Topics

2465b NVRAM replacement -2020?


Mark Hatch
 

What is the current state of the art for replacing the 2020b late model NVRAM?

0. Looks like the Dallas DS1225AD is a drop in but requires extracting the configuration data. (Yup, I get that my 2465B probably needs to be recalibrated anyway, but it is probably "good enough" for me as is ;-) )

1. It looks like amateurs like myself will destroy the DS1225 either thru the de-solder process or during the read process. So that sounds like a bad bet to count on taking the old one out, reading it, and then proramming a new one.

2. Saw that only a small portion of the NVRAM is stored, so apparently possible to read that out (seem to remember it was L02?) and wedge it into one of the existing configurations out there and burn a new NVRAM (seems like that would be error prone...). (I have option 05/TV, so that is not a positive here). (seems iike it should be easy to write a small program to do this...)

3. Also came across some 6800 assembly programs that might allow the configuration to be extracted. It looked like this required a swap of the existing firmware ROM's. Didn't understand how the data was saved and then transferred to the new NVRAM.

4. Also saw where somebody suggested reading out the config thru the GPID bus, but didn't see any reports of success.

5. Found it interesting where people took a dremel to the NVRAM and added an external battery, but that requires de-soldering with the risk of data lost. (see#1...)



I have no problem (yet). So I could let this ride a little. But while I have the A5 Board sitting in front of me...

My date code is 9616 (1996 16th week?) on a DS1225Y-200. So probably better than most, but still on borrowed time...

Suggestions?

Thanks

Mark


Vince Vielhaber
 

On 01/18/2021 10:09 PM, Mark Hatch wrote:
What is the current state of the art for replacing the 2020b late model NVRAM?

5. Found it interesting where people took a dremel to the NVRAM and added an external battery, but that requires de-soldering with the risk of data lost. (see#1...)
I did it on a 2440 but used a soldering iron to heat the potting and get it away. There are pics of where to dig, I think, in the files section. When I did mine, I added a wire to the ground pin on the outside and the positive wire to the point shown in the pics. No loss of data. I highly recommend the use of a desoldering station to remove it tho. Actually the 2440 has two NVRAMs I had to do.

Vince.
--
K8ZW http://www.metalworkingfun.com http://www.hamradio.fun


Sam Reaves
 

Wasn't there a FRAM memory chip at some point that could replace the Dallas
part? I seem to remember reading about it. I may have even bought some
chips to convert one of mine at some point but too many projects got in the
way.

Sam Reaves
ARS W3OHM
Owner and Moderator of:
LeCroy Owners Group on Groups.io (Current and Future Group)
Electronics and Mechanical Hardware Design Engineering Manager
Andritz Rolls Global Research Center (RETIRED)


Sam Reaves
 

Well it seems there are parallel FRAMs from Infineon/Cypress. Would this work? Mouser stocks them. You would have to use a SOIC to DIP adapter board. The datasheet says standard 8K X 8 pinout so it could work,

https://www.mouser.com/ProductDetail/Cypress-Semiconductor/FM16W08-SGTR/?qs=pA5MXup5wxE3mTcqcHRxDg%3D%3D

Sam
W3OHM


Ozan
 

I did FRAM replacement on my 2467B. I didn't bother reading the old NVRAM because I was going to calibrate anyway. There is an excellent documentation for different options (google finds it easily). One concern is scope writes to memory periodically and it would wear down the FRAM. With a battery backed RAM wear is not a concern. Author of the document calculated the wear of the FRAM and concluded it to last many years even in constant use. I don't power up my scope more than few hours a week; therefore, long standby life (no battery to worry about) is more important than wear for me. I put my FRAM on a socket so in future I can replace with another FRAM or another compatible part.

Ozan


Mark Hatch
 

Ozan,

I forgot about the FRAM option. But I was confused on its advantage vs the Dallas 1225AD. Seems like in both cases you need to capture the prior configuration data(unless like you it was of not value). So why not use the Dallas chip?

Thanks

Mark


Ozan
 

Hi Mark,
I wanted to eliminate the battery and not worry about it. In my use model FRAM would last pretty much forever (data retention is 151 years). Although Dallas chip battery seems to last longer than 10 years. Date code of my 1225Y was 9417 (1994 17th week) and it was still functional in 2019.

