Author Topic: HP 8566B Spectrum analyzer, trouble with band C (5.8 to 12.5 GHz)  (Read 8869 times)

0 Members and 1 Guest are viewing this topic.

Offline Berni

  • Super Contributor
  • ***
  • Posts: 2763
  • Country: si
HP 8566B Spectrum analyzer, trouble with band C (5.8 to 12.5 GHz)
« on: September 19, 2016, 03:55:07 pm »
Some time ago i bought a used HP 8566B. Quite the neat beast of a spectrum analyzer.

But i noticed things being odd within a certain frequency range so i used a Gigatronics 7100 synthesizer (10MHz to 20GHz) to sweep the frequency response in that area and noticed that it is not showing correctly. In the photo you can see a 8 to 15 Ghz sweep using max hold. It clearly shows how the frequency response is wonky but as soon as you make it past 12.5GHz where it goes in to the next band it becomes perfectly fine again. :--

I had a go at calibrating it out by following the adjustment procedure for that band, but its pretty difficult to do when you don't have a continuously sweeping signal generator. None the less i found that twiddling the frequency response adjustment pots for that band did not fix it. The 20dB of error on the screen also seams like its too far out to just be calibration.

Since all the other bands are fine i went looking at the block diagrams to figure out what is unique to that particular troublesome band, but the ones i found online are too fuzzy to make out any of the text and me being more of a digital and analog guy rather than a RF wizzard i decided its too much trouble to guesswork, especially with how ridiculously complicated this thing is. So i ordered a digital copy of some proper high resolution scans of the manuals on ebay. In the mean time i thought id post about it here in case anyone knows off the top of there head what it could be. I am hoping the issue is not in the RF path because i really am not looking forward to disassembling that rats nest of hard line coax.

Some help would be greatly appreciated.
 

Offline rastro

  • Frequent Contributor
  • **
  • Posts: 380
  • Country: 00
Re: HP 8566B Spectrum analyzer, trouble with band C (5.8 to 12.5 GHz)
« Reply #1 on: September 19, 2016, 10:55:02 pm »
Hello Berni,

How are you verifying your Gigatronics 7100 synthesizer?

-rastro
 

Offline cncjerry

  • Supporter
  • ****
  • Posts: 974
Re: HP 8566B Spectrum analyzer, trouble with band C (5.8 to 12.5 GHz)
« Reply #2 on: September 20, 2016, 02:49:03 am »
What's your reset or power on trace look like?
 

Offline Berni

  • Super Contributor
  • ***
  • Posts: 2763
  • Country: si
Re: HP 8566B Spectrum analyzer, trouble with band C (5.8 to 12.5 GHz)
« Reply #3 on: September 20, 2016, 03:53:06 pm »
Thanks for the quick response.

I checked the flatness of my synthesizer using a HP 8472B power detector. To be sure i checked it again for the same 8 to 15 GHz sweep and made a screenshot of it. The sweep was set up to be roughly 100s so it correlates nicely to the width of the screen and its filled in because the synthesizer jumps to zero power when stepping to the next tone. From what i see there everything is fine. Also included a photo of the setup, showing the detector plugged in on the end of the cable that i used for getting the signal to the sick spectrum analyzer.

I also attached a photo of what the noise floor looked like before i did anything . Since i been twiddling the pots to try to fix the response it now looks slightly different but its not hugely far off.
 

Offline rastro

  • Frequent Contributor
  • **
  • Posts: 380
  • Country: 00
Re: HP 8566B Spectrum analyzer, trouble with band C (5.8 to 12.5 GHz)
« Reply #4 on: September 20, 2016, 04:52:31 pm »
Looks like you posted moments before me....After some thought, your signal generator is probably OK as the problem is delineated at the band-switching points.

Some additional insight:http://www.hpl.hp.com/hpjournal/pdfs/IssuePDFs/1979-08.pdf

Both 'bias' and 'pre-filter tuning' to the YTX can influence RF level response.

Investigate YTX Pre-filter tuning:
Try injecting a 12.1 GHz signal to center frequency with bandwidth ~ 1MHz and executing 'PRESEL PEAK'.  This should search for optimal filter setting through fine tuning of the YTX pre-filter.  If pre-filter tuning is good there should be only a small change in amplitude after peaking.  When the SA is peak searching you should notice changes in the signal level; then it should pick the optimal value to the high level.

