Category Archives: Test Equipment Repairs

For some, it is like solving crossword puzzles: fixing defective test equipment. Preferably, mid-70 to early 90s vintage.

HP 4140B pA Meter/DC Voltage Source: some incorrect assumptions, but finally, a repair

Shortly prior to my departure from Japan I started repair of a HP 4140B. A very desirable instrument for semiconductor characterization. The issue remained that output “A” didn’t provide correct voltages, probably due to some issues in the amplifier.
Two month later, I have returned to Germany, and a spare LF256 J-FET opamp had arrived, so I thought it would be a quick fix – but to no avail.

Fitting the LF256 to the board – I usually leave some part of the old wire in the Teflon isolator, because I want to avoid soldering/melting the solder in the Teflon part – it is all difficult to clean up afterwards, so I prefer to solder the new part to some leftover wire, rather than contaminating the isolator.

But- with the new opamp installed, same symptom, no proper output, current limit function of channel A blinking, but the input of the opamp is good. So it must be something else down the chain. Another look at the schematic…

There is an analog switch, followed by a discrete linear amplifier with a dual J-FET input stage.
After some study of the analog switch (cutting a trace and checking it), the switch appears fine. Next in line, the dual J-FET, and in fact, this is dead – found it by measuring the E-B and C-B transition voltages with a diode tester (instrument powdered down and board removed!!), and the transistor around the FET shows largely different values compared to the working B channel. It is 1855-0049 HP part, available in some single piece quantities but expensive!

Looks still shiny and new, but it isn’t working.

Studying some NSN databases, found at least some data of this part, which had been manufactured in equivalent versions by some other manufacturers as well, probably in the 1970s.

It is a rather not so special depletion n-type J-FET. But it is a dual FET part, and while single J-FETs are no problem to get, dual FETs are rare specimens.

Even in their long past days, these didn’t come cheap… maybe something like 40 EUR a piece in today’s money.

So we need to do further study, and there are essentially two kinds of dual FETs – some that have a specially made dual die, with both FETs on one chip and coupled in various ways to keep them from drifting apart with temperature, etc., and the other kind, which is merely just two reasonably matched separate FETs in one case, for convenience more than anything (and for thermal match).

Screening through my inventory I found these 2N5457 FETs which have pretty similar electrical characteristics, in particular, zero-gate-voltage currents.

The parts I have are all quite uniform so there is no need to select a special pair.

With such replacement with similar parts, rather than identical parts, I think it is a good idea to take no risk, so I took the B-channel dual FET and transplanted it to the A channel. And the B channel, which is anyway only a secondary function of the instrument and doesn’t allow the same fast ramps and functions like the A-Channel, it will be definitely good enough to install the two FETs separately (closely together), rather than the original part.

The dual FET of the B channel replaced by two 2N5457.

The B channel dual FET 1855-0049 transplanted to the A channel.

With these repairs, the instrument powdered on just fine, and the output voltages were spot on without any need for alignment. Even the zero bias setting if the LF256, no need to adjust.

Induction range repair – just a couple of IGBTs, and a 20 Amp fuse

Recently, the induction cook-top of my SMEG range failed, leaving me with potentially expensive repair options quoted at above 1000 EUR, or to do some investigations myself. Surely, the latter option applies in my case. So, after receiving two spare IGBTs by mail, and a high current fuse that matches the “repair option” fuse holder of the cook-top, it took just a bit of soldering to get the thing up and running again.

Mounted the IGBTs, an easy job compared to the tedious mounting of all the coils and cables. It is not quite a service friendly design, and there are many sharp edges that can damage cables and your skin, so better wear gloves and handle everything with care.

After cooking on the range for a while, there is absolutely no problem at all, it’s a 20 dollar fix, if you don’t count your own time – maybe about two or three hours, mostly, to take out the electronics and put them back in.

Micro-Tel SG-811 Swept Signal Generator: another dead LH0021, and a design issue

After only a few hours of use since the last repair, the unit started to play up again. Simply, no output on all bands. A quick check revealed the issue, fortunately, no failure of the power supply, but again, the driver board for the oscillators, and again, the LH0021 power opamp isn’t delivering current.