Nothing wrong with either choice, Dallas chip is a drop in replacement and easier, FRAM needs soldering an adapter board.

If you want to keep a record of current calibration data there is a test menu that displays the calibration coefficients. You could take pictures or take the video while going over the pages. Unfortunately how this data maps to RAM contents in not documented. There was some activity to reverse engineer I don't know if anyone figured out yet.

Ozan


Roy Thistle
 

On Mon, Jan 18, 2021 at 09:18 PM, Ozan wrote:


One concern is scope writes to memory periodically and it would wear down the
FRAM.
For the FM16W08-SGTR, I reckon it's 2x10^9 writes per day for 137 years, with data "retention" for at least that long, at 65 C ... at least according to Cypress.


Mark Litwack
 

... 4. Also saw where somebody suggested reading out the config thru the GPID bus, but didn't see any reports of success.

It has been done successfully. The cal data can be both read and written. Here is a post that describes the GPIB command to do it:

https://www.eevblog.com/forum/testgear/tektronix-2465b-oscilloscope-teardown/msg1167050/#msg1167050

Since that was written, I've seen success reports now on all 24x5/24x7 models (plain, A, and B). Unfortunately, I haven't seen that anyone has written an easy-to-use wrapper for the commands, so it's still a manual procedure.

A good backup before messing with any of this is to take a video of the screen while flipping through the cal constants in EXER 02.

Also, on the FRAM option, it's widely reported as successful, but there's been at least one instance of it NOT working. The issue was tracked down to the power-down sequence of the FRAM not being supported properly by the scope which was designed for the characteristics of the DS1225. The fact that the FRAM works almost all of the time is a matter of good luck, and not because it was a carefully evaluated drop-in replacement.

-mark


Mark Litwack
 

... If you want to keep a record of current calibration data there is a test menu that displays the calibration coefficients. You could take pictures or take the video while going over the pages. Unfortunately how this data maps to RAM contents in not documented. There was some activity to reverse engineer I don't know if anyone figured out yet.

The 256 cal data values are mapped to NVRAM locations 0x1e00 - 0x1fff, and all are 16-bit values stored in big-endian format. They are the same values as shown in EXER 02. Writing the EXER 02 data directly into the NVRAM works as a restoration method.

-mark


tekscopegroup@...
 

Suggest you take a look at this thread at the EEVBlog, which is pretty much the bible of 2465 series scope restoration and repair, including NVRAM replacement. Its a bit long (currently at 69 pages) but its good reading and it will teach you many things and how-to's about this scope. I read through it in a couple of days, and the new knowledge actually gave me enough confidence to get a scope off ebay knowing that I'd be able to deal with any potential problems and the obvious things that always need to be done to a 30+ year instrument like this.

https://www.eevblog.com/forum/testgear/tektronix-2465b-oscilloscope-teardown/

The FRAM chip worked perfectly for me, I have a 2465B and used a FM16W08. To pull the chip off the board in my case just required a 30W soldering iron with a good tip, and a good (clean) solder sucker (aka "Soldapult"). Be sure to "wet" each pad with a bit of fresh solder before attempting to clear the hole, and you can also add a bit of rosin flux. Try to push each chip pin towards the center of the hole while the solder is melted as that will facilitate clearing the hole on the first attempt. There is one pin, I think its 14, which is ground and that requires additional heat to be cleared as it connects to more copper foil.

Other people swear by fancier desoldering tools to get this job done, I'm sure it will but really don't think they are absolutely necessary in this case if you have the required experience and just some very basic desoldering tools that are in good shape. But don't try this unless you are already skilled with soldering and desoldering techniques. This is a multi-layer board and pulling one of the through-holes would surely be a big problem.


Mark Hatch
 

Mark,

The GPID method seems to be a neat method. It look like you can save the values, install a new NVRAM, then stash the old values in. (except that a grid to usb is a $150 item... PCI board + GPID cable are still ~$75 wish)

The one confusing piece is that the article talked about 100 (decimal) slots (they look like 16 bit #'s, many >1024 but all < than 16k. (I read you have to store twice since 0-99 and 100-199 are duplicated.

However, elsewhere, I see mention that the locations are 0x1e00 - 0x1fff (256 decimal) 16 bit numbers. So 256 <> 200... Are their extra bytes in the NVRAM at the end that are not used?

