Repair: HP 8753C Network Analyzer

[hendorog]

I've been messing around with the correction adjustments trying to clean up the spurs on the source. It had a lot of spurs, the main ones are multiples of 25.5kHz.

To do this I needed to adjust the trimpots along the top edge of the A13 Frac-N board. The adjustment is called "Fractional-N Spur Avoidance and FM Sideband Adjustment" in the Service manual. It's an automated test and requires a heap of HP gear none of which I have.
So instead of that I just cautiously tweaked one of the pots in CW mode with a SA connected, and watched to see what would happen. When I saw the spurs dropping I carried on and did the others.

I also did the Source Spur Avoidance Tracking adjustment, which is simple. Just monitor a trace on the VNA itself while tweaking the obvious pot on the A3 source. Need to connect a test point and GND to the Analog bus using jumper leads. This wasn't that far out of adjustment and I don't think that made any real difference to the source trace that I could see.

And the Cavity Oscillator Frequency Correction Constants, which is used to correct the frequency of the Cavity Oscillator for aging. It involves identifying a particular spur of the oscillator on the screen. This is pretty difficult unless you have an appropriate filter. I don't so intend to make one and repeat this adjustment.

Forgot to save before pix which was dumb. Unfortunately I wasn't able to reduce the 25kHz spurs, but there were other spurs which have pretty much dropped into the noise.

It would be good to know if this is a typical source compared to other 8753 owners. I did find an old post on the Keysight forums of someone observing the 8753A model is typically cleaner than the 8753C model, which often has spurs. There wasn't any info about reducing them however.

Edit: Added images showing the source when unlocked (Source Tune mode). Note there is 27dB 37dB of attenuators in the path in these pics.  When looking at it from a wide span (last image) it appears normal. The narrow span in realtime mode looks pretty bad and I'm wondering if something is modulating it somehow. There is a regular 'glitch' which shows to the left of the fundamental. Any ideas?

[TurboTom]

I just checked with my 8753C and found a very similar spur spectrum. I didn't touch any adjustments on mine so I don't know if this can be improved significantly -- but I doubt it. After checking the 1GHz (btw, mine is slightly low so it may be a good idea to use an external reference if something really accurate needs to be done) for the spurs (first and second screenshot), I decided to check for harmonics, first with a generator amplitude of 0dBm (nothing visible above the SA noise) and then with the generator power increased to 20dBm -- surprise, surprise! But I guess that's simply the price to pay when cranking so much power out of the poor hybrid...

For what it's worth, my VNA had been calibrated the last time in 2003 and wasn't touched since (I broke the seals).

I hope this information may be a little useful 

Cheers,
Thomas

[hendorog]

Great stuff, thanks Thomas!

Any chance you could try it unlocked in Source Tune mode too?

If you can, the key combo is:
[SYSTEM][SERVICE MENU][SERVICE MODES][SRC TUNE]

Then look at it with a narrow span and see if it looks anything like mine.

Cheers,
Roger

[TurboTom]

Roger -

certainly it looks similar like yours (or it would if my SA would feature this kind of intensity/color grading of multiple sweeps)

Actually, this is no surprise since in this test mode, the YTO is not locked but rather free-running, just controlled by the pretune voltage. Hence it drifts and all the noise in the pretune circuitry (since not compensated by the PLL feedback loop) manifests itself as phase / frequency noise. I think this test is simply meant to check if the pretune circuitry puts the YTO somewhere close to the preselected frequency so the lock time for the PLL would be as small as possible. As long as this is approximately correct, all the jutter and drift aren't really relevant. I might be proven wrong, though...

All the best,
Thomas

[hendorog]

Thanks Thomas- yours looks more stable than mine, but at least it does look quite similar and so is probably normal.
My SA has some nice features and some not so nice ones, but it suits my usage quite well I do like the look of that new Siglent though.

Even without the PLL lock I think it could be more stable than this. I'm just not sure if its done by design or not.
I was wondering if the modulation here could be related to the spurs in the PLL locked output.

I tried it on zero span on my SA with FM demodulation and that shows a regular pattern which is interesting. The downward pulse visible in the top trace is the 'glitch'  which is in the earlier pics with the persistence.