Investigate YTX bias:
No input signal required.  Set SA for zero span to eliminate sweeping.  Monitor YTX bias input while CF between 12.1 and 12.6 (frequency clearly inside next band).  Increment CF in 50MHz steps starting at low frequency.  Is there a stair step that correlates with the picture of the frequency response you posted?

-rastro

« Last Edit: September 20, 2016, 05:00:30 pm by rastro »
 

Offline rastro

  • Frequent Contributor
  • **
  • Posts: 380
  • Country: 00
Re: HP 8566B Spectrum analyzer, trouble with band C (5.8 to 12.5 GHz)
« Reply #5 on: September 20, 2016, 05:44:48 pm »
Quick word of caution when measuring the YTX bias levels. 

The bias is applied to the YTX mixer via the ACLU.
The mixer inside the YTX is static sensitive - if it is damaged the YTX needs replaced.  Use static caution when handling the YTX or it's connectors.

A good place to measure the YTX bias is at the ACLU.  This is mounted on the "totem pole" under another RF modules.  The bias connector is a single wire on the narrow side of the ACLU.  The other wire connectors are power.  This wire should be accessible without removing the top module.  You should be able to monitor the voltage at the wire (disconnected) since you are just looking to see what the bias is driven to.  Again use static precautions around the exposed ACLU bias pin.
« Last Edit: September 20, 2016, 11:26:42 pm by rastro »
 

Offline Berni

  • Super Contributor
  • ***
  • Posts: 2763
  • Country: si
Re: HP 8566B Spectrum analyzer, trouble with band C (5.8 to 12.5 GHz)
« Reply #6 on: September 20, 2016, 08:11:18 pm »
I had a more serious go at calibrating out the problem with the pots and it is kinda better but still not perfect (Photo below).

I had a go at your troubleshooting tips and here is what i found:

Preselect peaking i fed a 12.1GHz -10dBm signal in to it, looking at it was showing about -20dBm, then after pressing the preselect button it jumped around a bit before ending up at -10.4dBm. Much closer to what it should be (My calibration for gain was a bit high for this band)

I hooked a multimeter on to that YTX bias pin and moved around some. I got about 0.98V in most of the problematic band but when i crossed 12.5GHz in to the band higher up it jumped to 1.39V. Is that acceptable?
 

Offline rastro

  • Frequent Contributor
  • **
  • Posts: 380
  • Country: 00
Re: HP 8566B Spectrum analyzer, trouble with band C (5.8 to 12.5 GHz)
« Reply #7 on: September 20, 2016, 10:46:23 pm »
It appears that the pre-filter is not tuning correctly towards the end of the 5.6-12.5GHz band (inital freq response picture).  If the system is tracking correctly the  "PRESELECT PEAK" should not have that large of impact - maybe a couple of dB correction (certainly not 10dB!).  Since the  "PRESELECT PEAK" correction did bring the signal level to close to where it was expected suggest that the RF path and other leveling factors are probably OK for the most part.

The good news is that the "PRESELECT PEAK" could correct the pre-filter tuning error.  At this point you should focus on the pre-filter tuning. 

For comparison  try running the same  "PRESELECT PEAK" test using a 12.6GHz signal.  You'll probably see only minor adjustment to the signal level if any.
Also you may want to try doing  "PRESELECT PEAK" at different points on the 5.6-12.5GHz band.  I suspect you'll see the  "PRESELECT PEAK" will fix the dips in your RF level response in the problem frequency areas. 

I don't know what you are changing with your "adjustments" - you are not clear on this.  Your systems problems can start moving around if you keep making adjustment to different sub-systems without understanding where the problem lies.  You can complicate troubleshoot by continually changing adjustments in different sub-systems.  Determine the problematic sub-system and focus on it first. 

Determine if there are any HW problems with the pre-filter tuning .  Then ensure that subsystem is correctly adjusted.  Then go back through any bias setting you may have changed to fix/compensate for prior poor pre-filter tuning.