And thanks for the pointer to the eevblog. Think I am up to page 10 and lots to go...

regards,

Mark


Chuck Harris
 

I think you are going to have to elaborate on your assertion.

For instance, who hasn't carefully evaluated the FRAM for use
in the 2465B? Tektronix surely didn't, but then the FRAM
didn't exist when they were developing, or maintaining, the 2465b.

Hugo Holden did, and I independently verified his work.

He found that the FRAM is fully compatible with the way the
2465B uses the NVRAM.

I have used FRAMs on dozens of customer's 2465B's without a
single failure.

There is no power down sequence for the 2465B scope.

When the scope's power is turned off, the scope's cpu continues
to run until its little brains get scrambled from the voltage
being too low. Before it gets to that point, the NVRAM locks
out the CE/ line, to prevent the CPU from writing garbage over
the NVRAM's data.

Depending on the version, the NVRAM's lock activates when VCC
drops 5%, or 10%, below its 5.0V nominal voltage.

The FRAM does the same thing.

The only issue with using an FRAM to replace the NVRAM, or a regular
RAM, is the way it latches the address into the chip. The chip wants
to see the CE/ cycle low every time a new address is latched. It will
not work if CE/ remains low through multiple address changes, ignoring
all new addresses after the first that occur while CE/ is held low.

Logic analysis of the 2465b's A5 board's use of the NVRAM shows that
the 2465b always activates CE/ after it latches the address, which is
exactly what the FRAM wants to see.

Some EPROM programmers do not cycle CE/ for every read or write, and
this can result in bad programming. For my old GTEK 9000A, it mungs
the first and last address in the chip, but all the rest are faithful.

Since the 2465B rewrites all addresses in the NVRAM, with the exception
of the calibration constants, every time it is powered up, this is not
a problem.

Search out a paper written by Hugo Holden, called: "Tektronix 2465b
Oscilloscope calibration issues: Re-Powering the Dallas DS1225, or
using RAMTRON FRAM or AUTOSTORE RAM or MRAM as DS1225 Emulators."

Hugo does a nice job of explaining why a FRAM is suitable for the
2465b.


-Chuck Harris

Mark Litwack wrote:

Also, on the FRAM option, it's widely reported as successful, but there's been at least one instance of it NOT working. The issue was tracked down to the power-down sequence of the FRAM not being supported properly by the scope which was designed for the characteristics of the DS1225. The fact that the FRAM works almost all of the time is a matter of good luck, and not because it was a carefully evaluated drop-in replacement.

-mark






Mark Litwack
 

Hi Mark,

That post about the GPIB commands was referring to a 2465(plain) scope, which uses an ER1400 EAROM for its non-volatile storage. The EAROM is organized in 100 entries of 14 bits each, which are mapped in the GPIB command to locations 0 to 99. There is another ER1400 EAROM on the optional buffer board which appears at locations 100 to 199. I don't know for sure, but locations 200 to 255 are probably undefined for the 2445/2465.

On later 24x5/24x7 scopes, the EAROM was replaced by NVRAM solutions with either an internal or external batteriy. On those scopes, the cal data is stored in 0x1e00 - 0x1fff in the NVRAM and is mapped to a continuous block on the GPIB command from locations 0 to 255.

The entire block of data from 0x1e00 - 0x1fff is not all used for cal data storage, but since its detailed use is opaque for us mere mortals, it's best to treat all of it as a block that should be preserved.

Even though the NVRAM supports 16 bits, only the lower 14 bits are used for cal data as a vestige of the old EAROM. The rest of the NVRAM outside the block of 0x1e00 - 0x1fff is used as regular SRAM by the scope's OS and the contents do not have to be preserved.

Sorry for the confusion. I hope this makes more sense now.

-mark


Mark Litwack
 

Hi Chuck,

My assertion comes from a series of posts where a user was having issues with a FM16W08 FRAM, embedded in the gigantic thread over on eevblog:

https://www.eevblog.com/forum/testgear/tektronix-2465b-oscilloscope-teardown/msg2003207/#msg2003207

In short, the issue was that the DS1225 memory support circuitry does indeed lockout spurious operation down to the power-down level of the DS1225 as you state, but the FRAM remains operational for a good bit beyond that, down to 2.7V. The memory support circuitry does not guarantee any predictable operation below 4.5V, let alone down to 2.7V.

