Maker Pro
Maker Pro

SDRAM problem

From time to time I must repair an equipment that has its own BIOS, CPU and SDRAM
This SDRAM has 256MB together and consists of 4 SDRAM modules(IC).
Sometimes equipment does not work.Then there are several possibilities for that.
Faulty CPU or BIOS or SDRAM.
Is there a method how I can find out that SDRAM is faulty? If so which module of those 4 SDRAM modules might be faulty?
Each SDRAM module has 66-pin Plastic TSOP(II) and is soldered on PCB.
Thank you for help
 
Just a quick couple of questions:

1. Is the SDRAM Manufacturer always the same, and using the same SDRAM chips?
2. Do you have a spare unit you can use to test, or part of one maybe?

Matty-
 
@MattyMatt:There are more manufacturers such as ELPIDA or HYNIX. But they are compatible.E.g when Hynix it is
HY5DU121622DTP-D43
Yes, I have spare units for testing.
Do you have any idea?
SDRAM modules are soldered, so it is not possible to use an equipment for testing all modules at a time
 
Hi,

If the bios and the cpu are socketed maybe you can remove them and find a way to test the SDRAM's with the suggested tester.

Olivier
 
But how can I find out a way to test the SDRAM. That is my question.
Imagine a computer like device.You do not want to desolder SDRAM, you do not have schematic of the mainboard.You can find out( by measuring) wires( path) from CPU to BIOS and from CPU to SDRAM.
Or how at least find out that CPU sends any address signals to SDRAM, so that I will know at least path between CPU and SDRAM is OK.It would be the first step.
 
Jane,

To test the RAM you must :

-Select it
-address it
-write it
-read it back

Exhaustive test involve write/read on all location's so.

Olivier
 
OLIVE2222 , yes, I know, but it is theory. Practice is rather more difficult.
Imagine I have 4 pcs of 66-pin Plastic TSOP(II) SDRAM modules - soldered.
I would have to either desolder them and use a memory tester, or replace the original BIOS with a changed BIOS that will know how to do all the actions, you say, when testing memory.
 
Jane, sure in practice you must adapt an existing tester via the CPU socket (if any) or an extension bus (if any), find maybe a way to pick-up the /CS signal... or indeed made/modify a bios or at least doing a test program to test the 4 pieces of SDRAM modules.
Do you have the board schematic, and a picture maybe?
Also you talk about time to time repairing, so it may happen that it' s take more time and money to build up a setup comparing to just replace all modules without even know if they are good or not. What you can also maybe do is to :
-unsolder the 4 modules
-sold 4 working ones
-test unsoldered modules apart
- keep the working modules for the next fixing

Olivier

EDIT,OK no schematic available
 
Last edited:
Olivier,thank you for the reply.
Yes, maybe it can be easier to desolder/solder them .But what if I solder them back and still the device does not work . Maybe I did not solder them properly or ( which is more dangerous)a path( below CPU) is disconnected.Because it is below CPU I can not check resistance value but I would need to check a signal from CPU to SDRAM's address bus.
Do you have any idea how to measure/test that there is a good path between CPU and SDRAM?
Do you think oscilloscope would be OK or I would need a logic analyzator?
Thanks
 
Jane, Maybe better to do something like first cheeking the CPU sanity by locking at key signals like R/W, clock and few data bits. If it 's look OK you can check all signals before and after buffer (including control and direction pins ), gates .. then you can look at key sdram signals /CS, /WE being specific for each IC. All raw/columns and bank signals are probably common for the 4 IC.
All this will be easier with a logic analyzer but a scope should be enough.

The problem can be that the Bios doesn't loop and so don't let you time to do measurement.
In this case having a large logic analyzer will be better, as you can probe (and log) many signals a time during the boot.

Olivier
 
Olivier ,
thanks for the reply.YES, you are correct.Bios doesn't seem to loop.And that is also probably my problem.
I can measure clock generated by CPU .Always .It is always available, eventhough device does NOT start.( also for the reason that SDRAM is faulty or path from CPU is damaged). But I can not measure any other signals with SDRAM, if the device does NOT start.The reason will be probably that BIOS doesn't loop.
But what solution would suggest then?
Thank you for help
 
Jane,


-What's the used processor?
-It's the BIOS socketed?
-It's the CPU socketed?
-Is the board have an extension bus?
-What's making this board so specific?
-is the board manufacturer still is business?
-Can you please post board pictures?
-Do you fell that, with a modified BIOS, the board can run with 1 RAM chip only
-What's the cost of your labor+part comparing to let a dedicated external company rework all the board (by batch maybe)?
-Can you expect a break even for a logic analyzer buying or renting?
-is the board part of a large expensive system or is almost the system by itself?

Olivier
 
Olivier, thanks for the reply.
And here are answers

-It's the BIOS socketed?
No, but it is not a problem to desolder BIOS chip and solder a new one

-It's the CPU socketed?
No, but it is not a big problem to desolder it either

What's the used processor?
Broadcom BCM7401

-Is the board have an extension bus?
I do not think so, but not sure. How can I identify such extension bus?

-What's making this board so specific?
We repair these products

-is the board manufacturer still is business?
Yes and there are also more companies that make copies

-Can you please post board pictures?
Here it is
350865x.jpg


2 more SDRAM modules are from the oposite site of the mainboard.CPU is under the fan

-Do you fell that, with a modified BIOS, the board can run with 1 RAM chip only
I do not thinks

-What's the cost of your labor+part comparing to let a dedicated external company rework all the board (by batch maybe)?
It would better for us to repair themselves

-Can you expect a break even for a logic analyzer buying or renting?
Hard to say

-is the board part of a large expensive system or is almost the system by itself?
No

Thanks for your help
 
jane,

Ok they are dreambox priced at around 200$ and using a very dedicated processor.
Bought a development chain and acquiring skills to develop a test program can probably not be done profitably for such components. I guess myself how you can desold and resold such processor without big problems? I also find out that the board under side have few regulation circuits, to be checked first. With such street price (or even more expensive) you can't take hours to repair them nor you can invest a lot to build nice debugging tools. So maybe better to
made a repair flowchart based on your experience and the components datasheets. Also clearly define tespoints on the PCB and give a go no go for repair. If the failure is well located (looks to be the ram very often) just replace them all (and maybe test them apart for reuse).

Olivier
 
Olivier,thanks for the reply.
If the failure is well located (looks to be the ram very often) just replace them all (and maybe test them apart for reuse).
But what I need to know first( before replacing ALL SDRAM modules)
is to find out that it is really SDRAM itself that causes problem (and then replace ALL SDRAM modules) or if it is a problem of paths(PCB) or even CPU problem?
Do you have any idea?
Thanks
 
Jane,

==>
made a repair flowchart based on your experience and the components datasheets. Also clearly define tespoints on the PCB

That's the only way i see.
Based on components datasheet you can made a partial reverse engineering and drawing a kind of schematic/building bloc mix with all sensitive signals and supplies to be tested. This again a working board.

they are no magical solutions.

Olivier


EDIT: Also to board may also be wrongly re flashed (I guess that it's should be possible via a bootloader)
so you may try to repair board having a nonworking program
 
Last edited:

Similar threads

Top