I respectfully question your approach to "calibrating out the problem".  Usually when I see performance that is significantly out I don't assume it is just an adjustment that will fix it.  Granted you may need to see if an adjustment has any effect but I would want to be sure I can re-establish it to it's original setting.  As a guide line: If it a little out maybe it's an adjustment; if it's significantly out it's likely an HW failure or operator issue.

-rastro
« Last Edit: September 20, 2016, 11:28:49 pm by rastro »
 

Offline Berni

  • Super Contributor
  • ***
  • Posts: 2763
  • Country: si
Re: HP 8566B Spectrum analyzer, trouble with band C (5.8 to 12.5 GHz)
« Reply #8 on: September 21, 2016, 06:50:39 am »
The pots i been touching the pots that are the ones talked about in "HP 8566B Performance tests and adjustments manual" in the section for frequency response adjustments(page 3-190). These are pots A6A12R84 C, A6A10R15 VC, A6A10R25 GC that have to do with the YTX and IF tracking and the pots A6A11F54 C1, A6A11R72 C2 for band flatness. I was careful not to touch the adjustments for other bands that are known good. For that reason i stopped flowing the procedure at step 99 as the those YTX linearity adjustments sound like they affect all bands. I can easily get back to the original settings for band C on the band flatness pots as they are one turn and have a direct effect on the waveform. The YTX and IF pots are multiturn so not easily adjusted back, but the adjustment procedure for those is easy to do where you just tweak them to get the signal on the screen to be as large as possible and they didn't seam to need adjustment since where i found the peak response was about where they used to be.

Also just to be sure i pulled out and reseated the 3 cards with the adjustments from before just to be sure its not a bad backplane contact issue (Contacts still have a nice shiny gold appearance on one card i pulled out fully to check)

The issue was likely there already when i got this unit so i don't know in what way the problem came around. Since the particular band mostly worked i thought it just needed a tweak rather than something being wrong.

When i get home from work i will have a deeper dive in to the service manuals and try to track down the fault more precisely. Thanks for the tips since it will certainly help narrow down where to look.

 

Offline Berni

  • Super Contributor
  • ***
  • Posts: 2763
  • Country: si
Re: HP 8566B Spectrum analyzer, trouble with band C (5.8 to 12.5 GHz)
« Reply #9 on: September 21, 2016, 08:29:39 pm »
I spent some more time troubleshooting it. Seams like the preselect button does indeed fix all the amplitude issues over the problematic band. Doing so shifts the YTX tuning voltage as seen on a test point by a few 10s of mV.

After digging trough the schematics i focused on some JFETs that could explain the issue if they went a little half dead. To get some signals off them i used some raw fork contacts from 0.1in female headers stuck in to female jumper cables to tap off some signals while the card is in the cage. The voltages i seen while adjusting the front panel controls seamed to be ok.

The next thing i want to check is a block called "YTX Linearity correction" on board A6A10. That's the part i didn't want to adjust during calibration so i stopped. From the looks of it its the part that was supposed to compensate out that droop in YTX tracking that i am seeing. I ran out of time to do it, but the plan was to check the diodes and opamps inside that circuit. If things look ok then i am thinking of adjusting that pot (A6A10R40 LB1 and A6A10R31 LR1) anyway and hope for the best.

Thanks for pointing me in the right direction, we are definitely on to something.
 

Offline G0HZU

  • Super Contributor
  • ***
  • Posts: 2576
  • Country: gb
Re: HP 8566B Spectrum analyzer, trouble with band C (5.8 to 12.5 GHz)
« Reply #10 on: September 21, 2016, 09:05:19 pm »
Quote
I respectfully question your approach to "calibrating out the problem".  Usually when I see performance that is significantly out I don't assume it is just an adjustment that will fix it.  Granted you may need to see if an adjustment has any effect but I would want to be sure I can re-establish it to it's original setting.  As a guide line: If it a little out maybe it's an adjustment; if it's significantly out it's likely an HW failure or operator issue.

I agree. I would exercise a bit of caution here and maybe study/research the analyser operation to try and learn how to prove the health of the various sub circuits associated with this issue. Otherwise, if you just start trying to tweak out such a large anomaly, you may well end up 'jelly wrestling' as you revisit each trimmer again and again.

