MiSTer FPGA Forum

Now I have build some 128m sdram v2.9，and I have test about 10 of them，the results were excellent！
Each was tested for at least 40 minutes，and all of them can run above 141mhz，5 of them keep 145mhz，one can keep 147mhz for 30minutes，I think it's a big improvement over the previous version，here is the pictures [/img]

Aren't you limiting your testing capabilities as you are using the AS4C32M16SB-7TIN instead of the AS4C32M16SB-6TIN?

aberu wrote: ↑Mon Jun 07, 2021 2:35 am Aren't you limiting your testing capabilities as you are using the AS4C32M16SB-7TIN instead of the AS4C32M16SB-6TIN?

I don't know if the -6tin can run faster, because I can't buy it，but the sdram v2.5 with -6tin of my friends can't run more than 150 MHZ in their hands，if have the sdram with -6tin，could you tell me your results，thank you！

I bought 6-tin a long time ago to make my own SDRAM and got impatient with my learning progress so I bought one from misteraddons.com again. I ordered the 2.9 pcb so I might try it out myself, but the soldering is near the limits of my experience.

aberu wrote: ↑Mon Jun 07, 2021 2:07 pm I bought 6-tin a long time ago to make my own SDRAM and got impatient with my learning progress so I bought one from misteraddons.com again. I ordered the 2.9 pcb so I might try it out myself, but the soldering is near the limits of my experience.

yes,It is not easy to solder a sdram , I'm looking forward to your testing with -6tin!

Hi there, I've built one v2.9 SDRAM board using the 6TIN variety of SDRAM ICs. My board is stable at 141MHz, however I have only run the memtest core for around 30 minutes and without any active cooling. It seems that signal quality is the limiting factor.

Hi,

Did any of you use hi quality capacitors? I've seen on some websites like they advertise using '
Taiyo Yuden Capacitors'. I guess that's important in terms of long time temp stability.

I was looking for all components and they are scattered over mouser/DigiKey, LSCS and ali.
Motherboard made on Jlcpcb?

Would be great if someone would share BOM to compare.

I just made sure not to use no-name capacitors from shady sources. IIRC the caps I used are by KEMET and Vishay. I agree that getting all parts is a messy task. Also it is not entirely clear which capacitance values to use. I followed the suggestions made by MickGyver given in Post 18 of this thread: viewtopic.php?f=8&t=2480&&p=. This means as opposed to the values in the schematic (1uF for all caps except the regulator input supply bypass cap, which is 10uF).

I had the PCBs made by JLCPCB, choosing "JLC7628" as the stackup for controlled impedance and HASL as the finish.
SDRAM, single-gate inverter and all capacitors were sourced from Mouser, but DigiKey or others would also suffice. For the caps I made sure to use X7R MLCCs with a sufficient voltage rating. I think the 10uFs were rated for 25V and the smaller caps for 50V, which is much more than needed.
I had to buy the 40-pin connector and LM1117-3.3 from AliExpress. The connector is hard to find, and the regulator is in a non-standard package and was sold out at LCSC. Luckily, everything worked out in the end.

Remember that you need to use hot air if you do not own an especially thin soldering iron, otherwise you will not be able to solder the 40-pin connector.

thank you, and layer sequence is:

TOP: GTL
Inner1: G1
Inner2: G2
Bot: GBL

Yes, this is correct. This is the standard Altium gerber layer sequence (Sorgelig uses Altium for PCB design).

A second remark regarding capacitors: Using only 1uF caps as marked in the schematic should be fine, as the regulator will do the heavy lifting regarding power supply filtering. I originally planned on putting together a v2.5 SDRAM board when v2.9 released, and I already had the v2.5 caps ordered. The SDRAM module will very likely work either way.

deadsnare wrote: ↑Wed Jun 23, 2021 12:03 pm Hi there, I've built one v2.9 SDRAM board using the 6TIN variety of SDRAM ICs. My board is stable at 141MHz, however I have only run the memtest core for around 30 minutes and without any active cooling. It seems that signal quality is the limiting factor.

I think 141Mhz is a little low ，I have build 200 V2.9 SDRAM and all tested for at least 60min，there are 20%are 150mhz，70%are 145-149mhz，only 10% below 145mhz

RetroCastle wrote: ↑Sun Jul 11, 2021 8:07 am I think 141Mhz is a little low

can you explain why a 141 Mhz pass is ‘low’? a little low for what?

my take is that none of the current cores require faster than the maximum clock speed shown on a mister wiki page of 126 Mhz? if a memory module fails to perform tests at high frequency, how is that a sign that the memory will down clock and work with most other cores that require much much lower clock speeds to run them? i thought this was one of the reasons jotego released the two memory test cores that test at 48 Mhz and 96 Mhz

just here to learn happy to be educated.