Tektronix 2465B oscilloscope teardown

[e-pirate]

Hi, folks.
Some time ago I started my work on replacing NVSRAM (the DS1225) in my Tek2467B and ended with a success replacement with SIMTEK STK12C68. The STK12C68 is a fast SRAM with EEPROM and a schematics to store/restore data to/from EEPROM on power failure condition. It combines both infinity read/write cycles of SRAM and 100-year data retention in EEPROM without need of any external power (and that solves battery problem). As data stored to EEPROM only at power-off condition, this specs simply make STK12C68 lifetime grater then scope's.
The original idea belongs to Mr. Hugo Holden, who consulted me during my work. His fears goes to store operation, hi told me that hi was able to reproduce condition at witch scope failed to boot and data in  STK12C68 corrupted after very fast unit switch on-off cycles. However, with my STK12C68 I was not able to reproduce that behavior in unit. I did a series of unit power no-off-on-off as fast, as I was able to kick power button and then varying time delays. But that didn't convince no me, neither Mr. Holden. So, he suggested another test that I performed: I powered STK12C68 from a function generator, varying both duty cycle (10%-90% for square wave) and frequency from 0 to a couple hundred Hertz (that will never happen in real life) and even under that test contents of STK12C68 remained intact. So, this NVSRAM confirmed it usability. It work perfect in my scope: everything stored correctly, front panel and all settings stored/restored during reboot perfectly. Even after power button kick test.
The IC requires an external cap to store energy for power off EEPROM store cycle. I use the smallest recommended value: 68 uF tantalum, bypassed by 0.1 ceramic one. Besides, I use 0.1 ceramic to bypass the IC itself and 10K resister between pin 27 and VCC as recommended in datasheet. Previously I replaced DS1225 with a round pin socket. My current "adapter" is just a wire soldered STK12C68 to two raws of pins inserted to socket and some caps and resister soldered to all this mess and covered by a tape so it will not short to cover. But that work grate for testing.
P.S. The programmer I use for all my memory operations is Vellon VP-380. And I did a quick test reading an aged DS1225 (mine is '91 production code) without external parallel battery - and I can confirm that this universal programmer can read this old NVSRAM secure without corrupting contents.

[David Hess]

I had good results replacing the two DS1230 embedded lithium battery NVSRAMs in my 2440 with EEPROM based STK16C88s.  The STK16C88 has the advantage over the STK12C68 of not requiring an external capacitor so they were drop in replacements.

[e-pirate]

The STK12C88 is, probably, superior to STK12C68 in terms of external elements and cause it 4 times STK12C68 size, so using just a some jumpers we can get 4 storage banks. Yah, that grate! But unfortunately, I have no available this ICs on easy

[iDevice]

I have no available this ICs on easy

Really ?
http://www.ebay.com/sch/i.html?_mPrRngCbx=1&_nkw=stk16c88

[David Hess]

It looks like Cypress is discontinuing their EEPROM based NVSRAMs.  They still list them but their distributors do not have them available anymore and Cypress does not sell them directly.

[miguelvp]

And if you don't trust eBay:

http://octopart.com/partsearch#!?q=STK16C88
http://www.findchips.com/search/STK16C88

Maybe is because you typed STK12C88 instead and that would have found nothing.

Edit: in any event, not a lot of stock left and some that have stock don't list the price. My bad, if you click on the part they show the price at the distributors site

[e-pirate]

There are STK16C88 available at my favorite semiconductor store, but they cost 1200 RUR a piece, where STK12C68 cost 220 RUR. Should I get more expensive part only to save a cheapy cap and a couple of resistors.. nop. And unfortunately, delivery prices to Russia are almost always unreasonable, at least from EU and US. But thats a different sad story

[David Hess]

It looks like Cypress is discontinuing their EEPROM based NVSRAMs.  They still list them but their distributors do not have them available anymore and Cypress does not sell them directly.

I was wrong about this.  Cypress is discontinuing EEPROM based NVSRAMs in DIP packages or at least the commercial ones in favor of surface mount packages including SOIC.  They use different part numbers for the surface mount packages because they all require an external capacitor so they are not direct replacements for JEDEC SRAMs but with a surface mount package, a DIP sized module could be produced which includes the capacitor.  Most are 3.3 volt or lower but some are 5.0 volt.

[Mr Simpleton]

Having replaced the leaking caps, cleaned the A5 board it was the big time to remove the Dallas RAM. Everything worked like charm, and I put in a quality socket for easy replacement.
Now having a new Dallas chip, I started of playing with my old ALL-07A programmer, and to my surprise it could not read/verify this chip. Odd... everytime I do access this new chip I do get a new checksum.
Lucklily the old RAM still works and sits in my "scope".