I've got an HP8566A/B here (A converted to B) and I'll see how mine looks with a decent sweep generator across 8-15GHz. It won't be tonight, but maybe tomorrow. My bench is a bit busy at the moment and the sweeper is very big and heavy...
 

Offline rastro

  • Frequent Contributor
  • **
  • Posts: 380
  • Country: 00
Re: HP 8566B Spectrum analyzer, trouble with band C (5.8 to 12.5 GHz)
« Reply #11 on: September 21, 2016, 11:33:59 pm »
Berni,

Hold off on adjusting  "YTX Linearity correction".  This will probably affect all bands and you don't want that. 

-rastro
« Last Edit: September 21, 2016, 11:41:46 pm by rastro »
 

Offline rastro

  • Frequent Contributor
  • **
  • Posts: 380
  • Country: 00
Re: HP 8566B Spectrum analyzer, trouble with band C (5.8 to 12.5 GHz)
« Reply #12 on: September 22, 2016, 01:16:02 am »
Breni,

You had a good idea studying the schematics and looking at the JFET signals.  You need to stay with that type of approach.

I'm concered you want to do further adjustments and 'hope for the best'.  It might be better if you set the tweeker down and step back to regroup.  Hit the books and understand the system a little better.

You have already de-tuned several subsystems.  They didn't really fix the problem and are now out of alignment - stop digging a hole for yourself!
Subsystems Impacted:
1. Segment Amplifier Band-C (A6A11F54, A6A11R72)
2. IF Offsets Band-C (A6A12R84)
3. Band-C Band Step Gain (A6A10R25)
4. Band-C YTX Diode Bias (A6A10R15)

You need to get a better understanding of how the pre-filter tuning works.
1. Look through the HP Journal I sent - there's a section on the YTX by the engineers that designed it.
2. How is the normal pre-filter on the YTX tuned?
3. How does the "PRESELECT PEAK" work?
4. How is band C similar to the other 3 bands?  How is it different?
5. What signals can you look at.  The flow charts have plenty of references to TP's.
6. Do you understand the fundamentals of the YTX?
- It is a combination YIG filter and mixer.
- It really only tunes about 4GHz.
- Different Bands are derived by looking at different harmonics out of the mixer.

Knowing that the YTX's YIG has a narrow range then the other bands have to get tuned over the same range - why do they work and not band-C?
You could also leverage this fact.  For example the zero span setting of 12.1GHz must have a corresponding filter tuning frequency on the band-D.  I would imagine it uses the same LO frequency.
I wonder what the corresponding YIG tuning level looks like around those corresponding pairs of frequencies?  But all bets are off if the working bands get screwed up.

- rastro


 

Offline Berni

  • Super Contributor
  • ***
  • Posts: 2763
  • Country: si
Re: HP 8566B Spectrum analyzer, trouble with band C (5.8 to 12.5 GHz)
« Reply #13 on: September 22, 2016, 06:02:11 am »
Those JFETs that switch bands on the YTX are the first thing i found to be unique to the YTX tuning. There are more JFETs that switch the offset per band for it but those just produce DC levels so they shouldn't go weird as the sweep moves. The way i tested the JFETs in there is i looked at the voltage across drain and source to make sure the FET opens up fully and checked the resistor divider in front that none of the FETs drag it off.

I did look in to the topology of the RF path so that i would know what is going on. Really neat trick they used with this harmonic mixer and tunable filter. I thought there would be 22 GHz flowing trough more of it but its just straight in to the mixer and sub GHz out the other end. I do now i see why they are so serious about not putting any DC in to the input of this analyzer as it goes almost straight in to that mixers sensitive diode. Its certainly quite an odd mixer in my opinion with the LO and output sharing a port, didn't know there are harmonic mixers that mix multiples of the LO, it does makes it pretty clear why the tuned filter needs to be in front, otherwise there would be a whole zoo of tones coming out as well as the LO going straight back out the front connector.

The preselect peak button is what i used to narrow down my search in the first place. I found the DAC that gets tweaked when you push it so i checked what signal that changes. The only part that was band specific and had a changing signal was the JFET switched divider for each band. Since that didn't show any abnormal behavior i started looking broader and found the YTX linearity correction circuit on the next board in the card cage(A11 i think). It was not instantly obvious that its band related since it does not have any band switch signals coming in. Later on i figured out that the pots on it are supposed to be set to kick it in to action at certain points along one of the bands. Testing the semiconductor parts in there is my next plan of action.

