Geting rid of the battery in the HP3457A! Part one. RAMTRON FM16W08. Disclaimer:
The following text is not a call to repeat the above actions, but merely describes the steps taken by the author himself. The author of the specified text does not give any guarantees of the operability or usefulness of the modifications for any instance of the device. The author assumes no responsibility for the actual or alleged damage caused to your property, data or health as a result of reading the following text. The reader should understand that he undertakes any of his actions at his own peril and risk, fully aware of the likelihood of negative consequences.
The lithium battery
BT601 in the 3457 performs 2 functions - provides backup power to the chip of RAM
U603 storing calibration and user data, and serves as a reference voltage for the old version “
WatchDog”.
One of the functions of the "watch dog" is the forced holding of the microprocessor in the "
RESET " state during transient processes in the power supply unit or under a greatly reduced power supply so that its unpredictable actions in the conditions can not damage the contents of
RAM .
Accordingly, before removing the battery, you must:
1) Replace the U603 chip with something non-volatile, for example
FM16W08 or STK12C68:
2) Ensure the regular work of "WatchDog" without a permanent battery power supply.
Each of these tasks can be solved in a variety of ways depending on the preferences of the owner of the device. For example, for item 1, there are options for minimal intervention in the original board (then the necessary decoding of the control signals will have to be performed on the additional module) and
minimum labor without the use of additional components (but then several sections (cutting of wires) will have to be made on the board) .
For item 2, it is possible to install an additional specialized microchip that does not require a battery for operation, or
install an additional node that generates the same voltage as the battery .
I chose the second, highlighted options for both items.
Two boards were upgraded: at the end of January 2017 FM16W08 was installed in
03457-66511 (A11) and STK12C68 was installed 3 months later in
03457-66531 (A31) .
1. FRAM in 3457.03457-66511 (A11) cirquit diadram.
Red color for cutted wires,
Dark blue color for added wires and names of signals,
Green color for added parts for watchdog.
Line at top of the signal's name mean "Active LOW", the same mean "tilda"simbol "~" before name of signal in the text:
Original chip U603 (
TOSHIBA TC5564APL-15 [1]) has one control input more than FM16W08 [2]. For right addressing we have join the signals
~RAM_EN1 and
~RAM_EN2 in one signal
~RAM_EN.
Additionally we must have one falling edge of signal
~CE for each new address at FM16W08. For that we can use 2OR
U624B, in the original was unused.
The inverted signal
Q from the 35th pin of the processor
U601 is fed to the second input of this element. In the processor's datasheet [3] said that this signal confirms the presence of a stable address on the Address Bus that can be snapped on the rising edge of this signal.
The falling edge that we need is obtained by taking the signal after the inverter
U634D .
In the scheme
red crosses denote the conductor slices,
in blue new connections and signal designations introduced for better understanding.
The green color is applied to work related to "
WatchDog ".
ATTENTION! You must have ability to read vintage SRAM TC5564APL-15 and write data in the FM16W08. Many programmer can't work with new version of FM16W08. Please check your programmer. Lost of cal data mean full calibrating of 3457. If you havn't ability to calibrate it, then change nothing in the DMM!Then perform the work step by step.
All work must be performed on a completely de-energized device, with the mains cable disconnected from the socket on the rear panel.
A soldering iron with a grounded wire can not be used for obvious reasons. It is acceptable to use a battery-powered soldering iron or hot air. In extreme cases, you need a soldering iron with a low-voltage power supply with a ground through the resistor.a) ATTENTION! Items a, b and d are CHANGED for content retention and subsequent correct reading. The new procedure is set out in the second part of the article (item 1...8 ), be guided by it.Prepaire additional supply for U603 and solder it to the pins 14 and 28 with right polarity:
b) Carefully desolder U603. Pay attention if you overheat pin 14 or 28, additional supply may be lost contact. Exactly this was happened in my case.
c) Install the socket
DIP28.