By removing the wire that connect the LH0021 output to the YIG tuning coils, and feeding current from an external supply, all working great – fortunately. With no other stock at hand, I decided to move the power amp LH0021 from the (not normally used) filter driver board to the oscillator board, these boards are essentially the same design.

After that switch, it worked again – but only for another 30 minutes, then it failed again, another LH0021 burned out. How can it be??

No, I took all apart, including the mica washer, suspecting some short through the mica or similar issue (the heatsink is ground, but the case of the LH0021 is output). The mice is OK, but there is an issue with the screw hole and its plating. On the top side, is is plated as much as that it contacts the heatsink just slightly, probably, when it expands with heat, it causes the deadly short. Noteworthy – the driver board has the top side of the screw hole completely unplated!

Anyway, too many defective LH0021 yet, and this time I couldn’t find a cheap source. And not willing to pay USD 20 and take more chances for these parts to fail.

As luck would have it, there are some audio amps in my stock, quite common in lower-cost stereo amplifiers. About 20 W audio power, in an easy to use TO220 case, and despite being obsolete, these are ubiquitous, and low cost.

Normally, these are AC-coupled at input and output, and I didn’t find much reference to DC coupled uses. So I set up a little test circuit, and in fact, it provides a nice power opamp (unity gain stable).

Furthermore, the TDA2030 has both short-circuit and thermal overload protection. I wouldn’t call it indestructible, but chances are, that it would survive some adverse conditions.

Only trouble, there is no good space to mount the TO220 case to the heatsink. But a temporary setup will do for now.

With no other change of the circuits, all seems to work well, and also the frequency response seems OK. The LH0021 has about 15 kHz bandwidth, this can be easily met by the TDA2030A.

And in fact, it works well in the SG-811. All working and no need to align anything. Still, I would like to make sure the device has a proper heatsink. So, from a piece of scrap aluminum alloy plate, I machined a heat distribution plate, about 10 mm thick.

That’s the ready-machined parts, degreased with a bit of alcohol.

The distributor mounted well to the board, I cut threads into the metal block, so it is easy to affix to the board without any additional holes or modification.

To be not again trapped by some strange things, I also did some testing of the inrush current, power-on behavior and such (a current spike or reverse voltage may also damage the power amplifier). Also, mounted two more caps to the rails, and a dual-diode 48 V limiter.

However, the startup of the 18 V rail is good and clean.

Same fix applied to the filter board – there is enough space to fit the amplifier without any trouble.

Running at 18 GHz for a while, the temperature stabilized at about 60 degC, well in the range of good working conditions. A few hours later, the SG-811 is still working. So, with some luck, hopefully, a permanent repair.

HP 4140B pA Meter / DC Voltage Source: Special low currents, special connectors, and various FETs

It is another great auction score, a HP 4140B meter, used widely in the semiconductor industry and automatic test stations. Also handy in the lab to test all kinds of diodes, Zeners etc.

It has two +-100 VDC voltage sources, and a ultra-sensitive pA meter built in.

The pA meter seems to work, but one of the voltage source current limit LEDs flashes, although nothing is connected. This will need some repair. The other voltage source is working just fine, so there is no issue with the control board or DAC at least (one DAC is sourcing the voltage for both voltage outputs).

The current input is using some very unusual and high value range resistors… megaohms, gigaohms! Rarely seen before…

The range resistors are switched by reed switches, but not very common design. The coils are actually at the underside of the board, and no physical contact to the reed case, which could lead to leakage currents in the picoamp range.

There are some (plated) iron rods going through the board. These will get the magnetic field to the reed contacts.

These precision resistors, they don’t seem to come cheap. Maybe HP got a discount at the time… at least it doesn’t appear recommendable to start building such pA meters from scratch yourself… rather get some old used units.

The input assembly uses a dual FET to sense the null current, and the FET is a U401, rather common device. Maybe some nice experimentation or null detector can be done in the future with such designs.

The FET is mounted in the board, within a ground plane, and shielding between and around.

From the top, although there is not much heat generated, generous utilization of space, it could probably made fit to 1/4 of the volume?

