Wavetek 2520a 0.2-2,200MHz RF sig gen repair

[dazz1]

Hi
I am trying to repair an old Wavetek 2520a sig gen.
I have narrowed the fault to the main board.  This board drags the +/-18V supplies to near zero.  There is an almost dead short on the board.
The +18V regulator  on the power supply is hotter than the -18V regulator.  The regulators are based on LM395 so I can run the power supply into a dead short. 

The +8V supply feeds the +5V regulators on the main board.  That is OK.

The +/-18V supply feeds the +/- 15V regulators on the main board.  The input to the 15V regulators is near zero, but the regulators are not even warm.   The outputs are less than 0,5V.   None of the ICs I can touch are warm (there are some under a shield).  There are no burn marks or signs of the release of the magic smoke. 


I have a scan of the service manual but it appears that two schematic drawings are missing.  There are parts on the board that have no purpose.

I don't have a thermal camera to trace the current flow.
I have a Fluke 77 so not ideal to track the voltage drop along tracks on the double sided PCB. 

Does anyone have a scan of the missing schematics?
Does anyone have suggestions on fault finding this board?

Dazz

[wn1fju]

Your description confuses me.  You say that the input to the regulators is near zero.  So why would you expect the output of the regulators to be anything other than near zero?  That doesn't necessarily imply a short on the main board.  If you had said the input to the 15V regulators was the normal 18V and the output was zero, then I would agree with you.

You should first unload the power supply by disconnecting the various boards and make sure all the power supply voltages are correct.  On my similar Wavetek 2510 unit, the Molex connector from the output of the main transformer to the rest of the power supply had corroded and the power supply didn't work correctly.  (I ended up removing the connector and soldering the wires since no amount of connector cleaning seemed to cure the problem).

There very well may be a short on the main board (or other boards).  Invariably it will be a capacitor hanging off the power rails.  You can start lifting one lead of the capacitors while watching an ohmmeter on the rails.  When the ohmmeter jumps away from zero, you've found your short.

[dazz1]

Hi
I started by isolating the power supply.  With an open output, the voltages are all good.  The LM395 transistors on the power supply limit current to about 2.2A.  I can safely connect the power supply to a dead short.  Good for testing purposes.   The power supply is fully operational. 

I then isolated the fault to the main board.  When the main board is connected, the +/-18V supplies are near zero.  As a result the outputs of the +/-15V regulators on the main board are also near zero.    Something on the main board is taking about 2A.  None of the chips have a high enough supply voltage to operate, so there is no signal to trace and no function to check.

The 5V supply on the main board is OK so all of the digital logic devices are also OK.   None are exposed to the +/-18V supply.

Temperature measurements on the power supply shows that the +18V regulator is running hotter than the -18V regulator.    This could simply be because of the design loading on the +18V circuits is more than the -18V circuits.

There are several hundred parts on the main board so randomly lifting parts would be the option of last resort. 
I have started by marking the components supplied from the +/-18V including the +/-15V regulators and attached parts.  Without the 2 missing schematics, this is difficult to finish.

Ideally I would have a thermal imaging camera to follow the current along the heated tracks to the faulty component, but I don't have one. I am considering converting an old webcam to infrared.

Today I am going to try and trace voltage gradients along the PCB tracks, but I don't have the PCB drawing or a really sensitive volt meter.  I will pay particular attention to capacitors.  There are no electrolytics on the main board.

Dazz

[wn1fju]

You're right, I apologize.  I just quickly looked at the mainboard schematic (at least the pages that aren't missing!) and saw the 15V regulators.  It didn't occur to me that the +/-18V was also propagated throughout the board. 

There may be no electrolytics, but it looks like there are quite a few (1 uF, 10 uF) tantalums which are notorious for going (near) short.  That would be my first guess.

[dazz1]

Hi
The regulated +/-18V from the power supply is applied to 7815/7915 regulators on the main PCB.   There are two separate +18V regulators on the power supply, presumable because someone really wanted to use the LM395 transistors, and one couldn't supply enough current.   The power supplies are double regulated.  Definitely over engineered.

I am having a go at hacking a old webcam to make an infra-red camera.  It that works, I should be able to "see" where the current is going.

Dazz

[tautech]

As wn1fju says Tants !
DO NOT overlook them !

[dazz1]

Hi
Visual inspection revealed one Tant split in half that was missed on the initial look over.  Clearly not the source of the short, but if one has gone, the others are probably near or past failure.

I tried a webcam hacked to block visible and allow infrared.  The infrared filter was replaced with crossed layers of polarizing film.  Not surprisingly it doesn't work at light frequencies low enough for fault finding.  It does show sub-visible infrared which equates to a temperature to just below red hot.  It does not detect a soldering iron at 350 deg C.  I would need a real infrared camera for that.

There are about 30 tant caps shown on the parts list, but naturally these are all on the missing circuit diagrams.

Dazz

[tautech]

Never 2nd guess Tants.
In Pt 1 Defpom replaced a split and burnt one but still had a short on a rail. Guess what he found:

[TheDefpom]

Just chuck your meter into diode test mode and put it across each tantalum, that will tell you if one is shorted out completely.

If you get any odd readings, reverse the probes and retest, if it is a short you will get the same reading again, if it is just some other component effecting the reading, then you will likely get a different reading, or even see it changing as a capacitor charges up.

[wn1fju]

Although I actually have the thermal cameras and micro-ohmmeters, I've often found it simpler to hook up an ohmmeter across the shorted power rail(s) and start lifting one end of the tantalums until the short disappears.  You can usually do this from the component side of the board without major damage or the need to pull the board out.  It can be a little difficult, however, if they folded the leads on the foil side. 

The last time I did this was on an HP 1345A digital display module (used in many vintage HP pieces) that had shorts on both 15 and -15 volt rails.  There were about a dozen tantalums (those little yellow cylindrical jobs) and I think it took me all of about 5 minutes to find the two shorted ones.  What is funny is that after I replaced the two bad ones, I powered the display up successfully and observed that the graphics/text was quite fuzzy.  Turns out I forgot to solder the lifted leads of the other capacitors back to the board!!!

[dazz1]

Hi
I tried lifting the 4x  10uF 25V Tants on the 18V supply.  One was blown apart. The other 3 are OK.  With these tants connected to air, the short is still there.
There are still 26x  1uF 25V tants on the 15V supply. 
The search continues.

Dazz

[Chris56000]

Hi!

Could you please possibly email me a copy of, or provide a link to, the incomplete manual please?

If one of these generators comes up in the future that I can buy than I can draw out the missing sheets again!

Chris Williams

[dazz1]

Hi!

Could you please possibly email me a copy of, or provide a link to, the incomplete manual please?

The manual is too big to email or post.
Based on the figure numbers (different to the drawing numbers), it appears the missing drawings were not included in the original manual.
Reconstructing the missing drawings would be made more complex because some parts of the circuit are fitted but not used.

Dazz

[dazz1]

Hi
I discovered that the main board has connectors on the bottom side that connect through to PCBs inside the enclosures on the other side of the equipment.

When I pulled the main board the short disappeared. 
I isolated the short to a Tant in the Oscillator module. 
When I refitted the mainboard and applied power, a Tant on the main board released the magic smoke. 
All of the failed 25V Tants are on the +18V rail.   I have replaced them with 35V Tants.

So now all the power rails are at the correct voltage. 

The self test and auto-calibration has failed. 
There are multiple control loops so now I need to work my way around these to find the next fault.

Dazz

[dazz1]

Hi

A local RF company http://www.rftest.co.nz/  with a ton of test equipment and expertise let me use a Thermal Imaging Camera on the main board. 

This was before I knew that there were still connected boards below the main board.  The thermal imaging camera was very sensitive and easily showed temperature differences of 1 degree C.  It did sense the slight temperature variations of current carrying tracks and components.  It correctly showed no hot components on the main board.

It was the camera that convinced me I needed to look harder at the equipment to trace what turned out to be a shorted Tant cap on another board.

Dazz

[wn1fju]

Glad you finally found the shorted tantalum.  Had to be that.