d) You can set old U603 in the socket and check workability of DMM. Next copy info FM16W08, that must be installed on the SO28/DIP28 adapter.e) Find
pln 20 of U603 and cut the wire:
f) Find
pins 4 and 5 of U624 and cut the wires:
g) Find
pin 11 of U606. Desolder and pull out this pin from whole:
h) Desolder and pull out any pin of
R647.
i) Make next connections with wires:
pin 11 of U606 with pin 12 of U634 (joine of
~RAM_EN1 and
~RAM_EN2)
pin 4 of U624 with pin 11 of U634 (
~RAM_EN)
pin 5 of U624 with pin 13 of U624 (
~Address_Valid)
pin 6 of U624 with pin 20 of U603 (
~CE)
j) Insert the adapter with FM16W08 into socket
U603.
k) That's all for memory.
Next step -
2. WatchDog.To simulate the battery, a node was assembled on
TL431 [4], the circuit was drawn
in green on the main circuit. With the specified ratings, the reference voltage was
3.4852 V against the battery's own voltage
3.4776 V .
However, what about the fact that the battery gives a constant voltage, but the simulator does not work at the moment of switching on? In other words, how quickly does the circuit enter the stabilization mode after the power is applied?
Datasheet is quite optimistic:
From 0.5 us for the reference voltage to about 1.8 us for the output. It sounds good…
We will check.
We will set up the functional generator for a meander with levels 0 and +5 V, we will submit it to the simulator and see it with an oscilloscope. First, set the frequency to
0.1 Hz , simulating one switch every 10 seconds:
Approximately 20 microseconds. The oscilloscope shows 17.1 us, because it measures between the levels 0.1 and 0.9 of the amplitude value. It is interesting, how will the circuit behave when the frequency of "switching ON" is accelerated? We put on the generator in series
1 and
10 Hz :
17.5 and 15 microseconds. Apparently, the internal capacitors do not have time to discharge completely at high switching frequencies, so the time to enter the mode is reduced.
We will accelerate further. 1, 10 and 100 kHz:
Aha, 12.5 us for the first two frequencies and
failure at 100 kHz - does not have time to get the required voltage. Let's check, at what frequency is it still possible:
27 kHz and the same 12.5 us.
Take the worst time - 20 us.
And what time reserve does the scheme give us? After all, the entire WatchDog scheme is not able to start instantly, it also needs time to enter the operating mode. In the original, the time reserve for transients is provided by the RC chain
R635 +
C635 . The diode
CR633 provides the capacitor
C635 a path for fast discharge when the power is turned off. The circuit shows
R635 = 20 kOhm and
C635 = 15 uF .
We will construct the model of this scheme with the "aged" parameters. Suppose that the capacitance of the capacitor has decreased threefold for many years, to 4.7 uF. This will reduce the time reserve. Let the resistor also grow old and reduce its resistance to 15.6 kOhm - this further reduces the time reserve.
We connect the RC-chain layout with the specified parameters in parallel with the battery simulator, enter the signals into the oscilloscope and look at the frequency
1 Hz . First, a slow sweep to see the charge of the capacitor:
Excellent. The capacitor needs 15ms to charge 1V. TL431 by this time is almost a thousand times ready (20 us, remember?). Let's increase the beam speed:
The same picture. Even faster:
Now the transient process TL431 can be seen, but the voltage on the capacitor does not have time to grow by one iota. Conclusion -
the constructed simulator of the battery will perfectly cope with the work and will not make any instability when switching on the device . We remove the RC-chain layout and proceed to installation.
l) Desolder the battery, starting with the negative output - there is one pin. On the positiveterminal - 2 pins, desolder in stages.
m) Solder in the board three racks of irradiated copper wire.
One - into the opened hole of the positive battery terminal (to it will be connected
the output of the battery simulator, ie the cathode TL431).
The second - to the terminal of the CR636 diode closest to the battery (it is
+5 V , will be fed to the simulator input, to the resistor 510 Ohm).