The defect of the voltage source, it could be easily traced to the A5 board. This has a track and hold circuit, with a FET input opamp. The 4140B is one of the few instruments that I only touch with gloves inside! Better don’t leave residues and fingerprints on these gigaohm resistors and teflon standoffs.

Turns out the input to the amplifier is good, but the output is defective. A simple LF256H opamp, quite a common part.
Waiting for the spare… but pretty sure that replacing the opamp will fix the A5 board.

Another difficulty, the main connector. Originally, the 4140B came with a set of cables and a connector assembly, but this is mostly lost in some drawers of the previous owners.
So I did a test with a rather temporary assembly, but it is showing the correct currents, so all is good in general.

Finally, I found a cheap triax cable assembly.

The connector, it is gold plated inside, and better don’t touch!

HP 54750 Digitizing Oscilloscope: a CARE package, and a scratched disc

Faster than expected I got the spare power supply from the US, it is in good shape, not dusty or anything. Well packaged.

First, I studied the circuit and the burned/unreadable diodes, because I will be trying to repair the defective supply later, just to have a spare. The diodes are 3.9 V and 11 V Zener diodes, fair enough.

Also dissected some special thermal fuse protected resistors, 22 Ohm, about 5 Watts, and a 130 degC thermal fuse in a ceramic package, an inrush current protector.

Probably going to replace these with discrete 22 Ohm resistors and thermal fuses. Actually, both the thermal fuses and the resistors were shot.

After fitting the power supply, some issues. The instrument starts up, and the screen initializes, showing a gray square, but nothing else. It just doesn’t boot up. Fiddling around a bit, I thought that maybe the battery protected memory got corrupted, or some other issue, so I set the dip switch to force-update and rom-unprotect, and started it twice, without actually loading firmware, but hoping that it would set some bytes or something to make the machine start at least. And it did. Also took out the video and cpu cards, reset all connectors. But finally I believe it was just some memory hickup.

Accidentally, found a stamp – made in 1996 – fits the datecode of the semiconductors.

The machine has been on for quite a while, no wonder the power supply eventually gave in. For precision timebase and jitter measurements, it is recommended to leave these instrument always on, or run it several hours before the critical test (3 picoseconds/div resolution, we are talking about mm distances at the speed of light…).

The firmware is a bit dated, and with the startup issues (that actually completely resolved once it started), I decided to update the firmware. Easy work with a 3.5 inch disc. Trying and trying – always getting read errors on my USB 3.5 floppy drive. EEE??? Some inspection – there is a big scratch in the disc.

I had been harsh to this disc, the only one in my possession in Japan, by storing it in a box with electronic parts and all kinds of things and dirt could easily get into the disc. In the late 80s, I kept these discs in a specially design box, etc.

Now, where to get such a disc it the city of Ube, Yamaguchi, Japan? A quick ride to the recycle store (2nd hand store) – no discs for sale, but I found an old network card, including an unused driver disc! 200 yen!

=”490″ class=”alignnone size-large wp-image-4682″ />

The disc is working just fine, and the update proceeded with no problem – version 7.12.

Final critical tests of the 54751A plug in – the samplers are working great, no issues at all!

The new firmware disc, I will keep it in a well sealed ZIP bag, for later use!

HP 54750A Digitizing Scope: it ain’t work with no power

At the famous Yahoo auction, in Japan, I found this great instrument, a 54750A with a 54751A plugin. This is no less than a 20 GHz bandwidth(!!) sampling scope. I used to own one before, but have sold it some years back. At the time of introduction, these were the top instruments, in the 1995 to 2000 years. Still today, a 20 GHz scope doesn’t come cheap!

Normal issues with such unknown, non-working units are broken plug-ins, shot samplers, or other various difficult to fix issues. The samplers tolerate no ESD, and no more than 2 Volts! So please don’t let just any kid play with it. Not sure what happened at some companies in the past, when the engineer shot the 10 kUSD plug in… the whole unit traded for about 50 kUSD at the time, not too long ago.

This unit, it just won’t power up at all. No sign of any activity. So I took out the plugin, and started troubleshooting. The power supply.

These supplies were purchased from YOKOGAWA, a really high end supplier, and there are protection circuits for all voltages, including interlocks for the fans! But this unit has no need for any interlocks, because there is absolutely no power.