I will tell you what was wrong with my unit when it quickly failed the auto-calibration and threw an error.  The main loop uses a VCO mixed with a quadrature DDS pair, then divided down to the phase detector and loop filter.  The DDS circuitry was outputting zero - not good!  I substituted my own quadrature signals from an external source and then everything worked.  So clearly the DDS was malfunctioning.  Wavetek uses a ROM-based look-up table that contains a (digital) sinewave.  Turns out that the ROM was not being addressed correctly by the bank of 74HC583 BCD adder chips.  And indeed, a logic analyzer showed a stuck bit on one of the chips.  A new 74HC583 fixed the unit.

Also, I found that the main board is really wedged tightly around the metal bulkheads surrounding it.  I had to pry the board out with a screwdriver (ouch).  And I found that Wavetek was fond of placing the 3-terminal regulators at the perimeter of the board, right at the bulkheads, potentially shorting them.  I found a couple that were almost touching so I moved them back a little.
 

[dazz1]

Glad you finally found the shorted tantalum.  Had to be that.

I will tell you what was wrong with my unit when it quickly failed the auto-calibration and threw an error.  The main loop uses a VCO mixed with a quadrature DDS pair, then divided down to the phase detector and loop filter.  The DDS circuitry was outputting zero - not good!  I substituted my own quadrature signals from an external source and then everything worked.  So clearly the DDS was malfunctioning.  Wavetek uses a ROM-based look-up table that contains a (digital) sinewave.  Turns out that the ROM was not being addressed correctly by the bank of 74HC583 BCD adder chips.  And indeed, a logic analyzer showed a stuck bit on one of the chips.  A new 74HC583 fixed the unit.

Also, I found that the main board is really wedged tightly around the metal bulkheads surrounding it.  I had to pry the board out with a screwdriver (ouch).  And I found that Wavetek was fond of placing the 3-terminal regulators at the perimeter of the board, right at the bulkheads, potentially shorting them.  I found a couple that were almost touching so I moved them back a little.

Hi
I read about your fault/repair on your website:  https://emperoroftestequipment.weebly.com/ 
You are lucky.  There is no way I would get approval from SWMBO to have a fraction of your collection.

I suspect I have a similar problem to your Wavetek.  The autocal for the main (coarse) loop should be able to get the oscillators somewhere close to their respective ranges.  The autocal faults on the first osc so either the first osc has failed or (more likely) one of the control loops has a fault. 

I get Error 1 on Osc 1, but I cannot find an explanation of the error codes.  No helpful.

The Wavetek also has an original Dallas 1225Y NVR with internal battery.  This is now 32 years old and may have passed its use by date 
If I can get the Wavetek working again, the plan is to replace the DS1225Y with a new DS1225AD, the modern equivalent.   


Dazz

[dazz1]

Hi
Looks like I could also have a fault in the DDS section.
I am not seeing a sine wave output on the output of IC215, an op-amp found on Schematic 4 of 4 : Digital Synthesis.  I am seeing -1V DC.

Before I dive down that rabbit hole, I need to check the inputs to that whole section.   I don't have a logic analyser so in-circuit checking of chip functions will be harder than it should be.
I need to start by checking that the Data, Clock and Strobe inputs are OK.  If the input is rubbish, the output will be to.

I am seeing clock signals on the Buffer Registers (74273).  I just haven't looked deeper yet.

Dazz

[dazz1]

Hi
I think I am making some progress.
I have not found a fault in the DDS.  Re-checks showed all the right wave forms where expected.

The Operator Manual includes info on the Auto Calibration feature.
This goes through a process of finding the upper/lower voltage ranges to control each of the 4x VCOs. Within each voltage range, the WaveTek finds the control voltage necessary to set a course frequency.   

The AutoCal fails on the first VCO.  When set to operating mode, the display indicates that stored settings are not being used.
I did think this could be due to a failed battery in the Dallas 1225 NVR (32 years old) but the user input settings are remembered between power cycles.  The NVR still seems to be working, but probably not for too long.

There is no serial port available for loading the NVR but there is an option to set factory defaults plus AutoCal.  I think both options were used in the factory to calibrate the Sig Gen (along with some other specified test equipment).  I think the user AutoCal is the same as the initial factory calibration.  If so, it means that I can replace the NVR with a DS1225AD, the modern version of the DS1225Y.  The NVR would then be loaded with the factory defaults and AutoCal.


The VCO control voltage is set by a DAC.  The DAC output is static during AutoCal.  I would expect the DAC output to be hunting up and down as a search is made for each coarse tuning point (35 points for the first oscillator). 

The data side of the DAC is also static during AutoCal.

Both the DAC input and output are also static when the frequency is changed by the operator. 

The input to the DAC is a 74HC595 serial to parallel register on the main board. 
The serial data comes from the 6502 microprocessor. 

So the next line of investigation is down the data side of the VCO control DAC.  This is where the  Logic Analyzer I don't have would be ideal.  I do have some Raspis and Arduinos so I may end up writing some simple code so they can simulate serial to parallel converters.   They could read the data and I could then compare the simulated output to the hardware output.  Any difference would reveal a fault.

I look with envy at the range of high quality logic analyzers available on e-bay, but the cost and risk of shipping them here would be more than SWMBO would ever approve. 
So the hunt continues for the next fault without the right tools to do the job.  I will just have to make do with what I have.

Dazz

[wn1fju]

Sounds like you are getting close!

Although I do have a logic analyzer, and actually used it to fix my Wavetek unit, I've often found it simpler just to use a couple of scope probes and look for I/O lines that are not changing or are at an intermediate undefined logic level.  Sometimes one gets lucky.  It's usually a big hassle to set up the logic analyzer and get all the clips attached to the chip.  I use it only as a last resort.

Yes, there are benefits to living here in the USA where there is an abundant supply of test equipment on eBay with reasonable shipping costs.  I won't go into the disadvantages of living in the USA as I don't want to start an argument!

[dazz1]

Sounds like you are getting close!

Although I do have a logic analyzer, and actually used it to fix my Wavetek unit, I've often found it simpler just to use a couple of scope probes and look for I/O lines that are not changing or are at an intermediate undefined logic level.  Sometimes one gets lucky.  It's usually a big hassle to set up the logic analyzer and get all the clips attached to the chip.  I use it only as a last resort.
...

Hi
The 74HC595 is a serial to parallel register.
The plan is to use the data write line connected to the external trigger of my ancient analog scope.  That will give me the Ch A and Ch B scope inputs to probe data and other pins to look for digital activity. 

The register outputs are all Lo so if I see any serial Hi data written into the register, I will know it is faulty.

Logic analyzers are really cool tech but with chip level integration and serial data streams the norm, they are not so useful any more.  I still want one.

Dazz

[dazz1]

Hi
My testing of the circuits and devices that convert serial commands from the 6502 to parallel inputs to the DAC that controls the course frequency did not reveal any faults in the digital circuits.
I did find that some of the pins on a PCB interconnect are not the same on the circuit diagram and the actual PCB.     These differences probably affected my test results. 

The photos show the test setup to test the serial to parallel devices.  I connected the ext trigger to the data strobe input to the serial/parallel 595 device.  The A ch is the serial data.  The B ch is one of the parallel outputs.  It almost made a 2 channel logic analyzer.    I was able to confirm that the serial input matches parallel output.    This confirms that the 6502 has control of the inputs to the coarse frequency DAC.


The inbuilt diagnostics includes DAC test routines but the documentation only discusses the test showing the right result on the LCD.  I found that the DAC tests also write to the respective DACs.  So I can check the DACs for faults with the DAC test routine. 

The problem with fault finding on the 2025a with the limited test gear I have is that the 2025a requires a series of circuits to work in multiple loops.  Finding a fault requires proving that each circuit works using the previous one to provide a test signal.

The AutoCal software routine uses a binary input (output from a comparator) to confirm the course frequency range.  That binary input is at the end of a chain of circuits in the loops.

The search for the fault(s) continues.

Dazz

[tautech]

A DSO would be much easier to fault find this. 
I'm down your ways in a couple weeks if you want to have a look at something.......

[dazz1]

A DSO would be much easier to fault find this. 
I'm down your ways in a couple weeks if you want to have a look at something.......

Hi
A DSO and all the other test equipment I don't have would make this easier. 
Rationally I should throw the 2520a in the bin.  The value of the time to fix it is far greater than the value of a 32 year old piece of test equipment.  It is a solid instrument built by engineers, not the marketing and accounting dept.  It was built to be fixable.

So far, I am still making progress, very slowly. 

If you want to drop by, that would be fine, but be prepared to be underwhelmed by my ancient and inadequate range of test gear. 