The third one is to the nearest
earthed terminal of any component (for example, C641). It will be earth. We mount the simulator of battery.
n) The upgrade is complete. Assemble the device, run the tests ... Now we can use the HP34357A for joy.
On the sweet - the original scheme
A11 in one picture (in the original service manual on 4 sheets):
I express my gratitude to the participants of the eevblog.com forum:
wolfalex for the fruitful idea [5] and
bingo600 for hinting a reference to this idea.
Literature. 1)
TC5564APL-15.pdf2)
FM16W08.pdf3)
The MC6809 CookBook (TAB BOOKS Inc).pdf4)
tl431.pdf5)
https://www.eevblog.com/forum/projects/3478a-fram-modification/ http://wolfalex.bplaced.net/te/hp3478a/fram/hp3478a_fram.htm28.01.2017. TEKTRON.
Part two. STK12C68. I promise nothing and I do not give any guarantees! You take your actions at your own risk.Over the past 3 months, the above-described upgrade has proven to be operational in 24/7 operation. A full calibration was performed without any problems.
It's time for a similar modification of the second, control 3457 unit. The measuring path of this device is left in the original state, only the backlight of LCD was added. For some reasons (more at the end of the text), it was decided to install it not RAMTRON, but STK12C68, developed by SIMTEK.
Since the control device, it is necessary at all costs to transfer the calibration parameters without losses. One of the days was spent on testing different ways of writing and reading the original TC5564ALP. As a result, the main requirements became clear. First, the backup power wires should not be detached from the chip even when the solder melts .. Second, the backup power must be increased to 5 V and ensure the operation of the chip regardless of the programmer. Third, the power must be blocked by the capacitor. Fourth, we must not allow direct contact of backup power with any contact of the programmer. During downtime, the programmer connects all of its pins to the ground potential, making no exceptions to the power output.
It was necessary to go on complication of the reserve scheme. The procedure for extracting the chip from the board is also complicated, it is absolutely necessary. Here's a workable scheme:
The backup power voltage is increased to 5 V (4 nickel-metal hydride batteries in series).
R1 - for protection against extra currents (batteries are very powerful ...). After installing the chip in the programmer, R1 is shorted by switch S1.
The rectifier Schottky diode VD1 decouples the programmer from the GB1 battery. An anode of VD1 is soldered to a piece of single-stranded irradiated copper wire 0.5 mm in diameter, an insulating tube is put on it and this wire will be connected to the programmer.
Since the additional EN1 input pin (pin 26) is active at a high level, it is connected to the PROGRAMMER's power output (more precisely, to the newly made additional lead from the anode VD1) so that the contents of the memory are not damaged during all preparatory manipulations.
The same purpose is to pull to the ground through R2. C1 - standard blocking for high-speed digital devices.
All this is mounted around the chip on the wires at once, looks not so good, but performs its function. This is the main thing.
The numbering of the outputs of the programmer is shown conditionally, by the number of the chip legs. Actually the connector of the programmer has more contacts, but the extra ones are not used.
We solder U603 in the following order:
1. By any available method (vacuum, copper braid), remove the solder from the contact holes under all the leads,
EXCEPT 14 and 28. With the tweezers tweezed each of the leads and make sure that it is free. For accuracy, you can consistently pick up the chip from the 1st and 15th pins and gently, VERY gently, literally for a millimeter to pull each of the sides of the chip from the board. The distances from the remaining soldered pins must move along the hole.
2. The wires (0.07 sq.mm. PTFE insulation) from the reserve power cassette (in which we do NOT insert ithe batteries yet) are peeled off approximately 8 mm, slightly bent and put the UNDER the chip into the gap between 13-14 and 27-28 pins respectively for minus and plus wires. The task is to make the wire a full turn around the output near the body itself, the remainder of the bare section should be wrapped around the same wire as possible. The connection is thoroughly soldered.