[hendorog]

I finally pulled out the CRT and found a wee surprise.

The penny dropped - the bang and the smoke back when it first started up was actually this cap going off.
When that happened the CRT was extremely bright until I turned it off. Also it was difficult to get a linear brightness out of the calibration.

I can get 47uF or 10uF 250V caps locally but not 22uF. So a choice: Bodge it or wait?

Onwards with the bodge then...

[hendorog]

This is pretty cool. I have been wanting to test out the 8753 in 6GHz mode. Even with the 006 option installed the VNA RF source only goes to 3GHz. So to get 6GHz you need an 85047A test set which has a doubler which switches in for 6GHz mode and out for 3GHz mode. These test sets are crazy expensive on eBay, and cost more than what you can get a 8753C for.

So I've been working on writing some PSoC code to respond to the VNA like a test set would. Then I can put a basic test set together using cheap doublers and couplers and use the VNA up to 6GHz. I just got the PSoC working, and using a mixer as a makeshift doubler was able to play with 6GHz mode. Although the mixer didn't suppress the fundamental so it was difficult to reliably get it to lock though. 

Well I discovered that the extra hardware that isn't required, at least for playing around, as there is a POKE which does the same thing.
The A9 jumper needs to be in Alter position for this to work according to the doc.

The address is: 1314412
POKE: 5 for 85047
POKE: 1 for 85046
POKE: 0 for no test set.
Press PRESET afterwards.

One of the pics shows a new menu option which appears allowing the choice between 3 and 6GHz sweep modes.

It may be possible to set this up as a programmed Sequence, so it can be called up at the press of a button.

*** TRAP for young players: Turning on 6GHz mode automatically sets the source power to 20dBm. So don't do this with RF Out connected directly to R or through a splitter without additional attenuation ***

I found this in the 8753D service guide: 08753-90406-d-serv.pdf [8753D Option 011 Network Analyzer Service Guide Manual]
http://www.siliconinvestigations.com/hprep/hp8753/08753-90406-d-serv.pdf

I'll still use the PSoC as I'm planning to make a test set for using up to 6GHz.

Analyzer is back in action again, cap in the display replaced with a 10uF 400V for now. Its not the correct value, but I figured less capacitance was the lesser of the two weevils.

I also replaced the supercap with a larger 1F one as I thought it was faulty. The display colour resets to default after restarting. Turns out that this still happens so the original was fine. The user settings I had saved were all lost as expected.

Lastly there is a pic of measurement of the voltage across the source pins ID- and ID+. This measurement should be within the range printed on the label attached to the source - in my case it is OK as the range is 380mV to 485mV. If this is too high or too low it likely means there is a problem with the source wire bonded microcircuits - hard and/or expensive to fix. So this would be a worthwhile test for someone looking to buy one of these.

[tautech]

I finally pulled out the CRT and found a wee surprise.

The penny dropped - the bang and the smoke back when it first started up was actually this cap going off.
When that happened the CRT was extremely bright until I turned it off. Also it was difficult to get a linear brightness out of the calibration.

I can get 47uF or 10uF 250V caps locally but not 22uF. So a choice: Bodge it or wait?

Onwards with the bodge then...

What's all these then?
http://nz.rs-online.com/web/c/passive-components/capacitors/aluminium-capacitors/?searchTerm=22uF+250V

[TiN]

Thank you for tips and sharing. It would be also great to see your progress in test set build, if you decide to go thru and worklog that project.

[hendorog]

I finally pulled out the CRT and found a wee surprise.

The penny dropped - the bang and the smoke back when it first started up was actually this cap going off.
When that happened the CRT was extremely bright until I turned it off. Also it was difficult to get a linear brightness out of the calibration.

I can get 47uF or 10uF 250V caps locally but not 22uF. So a choice: Bodge it or wait?

Onwards with the bodge then...

What's all these then?
http://nz.rs-online.com/web/c/passive-components/capacitors/aluminium-capacitors/?searchTerm=22uF+250V

Hah thanks that will be handy, I didn't even think about them!

