8645A Agile Signal Generator: power-up, lots of mechanical repair – finally, calibration, and some first tests

2014/11/02 Simon Leave a comment

The 8645A has been a real challenge. Not electronically, but with all the mechanical damage.

Most of the defective parts are now fixed or replaced, the last one missing:

The front connector and bracket, badly damaged – this will need to go back with me to Germany, don’t have the right tools here (anyone out there with a metal workshop, in the greated NYC area, let me know!). Will see if I can bend this back, or machine a replacement from aluminum alloy. Also, one of the screws is broken off and stuck in the thread – all pretty laborious to repair.

The temporary set-up – connected a SMA f/f adapter, to get the signal out.

Electronically, major success! Not going into any details for now, explaining the inner workings of the 8645A would take hours. But, it is now back to a stage where power can be supplied to the circuits, and watch out, this is the first result:

… after several minutes, still no change – checked the manual – and the self calibration can acutally take 5 minutes or more, so I am patiently waiting. Finally, this result:

Sounds good.

Went quickly through all of the basic functions, and the machine seems to put out almost +20 dBm over the full band, stays perfectly phase locked, and, as far as I can tell with the 8565A, it is pretty clean.

Machine under test –

– showing a ~1.7 GHz signal, FM modulated, 30 kHz depth, 1.5 kHz audio.

For now, repair-final check on hold, with more tests once the connector is back in place.

The many hours spent so far, certainly worthwhile,for a machine that cost about USD 39.5k in 1990, close to 90k nowadays… all its major capabilities, well described in the HP 1990 catalog.
8645a hp catalog 1990

Micro-Tel MSR-904A Microwave Receiver

Micro-Tel MSR-904A Microwave Receiver: a wire, another 2N2905A – repair completed (!) – sensitivity test passed

2014/11/02 Simon Leave a comment

After quite some heavy work with the MSR-904A, we are close to completion. The last defect, supposedly last, was an issue with the 0-9 V internal tuning voltage – could not get it to change from 0-9 V even with all adjustments suggested by the manual, and looking at the schematic – no other way to adjust. There can be drifts of some resistors over time, but all checked and very much as spot-on is they can be.

So, the issue comes from a part of the AFC circuit – there is some gain switching with a 4051 CMOS switch, and as it turns out, this was missing the -6 V supply voltage – causing a positive voltage at the output, which upset the tuning assembly, even with AFC disabled. The wire, not sure if it was broken when I received it – well, easy fix, but took me the best part of 2 hours to find it.

With the machine now ready to be put back in service and fully operational, a few hours warm-up, and all the oscillators and YIG filters fine-tuned, I slipped with the screw driver – a spark – and the 250 MHz amplifier was dead. Fortunately, no major defect, just the 2N2905A that switches on the power for the 250 MHz amp. Added a “new” one, 1984 vintage…

Finally, closed the lid, to keep my screwdrivers out, and did a quick check of sensitivity at about 1.7 GHz (because I have a really well calibrated 0.1-2.1 GHz source, a HP 8642B here already on the bench and did not want to carry around heavy microwave synthesizers…).

Here, the result, which I would call pretty much satisfactory. Doesn’t get much better unless you cool down the receiver with some liquid helium.

The machine, in all its beauty.

HP 8645A Agile Signal Generator

HPAK 8645A Agile Signal Generator: pushing it back in shape, Chinese spares

2014/11/01 Simon Leave a comment

The 8645A – see earlier post, reached my workshop in a really bad shape, electronically dead, and severely damaged. For any reasonable effort, damaged beyond repair.

However, I do this for a hobby and there is no rush, and I just can’t take this thing apart for scrap metal – the case suffered damage, but the inner parts, no major mechanical issue.

For most of their 1980-1995 equipment, HP used a so-called “System 2” modular case, mainly, die-cast aluminum alloy. The 8645A is not much different – the bottom and top lid are different (to provide better shielding than the regular System 2), but the front frame, it is the same.

Looking around at xbay – found a complete front system 2 front assembly, System 2, from a 8782B Vector Signal Generator. The Chinese vendor, he seems to make a cut by taking apart equipment, and selling it for parts – it wasn’t quite cheap, but still, for a 8645A, to get it back into service, it seems pretty much worthwhile.