The YTX is clearly tuned a tiny bit low near the end of band C. That YTX linearity adjustment circuit being wonky could explain it as correcting for that is pretty much the sole purpose of that block.

Should i be cautious about resistors changing in value? From my understanding modern resistors are very reliable when not exposed to heavy loads or huge transients. My repair experience is from fixing gear more modern than this dinosaur, nor have i ever component level troubleshooted anything this complex before. Things like audio amplifiers, programmable PSUs, non RF signal generators etc are much simpler.

Oh and i forgot to bring up that there are clear signs of someone being in this A6 card cage before. The metal cover that goes on top did not have all the screws and one of the cards is missing one of the plastic tabs for pulling the card out. So its possible someone tried to fix this before me, was fixing something else and caused this, or a card has been replaced.

 

Online KE5FX

  • Super Contributor
  • ***
  • Posts: 1166
  • Country: us
    • KE5FX.COM
Re: HP 8566B Spectrum analyzer, trouble with band C (5.8 to 12.5 GHz)
« Reply #14 on: September 22, 2016, 06:34:01 am »
Should i be cautious about resistors changing in value? From my understanding modern resistors are very reliable when not exposed to heavy loads or huge transients. My repair experience is from fixing gear more modern than this dinosaur, nor have i ever component level troubleshooted anything this complex before. Things like audio amplifiers, programmable PSUs, non RF signal generators etc are much simpler.

You don't need to worry about the resistors, but you should get an ESR meter and check all of the axial-leaded tantalum electrolytics.   Probably not related to your problem here, but there's a good chance you'll find one or two open.
 

Offline rastro

  • Frequent Contributor
  • **
  • Posts: 380
  • Country: 00
Re: HP 8566B Spectrum analyzer, trouble with band C (5.8 to 12.5 GHz)
« Reply #15 on: September 22, 2016, 02:09:11 pm »
Good observation on the A6 card.  Is there any signs of rework like soldering?  Still it may or may not be bad.  Clearly someone was careless with putting back all the hardware.  Since all other bands work there is only a few places to evaluate that is unique to band-C.

Regarding degraded resistors:  It's always possible although I'm sure HP used top shelf components.  Generally I expect to see problems with electrolytic and semiconductors before drifting or failed fixed resistors.  It's better to use measurements at a systems level to identify a problem subsystem first then start looking at individual components.

"YTX linearity adjustment circuit being wonky":
This subsystem is utilized by all pre-filter bands B-E.  If it "wonky" why doesn't band B,D,E exhibit the same drop in RF level?  Does this theory really make sense?  Shouldn't any subsystem shared by bands B-E should be low suspect?  These other band are working correctly - right?

Get better characterization of the problem behavior:
1. Is the low amplitude points caused due to the pre-filtertuning too high or low?  Both situations will cause a drop in RF level.  How did you objectively establish that the YTX is actually tuned low rather than high?  Is it consistent across the problem frequencies?  You know you can manually control the pre-selector DAC with <SHIFT><=GHz + dBm dB>....you can see the RF signal level vary as you move through the 64 level range.  At 12.1 GHz my system reads "33" which is about center of the range (which is probably ideal).
2. Did you figure out how the "PRESEL PEAK" ties in with the normal tuning?  This is pivotal to troubleshooting this issue - don't you think?
3. How are you going to measure YTX tuning?
4. There are other diagnostic features on the system you could also leverage.  For example: Band lock, DAC setting, Step Gain level, and LO setting.