Reading about misfortunes with MiniPro programmers, what would be the way to go?? Batronix seems to have a low-cost programmer just slightly 2x the price of the MiniPro...

I am tempted to put some glue logic around an Arduino and have a one-off construction for this project. Would be fun, but not sure I have the time. Clock is ticking in my old Dallas chip 

[Precipice]

Now having a new Dallas chip, I started of playing with my old ALL-07A programmer, and to my surprise it could not read/verify this chip. Odd... everytime I do access this new chip I do get a new checksum.

There's not an RTC in there, is there? (Sorry, checking back in this thread, I couldn't see which Dallas chip you were wrangling).
Even if not, it might be worth saving a few dumps, and seeing if there's anything interesting going on.

[David Hess]

Now having a new Dallas chip, I started of playing with my old ALL-07A programmer, and to my surprise it could not read/verify this chip. Odd... everytime I do access this new chip I do get a new checksum.

Lucklily the old RAM still works and sits in my "scope".

I had the same problem reading the Dallas NVRAMs in my old programmer which should have worked and did with the replacement memories.  I never tracked down the cause.

[Mr Simpleton]

@David
What did you end up doing then?? Give up and calibrate or managed to transfer the data to a new chip? If so what type of programmer?

[David Hess]

@David
What did you end up doing then?? Give up and calibrate or managed to transfer the data to a new chip? If so what type of programmer?

It was with a 2440 oscilloscope which has a much less onerous external calibration procedure than the 2465B so I just recalibrated it.  If it had been a 2465B, then I would have bought another inexpensive programmer that was specified to work with the NVRAMs, built something, or contacted someone with the right kind of programmer.

What was weird is that my programmer worked with the replacement parts which used a different technology but not the Dallas NVRAM parts.

[nctnico]

Probably due to solder on the pins. That doesn't go well with ZIF sockets. If you solder the original NVRAM into a socket with turned pins then you'll see the programmer reads it just fine.

[David Hess]

I cannot say that is impossible however the pins were cleaned of solder and flux and the same behavior occurred with both NVRAMs.  It looked like one of the lower address lines was not being latched properly and each complete read produced slightly different results.

The programmer was ISA based but running on a considerably newer system which had an ISA bus so I suspect there was a timing issue.  The newer NVRAMs were considerably faster, 45ns versus 120ns, and worked fine.

[Mr Simpleton]

The programmer I was using, listed the 1225 as supported, and still nothing. My take is there is a speed issue. Think the simplest thing would be for me to build a duplicator.
Use a address counter some glue logic and two 28 pin sockets. Make sure that the WR-pin is strapped to high to secure no write on source. Should be simple. Only thing is I will not have the data saved on file. Never mind...

[David Hess]

The programmer I was using, listed the 1225 as supported, and still nothing. My take is there is a speed issue. Think the simplest thing would be for me to build a duplicator.
Use a address counter some glue logic and two 28 pin sockets. Make sure that the WR-pin is strapped to high to secure no write on source. Should be simple. Only thing is I will not have the data saved on file. Never mind...

Enough people have reported problems reading old Dallas NVRAMs that there may be something else going on.  One hypothesis I have read is that the heat of soldering interacts with the module in some way to cause problems.  Maybe this only occurs when the lithium cell voltage is low.

To me this suggests that cooling or heating the NVRAM before using it in the programmer might allow recovery of the data.

I also know that there is an issue where if the internal battery backup voltage is low enough, the internal logic which protects the memory from invalid writes fails.  The data is retained in this case but the memory becomes inaccessible even with external power applied.  I do not know how this would cause problems with a programmer trying to read the contents though when the memory still works in the original circuit.

[Mr Simpleton]

What puzzles me is that my programmer (supporting 1225) failed on a brand new chip... just wanted to verify programmer operation before plug in my precious old chip!

The only thing I can come up with, except for a faulty chip, is that there are timing issues, programmer acting too fast some signals.

[Howardlong]

I sold a 2465BCT on ebay about five years ago, as I also had a TDS2024B at the time and couldn't justify having both: I was doing paid-for work in software only at the time. I sold it for pretty much the same as what I'd paid for it eight or nine years earlier.

I've regretted getting rid of that 2465 ever since. So a couple of weeks ago I picked up a 2467B with the CT option, for about thes same price as I'd paid and sold my 2465 for. I so love these scopes, it's been like catching up with a long lost friend.

Unfortunately I've been so spoilt by DSOs in the intervening period, I had to refer to w2aew on tips for using the delayed/dual timebase, I'd almost completely forgotten how to configure this on a 2465/7. Ten minutes of w2aew video beat trying to make sense of the written manuals. It's my go-to scope now, in pole position right next to an Agilent 54831D on the bench.

