128MB SDRAM v2.9

[eightbitminiboss]

There is a better test also by Jotego

https://github.com/antoniovillena/MiSTe ... p?raw=true

If your SDRAM module passes all 8 rbf you won't have future problems with your module

Could I get an explanation of how to run these tests? Is it just running each individual core and run the ghouls rom and if it passes the boot up tests, it's considered a pass?

Yes. If passes all cores with ok on boot up tests the memory is ok

Cool, thanks for clarifying. Even went as far as doing cold reboots in between cores and everything showing up as OK. Looks like I'm in the clear.

[OriginalXOR]

Is anybody selling these new versions yet?

MisterFPGA.co.uk appears to be selling the new version.
https://misterfpga.co.uk/product/mister ... mb-module/

Not had any issues with my previous module to be honest, so will need a good reason to open up my Aluminium Mister case and replace with this one.

Should be noted that Nat also does some good testing prior to sending anything out (see above link), and everything I've recieved has been top quality in every respect: for workmanship to packaging for post to support services. Highly Recommended

[RetroCastle]

I have build some SDRAM v2.9.and it test very good，I think it's a great improvement over the v2.5，but I found a strange place on the v2.9 board，there is a top solder on the top layer，I don't know what it's used for and It's very close to the pin 29. I think it will be a Hidden danger,because the pin 29 is connect to the +5v of GPIO.

QQ截图20210609090421.jpg (177.64 KiB) Viewed 9949 times

QQ截图20210609090248.jpg (77.67 KiB) Viewed 9949 times

qq_pic_merged_1623200745717.jpg (83.8 KiB) Viewed 9949 times

[badvision]

Looks like 2.9 is starting to show up on AliExpress.

[aberu]

Please excuse my ignorance, but having only just bought a v2.5 128Mb SDRAM, why is a new revision necessary? In layman's terms, Is there something wrong with the 2.5?

I think they are prone to being faulty. Problems were discovered when Jotego released his CPS1.5 or CPS2 beta and around 20% of users had problems if I remember correctly.

It wasn't 20%.

I took it as well, I had no problems, ran his test for over 30 minutes, with a 128MB SDRAM from misterfpga.co.uk and a 128MB SDRAM from misteraddons.com btw.

kDL0hoMFyD.png (157.68 KiB) Viewed 9405 times

339 participants total in the poll.

For 95.8% the test passed successfully. Also, this is ignoring the sampling bias effect no doubt present here since people who did have issues are going to pay more attention to his tweets and polls, and want to respond.

For hardware that is assembled by loads of different people, with different varying quality of components, using any one of the 3 main options for SDRAM components (7TCN/7TIN/6TIN), the percentages are not bad at all. This isn't a professionally assembled and intensely QC'd product mfg'd by a single company or something.

[hiddenbyleaves]

ooops I meant to say 2%.

Sorry bout that. My own sdram from misterfpga.co.uk never had a problem and when I entered my own test results I remember seeing the results at around 2%.

I hope my typo didn't freak anyone out.

[aberu]

ooops I meant to say 2%.

Sorry bout that. My own sdram from misterfpga.co.uk never had a problem and when I entered my own test results I remember seeing the results at around 2%.

I hope my typo didn't freak anyone out.

It was 4.2%, but from "official" sellers it was a little over 2%, yeah.

I have seen people say 20% before awhile ago so I made sure to point it out here in the way I did because I didn't want people to panic either

[pacoarcade]

I talked to Jotego about version 2.4/2.5 and he said the RAM chips on the board are suffering due to working out of specification, but most probably the DE10-Nano is safe. Even when my SDRAM 2.4 works stable I'd like to build a pair of 2.9 modules but BOM is not available.

The LDO used is in a nonstandard package. The LM1117 does not come in SOT89. Looking for this I only find Chinese brands. It’s not the first time I see the Mister project using these kinds of low quality/fake components.

huh? HTC Korea is a Chinese Brand?

https://lcsc.com/product-detail/Dropout ... 26021.html

https://lcsc.com/product-detail/Dropout ... 26020.html

Are both of them valid?

About U3 (LVC1G04) I'm not sure what is the footprint and exact part (SOT-23-5) I should look for since it's not specified in the schematic. IF SOT-23-5 is correct I'm considering part SN74LVC1G04DBVT

[aberu]

I talked to Jotego about version 2.4/2.5 and he said the RAM chips on the board are suffering due to working out of specification, but most probably the DE10-Nano is safe. Even when my SDRAM 2.4 works stable I'd like to build a pair of 2.9 modules but BOM is not available.

The spec in terms of the datasheet are not being 1:1 matched for a few reasons.

1. Limited number of pins makes it so the channels are being multiplexed.
2. Some pins were specifically not used because the FPGA has limited IO.

As I said in the reply just above yours, his test showed around 2% of people had issues with their SDRAM from reputable builders. People overall had a 4.3% failure rate with Jotego's test rbf overall, so it looks like people who buy from non-reputable builders had a lot higher likelihood of having problems. Additionally, afterwards Jotego got some advice on how to refine and improve his memory controller implementation, and it seems like this is not a major issue anymore.

https://lcsc.com/product-detail/Dropout ... 26021.html

https://lcsc.com/product-detail/Dropout ... 26020.html

Are both of them valid? not sure if it's SOT-89-3 the footprint I should look for and it's not specified in the schematic.

https://en.wikipedia.org/wiki/Small-out ... or#SOT89-3

download/file.php?id=2630

It should be SOT-89-3 since it has 3 leads and 1 solder point.