Dazz

[tautech]

A DSO would be much easier to fault find this. 
I'm down your ways in a couple weeks if you want to have a look at something.......

Hi
A DSO and all the other test equipment I don't have would make this easier. 
Rationally I should throw the 2025a in the bin.  The value of the time to fix it is far greater than the value of a 32 year old piece of test equipment.  It is a solid instrument built by engineers, not the marketing and accounting dept.  It was built to be fixable.

You can certainly afford to spend some amount of time on it as 2.2GHz sig gens are not cheap. 

Have you tried to use the TG on your SSA as a RF source ?

So far, I am still making progress, very slowly.

Repairs can take a disproportionate amount of time however the personal development is often worth the effort and stands us in good stead for decades to come. 

If you want to drop by, that would be fine, but be prepared to be underwhelmed by my ancient and inadequate range of test gear.

Everyone starts with old gear so never be ashamed of what you have if it does the job.
I'll drop you an email when I have some definite dates to check our contact details for you are still current and make some plan to drop by.

[dazz1]

Hi

I may have found the source of the next fault.
The built in diagnostics routine for testing the DACs helped confirm that the DAC/OpAmp combination IC506 DAC1020 and IC513A  5532 shown on the lower part of the schematic are not responding to the digital inputs.    The VCO drive voltage should range from 2VDC to 12VDC.  I am seeing a constant 1.2 DCV.

I am reading 0.7V DC across pins 2,3 of the Op Amp, which points to that being the failed part.  I am going to look for a suitable op amp in my parts bin.  That is the simplest, cheapest, fastest way forward right now. 

I have considered using the SSA as a sig gen but my main effort has been towards the digital and low freq (<10MHz) circuits.  I haven't even measured the outputs of the 4x VCOs to see if they are working.

The main reason I have been working on fixing the Wavetek 2520a is to gently exercise the brain as part of my recovery from a major road accident in Feb 21 that nearly killed me.    That is also why it has taken so long to get to where I am fixing the 2520a.   Doing diagnostics on some old test gear has been therapeutic.   

Dazz

[dazz1]

Hi
Achieved a major step forward today.
I replaced a suspect op-amp that is between a 12bit DAC and the VCOs.  The DAC sets the coarse control frequency in the main control loop.  The op-amp outputs the VCO frequency control voltage.
Another DDS loop sets the fine frequency.

I also replaced a number of tant caps.  The same type and voltage that have already failed.  That should improve future reliability.

After replacing the op-amp, the AutoCal routine No.19 for tuning the VCO frequencies passed.  This proves that the Sig Gen is functioning. This does not eliminate a fault somewhere degrading the RF output.

Next step is to do performance testing and calibration to prove full operation.

Dazz

[tautech]

Cool Dazz, your therapy is working ! 

[dazz1]

Cool Dazz, your therapy is working ! 

Hi
I know I am not operating within spec, but the Wavetek 2520a just might be.

Dazz

[dazz1]

Hi
I connected up the Spec Analyzer and got a good output signal during AutoCal.  I left the 2520a to run for a couple of hours to stabilize and ran AutoCal again.
While I was running AutoCal, the output RF went to zero and the AutoCal routine failed.

I have another fault but at least I know that it was working.  The symptoms are identical to the previous fault so there may be a deeper issue.
First thing I will look at is the power rails.  It is possible that one of the many remaining Tants has released their magic smoke.

Dazz

[dazz1]

Hi
I removed and isolated the main board.  That takes about half an hour.   I applied power and the positive input power is shorted.  None of the +/-15V voltage regulators or caps are warm.  It is most likely a shorted cap on the input side of the voltage regulators.  I have already replaced 4 but there are plenty more to fail.

I have decided to go through and replace all tant caps on the +/-18 volt supply rail.    This includes replacing the no-name-brand new tant caps I sourced from the local hobby electronics shop. 
You may be wondering why I didn't do this from the start.

When I started, I didn't know if the 2520a was fixable.  It is old enough to include obsolete irreplaceable parts.  The eprom code and LCD displays cannot be replaced.
Now that I have had the 2520a running for at least a couple of hours before faulting again, I know that it is fixable. 

I also plan to replace the battery backed ram.  At 32 years of age, it should not still be working.
I am trying to find someone somewhere who can burn the eproms.   That process would give me a backup of the eprom contents. 

It is going to take me a while to get the parts and complete the install so this topic will probably be silent for a while.

Dazz

[TheDefpom]

How many EPROM's are there? I could only see one from the photos and diagrams you posted, and that was a 27C64.

[dazz1]

How many EPROM's are there? I could only see one from the photos and diagrams you posted, and that was a 27C64.

Hi
There are five.  1x 64 and 4x 128.
I have not been able to find an accessible eprom programmer.  Even if I did, I suspect find a PC that can run the code would be a problem.

Dazz

[tautech]

How many EPROM's are there? I could only see one from the photos and diagrams you posted, and that was a 27C64.

Hi
There are five.  1x 64 and 4x 128.
I have not been able to find an accessible eprom programmer.  Even if I did, I suspect find a PC that can run the code would be a problem.

Dazz

Do you have a W10 PC ?
If so Defpom and I can make it happen.

[dazz1]

Hi
W 10PC = Windows 10 PC ??   Yes I have several. 
Code that I wrote for 8086 PCs in the 1980's and 1990's  won't run on Win10 without some retro conversion software.

Dazz

[tautech]

Yep Windows 10.
Sent you an email you might like to study. 

[dazz1]

Hi
I found the next fault.  Another shorted Tant cap.
The Wavetek seemingly has hundreds of them (only slight exaggeration).   I narrowed down to the main board (there are about 10pcbs)
I used the finger test to conclude that none of the voltage regulators were supplying a dead short.
This left just 5 candidates, one had already failed and been replaced.
After lifting the 3rd cap, I found the faulty one.

All of the failed caps have been on the +18V supply. I have now replaced all of the Tant caps on the +18V DC supply.  These were all rated 25V.  The new caps are 35V.

The power supply includes a soft start feature.  I am going to see if this is working properly.    If it is I will look at modifying it to make it a little softer to reduce the surge current seen by the Tant Caps.  Surge current is a known killer of Tant caps.

The 2520A has just passed the autocal process.  This is a good indicator that the 2520A is functional. 

The next steps, in no particular order, are to:

• make a copy of the eproms holding the firmware.
  
  erase and re-burn the eproms
  
  replace the NVRam with a DS1225AD modern version
  
  investigate the soft start, and modify if required
  
  do a full calibration

If there is any interest, I am thinking of doing a video on the repair.

Dazz

[wn1fju]

Almost sounds like Wavetek got a marginal batch of tantalums when they assembled your unit many years ago.  I've had a few (non-Wavetek) pieces where there were an excessive number of bad tantalums, all of which suspiciously were the same uF value.  It was much more than an isolated bad tantalum that one might expect.  Who knows?

As I said in my earlier post, I had a big problem with the connector between the main transformer and the rest of the power supply.  There clearly was enough arcing going on to burn the contacts.  You might inspect your unit just to be sure. 

So far, I've not had any tantalum failures in my 2510A.  But then again, I've probably only turned it on five times in the last year.

[dazz1]

Hi
I suspect that 18V was just enough to kill the 32 year old Tant caps.   The same make/value Tant caps on 15V have survived.    The only difference is +3V and the 7815 regulators in the path of the surge current.  The same tant caps on -18V are OK.
There are 2x +18V supplies, and the failed Tants are connected to the same +18V supply.    I am going to test the hypothesis that the soft start is only fully active on one of the +18V supplies.

Dazz

[HandWave25]

I am going balder by the minute with a Wavetek 2505 (.1-550MHz) Sig Gen. I am guessing it is similar to the 200MHz one. I have a "tripped" text on the "Level" display and I get no RF out. Can it be reset somehow? or where do I look to fix? Can't find a manual anywhere on-line. :-(      rodg

[dazz1]

I am going balder by the minute with a Wavetek 2505 (.1-550MHz) Sig Gen. I am guessing it is similar to the 200MHz one. I have a "tripped" text on the "Level" display and I get no RF out. Can it be reset somehow? or where do I look to fix? Can't find a manual anywhere on-line. :-(      rodg

Hi
I don't have any info on a 2505 but if the circuit is similar, then the attached may help.
The reverse power relay should be reset by cycling the power.