It arrived already after about one week.

Not in best shape, but usable.

This assembly also has quite a few BNC connectors, and a precision N connector, all very handy to fix the replace the snapped-off parts.
As it turns out, the N connector is of a different length, but never mind, I have some more spares around.

The front panel assembly, it is apparently one of the weak points of the 864x series.
Two layers of rather soft aluminum, a ~0.8 m plastic sheet that also acts as filters for the fluorescent displays, and a thin outer layer with the labels.

This is quite easily pushed in, if force is put on the BNC connectors – it is always best to safely secure the instrument at the corners during transport, and to attach handles.

The layers are held together by glue, and it seems, HP has used a very sticky, but not permanent type – with great care, it can be separated. Key thing is to keep the sticky surfaces clean, and they are best protected by a polyethylene foil.

To flatten things out, a carefully adjust iron is a handy tool – with the plastic sheets put in between two layers of non-stick paper.

The more laborious part – the flattening of the aluminum panels, with a hammer, and a piece of wood, and about 2 hours of work.

The result – all things glued back together, pretty happy with the result.

Sure, there are some little blemishes, well, fair enough!

Now we just need to get the inner workings going…..

Micro-Tel MSR-904A Microwave Receiver

Micro-Tel MSR-904A Microwave Receiver: three bad guys, an a revised YIG filter

2014/10/30 Simon 7 Comments

There are a few remaining items that need to be fixed:

(1) The frequency marker doesn’t work
(2) The AGC circuit is acting up a bit
(3) The 21.4 MHz IF chain – it is working but gain seems low
(4) The YIG driver – it has too much inherent noise when using the external input (EXT frequency mode). Added a 100 n capacitor (see earlier post) – but this interfers very much with the faster sweep rates, where full YIG driver bandwidth is needed. So I need to add a switch that takes out the 100 n cap, except when in EXT mode (not planning on any super fast sweeps in EXT mode, but definitely need low phase noise)
(5) The ‘signal strength indicator’ is working (a nice LED bargraph), but the offset can’t be adjusted properly, and it moves eratically, at times.

With the manual, the search defect (1) – marker that was so far difficult to track down has an end: the issue is one more of the CMOS switches, a 4052. Well, already the second of these switches which is bad. Maybe, I should just replace all the 4051, 4052 and 4053 switches of the units, as a matter of preventative maintenance. Well, we don’t want to fix things that ain’t broke.

Item (3) – nothing worse that disassembling the IF chain – by all admiration for this unit, the 2nd IF chain is not build for troubleshooting. But, following the instructions, first a check of the control assembly. And, quite to my surprise – a defective transistor switch, using a 2N2905 PNP transistor that provides +15 V power to one of the 24.1 MHz amplifiers – this is switched on and off depending on the BAND setting. The transistor – only had a 2N2904 around, same kind, but a bit lower gain. Well, it works – need to check stock back home in Germany when I get a chance.

These are the bad guys-

Item (2) – this was “just” a matter of adjustment. Adjustment isn’t all that straightforward – and the manual doesn’t cover everything. Needed to peak the 250 MHz amp/AM detector circuit – there is a tuned amplifier inside, which has and adjustment capacitor not described in the adjustment instructions – fortunately, pretty clear from the schematic.

Item (5) – turned out to be a 1458 opamp not working properly, replaced. Pretty easy – most of the integrated circuits are in sockets. Sorry, no picture of the defective 1458 😉

Item (4) – added a good Chinese brand SIP reed relay, PAN CHANG SIP-1A05:

Routed a wire with the EXT signal (low when activated) to the YIG driver board. This is now taking the 100 n cap out of the circuit, unless the EXT mode is active. Added a protection diode, to avoid voltage spikes damaging the TTL circuit driving the reed relay.

The 100 n cap and reed relay is installed parallel to the R4 resistor – same as for the C5 capacitor, which is activated in CW mode.

Micro-Tel MSR-904A Microwave Receiver

Micro-Tel MSR-904A Microwave Receiver: manual found!

