HP8920A Communications Test Set Repair

[AteMan]

Hi, I am looking for some pointers to repair a HP8920A Communications Test Set.

In summary the test set power up okay and passes the power on self-test.

However after a short period of time the test set reports Self Calibration errors for the Volt Meter (Error 223), Counter (Error 212) and Spectrum Analyser (Error 233).

If I try to run the self test from the Tests menu it reports a time-out for the volt meter, so I assume the others errors are the same.

Does anyone know if these errors could be caused by a configuration issue caused by the internal batteries, as they are bound to be dead?

Any pointers would be appreciated.

[TheSteve]

Sounds like a possible thermal issue. The service guide is reasonably decent for the 8920A. So look at everything that is failing and see what might be in common. It wouldn't hurt to reseat all of the boards as well.

[George Edmonds]

Hi

I have repaired HP8920A's.

First you could well have failing power supply electrolytic capacitors,  a common problem, the power supply is a PITA to remove due to the plastic power switch actuator that runs from the front of the instrument to the rear.  If you fail to remove the power supply correctly you will damage the actuator and possibly the mains swith in the power supply.

Second the cooling fan is in the middle of the instrument and difficult to gain access to.

Can provide advise if required.

George G6HIG

[AteMan]

Many thanks for the replies, I will have another  look tomorrow night and report back.

Trawling the web I managed to find the schematics for the PSU, so I should be able to measure the outputs / test it on the bench.

[George Edmonds]

Hi

The power supply must have a minimum load to operate correctly, the fan is a real pig, need to virtually dismantle instrument to get access.

George G6HIG

[AteMan]

Hi Guys,
            I have now confirmed that the PSU outputs are all good.

I have found the following:

If I manually run self-test, the test-set first reports a Real Time Clock failure, followed by a DMM timeout when trying to measure the reference voltages.

The real time clock appears to be working because the correct date and time is displayed on the service menu and this can be updated.

It appears that the internal 10MHz reference is not working and hence the frequencies derived from it. I don't have another 10Mhz source, so I can't tell if it will work with an external reference.

The investigation continues...

[George Edmonds]

Hi

I still  think that you may have a power supply problem, did you check the ripple/noise on the power supply in addition to the voltage?

It looks like you are in a catch 22 situation that can only be broken with the help of someone with a fair amount of test gear.

Regards

George G6HIG

[Wallace Gasiewicz]

You might want to try the groups.io  there is a hp8920/8935 Family group just for this equipment.
How do you know there is no 10 MHz reference? Did you check it on the BNC output ref or at the board?
I assume that you are unable to put out any signal of any type and you are unable to pick up anything on the spectrum, is this correct?
Possibly check power to the board with the ref oscillator.
George has a point about ripple on the power.
I have an 8935 and there is a config setting for int/ext ref freq. I do not think this is your problem, since you are running a self test.

[George Edmonds]

Hi

The groups.io address is

https://groups.io/g/HP8924-Family-Spectrum-Analyzers

I am a member and contributor

George G6HIG

[AteMan]

Hi,
    Thanks for the link and suggestions, I stumbled across the group the other day by chance.

On Monday I built a serial cable so that I could dump the calibration data using an application developed by one of the members, while I was at it I generated a C# to obtain the data over GPIB as well.

Luckily I can still communicate with the controller over Serial and GPIB. 

Unfortunately when I tried to upload the file to the group,  a message popped up to say that a moderator had to review it before it was posted and hence this is where it stopped.

On the repair front the next step is to measure the PSU ripple as suggested. Hopefully if I get time I can do this tonight and report back.

Will let you know....

[Wallace Gasiewicz]

Remember Redd Fox??
He obviously was a power supply expert.
His drink of choice was "Flapple"  (flat Ripple)

I could not resist

[George Edmonds]

Redd Foxx WHO?

George

[AteMan]

Hi Guys,
             I have some progress to report... Yesterday I manufactured a 9w DSUB (M) to (F) cable with break out so that I could measure the ripple on the supplies to the modules and check out the  serial bus from the controller, as this is used to set the control voltage on the 10MHz Reference Oscillator, which appears not to be working.

For the ripple I measured ~25mV P-P on the +12, -12 and +5V. Is this acceptable?

I confirmed that the clock and data to the Reference module were working (See attached). If I change the reference coarse and fine setting under the service menu I can see the bit pattern change. Apologies for the photo of the trace, the scope only has a GPIB interface and is no where near my PC.

Today I removed the Reference module from its metal case so that I could check that control voltage was actually being set.

It turns out the  12VDC supply +12V(F) to the VCO (U1) is missing, tracing it back I found that L2 was open circuit.

Measuring the supply pin on the oscillator it appears that it has developed an internal short (~34Ohms), which would have blown L2 on application of power.

This explains why the 10Mhz reference is missing and hence all the derived clocks. It also goes some way to explain the Self-Test failures, but I can only confirm this once I have located a replacement oscillator (P/N CTS KNIGHTS 970-4302-1).

Small steps....

[George Edmonds]

Hi

What options does your 8920A have?  There is an optional high stability 10MHz standard.

George

[AteMan]

Hi George,
                The label on the back states options 003, 007 and 102.

Regards

Adrian.

[George Edmonds]

Hi Adrian

The High performance OCXO 10MHz standard is option 001, you have the TXCO version.

The main source of 8920A spares is Amtronix in the USA, they want $450 for a high  performance standard.

Personally I think that there is a good chance that there is nothing wrong with your 10MHz standard, HP's small inductors are notorious for failing.  May I suggest that you power up the TXCO from an external 12v supply and monitor the current drawn.  As far as can be ascertained it should be between 15ma and 20ma.

Good luck

George G6HIG

[Wallace Gasiewicz]

I have had both the high performance and regular TCXO and compared them. I did not find any huge difference.
George's suggestion is a good one, after that there may be a volt regulator on the board that is problematic, or maybe some other fixable component that is in the oscillator circuit.
If the board does not draw too much current, you might look for the 10 MHz signal.
Wally

[AteMan]

Hi Guys,
            I have an update on the 8920 repair.... While waiting for the replacement L2 to be delivered I decided to try and power the 10Mhz reference independently directly from my bench power supply to see how much current it was pulling. On the first couple attempts the power supply when into current limit ~1A on the third attempt the current dropped to a few hundred milliamps and I could see the 10MHz output on the scope.

The good news is that after fitting the new L2, the 8920 now works and passes all self-tests except for the real time clock, which I still need to investigate as it successfully keeps track of both the date and the time when powered off.

I have attached a picture of the spectrum analyser tuned into a local radio station.

It appears that in this particular case the reference oscillator developed a partial internal short, which caused L2 to go open circuit. Applying the bench power supply at a higher current appears to have cleared it so it could have been some type of loose debris (e.g. solder ball) or a tin whisker from the shielding can.

[Vit G]

Hi!
In order not to create a new topic, I will ask here.
In my 8920A, the spectrum analyzer shows the signal level 3.5dB lower than that supplied to the input. The same for Ant and RF IN inputs. In power measurement mode, the reading is 1 dB lower than the input signal.
is this normal for this device or can there be any calibrations? I did not find clear instructions on how to do this in the instructions for calibrating the device. All tests that are offered in the service manual the device passes without errors.
 Need your help!

Photo:  10.5 dB input