A little bit of examination shows discoloration and defective diodes.

The diodes overheated, because they are short. These are primary side diodes.

Also one of the transistors around this area of the supply failed. A 2SC3866 high voltage transistor, nothing too special, but none at hand at the moment!

Some analysis of the circuit – it is clear that this is the startup (auxiliary) power supply that will start up the main power supply. Also the input current limiting resistors (2 units that have 22 Ohm and a 130 degC thermal fuse each) are blown, no wonder that there is no action at all.

So I tried to connect an external supply, but there seem to be some other functions and details, so I can’t get the unit to start up. At least the red LED that indicates the auxiliary power, it is lit. But nothing else. Better not to proceed without schematics – which aren’t available.

So I have two options – wait to be able to go to Germany again, where I have a 54720A mainframe that has the same supply, and do some tests and investigations on that supply, or, see if there is a spare supply available somewhere. Kindly enough, a US enterprise offered a spare for USD 25, a great price!!! Plus 70 USD shipping to Japan…. Well, I purchased it and now waiting.

The 54751A module – even if the mainframe won’t work again, it is great find. These usually go for USD 1.5k, working condition. And I do have spare samplers back in Germany, in case it is a damaged unit (but as the mainframe has a power supply failure, I suspect that the module is fine, and that the unit was taken out of service because of the non-working power supply).

Some current offers…

HP 3335A Synthesizer/Level Generator: Unlock issue, soldering issue?

The recently fixed 3335A, it still has some issue with the reference input. From time to time, or when touching the BNC cable connecting the 10 MHz OCXO output with the reference input, the *UNLOCK* light comes on.

First I thought it has to do with the low cost and cheaply made BNC cables, but even with very good BNC connectors, the issue persists. So, let’s open the thing up another time. I don’t like partially fixed instruments – even though this error could probably be solved by just not touching the cable….

Fortunately, the BNC connector has a long cable internally, and can be disassembled without destroying the crimping.
Checking it, the center wire has no connection. Looking at it in magnification, it seems that only ever the very tip has been in touch with solder, but inside of the golden BNC center connector, there is no solder. So it is clearly a HP manufacturing issue, someone didn’t solder it quite right.

After soldering it again, with plenty of solder, and removing the excess, all is working well! Finally!

HP 3335A Synthesizer/Level Generator: a good fake transistor? at least, a working unit!

Recently, I got a very nice but non-working HP 3335A, with a defective power supply.
After a little wait, the spare PNP power darlington transistors arrived. 2N6052, pretty general purpose power transistors.

The look very shiny and heavy chromium plated, I got 4 pcs, and two had some marks of screws, and 1 had irregular connection wires. Doesn’t look like a genuine Motorola, but well, it is no rocket science anymore to make PNP darlington transistors.

After installing the transistors (I checked the contacts on the board, they seem to be good), some quick test with the load transistor showed perfect regulation of the power supply both on the -15 V and -5 V rails.

With the module assembly connected, unfortunately, the -15 V rail goes up to about -3 V, by current limit regulation.

No other way to find out than by checking module by module. The 3rd last checked turned out to be the faulty one – A7 assembly, mixer.

A little shorted 2.2 uF cap…

At the far lower left.

Replaced it with a 10 uF ceramic multi-layer capacitor.

To check the quality of the transistors, I sacrificed one piece, and cut it open – seems very solid inside, with copper heat spreader, well-mounted die, good bonding wires. So even if it has not been made by Motorola, it seems like decent quality.

Now… doing a proper test…. several hours on, to see if there are other weak links. The attenuator at least is working, and the output is good!

Mettler AE 163 Dual Range Analytical Balance: Swiss Made equipment, in Japan

Regularly screening through Japanese auction sites on the lookout for some gems, I found a great AE 163 Dual Range analytical balance, completely non-working condition. No display at all. From the picture it looked like a rarely used clean unit (be careful when buying some old lab equipment, some might have quite some damage by chemical vapors etc.). I scored it for 7 EUR, great!! Plus another 20 EUR in shipment charges, but at least it was packaged very well and arrived with no damage in transit.