Wavetek give away an on-line book with all the Operator manuals   https://www.wavetek.org/.    Probably only of some value to you. 

 
Dazz

[dazz1]

...    I am going to test the hypothesis that the soft start is only fully active on one of the +18V supplies.

Dazz

Hi
Testing shows that both soft start circuits are working properly.  The RC time constant is 39ms, which is slow enough to make any surge current more like a crawl in electrical terms.
I have looked at the Wavetek 2520a output on the spectrum analyser and an initial look indicates the 2520a is functional.

Dazz

[tautech]

Cool.
All that remains is to backup the EPROMs then so will see you next weekend to help with that. 

[dazz1]

Hi
I have been looking at the output of the 2520a to see if it is working OK.  This is really the first time I have used the spectrum analyser connected to the Wavetek.  All of the faults found have been well within the freq range of my ancient oscilloscope.
The output looks like it is well within factory specs.  In all cases looked at so far, it exceeds specs.  In all cases, there is a 10dB attennuator between the 2520a and my spectrum analyser. 

I have found another fault.  The nvram does not retain memory.  The one good thing about this is that the failed nvram has not bricked the 2520a.  It appears that the calibration programs generate everything stored in the nvram.  Without nvram, there is no point in doing a full calibration.  The settings would be lost on power down.

So the next step is to buy new eproms and a new nvram.

Dazz

[dazz1]

Hi
One of the calibration routines steps the control voltage on the 4x VCOs.  The nvram then stores a lookup table.  When the user selects an output frequency, the control system does a look up to get the rough frequency, selects the right VCO and sets the control voltage.  This is open loop control.  The DDS then uses closed loop control to fine tune the output frequency.

The spectrum monitor output shows the No.19 routine stepping through the control voltages to generate the lookup table.

Dazz

[dazz1]

Hi
Here are some outputs showing the harmonics generated by the 2520a at different selected frequency outputs.    They are consistent across different frequencies.

Dazz

[dazz1]

Hi
I found another fault. 
The AM/FM internal modulation isn't working.  I quickly traced the problem to the op-amp pair that converts the square wave digital signal to a sine wave. 

There are two things that helped narrow down the fault.
1. Binary division of the signal chain.
2. Parallel operation of the 400Hz and 1000Hz signal chain.

There is a long chain of components that process the 400Hz and 1000Hz signals that modulate the RF output.  Rather than following the signal path from one end to the other, I started in the middle of both.  This just happened to be at the junction between the digital and analog signals.    If I had found that the digital signal was faulty, I would have jumped half-way toward the input/origin of the modulation signals.    I did find the signals were OK so I jumped half way towards the end of the signal path. Repetition of this binary division along the signal path quickly isolated the fault to IC107 (Fig 4.2), a dual 5534AN low noise op amp.  I have already replaced another of the same op-amp type elsewhere.

I also compared the 400Hz and 1000Hz signal paths because they are the same except for frequency.

The signal path is spread across two circuit diagrams, not including the control signals originating from the 6805.    Following the signals from end to end would have taken ages and been very inefficient.   In contrast, binary division to isolate the fault was quick. 
Comparing the 1000Hz and 400Hz signals confirmed the diagnosis beyond doubt.

Both filter inputs include 1uF tantalum Caps (C171 and C172).  These are DC blockers and I checked these for shorts.  The same component type has failed on the 18V rail, but none have failed on the 15V supply.  They are OK.

I think this brings the fault count to about 6. 

I am aware that AM/FM modulation isn't used much these days but it would annoy me if I didn't fix something that I can fix so easily.  The Wavetek 2520A includes an external modulation input which opens the possibility of other modulation techniques being applied to the RF output.

Dazz

[tautech]

Nice.
You wouldn't expect problems from tants in a coupling role whereas across a rail decoupling is another matter.

[dazz1]

Nice.
You wouldn't expect problems from tants in a coupling role whereas across a rail decoupling is another matter.

Hi

But it only took a few seconds to check that they were OK.  If one was faulty and not checked, I would have dived down a rabbit hole.

Dazz

[dazz1]

Hi
Although I found a fault, it wasn't the right one.  Attached are images of the 400Hz and 1000Hz modulation signals from different dual op-amps.   Both signal paths show a DC shift but the 1kHz signal looks better than the 400Hz.  I thought the problem was the dual opamp in the 400Hz signal path but this turned out to be wrong.   

Although the 1kHz signal was there, I did not see any AM modulation on the spectrum analyzer when I had it connected up.  Although I found a fault, I looked further down the signal path to see if there was another one.  I looked halfway along the down stream side of the signal path and did not see a signal.    It didn't take long to figure out that all of the analog devices were not working correctly. 

It turns out that by choosing the meeting point  of the digital and the analog signal I had landed right on top of the first symptom of a fault. What I failed to do was re-check the +/-15V voltage regulator output to the analog devices.  When I did check, I found that the +/-18V supply (that I had just fixed again) was all good, but the +15V regulator output was about 0.1V.     There a few 15V regulators spread across the PCBs so the 2520a mostly works OK.  It is just that the modulation function isn't working.

 It is almost certain that I have at least one shorted Tant cap on the +15V supply.    That is a first for this 2520A.  It is not that I have never checked the 15V supplies.    I have previously checked  the 15V regulators and they were all cold with good outputs.  Now one of the +15V regulator is running hot.  This is a fault that has developed since I started fault finding.

Binary division doesn't work for finding one faulty cap in a whole lot of parallel connected caps.    It looks like several designers have been responsible for the main board.  Each designer has added 15V and 5V regulators to power their circuits.  This reduces the number of capacitors I will need to lift and test to find the faulty one.


Dazz

[tautech]

See you Thursday arvo with the rescue package.

[dazz1]

Hi
I identified 5x 1uF Tant Caps connected to the +15V regulator output.   At least one of them had a short.
I went through and destructively checked each cap.  I cut a lead and lifted each cap for testing.
Naturally it was the last cap that I tested that had the short.  This positively confirmed I had identified the faulty component.

It takes me not very much time to remove the main pcb having done it a few times. 

I am now replacing all 5 tant caps.  If there are no further faults, the 2520A should be fully operational again. 

Dazz

[tautech]

I've seen Defpom replace caps only soldering the top side pad but providing sufficient dwell time to let the solder wick through to the pad on the other side so to not need to remove PCB's that might be difficult to remove.

Guess you got my mail explaining I how won't be done until Welly sorts out their Covid situation. 
2 weeks I guess should have it sorted.

[dazz1]

I've seen Defpom replace caps only soldering the top side pad but providing sufficient dwell time to let the solder wick through to the pad on the other side so to not need to remove PCB's that might be difficult to remove.

Guess you got my mail explaining I how won't be done until Welly sorts out their Covid situation. 
2 weeks I guess should have it sorted.

Hi
The leads are bent over on the underside.  I only have a rubbish single shot solder sucker, so clearing holes is a 2 sided operation.   Taking out the main pcb and cutting the leads was going to be easier that trying to do it all from the top. 
I have now got to the point if I find one failed Tant Cap, I replace all on the same power supply.  Another reason why I did destructive testing.    I am trying to avoid replacing every Tant Cap.  They are everywhere.

Testing converted a hypothesis (that a Tant Cap had failed) into a fact. Similar symptoms could have been produced by a failed device.

I'm looking forward to meeting with you but Covid disruptions are now a normal part of life.

Dazz

[dazz1]

Hi
I think I have almost cleared the last fault.  I still have to replace the Dallas nvRAM.   Attached are images from the Siglent spectrum analyser with FM and AM modulation.  I used the internal signal (400Hz or 1000Hz) source, and the external 1000Hz modulation feature.  The external sig-gen is an Elector XR2206 based design.   OK but not a pure sine wave generator. 

The attached screen shots are functional tests only.  I was just looking to see if I have more faults to deal with.

Internally, the 400Hz/1kHz oscillators produce a 10V signal at full scale.
The external modulation signal input only requires 2.5V to match the internal signal.  There are a couple of images showing a 5V input signal, over modulating and over driving the Wavetek.

The modulated spectrum is quite messy as the modulation depth rises but I probably wouldn't use the modulation feature.  If I did, it would probably be good enough.
I haven't looked to see if there is another fault. 