[tymm]

Hi,

Wondering if anyone has attempted to hand-copy cal constants using the "Calibration RAM Examine" thingie in the exercisers (or has at least verified that they are stored as shown in the RAM)...

Here's the blurb (from page 6-17 of the service manual):

Code: [Select]

CALIBRATION RAM EXAMINE (Exerciser 02). This routine allows the operator to examine the contents of 256 decimal locations, 00 (Hex) through FF (Hex), in RAM. When entered, the Exerciser displays the contents of RAM location 00 (Hex) on the top line of the CRT display. One hundred and seventy calibration constants reside between addresses 01 (Hex) and AA (Hex). Calibration constants residing between 01 (Hex) and 6E (Hex) should have odd parity as explained below. The remaining locations may be of either parity. The readout display line has the following format:
AA DDDD P
The format is defined as follows:
"DDDD" is the 14-bit word stored at that location (13 bits of data and one parity bit).
"P" is a parity indicator for the data word: X indicates even parity; blank is odd parity.
Pushing the upper or lower TRIGGER MODE switch will increment or decrement the RAM address by 16 (10 Hex) respectively. Similarly, pushing the upper or lower TRIGGER SOURCE switch will increment or decrement the address by 1 respectively.
-- looks like it's time to replace the NVRAM on mine soon and hoping that a hand copy of the cal data might give a backup in case something happens during the procedure (I'm a little leery due to past experiences with data loss when moving Dallas NVRAMs from board to board; in that case I think one of the boards had residual voltage on the + rail & it was probably more user fail than IC fail... but still makes me hesitate).

best case would seem to be that those 340 bytes are just stored on the NVRAM starting at address 1 - and that if one takes a blank NVRAM and writes only those bytes, that nothing else will be required to make the scope happy (e.g. no other checksums or other data required for the scope to see that it's calibrated and work properly).

[auato]

Hi all.... instead I have this fault and unfortunatley I am not able to read the CRT anymore (the only evidence available is the internal self test @power-up which probably jams on test 05)

http://youtu.be/WNzYjhgLwis

[mikeselectricstuff]

What puzzles me is that my programmer (supporting 1225) failed on a brand new chip... just wanted to verify programmer operation before plug in my precious old chip!

The only thing I can come up with, except for a faulty chip, is that there are timing issues, programmer acting too fast some signals.

I have a vague recollection that some NVRs have some sort of "factory seal" to maximise shelf life, and are activated by a special command to wake them up - could this be it?

[c4757p]

The 1225 has this (they call it a "freshness seal" in the current datasheet, which makes it sound rather like food...), but it's activated simply by the first application of power.

[peter-h]

I have just read most of this thread, and I apologise if I have misunderstood something, but I have a 2465B (400MHz) on which all the illumination (except the CRT graticule variable illum) has just "gone funny".

Initially it all went blank when I turned on the scope today.

So e.g. I can't tell if ch1 is on

AC
GND
DC
GND
50 ohm DC

Then after about 1hr all the lights lit up, but ALL of them so I couldn't tell which of the options were selected

When I moved one of the buttons, it all sorted itself out and now seems to work.

I have had this scope for about 7 years and it has been perfect.

What is the likely cause? It looks like it might be temperature dependent.

[sparkybg]

Hi,

Wondering if anyone has attempted to hand-copy cal constants using the "Calibration RAM Examine" thingie in the exercisers (or has at least verified that they are stored as shown in the RAM)...

-- looks like it's time to replace the NVRAM on mine soon and hoping that a hand copy of the cal data might give a backup in case something happens during the procedure (I'm a little leery due to past experiences with data loss when moving Dallas NVRAMs from board to board; in that case I think one of the boards had residual voltage on the + rail & it was probably more user fail than IC fail... but still makes me hesitate).

best case would seem to be that those 340 bytes are just stored on the NVRAM starting at address 1 - and that if one takes a blank NVRAM and writes only those bytes, that nothing else will be required to make the scope happy (e.g. no other checksums or other data required for the scope to see that it's calibrated and work properly).

I am planning a full recap of my 2445B next week. The parts are on their way from Farnell, and I also ordered FM18W08 to put on the A5 board. The dallas NVRAM still works on my unit, but I suspect it is a matter of time for it to fail. The date code of most of the chips inside is 1991, so I suspect my unit manufacturing date is not far from this. The serial is B062263. I bought it from one the "famous" Israeli ebay shops with readout problem. The problem was the well known cap leak on A5 which I repaired some time ago. Some 74HC chips were gone. Now I will do full restorarion and replace the DAC chip also, because I suspect it has a leakage path under it on the PCB.

Anyway, I've got EasyPro 90B programmer which has both Dallas 1225 and 1608/1808 chips in it's device list and I hope everything will be OK with programming. I will back-up the calibration data with the exercise you mentioned and the will try to find out where is it in the NVRAM.