I do remember knowing about them years ago - I was more into software and automation then but was working with some EE's.

[tautech]

I finally pulled out the CRT and found a wee surprise.

The penny dropped - the bang and the smoke back when it first started up was actually this cap going off.
When that happened the CRT was extremely bright until I turned it off. Also it was difficult to get a linear brightness out of the calibration.

I can get 47uF or 10uF 250V caps locally but not 22uF. So a choice: Bodge it or wait?

Onwards with the bodge then...

What's all these then?
http://nz.rs-online.com/web/c/passive-components/capacitors/aluminium-capacitors/?searchTerm=22uF+250V

Hah thanks that will be handy, I didn't even think about them!

I do remember knowing about them years ago - I was more into software and automation then but was working with some EE's.

Bookmark this one too: 
http://nz.element14.com/

[hendorog]

Bookmark this one too: 
http://nz.element14.com/

Yep I knew about them, haven't used them though.

[hendorog]

Thank you for tips and sharing. It would be also great to see your progress in test set build, if you decide to go thru and worklog that project.

Glad it's of some help Yep I am planning to do that.

Here is some info about where I'm at:

This doubler PCB and MMIC's are on the way:
http://www.ebay.com/itm/Develop-PCB-Hittite-Doubler-HMC188MS8-with-2-MMIICs-/160794801260?hash=item25701df46c:m:mdEwzLLPGZOcXNZKX8dcbxw

And I scored (gambled on) these cheapish directional couplers with no spec. Given the part numbering, I was hoping these were out of the 3600 20GHz test set for the Wiltron 360 VNA.
http://www.ebay.com/itm/262433378082

However I've now found a parts manual and it's not the same part. D-20363 is the 20GHz coupler in the 3600 test set, these are on eBay and are not the same.

So I'm not sure what these are from and what they will be like. They will still work fine, since some qualities can be compensated by the correct type of VNA calibration, but there will of course be some loss in accuracy.

I'll make a resistive splitter for the R Channel, and will also try to make a bridge so the couplers aren't needed. I'm anticipating that the bridge will be the most difficult part of this and so having off the shelf couplers will provide a quicker path to getting it working initially. Will probably ultimately use Peregrine switches, but I've got some HP coax switches to get started with.

[hendorog]

Quick warning about this ebay sale which looks dodge to me.

This is being sold as complete but not working -from a new seller with only 3 sales:
http://www.ebay.com/itm/Hewlett-Packard-8752-Network-Analyzer-300kHz-3GhZ-/252476809778?&_trksid=p2056016.l4276

I'm sure it is the same device as this sale which just ended unsold from a different seller with 127 sales:
http://www.ebay.com/itm/HP-Agilent-8752C-Network-Analyzer-50-3GHz-/182183359815?_trksid=p2047675.l2557&ssPageName=STRK%3AMEBIDX%3AIT&nma=true&si=BbDQC%252BFdNe6m7UJDAfxrtrI8rGU%253D&orig_cvip=true&rt=nc

The original sale had a pic of the inside. It is not complete.
Major issue is the (expensive) source does not look like the correct one for the 8752 and the interconnecting cables are all missing.
Minor issues various parts missing like power button, feet, trim etc.

The thing looks like a parts unit and may just be full of broken bits.

[hendorog]

I have isolated the source of the spurs.

The clue was something I just read on the Keysight forums - that 25.5kHz was the sync frequency used for the external monitor. Aha, that is also the offset of these spurs from the CW.

Initially I thought my routing of the external monitor cables was the issue. Playing with that didn't help.
Then I tried disconnecting the CRT and blind keying into CW. Big improvement in all except the 100kHz spur which gets worse for some reason.
That spur has its own tuning pot but I couldn't reduce it any further. 

The key question is does this affect the device? I have noticed some very occasional glitches in the CW signal, which makes me wonder if the spurs are high enough to confuse the phase lock. Apart from that everything seems comfortably better than spec.

85044A T/R Test set has arrived. Its ex Colorado Uni and by the look of the dings in it, it has taken some punishment. I'm one 7mm to SMA adapter and one N to SMA short
so can't fully test it. However I tested each channel (R and A)  individually and it seems to work OK.