The envelope shape for FM might be distorted because the internal/external modulation oscillators aren't great.  I wouldn't want to dive too deeply into finding a fault unless I can get access to a low distortion audio sig-gen.

Dazz

[dazz1]

Hi
Ran the Wavetek 2520a and connected Siglent SSA all day to do a burn in test to flush out any more defects on the 2520a, and to stabilize the crystal osc on both instruments.
The 2520A is fitted with a 10MHz OCXO. 

The tolerance for the frequency is +/-2.5ppm.    This translates to +/-25Hz for the 10MHz reference.

The measured relative difference between the 2520A and the Siglent is 12.6Hz, but I don't know what the Siglent reference error is.    What I can say is that the OCXO is likely to be within tolerance, and if it isn't, there is a calibration and frequency adjustment procedure I can follow based on a more precise time reference.

Of course all of this points to falling down Alice's rabbit hole trying to find true time.

So all tests done to date indicate the 2520A of fully functional with all faults cleared except for the failing nvram.  A replacement is on order.

Dazz

[tautech]

Of course all of this points to falling down Alice's rabbit hole trying to find true time.

Come back Dazz ! 

[Zero999]

Wow I didn't know there was such as device capable of going from 0.2MGHz which is IR, through to 2200MGHz, in the hard X-ray region.

[dazz1]

Hi
For the benefit of others that may need to repair/restore a WaveTek 2520A, attached is a copy of all the eproms of my one.
These are version 5.1 of the firmware. 
The file names include the chip number on the PCB and the component id.

There is one eprom that stores the values of sine waves, for the frequency synthesis.
There is one eprom that is installed on the GPIB board, if this option is installed.  It probably is.
There are 3x eproms that hold the firmware on the board with the 6805 processor. 

The files are in intel hex format.

I have copied the eproms onto Atmel TMS27C256-R-70PU one time programmable devices.  They are almost pin for pin compatible.
I replicated the firmware from the TMS27C128 devices because pin 27 is address A14 or PGM depending on the device.

For the TS27C64 replacement by the TMS27C256, I soldered a link between pins 26 and 27 directly on the memory device.  This meant I could fit either the original or the replacement device without modifying the PCB.  It should be possible to link these pins on the pcb and still swap because on the 64kb eprom, pin 26 is NC.

I took a great deal of care to avoid risk of static discharge wrecking the chips.  Now I have a copy of the original eproms, the plan is to replace them with the otp eproms.  If I have a problem with those, I can revert to the original eproms. 

Thanks to Tautech for organizing the eprom programmer loan from The Def Pom.  They used to be so common but are now surprisingly rare.

I have programmed the new otp eproms, but I have not yet installed them.  The 2520A has developed a new fault.  Prior to removing the original eproms to copy them, I switched on the 2520A to check that it was OK.  This showed the LCD display flashing randomly.  The LCD display segments are driven by a number of driver chips, so whatever is causing the problem affects them all at the same time.  This points to a bad joint or an unstable power supply.  I need to investigate.

Doing a power-on check proved to be a really good idea, because I know that the fault was not caused by removing, copying and replacing the eproms.  The symptoms did not change.  If I had corrupted the firmware while copying, the 2520A would be bricked.  The amount of effort required to reverse engineer the firmware to find and fix the corrupted code would have been too much effort to justify.

I suspect this 2520A has been sitting around for years and things have deteriorated.
Applying power has revealed a cluster of defects.


Dazz

[dazz1]

Wow I didn't know there was such as device capable of going from 0.2MGHz which is IR, through to 2200MGHz, in the hard X-ray region.

Do you know how long it took me to edit my posts    

Dazz

[dazz1]

Hi
I did some fault finding on the flickering LCD display.    As usual, I started in the middle of the circuits that could cause the problem.  This just happened to be a single row idc interconnect flat ribbon cable between the main cpu board and the lcd display board.
I noticed that the first and last pins weren't there.    They looked like they had been cut off at the factory.  I remembered that one pin had always been missing but this is the first time I had noticed 2 pins missing.  Some of the connectors on the 2520A have pins cut and sockets blocked as a crude way of keying connectors.

After a further study of the circuit diagrams, I concluded that one pin was NC, and the other was an address line for a serial to parallel device that controlled the LCD display drivers.  A missing pin would leave the address line floating, consistent with the flickering LCD display symptoms I had seen.  I think I must have broken at least one of the pins while isolating boards to find other faults. 

Attached is part of the circuit diagram for the CPU board, alongside the diagram of the LCD display board.    The blue lines show the flat cable connections.  The red lines show signals disconnected by the broken pins.    Clearly these diagrams have been drawn by different designers.   The single row connector wasn't a great choice because it is relatively fragile, without a restraint, and non-polarized.    The photos show the problem.

By starting in the middle of the signal path between the CPU and the LCD, I happened to land directly on the fault.  I hadn't even checked the power supplies (normally step 1).    To fix this fault properly, I need to find a replacement flat cable, or replace the single row PCB sockets and cable with something different. 

Once I find and fit a replacement interconnect cable to get the 2520A fully functional again, then I will fit the newly programmed otp eproms.  Programmed thanks to the help of tautech and the def pom.

Assuming the otp eproms are correctly programmed, I will then replace the nvram with a modern version.

There is no connector that would allow external access to the nvram.  This makes it unlikely that the nvram has any factory data that cannot be generated by the calibration process in the manual.  I anticipate that I will not loose anything by replacing the nvram.
After the nvram is replaced, I will aim to fully calibrate the 2520A.  Initial testing indicated it should be within the original specs. 

I have lost count of the number of faults I have found and fixed on the 2520A but I hope they have all revealed themselves.


Dazz

[dazz1]

Hi
I have been looking for a replacement cable assembly to replace the one I broke,  This has turned out to be much harder that it should be.  The original cable has male connectors, but I could not find an available replacement.

I could have just made up a custom cable assembly with some wire and Du Pont style connectors and that would work.    Du Pont type connectors are used elsewhere in the 2520A so they would not look out of place if I used them.

I like to keep things looking factory and original.  I found this Santek cable assembly : https://nz.element14.com/samtec/idss-16-d-05-00-g/cable-assy-16p-idc-rcpt-rcpt-127mm/dp/3551322?ost=3551322
which has female connectors to mate with a pinned header.   So the plan is to replace the sockets on the PCBs with male headers to mate with the Santek cable assembly.

That will take a while because I need to order at least $50 of parts to avoid shipping costs.  Once I get the 2520A fully functioning again, I will continue with repairing the known outstanding defects and replace the eproms with the programmed otp eproms..

Dazz

[TheDefpom]

It looks like it is only using every 2nd conductor of that ribbon cable, so my suggestion would be to fit an IDC connector and use a long pin header to plug the two together, and cover/plug the unused holes in the ribbon plug.

OR

Swap the PCB's IDC socket to be a pin header, and fit a IDC plug on the ribbon as I mention above, or try to find one similar to the one you found, the one you found is a 0.1" ribbon spacing, so it may not clamp on the ribbon due to the extra conductors in between the pins.

[dazz1]

Hi Scot

Thanks for your suggestions.  I am basically going for the 2nd option.  It may not be clear from the e14 catalog but the Santek part is a complete cable assembly.  I just need to swap the female connectors on the pcb for headers. 
That is a good thing because the factory ribbon cables and connectors are rubbish.  They should have used standard 2 row, 16 pin keyed and locked connectors, like they have used elsewhere in the 2520A.

That I think is the cheapest and easiest solution.

Dazz

[dazz1]

Hi
I received a complete IDC cable assembly sample from Samtec. It was exactly what I needed but the cable ends were fitted with female sockets, for headers. 
The factory fitted female sockets were on the PCB.

I removed the sockets on the PCBs and replaced them with headers.  A much better arrangement than the factory version.  The headers and the new cable assembly looks factory original, but better.
I had to take the display/key panel apart to get to that socket on the PCB.  I fitted the new header on the other side of the PCB to make future access better.  No need to disassemble the front display/key panel.

Once I confirmed that the display was now working again, I started with testing the otp eproms.  This took quite a while because I swapped one eprom at a time, then ran a full set of diagnostics tests.  They all worked perfectly, so I have left them fitted in place.  Now that I have file copies of the eproms, I don't need the originals.   

The only fault left now is the nvram.   The battery isn't a battery anymore.  I need a proper solder sucker work station to remove the original nvram.  Once that is replaced, I will then be able to do a calibration, if I can get access to suitable test equipment.

