• Please review our updated Terms and Rules here

PSA: Those "New" Sun NVRAMs You See on eBay

glitch

glitch

Veteran Member
Joined
Jan 31, 2010
Messages
5,365
Location
Central VA
Picture says most of it:



I had a customer send in some M48T08 and M48T59 NVRAMs from some of their Sun systems recently. They'd purchased "new" NVRAMs from someone claiming to support Sun systems on eBay who'd sold a pile of them for $70-some USD. My customer wanted them rebuilt with one of our maintainable repair boards so they never had to deal with it again. Pictured above is what I found.It appears the seller digs out the old battery from the top with a Dremel or similar rotary tool, then pulls the tabs off, removing just the battery and leaving the tabs attached to the old NVRAM. They then put in a different battery and spot weld the old tabs to it. After this, the hole from the battery excavation is packed with some sort of filler putty, sanded flat, and blacktopped. They put a printed label over the top with the part number, rebuild date, etc.Note the deep digs in the top of the NVRAM IC. The epoxy on all four modules was also gummy, suggesting the NVRAMs were heated to make digging easier.

This is a closeup of the replacement battery:



Notice I'm not saying *new* battery...there are clearly two old spot weld marks, visible around the + mark on the above battery. So, either the seller is reusing old batteries (might explain why this one is already dead), or they're ordering batteries with tabs and ripping the tabs off. Either way, not how it ought to be done.

Hot garbage in my opinion, especially for the money. Save and rebuild your originals!
 
I simply got a new Dallas chip and went through the init steps you can do in the OBP. Easy peasy and wasn't hard to find them. No ebay fake nvrams needed and no battery replacement needed. :)
 
Some of the new ones don't work on old Suns so you're stuck with having to rebuild. I tried I think it was new M48T02 and M48T08s from Mouser when I refurbished 50x probably 5-7 years ago and got clock faults from power-on diagnostics, which prevented autobooting. The machines "worked fine" otherwise, but that's a pretty big limitation! I ended up designing repair boards to make reusing the old ones easier, since I had such a large number to go through.

I think the new M48T59s do work, I think that's what I originally put in my Ultra 10. I have a rebuild for that one now too, so I just rebuild the dead modules. Plus I like doing it :P
 
Takes 10-15 minutes to dremel it out, cut the lead to the internal battery, and soldier on a battery holder with wire leads. Double sided tape for the battery holder and done.
 
For "production" Ultra 5, Ultra 10 and E3500 boxes I've been using new devices.
For testing/lab stuff it's been Dremel+CR2032.

If the "new" devices start dying early or are DOA, then even the production kit will get Dremel+battery holder.

(Yes one of my end customers is still using this stuff because that's what was delivered as part of the solution back in the day, and you can't get modern kit that will run the versions of Solaris required).
 
I recently picked up an Ultra 1 which of course has a dead NVRAM battery. I whipped up a little adapter board for the SMD version of the M48T59Y since the "snaphat" battery packs seem to be available still... Hopefully this will be more reliable than the eBay "rebuilt" chips.
 
This is my solution:



I had to refurb like 50x SPARCstations and was having issues with new-made NVRAMs throwing clock errors. I didn't want to do battery-on-a-wire, and couldn't tape or glue it to the top of the old encapsulation like I usually did because of height issues under the SBus slots. The repair board is super fast to apply and is the same height or a little shorter, and fits under a SBus card just fine:



Using the repair board allows me to make two small cuts with a hacksaw on the short ends, and pop the entire encapsulation off with a screwdriver! Takes less than 5 minutes per NVRAM, which is important when you have 50x to do at once :P Here's a picture of the first lot of 50x 48T02s in process:



Same 50x with all the encapsulations off and ready for repair boards:

 
I really like this! I'm running one in my SS10. A one-time chore then all I need to worry about is sourcing CR1225 batteries.

How long does the CR1225 battery last, by the way? Have you run these long enough to find out?

