through the scripts). All components are exactly those in the BOM on the MiSTer Wiki (Windbond chips) and purchased from Digi-Key (they are not cheaper or subpar components).

I have sold over 60 of these SDRAM boards (XS 1.1) over a 9 months period without a single issue and I always test the boards for 1 hour at a minimum 160MHz. While I know that anyone can make mistakes, I have a lot of experience assembling these. This is why it surprised me to find that 4 different boards were giving errors at 90MHz only, while they passed without errors at 160MHz or higher. The two boards that I sent Ashenshards were also specifically tested at 167MHz for over 1 hour wihout errors.

I am curious if someone else can perform this 90MHz test with a Windbond XS1.1 board for at least 30 minutes with a full MiSTer update to get an independent data point. Thank you for your time.

IMG_4829_small.jpeg

IMG_4831_small.jpeg

[sawf01]

Hi guys,

I'd like to add to this if I may. I've run the memtest on my latest batch of xs 1.1 memory boards with the alliance chip and I see the same behaviour. After 5 minutes it shows 766 in red. It seems to run fine for a while and then adds a load of errors and then run fine again in a cycle.

I too have made large amounts of these boards that pass 167 every time for several hours. I have not however noticed any instabilities in cores though.

Hope this helps.

[merlin4782]

Hi guys,

I'd like to add to this if I may. I've run the memtest on my latest batch of xs 1.1 memory boards with the alliance chip and I see the same behaviour. After 5 minutes it shows 766 in red. It seems to run fine for a while and then adds a load of errors and then run fine again in a cycle.

I too have made large amounts of these boards that pass 167 every time for several hours. I have not however noticed any instabilities in cores though.

Hope this helps.

sawf01

Thank you for your input. I have tested boards that sometimes pass all speeds but fail at 90 Mhz after only 5-10 minutes, then I tested the same board again later and it only shows errors after 100 minutes. This same board also runs NES and SNES cores without problems. However, Ashenshards is having issues in game also despite his board passing without errors at all speeds except 90 Mhz.

[ExCyber]

I haven't run it for an extended period at 90 MHz, but I definitely see bursts of errors specific to 90 MHz with an XS v1.1 board from CBMstuff.com (Jim Drew) with a Winbond chip.

Also, sometimes when I switch to 140 MHz, the cycle counter doesn't increment. I don't see that with any other speed. This may be an unrelated issue.

I've also seen issues with the SNES core, but these are new. I found that I usually get a lockup with graphics corruption or a black screen during the first level of Battletoads in Battlemaniacs on SNES release 20190823. I also got a lockup in Death & Return of Superman, but that took a bit longer so I've tested with it less. I also noticed a few unexpected stalls that only last for a fraction of a second. So far I haven't seen these problems in 20190824. I'm returning to MiSTer after a bit of a break, so I'm not sure which versions I skipped, but I don't think I've seen anything like this in older releases of the SNES core.

I'm willing to bisect the Battlemaniacs lockup if any developers think that's useful, though I probably won't be able to do it until Monday at the earliest.

[Sorgelig]

You always can report the bug on github. If you can identify the exact commit broke this game, then it will be much better.

[sawf01]

Hi guys,

I'd like to add to this if I may. I've run the memtest on my latest batch of xs 1.1 memory boards with the alliance chip and I see the same behaviour. After 5 minutes it shows 766 in red. It seems to run fine for a while and then adds a load of errors and then run fine again in a cycle.

I too have made large amounts of these boards that pass 167 every time for several hours. I have not however noticed any instabilities in cores though.

Hope this helps.

sawf01

Thank you for your input. I have tested boards that sometimes pass all speeds but fail at 90 Mhz after only 5-10 minutes, then I tested the same board again later and it only shows errors after 100 minutes. This same board also runs NES and SNES cores without problems. However, Ashenshards is having issues in game also despite his board passing without errors at all speeds except 90 Mhz.

I would have thought if this was causing instabilities that we would have heard about it from more than one user? Maybe its something else causing the issues and the memory test is just an anomaly as Sorgelig said?

[Poobah]

So, keep in mind the 90 MHz thing and the SNES thing could be unrelated. From what I am reading, the SNES thing is a relatively new development, the memtest code, and the memory board designs haven't changed in quite a while.

I'd bet the 90 MHz thing is an anomaly in the memtest code. It could be one of the the magic numbers on the timing table is sub-optimal, and went un-noticed since most people test the high end.

[Ashenshards]

Delete double post. Sorry.

[Ashenshards]

I think the 90 MHz is unrelated. I'm on my third ram board and still having issues with SNES core. Yoshi island has a lot of glitches. Unless no one is playing enough to test there must be something wrong with my de 10 nano. I had to play through the whole 1st world quit and come back later before I got these glitches.

https://youtu.be/GSGlUrRPSNo

If I reboot the Mister 10 times the game will load without glitches, but if I run the game glitchy it erases the save.

[JBeau]

I think the 90 MHz is unrelated. I'm on my third ram board and still having issues with SNES core. Yoshi island has a lot of glitches. Unless no one is playing enough to test there must be something wrong with my de 10 nano. I had to play through the whole 1st world quit and come back later before I got these glitches.

https://youtu.be/GSGlUrRPSNo

If I reboot the Mister 10 times the game will load without glitches, but if I run the game glitchy it erases the save.

I read this and tried it myself. I have errors on 90MHz too. Must be the memtest. I played through the first world of Yoshi's Island and did not notice any glitches like in your video. Maybe a little flicker when the screen fades in, but that might be normal. I don't know. Tried restarting the core again and loading my saved game. Still no glitches. I running the latest version, updated through the updater script. Hope this helps.