So I seem to be approaching the end of fault finding/fixing on this old piece of test equipment.   I suspect it was sitting in storage for a long time, and accumulated faults while there.  Despite its age, it is still a useful  item to have.  Maybe it will still be going in another 30 or more years. 

The first photo shows where the PCB mounted socket was.  To unplug the connector required taking the front panel apart.   Now a new header is soldered on the other side of the board.

The second photo shows the front panel removed from the chassis and opened up to reveal the LCDs.  Under the LCDs is an electroluminescent panel that provides a feeble back light.  In 1989 that was high tech stuff. The LCD and inside glass panel were quite grubby so this was a good opportunity to clean them.

[dazz1]

Some more photos

The 1st photo is the new cable assembly in-place.  It is and looks better than the factory original.
The 2nd photo shows 4 of the 5 otp eproms installed.

All of the otp eproms are 256k compared to the original 128k.  The only thing needed to make these work was to fill up the 256k devices with duplicate copies of what was on the 128k device. 

The sine wave table for the synthesizer was stored on a 64k eprom.  For this, 4 copies of the eprom fitted in the 256k device. I also soldered a bridge wire directly across pins 26 and 27 of the 256k otp eproms. No changes required to the pcb.

[dazz1]

Hi
Just a couple of photos.

The wire link soldered directly to the 26-27 pins of the 256k otp eprom (follow arrow) to make it plug-in compatible with the 64k eprom it replaced.

Confirmation that all the otp eproms are working.  There are 3 ROMs with code in them.  They switched in/out depending on the mode.  It is possible that the 2520A could have passed all its self-tests in calibration mode, while the operating ROM was defective.
Note that the 2520A passed all the eprom (parity ??) tests so that possibility should be impossible. 

[dazz1]

Hi
I replaced the Dallas battery backed nvram today with a modern equivalent.  When the 2520A was powered up, it displayed error messages. Not surprising.
I went through the diagnostics routines which includes loading default settings.
Everything went swimmingly except for one small detail.  The "Options" self test failed.


I suspect that the nvram still retains some memories, which includes the options are installed on this particular 2520A.  When I swapped the nvram, the memory of those options remained on the old nvram. 
Given that there is no connector that would allow an external device to read/write to the nvram, and given the vintage of the 2520A, I suspect that the front panel keys are the only way to write to the nvram. 
If that is true, then it is likely that the ROMs include the routines necessary to make the options usable.   

My theory is that there is a secret command entered on the key pad that enables specific options.  It is probably a variation of the key commands to run the diagnostics and calibration routines.   Testing that theory probably requires disassembly of the firmware code.  I think that would be challenging to say the least.  Additionally, I don't know what options are now missing.  There are factory stickers on a couple of option keys.  One is for external phase modulation.  The other is to specify units of rad.  It is unlikely I would use either of those options, which I think are related and part of the same option.   

It would be nice to restore the 2520A to its full operation.


EDIT

I found a description on the Phase  Modulation feature in the Operator Manual, now attached.

[khs]

Here a 2520-SP782 with up to 2.7 GHz. All looks similar.
But it looks they have made some modifications regarding the RF attenuator...

It looks the schematic of the GPIB interface is a big secret, so there is no way to add the missing GPIB interface...

[dazz1]

Hi
Definitely different to mine.  Your firmware is ver 8.3, mine is 5.1 but the dates on the boards look older than mine.   Maybe yours was factory upgraded??   I wonder if any two of the 2520A are the same?

I have considered the possibility of modifying mine to have a higher frequency limit but reverse engineering the firmware would be just too much work.    Wifi would be within the 2.7GHz range but I don't know if the phase modulation feature I have (currently not enabled) would be useful.   It would require a computer ahead of the sig-gen to generate the base band signal, and a receiver.

It wouldn't be too difficult to reverse engineer the GPIB board because it is only 2 layered and I have posted a copy of the GPIB  firmware but figuring out how enable that option in the main firmware would take time and effort. 

[dazz1]

The problem I am facing right now is how to enable the Phase Modulation option.  Replacing the evram has disabled this feature, so obviously something stored on the factory evram was the magic key.
I have a number of options to restore the lost feature.  The most complex would be to reverse engineer the firmware and find out what test is done to enable/disable the features.
The simplest is to load an image of the factory evram contents onto the new evram. 

DefPom has made the later option viable because he was able to read the evram contents and send me the file.   Thank you DefPom.   
Remarkably, it appears that the 32 year old battery nvram, was still working. So simply writing the contents of the factory nvram to the new nvram should restore the lost Phase Modulation option.
I can then calibrate the 2520A and fully restore it to factory condition.

To do that I need a parallel ram programmer.  Not something I can buy cheaply anywhere.
So the plan is to make a ram programmer based on an Arduino Uno.  The Uno has everything I need except for a spare serial port to read the file from my PC.
I will need to use the one serial port for software development.   

The easiest way to read the file into the Uno is with a cheap Chinese micro SD card reader, which I have just ordered.    That will probably take around 4 to 12 weeks to deliver given the current shipping delays.  So I expect it will be a while before I make another post on this thread.

If copying the factory evram fails to restore the Phase Modulation feature, then reverse engineering is the only option left.    This shouldn't be as hard as it could be.  When the 2520A is switched on, it goes through self test routines.    That will tell me what test fails when the firmware is checking for added options.  There is also a key switch that selects a diagnostics rom.  You enter a two digit number and run a selected routine.  I suspect the lookup table for selecting the routine will include undocumented numbers for running routines such as enabling options. 

[khs]

Hi
Definitely different to mine.  Your firmware is ver 8.3, mine is 5.1 but the dates on the boards look older than mine.   Maybe yours was factory upgraded??   I wonder if any two of the 2520A are the same?

I have considered the possibility of modifying mine to have a higher frequency limit but reverse engineering the firmware would be just too much work.    Wifi would be within the 2.7GHz range but I don't know if the phase modulation feature I have (currently not enabled) would be useful.   It would require a computer ahead of the sig-gen to generate the base band signal, and a receiver.

It wouldn't be too difficult to reverse engineer the GPIB board because it is only 2 layered and I have posted a copy of the GPIB  firmware but figuring out how enable that option in the main firmware would take time and effort.

Here the FW 8.3 of my instrument.

Maybe it enables 2.7 GHz, because it looks the only difference is the RF attenuator control.

It can be the standard hardware was able to deliver 2.7 GHz but the standard RF attenuator was not working fine up to 2.7 GHz, so there was only the need to modify the frequency limits of the firmware and the RF attenuator. The additional pc-board with some flying wires looks a little bit quick and dirty so for me it's an additional hint there are minor modifications only..

I'm quite sure it's a factory update due to the modified frontpanel.

The  RF attenuator relays are controlled by the output pulses of the 74LS123 monoflops, so it looks the relays for the 2.7 GHz are a little bit more delicate compared to the standard relays.
 
You may try the firmware it after making a copy of the NVRAM, but you shoud expect your RF attenuator it not the best at 2.7 GHz .

If there is some time left it would be nice to have a picture of both sides of the GPIB board..

[khs]

The problem I am facing right now is how to enable the Phase Modulation option.  Replacing the evram has disabled this feature, so obviously something stored on the factory evram was the magic key.
I have a number of options to restore the lost feature.  The most complex would be to reverse engineer the firmware and find out what test is done to enable/disable the features.
The simplest is to load an image of the factory evram contents onto the new evram. 

DefPom has made the later option viable because he was able to read the evram contents and send me the file.   Thank you DefPom.   
Remarkably, it appears that the 32 year old battery nvram, was still working. So simply writing the contents of the factory nvram to the new nvram should restore the lost Phase Modulation option.
I can then calibrate the 2520A and fully restore it to factory condition.

To do that I need a parallel ram programmer.  Not something I can buy cheaply anywhere.
So the plan is to make a ram programmer based on an Arduino Uno.  The Uno has everything I need except for a spare serial port to read the file from my PC.
I will need to use the one serial port for software development.   

The easiest way to read the file into the Uno is with a cheap Chinese micro SD card reader, which I have just ordered.    That will probably take around 4 to 12 weeks to deliver given the current shipping delays.  So I expect it will be a while before I make another post on this thread.