IMG_8794.jpg
 
ankylo5aurus said:
How long does the CR1225 battery last, by the way? Have you run these long enough to find out?
Click to expand...
We characterized "up and running RTC" standby current consumption using our picoammeter and determined that 3-5 years standby runtime will be average with these CR1225 cells. The main thing is electrolyte loss from drying, essentially the batteries hitting shelf life limits. If the machine is in a cool, dry environment that does not see temperature swings, and the machine is powered on once, set up so that the RTC is running, and then never powered on again, you'll see around 5 years life. If it's in an unconditioned space like an attic, you'll see shortened battery life. Part of the attraction for engineers designing in the original potted modules was that the potting pretty well doubled standby battery life by greatly reducing battery electrolyte loss.

If the system gets run kinda regularly and lives in a conditioned space, battery life will be much longer. That first one I rebuilt in 2017 for my personal SPARCstation IPX has never had a battery change, but I run it a fair bit. We have customers who run these in cleanroom environments with close to 24/7 uptime (shutdowns for machine maintenance), we change the batteries on a maintenance schedule to ensure uptime and in those situations there's basically no aging on the battery capacity.
 
glitch said:
Using the repair board allows me to make two small cuts with a hacksaw on the short ends, and pop the entire encapsulation off with a screwdriver!
Click to expand...
Hmm I didn't realize it was that easy, it seems like a very sturdy case on those. I envisioned a much more involved process to mount that.

On my SS10 I managed to get a "new" (who knows) chip to work ok. Normally I find it easy to find a location to to tape down a battery holder (like IBM PS/2) but I can see how on a SPARCStation is pretty tight.
 
I think it's the 48T02 where the Dremel solution is mandatory. (Sun: IPC, SS1, SS2 + others). There's also a number of Dallas devices which have a hard coded serial number (seen on a number of HP-UX boxes, not sure if the serial number=hostid).

Otherwise for RTC/RTC+NVRAM devices the choice of Dremel or new is down to device availability/price.

Pure NVRAM is a different kettle of fish. If you have a 28pin socket with a vanilla 8k x8 battery backed SRAM with a JEDEC pinout, anything that's fast enough and fits will do the job.
 
mdog69 said:
I think it's the 48T02 where the Dremel solution is mandatory. (Sun: IPC, SS1, SS2 + others). There's also a number of Dallas devices which have a hard coded serial number (seen on a number of HP-UX boxes, not sure if the serial number=hostid).

Otherwise for RTC/RTC+NVRAM devices the choice of Dremel or new is down to device availability/price.
Click to expand...

Yeah, the new-made M48T02 and M48T08 parts available through legitimate parts suppliers (e.g. Mouser, Digi-Key, etc.) have issues with the clock section on old Suns. Before creating the repair boards, I was going to just order new, but I found that the new NVRAMs I was getting from Mouser gave a clock error during power-on diagnostics and kept the machines from autobooting.

There are some modules have have been EOL so long that NOS ones with good batteries are not obtainable, like the DS1742W. Those aren't very repairable because internally they're a circuit board, and sometimes the circuit board is on the top of the module with the battery on the bottom! For those, we've been able to design total replacements using a small PCB and Batten & Allen leadframe pins (like a BASIC Stamp).

mdog69 said:
Pure NVRAM is a different kettle of fish. If you have a 28pin socket with a vanilla 8k x8 battery backed SRAM with a JEDEC pinout, anything that's fast enough and fits will do the job.
Click to expand...

With a couple of exceptions, of course! Some of the NVRAM controllers signal low battery conditions by interrupting or modifying NVRAM reads and/or writes, e.g. NVRAMs based on the old DS1210 controller fail the first write to indicate low battery. Some also use spare lines (e.g. pin 1 on 8K devices) for extra functions, like brownout resets.
 
So my customer confirmed that these were $69.99 USD, that the seller claims they are a professional replacement suitable for production use, and that they died within 2-3 months of purchase.
 