2014/10/30 Simon Leave a comment

I had almost given up. It seemed that a full manual of a MSR-904A would be perfectly unobtainium. Well, most secrets of the MSR-904A were reverse engineered anyway, but there are some subtle details and adjustments – just a bit arbitrary to do without proper documentation and at a least a bit of insight into the thoughts of the designers of this marvelous apparatus.
One reason might be that these units were long banned from export and mostly governmentally owned. Eventually, the units reached the secondary market, but the manuals, they rest in some archives, or have long been disposed off by the agencies.

Now, to my greatest surprise – a copy of the manual has been located – in Germany. The paper copy, it seems to belong (or have belonged) to a certain H. S. in Webenheim, a place, a mere 72 miles form my German home.
Another fellow German must have scanned this, and again, another fellow, send it to me, electronically. Albeit, not for free, but, fair enough.

It is complete with all adjustments instructions, explanation of how it works, parts list, and SCHEMATICs.

Here, a block diagram of the IF chain – without the 160 MHz 2nd IF output option (by conversion of the 250 MHz 1st IF using a 410 MHz LO) installed in the unit I have for repair.

Screening through the manual, the complexity of the thing is pretty much amazing. All without any controllers or processors – but there is at least one programmable part, a PROM that has the filter switching matrix. Come on, this could have been done with a diode matrix!

AIOM - Analog Input Output Module

AIOM: updated schematics, differential input bias, high-impedance input protection and amplifier

2014/10/27 Simon Leave a comment

Some progress, with the universal analog input output module, analog response analyzer, or line tracer, digitizer – however you want to call it. Added a few features – a bias supply for the differential input, and a high impedance amplifier (and a bit of protection circuitry) for the single-ended input.

First the bias supply – to allow a wide range of input voltages, say +-20 V (even below ground and well above the positive rail), we need a resistive divider network, and this needs to convert a differential voltage (with absolute voltages centered around ground), to a differential 0 to 2.5 V input, centered around 2.5 V (the internal bias of the ADC). A little calculator is used to find the right resistors and bias voltage. About 2.65 V will do the trick, for the desired input range, and resistor combination. Bias current is just about 1 mA or so, easily sourced/sinked by a OPA703 opamp.

adc input bias calculator

This input is mainly designed to sense low impedance sources, e.g., current shunts or supply voltages (lead compensation, or similar configurations). So the ~50 k input impedance will be perfectly fine.

The general circuit, nothing really exciting about it – the AD7712 has a 8 MHz crystal, will run easily up to 1 kHz conversion rate. But mostly, it will be run at 50 or 60 Hz, to suppress any mains related noise, or even at 10 Hz.

A quick test showed that the USB communication is working (using a JY-MCU ATmega32L minimum board) – just waiting for some long waiting times and train travels to write a simple user interface, to control the outputs and the data acquisition by Windows GUI. For plotting and data analysis, I will resort to gnuplot and/or R, not re-inventing the wheel here. Maybe a simple preview screen.

HP 8569A Spectrum Analyzer

HPAK 8569A Microwave Spectrum Analyzer: some adjustments, an ultrastable 10 Volt reference, GPIB test- repair completed!

2014/10/26 Simon Leave a comment

The manual for the 8569A describes a series of performance tests – not all of them were completed, but most. Fortunately, the IF chain doesn’t need adjustment. Still, it took nearly three hours to get the YTF tracking, the A/D converter/digital display section and other display related circuits adjusted. The log amplifier was carefully tested as well, all perfectly in tune.

The most amazing find – the internal 10.0 V reference was found at 10.00004 V. That’s +4 ppm – most likely, better than the voltmeter I am using to measure this.

The reference circuit, according to the datasheet, is uses a regular HP/RCA low noise opamp of the late 1970s (selected 741), and a +-10 ppm/K 1N827 temperature compensated zener. As it turns out – the actual part used is a 1826-1058, an OP-02 equivalent (0.65 µVpp noise, 8 µV/K drift), much better.
Resistors in blue frame are +-10 ppm/K tempco. Others are regular, +-100 ppm/K resistors.
The circuit layout – very much refined, a marvel of analog engineering – guard traces all around!