If copying the factory evram fails to restore the Phase Modulation feature, then reverse engineering is the only option left.    This shouldn't be as hard as it could be.  When the 2520A is switched on, it goes through self test routines.    That will tell me what test fails when the firmware is checking for added options.  There is also a key switch that selects a diagnostics rom.  You enter a two digit number and run a selected routine.  I suspect the lookup table for selecting the routine will include undocumented numbers for running routines such as enabling options.

The aduino uno has 32KB Flash memory (ATmega328P), so maybe you can insert the NVRAM values as character (hex) array into your program and then write it into the NVRAM...

[TheDefpom]

I would suggest using the ESP32, it has a lot more memory and a lot more functionality than a basic arduino module.

[dazz1]

Hi
I chose the Arduino Uno because it interfaces directly with 5V logic and ram and,
because I have one already and,
the SD interface pcb is less than $1.

So the Uno is far from the best choice but it will do the job with minimal extra hardware. 
Apart from the Uno,  and the SD adapter, I will need some serial to parallel ICs.

This will be a single use project so nothing fancy required.  Worst case will be if I reverse engineer the firmware code.

Dazz

[LogicalDave]

Tantalum caps need significant voltage derating; using 25V rated tantalums on an 18V line is a formula for long term failure.
Most manufacturers' rule of thumb for tantalums is to de-rate the voltage 50%, especially for MnO2 cathode tantalums.

It sounds like it is worth staying away from these older sig-gens unless you're ready to re-cap them; thanks for sharing
the detailed info; I was considering buying one and will probably now pass.

Some good white papers
https://sh.kemet.com/Lists/TechnicalArticles/Attachments/25/Derating_differences_Ta_KO_AO.pdf:
https://nepp.nasa.gov/files/24745/2013_n240_Teverovsky_ESTEC_Derating_paper.pdf
https://www.avx.com/docs/techinfo/VoltageDeratingRulesforSolidTantalumandNiobiumCapacitors.pdf

[AndreH]

Hi all
I too have the Wavetek 2510 (.2 - 1100MHz)
I managed to invoke the auto calibration, but I don't understand the selections 10, 12, and especially 18. When I select 18, I receive error on codes LVL, UVU, which I don't quite understand since I don't have the manuals. I can get the instrument to go past its self check startup sequence and it then works fine, but I have to virtually 'kickstart' the instrument to get it to work.

[AndreH]

Hi all
I too have the Wavetek 2510 (.2 - 1100MHz)
I managed to invoke the auto calibration, but I don't understand the selections 10, 12, and especially 18. When I select 18, I receive error on codes LVL, UVU, which I don't quite understand since I don't have the manuals. I can get the instrument to go past its self check startup sequence and it then works fine, but I have to virtually 'kickstart' the instrument to get it to work.

[dazz1]

Hi all
I too have the Wavetek 2510 (.2 - 1100MHz)
I managed to invoke the auto calibration, but I don't understand the selections 10, 12, and especially 18. When I select 18, I receive error on codes LVL, UVU, which I don't quite understand since I don't have the manuals. I can get the instrument to go past its self check startup sequence and it then works fine, but I have to virtually 'kickstart' the instrument to get it to work.

Hi
You have a journey ahead of you.  Attached is the page from the manual explaining the purpose of the diagnostics routines.
Passing self-check at startup means not a lot.

Test 18 displays outputs from control loops.  If the outputs are out of range, you get error codes.  Chances are that you cannot complete a full calibration procedure to get the instrument within spec.  If you have followed this thread, you will know what you face.   

I cannot tell you to go to xxx component and replace that to solve your problems.  I can tell you:

• a operator and service manual are essential (too big to post or email)
• look for failed tantalums first
• take a copy of the nvram asap
• post a copy of the ROM and nvram contents for others that follow you (and I)

Almost everything in these instruments is replaceable, so it is likely that your faults can be fixed.

Good luck.

I am currently waiting for parts I need to build a nvram reader/writer.  I was hoping that the contents of the nvram would be created by the autocal procedure.  I was not quite right.  I suspect there is an undocumented keyboard entry command required to factory enable installed hardware features.    That would have allowed them to use the same nvram board to fit to any 2510 with any features installed.

[AndreH]

 Thanks Dazz1. I noticed the name Artek manuals and I immediately ordered a copy. I received both the OM and the SM and I look forward to resolving my issue.

On switch on the LCD display displays flashing 'gibberish' By accident I noticed when I touch pin 36 on the uP the generator begins ts startup sequence and the display normalizes. It then stops at code 22. After spinning the dial it then stops at code 80. After spinning the dial again it completes the sequence and the unit is ready for use.

Everything seems copacetic. Frequencies and levels are fine. It's just on switch on that I have this issue. Initially there were various frequencies from 200kHz to 1.1GHz that were totally out. As an example, when I dial in 10MHz the output shows 30Mhz etc. After loading the 'factory' setting everything was perfect. I removed the NVRAM and drilled a hole to the positive of the battery. Battery voltage on the NVRAM measures 3.157V. The NVRAM seems fine. I soldered a wire to this positive pin and will monitor this voltage over time.

Anyway, I'll see if I can determine the problem.

Best regards

Andre
South Africa

[dazz1]

Hi
I replaced my nvram with a modern equivalent, but I didn't pre-load it with the contents of the old nvram.  As a result, I lost a feature.

I suggest to remove the nvram and copy its contents, so when it fails, you can program its replacement.

The firmware on the ROMs is the one thing that definitely can't be 2nd sourced.  Closely behind that is specific contents of the nvram not regenerated by the diagnostics routines. 

[dazz1]

Hi
Progress on the repair of the Wavetek 2520 has stalled for months while I waited for hardware from China. 
The hardware and software were copied directly from https://sites.google.com/site/gogleoops/mega-nvram-prog.
The only tweak I made was to use a 28pin socket rather than a 40pin socket. 

I could have purchased a general purpose programmer, but I would never use it again.  I elected to make a single purpose nvRAM programmer based on an Arduino Mega board.  I doubt the nvRAM programmer will never be used again, but I expect the Arduino will find a useful place to be.

I also fitted an IC socket to the 6502 micro-processor board so I can remove/reprogram/replace the nvRAM if needed. 

I have successfully programmed the nvRAM but not yet installed and tried it.    I used an image from the original factory fitted nvRAM to copy across to the new DS1225 nvRAM.  The old factory fitted nvRAM battery was well past its expected life, so it is entirely possible that the original factory image is corrupted.    Fortunately the nvRAM records configuration and calibration data.  The firmware is now located in OTP EPROM to replace the factory fitted EPROM (UV erasable).

[dazz1]

Hi
First power-on test indicates that the repairs, including the programming of the nvram, are 100% successful.  I need to run full diagnostics and calibration to confirm, but it is looking good so far.

[dazz1]

Hi
OK so I still have at least one more fault.
I was able to get through all of the self-tests OK.  No errors.
I was able to calibrate the oscillators and output level control.

Further testing showed AM and FM modulation is not working.  It was, but not anymore.
The FM modulation sort of works, but there is a step change at a certain modulation frequency.
Enabling AM modulation does not produce side bands, but it does change the output level. 

The LCD display is showing what I would expect to see.   That indicates that the firmware is OK.  I haven't created a fault with the nvRAM.
The manual has a diagnostics section for modulation, so I will work through that.   


So, do I expect to need or use FM/AM modulation?  NO. 
Do I need to fix it for any other reason?  NO.
Am I going to fix it?? YES of course. 
It would just annoy me to know that it is broken when I know I can fix it. 

The block diagram, with annotated modulation signal flow, is attached.

[dazz1]

Hi
OK did some some fault finding today.  With the assistance of my trusted hairy assistant, I identified the problem as zero volts on the -15VDC supply to the circuits for AM/FM modulation.  The good news is that finding this fault shows the programmed nvRAM is working as it should.   

Previous experience shows this is most likely due to failed Tant caps.   When I went looking, my trusty assistant pointed out the toasty brown one in the picture.  Well at least I think he did.  Unfortunately, inspection of the circuit diagram showed that the toasted Tan cap was on the +15VDC supply. 

The time taken to lift and test Tant caps is about the same time as replacing it.  I have just gone out and purchased about 20 Tan Caps to replace the toasted +15V and all the -15V caps. 