Thanks for pointing out the NVRAM gotcha @glitch . The only "interference" between host and RAM I've seen from a "turn SRAM into NVRAM" device was to block writes as the main +5v supply starts to fail.
 
No problem! We've got nearly 30 different NVRAM rebuilds and replacements now...you'd think there would be less variety for such a conceptually simple thing :P
 
@glitch, sorry to revive this thread but I would like to try one of your GW-1742W-1 RTC modules to replace the Dallas DS1742W on my Tektronix TDS 3014 scope. You mention on your Tindie page for that module that:

"The GW-1742W-1 is 5V tolerant, like the original DS1742W, and can be read and written using 5V programmers such as the common TL866+. Programmers that do not support the RTC section of the module may throw programming or checksum errors."

I have a TL866II Plus programmer but I don't see the DS1742W in the Device List. Is there some other device I need to select to read the RAM in the old RTC so I can write the values to your replacement RTC module? I have seen posts on eevblog indicating they had issues trying to read the part with their TL866 programmers.
 
Last edited:
I've just bought a "New" Dallas DP12887A module from a UK ebay seller for a GA-586VX Socket 7 board. The Dallas chip is socketed in this case so that's lucky. I didn't buy a cheaper one from China, I thought a UK supplier might be better. What should I expect from the new one, would it be old stock and maybe exhausted already?
 
richarjo said:
I've just bought a "New" Dallas DP12887A module from a UK ebay seller for a GA-586VX Socket 7 board. The Dallas chip is socketed in this case so that's lucky. I didn't buy a cheaper one from China, I thought a UK supplier might be better. What should I expect from the new one, would it be old stock and maybe exhausted already?
Click to expand...
For STM chips like in Suns I have found the "new" on eBay to still be in working order. I wasn't sure if the clock didn't start until the chips were initialized, hence saving the battery, if they are remanufactured or newly made clones, or if there is some other story.

For Dallas DS12887 of eBay (not DP12887A) I think I found that they had dead batteries because I think I modified two of them to take CR2032 batteries and the order date for the battery holders is not long after I ordered the Dallas chips. Which leads me to another thing - at least on a PS/2 55SX it can fry the motherboard (planar) along with the clock chip if inserted backward. I learned that the hard way. I may not be remembering correctly and I may have found that the replacements didn't work at all and I modded the original chips with batteries. Wish I could be more help on that, I know I had to do the battery mod to get something working.
 
thisisamigaspeaking said:
For STM chips like in Suns I have found the "new" on eBay to still be in working order. I wasn't sure if the clock didn't start until the chips were initialized, hence saving the battery, if they are remanufactured or newly made clones, or if there is some other story.

For Dallas DS12887 of eBay (not DP12887A) I think I found that they had dead batteries because I think I modified two of them to take CR2032 batteries and the order date for the battery holders is not long after I ordered the Dallas chips. Which leads me to another thing - at least on a PS/2 55SX it can fry the motherboard (planar) along with the clock chip if inserted backward. I learned that the hard way. I may not be remembering correctly and I may have found that the replacements didn't work at all and I modded the original chips with batteries. Wish I could be more help on that, I know I had to do the battery mod to get something working.
Click to expand...
Ok Thanks for that. I have the original Dallas DS12887 module but thought not to chop it up. I'll put the new one on the meter when it arrives to see what kind of voltage it's giving.
 
richarjo said:
I've just bought a "New" Dallas DP12887A module from a UK ebay seller for a GA-586VX Socket 7 board. The Dallas chip is socketed in this case so that's lucky. I didn't buy a cheaper one from China, I thought a UK supplier might be better. What should I expect from the new one, would it be old stock and maybe exhausted already?
Click to expand...
I'm really confused why people are doing this with DS12887 or even DS1287 when new production DS12887+ is still available from suppliers like Mouser new and not old stock. Yes, they're almost $20 each, but they're new. Am I missing something?
 
You must log in or register to reply here.
Back
Top