It is run at exactly 7.44 mA, using the 10 V reference to set the diode current – to minimize the temperature coefficient. See this diagram from the 1N827 datasheet:

Seems that after 34 years of aging, it is perfectly stable now.

Back-up of the internal EPROMs (4x 4 kbyte!) stored – just in case they fade out over time.

Last step – test of the GPIB functions – and, no issues at all.

Writing to the machine:

And receiving plot data back (two signals, close to 100 MHz, -10 dB):

At its final place, for now, on top of a 3585A – total of almost 200 pounds of test equipment. Hope the bench won’t cave in.

HP 8569A Spectrum Analyzer

HPAK 8569A Microwave Spectrum Analyzer: tuning stabilizer/tickler repair

2014/10/25 Simon Leave a comment

After fixing the display and front panel related issues (attenuator, “0 dB” indicator), the only major defect remaining so far is the tuning stabilizer circuit. Once the span is set to below 100 kHz, it is activated – and sets the unit to zero span, rather than the span selected, permanently. Only when the stabilizer is disabled (there is a button on the front panel), the span is back to normal.

The source of this issue – most likely, the A14 tuning stabilizer assembly. Switched the board (part number 08565-60018) with the identical board of a 8565A – and, as expected, the zero span issue is gone!!

Now, we just need to fix the board.

The description in the service manual is actually pretty clear – the circuit uses a 1 MHz harmonic sampler, to keep the frequency of the LO stable. The initial lock is achieved by setting the analyzer to zero span, finding a frequency close to 1 MHz that represents a lock point. This frequency is controlled by a voltage, and a very long lasting sample-and-hold is used, than employs a special construction of a HP reed relay, and a PE/PTFE(?) lowest leakage capacitor – the white coil in the front, left, and the the capacitor, front, right – in the middle, a PTFE isolated post, going to a FET gate.

Certainly, a very special capacitor. Once this is achieved, error voltage of the sampler is used to steer the LO, and the analyzer is returned from zero to normal span mode.

The sequence of all these steps – controlled by a very interesting chain of transistors, resistors, and capacitors – nowadays, just a few lines of program code… but, nothing better than a handful of discrete high quality components.

The “control generator”, see schematic below – it is not controlling the sequence properly – and, again, a defective cap. 0.1 µF (!), tantalum – red frame; a few 100 Ohms – near short. No way that this can get charged, and let the sequence proceed.

Replaced it, with a regular 0.1 µF cap (why did they use a tantalum??), and: A14 assy working again. 20 kHz span, 10 kHz resolution bandwidth. Working just fine – at all frequencies up to 20 GHz, the highest I can check here.

The black sheep (date code: 1980, week 7):

We can’t argue that they didn’t skim on parts – at the time, certainly, one of the best 0.1 µF tantalums available.
Parts list specifies 150D series, MIL–PRF–39003 mil-spec qualified high reliability hermetically sealed cap:

HP 8569A Spectrum Analyzer

HPAK 8569A Microwave Spectrum Analyzer: front panel repair/re-build

2014/10/25 Simon Leave a comment

The easy of use by the three-knob operation mentioned before – it has a downside: a very complex front plate assembly. And this might be the reason why more modern units almost exclusively use single-axis rotary encoders, often only one single encoder to operate all functions, per unit. The multi-turn-coupled-limited-synchronized-geared movements, just too complex and too expensive, too heavy, too laborious to maintain.

The front panel controls – as delivered.

The 8569A, 8569B and 8565A all use very similar front assemblies – only seem to differ in the print and in the exact shade of the label colors. And these assemblies use plastic parts – knobs, wheels, and so on. After 30+ years, no wonder, aging has visible effect, and parts become brittle. Common defects are broken-off sliding contacts. Not a big thing, can be fixed (re-attach the contact with M1.2 screws).

The attenuator and “0 dB” warning indicator – these are located on the same rotary cam, same contact pair. And, you bet, the corresponding wiper contact – broken off-missing.