The specifications are better than most modern analytical balances ranging in the 3-4 kEURs, with 0.1~0.2 mg linearity, built-in calibration weight (accurate to 0.2 mg – very hand to recalibrate the balance after taking it to another place, or just to confirm that it is working fine), and these were the high end balances of the 80s, still in use today in various labs. I remember to use such balance during my time as a researcher at the University of Eugene, Oregon, a while back…

The balance has about 4 circuit boards, a display/keypad (an ingenious single bar keypad, easy to handle with gloves on, etc, without disturbing the balance), a control board that also has the main power supply, a sensor board for the force compensator, and a current driver board for the coil. These balances work by force compensation, i.e., there is a magnet coil that will compensate any weight you but on the balance by electromagentic force. And there is a pretty sensitive position detector (a light gate) to keep the regulation control loop going.

After some probing (there are no schematics unfortunately, but anyway, difficult to fix because there are mask-programmed controllers and custom ICs), found that one of the supply rails is down, shorted by some tantalum. 10 uF blue paint-dip type.

Decided to replace them all, including two 1 uF tantalums. Tantalums can last a long time, but some series tend to fail one after the other.

With quite little effort (also because of the nice serviceable design of the unit), all working again.

Here is a closeup of the force coil, it should have a coil and a strong magnet inside.

The light gate of the position detector.

These will also need to be replaced, 3n3 Y-rated capacitors, getting brittle after 28 years…

The balance also had an add-on, a serial interface. The circuit is quite complicated for its function, using mask-programmed CPU, but that used the be the most reliable technology at the time (and still working today).

Also with that interface add-on, replace the tantalum caps, and the Y-rated caps (mains is fed-through to the balance from this add-on module. Not sure why they added another set of Y-caps, as there is no mains related circuitry inside (2n2 value caps).

HP 3335A Synthesizer/Level Generator: a marvelous piece of engineering and precision

For several years I have been looking for a HP 3335A, which is probably the most precision level generator that is available. Some of the more recent devices struggled to keep up with the performance, so the 3335A is still used in some calibration labs, and has thus been quite expensive even until now. This time, in Japan, I found a non working unit for 10 EUR (!!!) including a precision OCXO reference. It is a very clean and late unit (about 1991), but not showing any signs of operation.

The front panel, it is pristine, with no scratch or anything. Maybe a rarely used instrument from some remote Japanese cal lab or university.

The key part, the attenuator, it is a marvel of engineering, don’t touch it without reading first the repair instructions.

The 1990 HP catalog entry, 0.05 dB absolute level accuracy, 0.07 dB flatness. Such performance did not come cheap, 13000 US dollars in 1990….

To achieve the frequency resolution, a fractional N loop was used, it is one of the first instruments that used such PLL technique.

The RF boards are in two metal enclosures, pretty heavy extruded aluminum.

After some tests, pretty obvious faults – the -5 and -15 V rails are dead. There is unregulated voltage present, so it must be something with the regulators, a classic design with darlington pass transistors, and opamps to regulate the voltage and current.

The strange thing, with the modules disconnected, the rails come back to the precise voltages.

Made a plug with a 27 Ohm resistor to load the rails a bit, and, they immediately drop to near-zero. Seems there is not enough capability to drive current.

Took out the pass transistors, and tested these by driving 1 mA of current – but no amplification or anything. Open circuit.

Opened up both transistors, and both have the same defect – the die is not attached to the case any more, somehow, time, heat or something destroyed them (overload is unlikely, because the current regulation loop and foldback is working).

The original parts 1853-0415, power darlington, seem to be equivalent to PMD13K60 of a mysterious Lambda semiconductor company.

There are NPN and PNP complementary sets available. It is a fairly standard darlington power transistor.

From the web, we get another proof, there HP installed the parts with the original part number, rather than re-labeled to HP number.

The replace, the 2N6052 seems to be a good replacement. So I ordered a few – waiting.

As a backup, also ordered some KD366B which seem very strong and well fabricated.

The NPN transistors seem still Ok, but who knows, they might have the same manufacturing defect, so I ordered some BDX87B (which are ST devices universally compatible with various NPN power darlington transistors up to about 100 V; nowadays, all these power darlingtons seem to have the same die inside).

Now, let’s wait for the transistors, stay posted!