Some Ref material:
http://www.siliconinvestigations.com/HPREP/HP8566-8/8566A-opr.pdf
Page
58 Information on preselector peak values held in memory for each band.
86 Some diagnostic modes
89 Factory preselector setting; Manual preselector tracking. (Troubleshooting?)
« Last Edit: September 22, 2016, 02:25:32 pm by rastro »
 

Offline Berni

  • Super Contributor
  • ***
  • Posts: 2763
  • Country: si
Re: HP 8566B Spectrum analyzer, trouble with band C (5.8 to 12.5 GHz)
« Reply #16 on: September 22, 2016, 07:24:50 pm »
From a visual look at the cards i can see some flux residue on A6A10. All the timm pots in the YTX Linearity correction block have that flux residue on the pins (R31 R34 R37 R76 R40 R41 R42 R70). Since i don't see a reason why someone would replace all of them makes me assume that HP forgot to clean it at the factory. I did have a quick check of the tantalum caps on there and they seam to be close to the values they should be (I at first dismissed them as foil caps due to the appearance but then noticed they do indeed have a polarity marked on them).

The reason i started to suspect this particular block is because pots labeled LB1 to LB4 set a threshold at what point in the turning range the compensation kicks in and starts to give extra current trough the tuning coil. However i checked all the semiconductors there and the diodes did have reasonable and similar forward voltages. I checked the opamps and they had 0.5mV input offsets once they kicked in at the corresponding tuning frequency(datasheet spec is 2mV). And i measured the input bias current to be about 60nA (datasheet was 80nA for all four together) by powering up the board outside of the cage and meshing the current with a bench multimeter. The opamps are seen kicking in at roughly 8Ghz 16Ghz 18Ghz and 22GHz.

The way i determine the effect of preselect peak is that i followed where the DAC goes in and it basically sums in to the band switch and IF offset signal that generate most of the YTX tuning signal, so a signal from the dac directly adds or subtracts the YTX turning signal (Almost as if the incoming sweep ramp was shifted up or down a bit). I determined that it was low because i looked at the YTX sense signal (A6A12 TP6) and i have observed the voltage there increase slightly when i press preselect. So by that the original tune of the YTX is at a too low frequency. I did quickly notice before that the YTX tunable filter has a very sharp band pass response.

I used your sugestion to check the preselect DAC values and for all the bands it seams to chose values between 29 and 36 exept the problematic C band. It chooses close to 32 up to about 10GHz but afterwards it starts to go up until it reaches at worst value 47 towards the 12.5 Ghz end of band. Seams to about correlate with where the waveform amplitude drops off. However once i step in to the next band it goes back to 31 again.

I also checked the list of correction data in memory (Shift + [Trigger]Line) and it all seamed to be within +/- 1dB and from what i hear that's good.

So yeah so far looking at the preselect DAC values it also seams to say that the YTX is tuning too low towards the end of the band. The only block having to do with YTX tuning on a band by band basis that haven't component level tested is the IF offset block on A6A12. Mostly because i as far as i see it is that if that part was broken it would have messed up the whole band not just towards the end of the band.
 

Offline G0HZU

  • Super Contributor
  • ***
  • Posts: 2576
  • Country: gb
Re: HP 8566B Spectrum analyzer, trouble with band C (5.8 to 12.5 GHz)
« Reply #17 on: September 22, 2016, 08:55:58 pm »
I tried sweeping this range on my HP8566A/B this evening and I got a poor result at first. I didn't take an image but the 12.5GHz+ range was low by several dB and the range below (i.e. your problem range) was also a bit low and had a slight slope on it but the other way around to yours.

So I was a bit worried about this and had a play with the preselector peaking and soon got it looking better. But I then decided to try the factory default lookup and this was even better. See below for my analyser after a short warmup running the default lookup table for the tuning. I'm now wondering if my random fumblings with Preselector Peak over the time I have had the analyser has produced some funny effects in the preselector slope. In the past I've used it with varying success at finding the peak tune. So maybe I was corrupting it in other places each time I ran it.

The factory default seems to be quite good considering this analyser is probably over 30 years old. Sorry about the image quality. This was taken with a mobile phone.

I let it sweep once on max hold. The sweeper is very flat across this range when tested into a decent power meter but I suspect that there will be some source/load mismatch uncertainty creeping in at the top end when the analyser is connected in place of the power meter. So maybe this is why there is some ripple up near 15GHz in the plot. I'm using a top quality (and very short) Gore cable to connect both instruments.