This we have achieved that even if this solder melts, the wire due to the loop we made does not come off the lead and will continue to contact through the melted solder. Be sure not to short-circuit the nearest terminal. Also, make sure that S1 is still open. Check the polarity of the cassette connection.
3. Now you can insert the batteries into the cassette and check if 5 V appeared on the 28th pin of the U603. If yes, then alternately we evaporate the 28th and 14th pins of the chip from the PCB and put it on the
ISOLATED surface.
4. Tweezers bend the pins 26 and 28 upwards so that in the future they do not touch the contacts of the programmer:
5. We mount in accordance with the above diagram. At me it turned out so:
6. We configure the programmer to work with the FM1608 and insert the chip that was removed from the device. Visually check that the 28th pin of the chip does not touch the corresponding contact of the programmer. Because if he ever touched - everything was lost, all the labors failed, and the calibration coefficients were lost. Here so.
7. If the 28th pin is all good (and + 5V is observed on the 14th leg), check that the 26th pin does not touch the CORRESPONDING contact of the programmer, which I named the 26th. If also here OK, we close S1:
8. We read the contents into the buffer and run verification several times..
If everything went smoothly, save the file to disk, but do not relax yet. We remove the chip from the programmer, but we leave it under reserve power until the modification is completed and a positive result is obtained..
9. Modify the board A11 / A31 (03457-66511 or 03457-66531)...
... in full compliance with the recommendations of the first part (points c, e ... i) with the sole exception - the Schottky diode CR636 DO NOT PULL! Since the new chip represents a significant capacitive load (68 ?F), without this diode there will be an infinite RESET. (There is an assumption that it can be replaced by a simple jumper, but it was decided to leave as is)
10. We install in the programmer the prepared STK12C68 chip and write down the information in it. To prepare, you need to solder a new chip on the adapter, meet the requirements of the datasheet (parallel connection of 68 ?F and 0.1 ?F between the 1st and 14th pins, and 10 k? between the 27th and 28th pins); Into the adapter DO NOT solder the contacts into the holes 1 and 26.
11. We modify WatchDog in full accordance with clauses l, m and n of the first part of the article.
12. We check the installation, insert the adapter with a new chip into the panel, mount A11 / A31 in the device, turn on and run the self-test. When the desired SELF TEST OK indication appears, you can uncork a bottle of milk and celebrate the event, but before doing so, do not forget to make several copies of the calibration information on different media.
A set of photos on the topic:
13. Reasons why for the HP3457A I found the STK12C68 preferable to FM16W08. In this device, starting with the control board version A11, RAM and CalRAM are physically combined in a single chip. Accordingly, all write accesses to RAM during normal operation, consume a resource of ferroelectric memory. This resource, although huge, but not infinite, and the record is executed in the same address, dynamically there is nothing relocated. But STK works differently, as long as the power is within normal limits - normal static RAM works and ROM resource is not consumed. As soon as the power began to drop - for 10 ms, the contents of RAM are backed up to the built-in ROM due to the energy stored in the capacitor. When the power is turned on again before the first external access, the RAM is restored to the previous contents by reading from the built-in ROM and then continues working as with the usual static RAM. The number of records/readings in/out of the ROM equals the number of switchings OFF / switching ON of the device. For this specific case (24/7 work), the choice is obvious. That's why two STKs were prepared, when the device is first opened with RAMTRON it will be replaced by STK (with the addition of the previously removed CR636 or a jumper instead). Identical refinements were needed to test the interchangeability of FRAM and STK.
I would like to express my gratitude to the participant of this forum
e-pirate for
useful information.
The translation of both parts is done by Google - translater with a slight correction, do not judge too strictly
. The original articles in Russian are located at:
Part 1Part 2.STK12C68 datasheet08.05.2017. TEKTRON.
P.S.
Full set of pictures of part1. 5,6 MB Full set of pictures of part 2. 7 MB.
Home
Help
Search
Login
Register