Some of you might be thinking that I should go through and replace every Tant Cap.  I have considered that but there are well over 100 of them on the boards.  I have decided to replace them one supply at a time.  If one fails, I just replace all the Tant Caps on that supply.   As I learn more about this piece of kit, I can isolate faults relatively quickly.  I have long lost count of how many faults I have fixed, but it is a lot.   It is likely that as I replace more failed Tant caps, I will get to the point where there aren't many more to replace. 

As an aside, I now think it most likely that the options are coded into ROM, and not enabled by some secret value loaded into nvRAM.  This could be confirmed by looking at the 6502 code that handles front panel button presses.  If there is a secret combination of button presses, they will be found there.  This means investigating or modifying the Wavetek would require reverse engineering the 6502 code.  Something I will not be doing.

[dazz1]

Hi
I replaced all of the Tant caps on the -15V supply plus the toasted one on the +15V supply.  I didn't bother testing the removed caps, I just replaced them.  After recapping the -15V supply was restored to normal.  So that fixed at least two more faulty components.

Now I am doing functional checks.  The internal AM/FM modulation works fine.  I just had to adjust the modulated signal amplitude.

I connected an external AF sig gen to the modulation input and that doesn't work.  The signal disappears in the circuit somewhere.  Initial testing indicates the problem may be digital.  A DAC is used to adjust the modulation level from 0 to 100%.  This appears to be stuck on zero so the DAC blocks the modulation signal.   The DAC is driven by a register with all output pins = low, so the fault is probably not the DAC.   I need to do some more testing to identify the root cause.

[dazz1]

Hi
I did some functional testing today.  The aim being to check that everything is working.
All testing shown here was at 500MHz with a 0dBm output from the Wavetek 2520.
The input of the spectrum analyzer was fitted with a Minicircuits  10dB attenuator  to ensure a 50ohm load for the Wavetek output.   This will have the effect of raising the RF noise floor seen on the spectrum analyser.

The first image is a 500MHz CW signal looking from 0Hz to 2.1GHz
There is one chirp above the noise floor, just to the left of the 500MHz signal.
 

Zooming into the main signal at 500MHz with a 10kHz span.   


Looking more closely at the main output at 500MHz with a 1kHz span.
The first peaks either side of the main output will be offset by the closed loop frequencies for the frequency control circuits. 


The images are supposed to be inline expandable thumbnails, but they aren't.

[dazz1]

Hi
Here is a the Wavetek 2520 output with internal AM modulation active.
In theory, there should only two frequencies, one either side of the main output. 
The spectrum is messy but still useable.

Note that the Wavetek and spectrum analyser don't agree on the frequency by 640Hz.  I don't yet know which one is the most correct. 

[dazz1]

Hi
The FM internal modulation produces a nice clean output when the Wavetek is setup for wideband FM modulation.

[dazz1]

Hi
I have proved that the internal FM/AM modulation functions work as intended.
I still need to figure out why the external FM/AM modulation isn't working.

[dazz1]

Hi
As I feared, the root cause of the problem with the non-functional external FM/AM modulation traced back to the firmware.
It turns out that the order of going from one modulation to another is important.  The attached drawing shows the permitted sequential steps in green, and the ones that don't work in red.
I can go from Internal AM to Internal FM, but I can't go diagnonally from Internal AM to External FM.    So all the hardware symptoms I saw were due to a firmware "feature".

The Wavetek also has an option button for external phase modulation, but this button does nothing I can see and this oiption does not show up in the self test.  There is a test that shows the options installed.  Also the manual indicates phase modulation requires an additional PCB module, which is not present.

The front panel has a covered hole for an option for pulse modulation of the RF output.  It also needs added circuitry.  I definitely don't have this option fitted.

So the plan is to soak test the Wavetek 2520 tomorrow to try and flush out any latent defects before I put the covers back on.   Then over the coming weeks, I will try and get access to enough equipment to do a basic calibration, focusing on  crystal frequency reference and output level. 

I think I started this repair about a year ago.  I never expected to find so many faults in one piece of equipment.  It has been a challenge but I learned electronics at around the period this was designed and built, so it has a familiarity to it.  It has definitely been hand built.  The circuit boards are dated and signed by real people.  No stick-on labels here.  Unlike old HP gear, it looks like Wavetek didn't quite finish design and development before it went into production.    The optional modules are solder-in not plug-in, and probably needed to be factory fitted, including new EPROMs with updated/modified firmware.

[tautech]

I think I started this repair about a year ago. 

You did, maybe a little over a year and July last was when I came to Wellywood and stopped by with Defpoms programmer.
Never did I imagine it would take you this long to get to where you are however life has this habit in getting in the way of projects don't it !
Keep at it Daz !

[dazz1]

Hi
Last night, I ran the built-in calibration. This sets the range of each of the voltage controlled oscillators.   This calibration means the Wavetek can select an oscillator, then apply a frequency control voltage that is close to the required control voltage.  Any remaining error is eliminated by the closed loop frequency control system.

There is another test where the Wavetek RF output level is checked at 60 points across the frequency range. I used my Siglent SA to check the levels.    At each point, the RF level is adjusted to a 0dBm output to a resolution of 0.1dBm.  Between points the Wavetek interpolates the output level.  As a result, the output level should be stable across the entire frequency range in a way that would be difficult to achieve with traditional analogue control methods.

Today I left the Wavetek running as a short soak test to try to flush out any latent defects.
The test was carefully supervised by His Majesty, the Right Royal and Honourable Sir Fluffy who showed true dedication and stayed the entire day.

I am now ready to put the covers back on.

The final step is to see if I can get access to equipment that I can use to verify the output level across the frequency range, and to adjust the reference oscillator.

[dazz1]

Hi
So the Wavetek 2520 project to complete all the repairs and get it back to a fully operational state ended today.
RF Test Solutions at https://www.rftest.co.nz/ kindly helped me do a limited calibration with a range of test equipment I could only dream about.

I was able to tune the 10MHz oven controlled crystal oscillator to within 0.28Hz, which corresponds to a 60Hz error at 2.2GHz. 
I ran the built-in calibration procedure to adjust the output level from 200kH to 2.2GHz to within +/-0.1dB at 0dBm output.
So the Wavetek is probably better than factory, because way back in the 1990's they probably didn't have the quality of test equipment that I had access to today.

So thanks to the two Dave's at RF Test Solutions.  I couldn't really finish off this project without them.    It has been well over a year since I started and now I can move onto something different.  I have already.

Now that I have restored the Wavetek, I just need to find a reason to use it.  At present it is a solution looking for a problem.

[Jeff eelcr]

Looking at your PCB I see a burnt ? near reg IC left side Middle
Jeff
This was in responce to post 7 picture 3.

[dazz1]

Hi
IC121?

[Jeff eelcr]

Glad you got it going nice peice of equipment enjoy it.
Jeff

[Abel]

Hi Dazz1,
I left a PM with you; if you do not get it in your Inbox, please check your Spam, you never know...
Best regards
AB

[dazz1]

Hi Dazz1,
I left a PM with you; if you do not get it in your Inbox, please check your Spam, you never know...
Best regards
AB

PM received.

[dazz1]

Hi
The Wavetek is backlit with a green electroluminescent (EL) panel that gives a nice even glow, but after about 35 years, it is looking a little weak so I started looking at options.  EL panels have a limited life which explains why the Wavetek has a switch on the back to turn the EL on/off.

Eventually I found 50cm x 3cm EL panels on Aliexpress.  I ordered the aqua colour because it is white when off.  It is also very close to the original EL colour.  I ordered it and fitted it.


It is very difficult to show how bright the new EL panel is but if you look carefully, the LCD contrast is relatively low.  Contrast improves with higher angle of view from slightly above.   The EL panel is almost too bright for the LCD, but not to the eye. 
Typical EL panels cannot be cut so I just left the surplus length to illuminate the interior of the enclosure.

Given that this was a one-for-one swap of new for old, this was a very easy mod.

The new EL panel came with an inverter that only produces 67VAC.  The Wavetek inverter provides 90VAC, but the LCD polarizing film reduces the light output.    I also replaced the wiring the the EL panel because this looked suspiciously like copper clad aluminium.    This turns to powder after a while. 

So not only is the WaveTek working, it now looks very cool with its new blue backlight.  As an added bonus, I think the external phase modulation option has started working after trying some GPIB commands.  I haven't tested it yet, but the button puts it into the correct mode.