Pulled the cover off it - it has little plastic zip slider things, no screws. Opened up the directional bridge and wow, it is a thing of beauty inside. Wish I had a better camera.





Top gold box is coupler/bridge, bottom silver box is variable attenuator. Connections on right are - top bias from VNA,  middle top A to VNA, middle bott R to VNA, bottom RF from VNA


Inside the gold box.


Old repair


RF input comes from the right via the choke, bottom is test port, left goes to reflected A output to VNA, top goes to termination resistor.





Tiny Termination resistor at the end of the line from the Test port.


13dB Compensation, to  cable strangely routed the long way around the front of the case, goes to R output


Splitter - top exits to variable attenuator, botton to fixed 13dB complensation


Pin coming in on centre right is RF input from variable attenuator. Heads left via the choke to the coupler Resoldered cap visible. Lower right is bias input.


Edit: Found a schematic for the 85044A Test Set

[dr.diesel]

I also replaced the supercap with a larger 1F one as I thought it was faulty. The display colour resets to default after restarting. Turns out that this still happens so the original was fine. The user settings I had saved were all lost as expected.

The supercap supports specific test offsets, user preset type of data, not instrument calibration correct?

Reading/saving EEPROM data might be my next goal. 

[hendorog]

I also replaced the supercap with a larger 1F one as I thought it was faulty. The display colour resets to default after restarting. Turns out that this still happens so the original was fine. The user settings I had saved were all lost as expected.

The supercap supports specific test offsets, user preset type of data, not instrument calibration correct?

Reading/saving EEPROM data might be my next goal. 

Yes, the EEPROM stores all of the instrument cal data - i.e. the results of the alignment tests in the manual. This is what you want to back up, although it doesn't seem to be _that_ difficult to run through the tests to re-create it.
 
The user preset data is definitely saved by the supercap, losing those is no biggie.

Some unexpected but less important things are reset after an extended time without power (i.e. after say 5 minutes) even with the supercap in place:
* User Calibrations <- I haven't worked out if these are supposed to be saved or not. There is an option to save the cal after you complete one, and PRESET then Recall seems to work. However when I power down and come back the next day and Recall the saved preset again it seems to be gone. Other settings are Recalled. Could be PEBKAC error.
* Display colours <- There is a separate save and restore function under the Display button which does work for this.

[KE5FX]

Some unexpected but less important things are reset after an extended time without power (i.e. after say 5 minutes) even with the supercap in place:
* User Calibrations <- I haven't worked out if these are supposed to be saved or not. There is an option to save the cal after you complete one, and PRESET then Recall seems to work. However when I power down and come back the next day and Recall the saved preset again it seems to be gone. Other settings are Recalled. Could be PEBKAC error.
* Display colours <- There is a separate save and restore function under the Display button which does work for this.

Calibrations aren't saved on the 8753A-C models, unlike the 8510C.  A real forehead-slapper on HP's part.  Not sure if they are on the -D/-E models.

You can save and restore the complete instrument setup/calibration state with a cheesy Windows app I wrote for that purpose.  Requires an NI GPIB adapter.

[hendorog]

Detailed info here - chapter 10
http://literature.cdn.keysight.com/litweb/pdf/08753-90901.pdf?id=807274

[KE5FX]

Detailed info here - chapter 10
http://literature.cdn.keysight.com/litweb/pdf/08753-90901.pdf?id=807274

The key part is on page 10-2: "Calibration data and memory traces are stored in volatile memory.  While this data will survive an instrument preset, it is lost when line power is turned off."

So you end up being able to recall the state after a power cycle, but not the calibration that went with it. 

[hendorog]

Detailed info here - chapter 10
http://literature.cdn.keysight.com/litweb/pdf/08753-90901.pdf?id=807274

The key part is on page 10-2: "Calibration data and memory traces are stored in volatile memory.  While this data will survive an instrument preset, it is lost when line power is turned off."

So you end up being able to recall the state after a power cycle, but not the calibration that went with it. 