Well, having rebuild some of these assemblies for other repair jobs before – the only way to fix this will be to take the assembly apart, almost fully. Actually, if all screws are kept well-sorted, not really such a big effort as is sounds. Some hours, but then it will be working again.

In addition to just the repair, all contacts and surfaces were thoroughly cleaned, by various means: DeOxit D5, a good eraser, isopropyl alcohol. For the painted surfaces and parts, use 40-50% isopropyl alcohol only, and a rather soft or medium-hard brush, to avoid damage to the paintwork.

The replacement cam (on the right) has two wiper contacts, the defective part, only one – the plastic fingers holding it down became brittle, and it eventually broke off.

Here, the result: the rebuild assembly. From earlier repairs, the remaining contacts should last – somehow, no all are susceptible to breaking-off; or it is just the amount of use they get – at least, I have some units fixed with small screws holding down the contacts, and there have since been working better than before. Therefore, I am not too worried about more contacts failing in the near future.

Note: as it turned out, also the 12 dB vernier attenuator encoder had a partially broken off contact – can be fixed with some epoxy glue, but I rather replaced the encoder, with a spare from the parts unit.

HP 8569A Spectrum Analyzer

HPAK 8569A Microwave Spectrum Analyzer: display fix, A6 assembly

2014/10/25 Simon Leave a comment

No display – this can be bad news, because spare CRTs are hard to come by. Let’s see what we can do here.

Overview of the assembly – red: XYZ driver boards; green: high voltage (and other supplies) for CRT; blue: digital circuits-video signal generator.

The digital part and XYZ driver – seem to be fine, can get a nice video signal when probing around.

Well, the HV power supply – no signs of life. You can easy check, just switch off the unit – there are two neon bulbs that discharge the CRT – but in the current case, all remains dark.

Removed the HV power supply assembly (A6) – fuse is blown! A quick look at the schematic:

Only three options, either the 26 V filter, the transformer, or the Q1 driver transistor. Driver transistors are a common fault – high voltage transients can damage them. But, Q1 is testing fine.

The transformer, looking good.

The filter – well, there we go, a classic fault, a defective cap.

This is a 50 µF, 50 V, Sprague (now: Vishay), 30D series. Date code: 1980, 26th week! “Made in USA”

This exact series – still available today, after nearly 35 years! These don’t fail so easily, but the A6 assembly, there will be some stress on these parts. The failed part is capacitor C2 – the part that is doing most of the filtering. Maybe it would be a good idea to put a 1 µF foil capacitor in parallel with the electrolytic caps, to absorb the spikes…

For the time being, just replaced the defective cap, and ordered some 105°C rated long-life caps. 30D series caps aren’t cheap. So I settled for a Kemet brand type. Will be good enough.

After some soldering, powered the thing on:

We have a working display!! And, it is actually a very good CRT – no signs of wear! Most likely, sitting in some lab at Iowa State University, as one of the labels on the case suggests, and then, stored away in a nasty shed or garage for years.

Now, repair can move forward – without a display, at would be only half as useful.

The culprit, and the dead fuse:

SimonsDialogs

8645A Agile Signal Generator: power-up, lots of mechanical repair – finally, calibration, and some first tests

Micro-Tel MSR-904A Microwave Receiver: a wire, another 2N2905A – repair completed (!) – sensitivity test passed

HPAK 8645A Agile Signal Generator: pushing it back in shape, Chinese spares

Micro-Tel MSR-904A Microwave Receiver: three bad guys, an a revised YIG filter

Micro-Tel MSR-904A Microwave Receiver: manual found!

AIOM: updated schematics, differential input bias, high-impedance input protection and amplifier

HPAK 8569A Microwave Spectrum Analyzer: some adjustments, an ultrastable 10 Volt reference, GPIB test- repair completed!

HPAK 8569A Microwave Spectrum Analyzer: tuning stabilizer/tickler repair

HPAK 8569A Microwave Spectrum Analyzer: front panel repair/re-build

HPAK 8569A Microwave Spectrum Analyzer: display fix, A6 assembly

SimonsDialogs – A wild collection of random thoughts, observations and learnings. Presented by Simon.