[pacoarcade]

About U3 (LVC1G04), the footprint looks identical to the one in sdram 2.4/2.5, and according to the wiki it's SN74LVC1G04DBVR (SOT-23-5).

[Nat]

The SN74LVC1G04DBVR inverter is fine for use as U3 when building the v2.9 SDRAM

[pacoarcade]

Thank you for your help.

For U4 I think it's HTC Korea TAEJIN Tech LM1117F-3.3 instead of HTC Korea TAEJIN Tech LM1117F-5.0 or LM1117F-ADJ linked in a previous post. The schematic specifies 3.3V so I'd choose LM1117F-3.3. From RetroCastle photo it looks it's effectively LM1117F-3.3.

[netbeui]

There is a better test also by Jotego

https://github.com/antoniovillena/MiSTe ... p?raw=true

If your SDRAM module passes all 8 rbf you won't have future problems with your module

Could I get an explanation of how to run these tests? Is it just running each individual core and run the ghouls rom and if it passes the boot up tests, it's considered a pass?

Yes. If passes all cores with ok on boot up tests the memory is ok

Hi, I ran this test and 7 out of 8 rbfs passed but JTCPS1_-520PS failed. It keeps restarting the screen with the 00 and other characters. Does this mean my RAM module is bad and what exactly would this impact on my MiSTer?

[antoniovillena]

Could I get an explanation of how to run these tests? Is it just running each individual core and run the ghouls rom and if it passes the boot up tests, it's considered a pass?

Yes. If passes all cores with ok on boot up tests the memory is ok

Hi, I ran this test and 7 out of 8 rbfs passed but JTCPS1_-520PS failed. It keeps restarting the screen with the 00 and other characters. Does this mean my RAM module is bad and what exactly would this impact on my MiSTer?

It's not perfect but if you have not problems with the cores it's ok

[netbeui]

Thanks antoniovillena! I tried a few of the larger Neogeo cores and the Neogeo bad apple demo and they all worked fine. Is there anything else I can test to confirm that the RAM is working properly?

[aberu]

Thank you for your help.

For U4 I think it's HTC Korea TAEJIN Tech LM1117F-3.3 instead of HTC Korea TAEJIN Tech LM1117F-5.0 or LM1117F-ADJ linked in a previous post. The schematic specifies 3.3V so I'd choose LM1117F-3.3. From RetroCastle photo it looks it's effectively LM1117F-3.3.

I think you are correct. My linking was not intended as a BOM, it was intended to dispell this notion that it was "cheap chinese junk" which is both insulting to the Chinese and inaccurate as the parts you can buy on LCSC are not limited to things manufactured in China.

Thanks antoniovillena! I tried a few of the larger Neogeo cores and the Neogeo bad apple demo and they all worked fine. Is there anything else I can test to confirm that the RAM is working properly?

You already did. The synthetic test kinda pushes it to the extreme. NeoGeo and CPS2 are going to be the most intense on the SDRAM for real world tests.

[netbeui]

I tried Super Street Fighter and I see graphical glitches. Before I buy another SDRAM module can someone take a quick look and confirm it's working correctly? Just want to make sure this is truly a RAM issue and not a problem with that core...

[aberu]

I've never seen that, so yeah it looks like your SDRAM might be the culprit and could be badly assembled. @jotego could probably confirm.

[chartbundle]

I'm new here, so I apologize. I've noticed problems finding the connector, but is the 2.9 connector just the same as the Samtec SMH-120-02-x-D (Double Row) I can't find the PCB to see if it's the same footprint or not. But appears to be normally stocking at US Mouser at least.

[Xbytez]

I can't find the PCB to see if it's the same footprint or not.

PCB drawing is here:

https://github.com/MiSTer-devel/Hardwar ... ds_2.9.pdf

The connector part number you provide is specified as having a 2.54mm pitch so would be suitable.

[chartbundle]

PCB drawing is here:

https://github.com/MiSTer-devel/Hardwar ... ds_2.9.pdf

The connector part number you provide is specified as having a 2.54mm pitch so would be suitable.

I wasn't worried about the 2.54mm, but the spacing between the first and second row pads. I'm still not 100% sure but it looks close.
For anyone interested the main page for the connector is here: https://www.samtec.com/products/smh with the exact footprints.
The part numbers will be SMH-120-02-x-D with x being the plating. L=Flash Gold, G=Gold, T=Tin. The D is important since it designates a dual row. the 02 does not mean that, so if you find something-S then it will be wrong.
As I said, Mouser has stock. Looks like Digikey doesn't stock but has qty 1 for sale with a 11 week lead time. Avnet is also non-stocking but shows qty 1 available in 2 weeks for Tin or Gold Flash, to 6 weeks for Gold.
Prices range from about $6 to $13 for qty 1.

Of course, now that I've checked all that I've decided to be lazy and just order a finished module from one of the vendors.