Thanks. The part that was threw me was that the cal and display settings actually survive a (relatively quick) power cycle.

Regarding your GPIB utils, do they specifically require the NI or Prologix GPIB controllers or will they generally work with most 'standard' controllers? The cheapest seem to be the HP copies on eBay at about NZD100.
The eye of Sauron has now swung in my direction and is scrutinising how much I spend on this stuff so I need to go as cheap as possible

[KE5FX]

The eye of Sauron has now swung in my direction and is scrutinising how much I spend on this stuff so I need to go as cheap as possible

Unfortunately, one does not simply use a Prologix adapter with VNA.EXE.  It will "work" but you'll see occasional bus lockups.  This is especially true with the GPIB-Ethernet adapter, or when using sweeps with larger point counts (>201).  There's a lot of additional advice on that subject here. 

This isn't necessarily Prologix's fault, but it's not clear whose fault it is.  Probably mine.   

Others have had good results with the Agilent 82357B adapters in NI488.2 mode, but I have almost no personal experience with those.  My recommendation above all others is the NI GPIB-USB-HS models that have been popular on eBay out of China lately.  Unlike the "Agilent"-labeled units, they are either not counterfeits, or they're good enough copies that I can't tell the difference.  Those are the only GPIB adapters I use at this point.

[hendorog]

The eye of Sauron has now swung in my direction and is scrutinising how much I spend on this stuff so I need to go as cheap as possible

Unfortunately, one does not simply use a Prologix adapter with VNA.EXE.  It will "work" but you'll see occasional bus lockups.  This is especially true with the GPIB-Ethernet adapter, or when using sweeps with larger point counts (>201).  There's a lot of additional advice on that subject here. 

This isn't necessarily Prologix's fault, but it's not clear whose fault it is.  Probably mine.   

Others have had good results with the Agilent 82357B adapters in NI488.2 mode, but I have almost no personal experience with those.  My recommendation above all others is the NI GPIB-USB-HS models that have been popular on eBay out of China lately.  Unlike the "Agilent"-labeled units, they are either not counterfeits, or they're good enough copies that I can't tell the difference.  Those are the only GPIB adapters I use at this point.

Great info, thanks. I'll start looking for an NI adapter.

[orin]

And the Cavity Oscillator Frequency Correction Constants, which is used to correct the frequency of the Cavity Oscillator for aging. It involves identifying a particular spur of the oscillator on the screen. This is pretty difficult unless you have an appropriate filter. I don't so intend to make one and repeat this adjustment.

Reviving an 8753C that came with a blown "A" sampler here.  I replaced the sampler with a used one (<$120 from ebay) and am going through the adjustments.

Most likely you can use a Mini Circuits SLP-1200 or NLP-1200 filter for the cavity oscillator adjustment.  They are reasonably priced - I just got the SMA version on ebay for about $25 shipped; there is one more available as of the time of this posting.  The specified HP filters are also available for not much more.

[hendorog]

And the Cavity Oscillator Frequency Correction Constants, which is used to correct the frequency of the Cavity Oscillator for aging. It involves identifying a particular spur of the oscillator on the screen. This is pretty difficult unless you have an appropriate filter. I don't so intend to make one and repeat this adjustment.

Reviving an 8753C that came with a blown "A" sampler here.  I replaced the sampler with a used one (<$120 from ebay) and am going through the adjustments.

Most likely you can use a Mini Circuits SLP-1200 or NLP-1200 filter for the cavity oscillator adjustment.  They are reasonably priced - I just got the SMA version on ebay for about $25 shipped; there is one more available as of the time of this posting.  The specified HP filters are also available for not much more.

Nice one on the repair - btw if you need a part which is too expensive on eBay, try posting on the Agilent/HP yahoo group, there are people there sitting on piles of parts. Or let me know and I'll send you the contacts of the guys I got parts from.

Thanks for the heads up on the filters that is good to know and I'll check them out. I think I got the correct spur, but hard to be sure and I'd like to check.
Sometimes it's not worth it as shipping to NZ can be expensive and slow sometimes - and I feel I shouldn't be lazy since I have a great way to test the filter