In addition, the FRAM datasheet is also very emphatic about /CE being held high when Vdd drops below 2.7V to prevent spurious internal operation. Again, not a feature of the DS1225 support circuitry since the designers did not have to be worried about it.

The user in the thread was experiencing spurious writes on power-down because of this, which happened to corrupt one of the cal locations.

I've read Hugo's paper a long time ago, and it spends a large number of pages on operational timing and FRAM internals, but does not consider the power-down control signal behavior *required* by the FRAM datasheet.

If you read further in the eevblog thread, the user was able to solve his issue with a pullup resistor, but speaking from a design perspective, the memory control circuitry would look different if it was a FRAM in there instead of a DS1225. In my opinion, as I said previously, it is working by a stroke of luck.

And, the 2465B does indeed read and write every address in the NVRAM on boot. To check that memory is ok, for each location it saves the contents, writes and verifies two patterns (0x55 and 0xAA), and then restores the original value. So there is an opportunity for corruption of cal data here too.

-mark


Mark Hatch
 

Mark,

Thanks for the clarification on the 0-199 vs 0-255. Makes great sense now. I like this approach as I can gather the data now, and not touch my current NVRAM until I have to... Who knows, perhaps I got the lucky "immortal Dallas"? :-)

Any suggestions on the GPIB interface? Looks like a USB to GPIB is >$125 and a PCI board + GPIB cable is <$75.

Regards,

Mark


Ozan
 

After reading Mark's (Litwack) post I reviewed the FRAM path again. Although there is a power-on-reset chip on the CPU board and it gates of CE on old (non-DS1225) version, I can't find the schematic of the DS1225 version to check if it is the case.

Spending more time on it I found another potential problem: CPU and its support circuitry are TTL level (at least for the old board), Cypress FRAM chip is CMOS level. These specs have margins so interface may work OK but it is not guaranteed.

I don't recommend FRAM path anymore.

Ozan


Mark Litwack
 

Hi Miark,

... Any suggestions on the GPIB interface? Looks like a USB to GPIB is >$125 and a PCI board + GPIB cable is <$75

I would go for a USB version since PCI machines are a slowly dying breed. If you want inexpensive, there are also a bunch of open-source DIY GPIB adapters, like the AR488.

I have an NI PCI-GPIB card that I actively use, and have purchased a Keysight 82357B USB adapter to start migrating away from PCI. I'm sure other list members can jump in here with more suggestions.

If you're in the USA and really into old hardware, I can send you either an HP 27209-60001 GPIB 8-bit ISA card, or an IOTech Micro488/EX Bus Controller (RS232<-->GPIB) for the cost of postage on a USPS flat-rate box. They're not doing anything here except collecting dust. Contact me off-list if interested. (I did say "old"!)

-mark


Mark Litwack
 

Hi Ozan,

... After reading Mark's (Litwack) post I reviewed the FRAM path again. Although there is a power-on-reset chip on the CPU board and it gates of CE on old (non-DS1225) version, I can't find the schematic of the DS1225 version to check if it is the case.
...

I did look at this on the schematics I had on hand a couple of years ago. The DS1225 (U2460) /CE (pin 20) is driven directly by PLDs: on the 2445B by PLDC20xx, and on the 2465B by PLS153A. /CE is not gated like on the design with the separate SRAM chip and battery, so the power-down behavior is at the mercy of whatever the PLDs feel like doing that day.

These are the only two models with DS1225 that I had schematics for, but I have no reason to believe others in the series with DS1225 chips are designed any different.

-mark


Steve Hendrix
 

At 2021-01-19 05:45 PM, you wrote:
... Any suggestions on the GPIB interface? Looks like a USB to GPIB is >$125 and a PCI board + GPIB cable is <$75
I'll suggest my KISS-488 GPIB to Ethernet interface. Not all that cheap, but it won't go obsolete when the USB drivers for those others no longer work with current hardware. You can view it either at www.hxengineering.com or on eBay or Amazon. I offer a discount to purchasers from this list; if you're interested, contact me directly offline at SteveHx@HxEngineering.com.

Steve Hendrix