I'm using my old HP83752A as the swept source.
http://www.keysight.com/en/pd-1000001888%3Aepsg%3Apro-pn-83752A/synthesized-sweeper-001-20-ghz?cc=US&lc=eng

Have you tried the factory default lookup table?
« Last Edit: September 22, 2016, 09:42:06 pm by G0HZU »
 

Offline G0HZU

  • Super Contributor
  • ***
  • Posts: 2576
  • Country: gb
Re: HP 8566B Spectrum analyzer, trouble with band C (5.8 to 12.5 GHz)
« Reply #18 on: September 22, 2016, 09:34:33 pm »
I did a few more tests and PRESEL PEAK appears to be a very risky button to press on my analyser. If I start with the factory preset lookup and then do a PRESEL PEAK at 2GHz then the net result on the rest of that sub range can be quite dramatic. I can easily get a 6dB slope on the first 2-5.8GHz band by just running PRESEL PEAK once at 2GHz.

I'm going to leave mine on the factory default setting for now because casual use of PRESEL PEAK can cause vary degrees of slope across a sub band.
This aspect of the analyser definitely needs investigating at some point. Maybe my analyser is in need of a bit of TLC here because I doubt this behaviour would have been seen when the analyser was new. But it seems fine on the default settings and so I'll leave it there for now I think.
 

Offline G0HZU

  • Super Contributor
  • ***
  • Posts: 2576
  • Country: gb
Re: HP 8566B Spectrum analyzer, trouble with band C (5.8 to 12.5 GHz)
« Reply #19 on: September 22, 2016, 10:01:23 pm »
In case you haven't seen it already, there is some useful background info about the YIG mixer in the August 1978 HP Journal.

http://www.hpl.hp.com/hpjournal/pdfs/IssuePDFs/1979-08.pdf

For anyone else reading this thread, it's worth having a skim through this edition of the HPJournal just to get a 'feelgood' feeling about how special this analyser was in those days and you get some info about the team of people who designed it. Sadly, this level of detail/info is generally not given for today's instruments.
 

Offline rastro

  • Frequent Contributor
  • **
  • Posts: 380
  • Country: 00
Re: HP 8566B Spectrum analyzer, trouble with band C (5.8 to 12.5 GHz)
« Reply #20 on: September 22, 2016, 11:23:30 pm »
GOHZU has an interesting point.  But even if you go back to factory adjustments band-C is still going to be out of adjustment from prior tweaking. 

I would also suggest you test 10-12 equally spaced frequencies across band-c.  For each frequency record the "current DAC value" then use the DAC adjustment to find optimal peak for that frequency, record this "optimal value" and then put the DAC value back to its original value and go to the next frequency.  Do not utilize "PRESEL PEAK" while doing the measurements. If DAC count is walking out towards the lower values as frequency increases it suggests that the nominal tuning slope for the YTX pre-filter is not steep enough.  I think this is what you are telling us but it would be nice to put some reference values or base line for later comparison.

Regarding " correction data in memory (Shift + [Trigger]Line) ".  This is based on the results of the calibration routine running at 100MHz/Band-A.  The values are mostly to do with the health of IF section in the display unit.  It also checks Step Gain and the step attenuator linearity based on 100MHz signal.  It is not specifically testing upper bands B-E.

Just to verify - you have checked the flatness of Band-B,D, and E as far as you Gigatronics 7100 can go ~20GHz?  You don't see more than a few dB variation in flatness or really the expect level once you account for cable losses.  On a normal SA I would expect to see some normal high frequency drop off across all the bands due to transmission loss as frequency increases - But you are showing/expecting a flat responses.  How are you compensating for this?
 

Offline Berni

  • Super Contributor
  • ***
  • Posts: 2763
  • Country: si
Re: HP 8566B Spectrum analyzer, trouble with band C (5.8 to 12.5 GHz)
« Reply #21 on: September 23, 2016, 06:31:37 am »
That is indeed some interesting reading material, love that they show what is actually inside these RF voodoo black boxes (Well gold plated black boxes). Especially that YTX that feels like it is the heart of this instrument. Because of the reputation of how amazing this thing was back in the day is one of the reasons i went and got one when i found one for a reasonable price. Yeah shame we don't get to see with the same detail the love that the hard working engineers put in to Keysights latest and greatest equipment.

Later on playing with peak preselect button i noticed that it seams to remember a bit of the preselect dac setting. If i leave the manual preselect DAC adjustment on the screen while i move the marker between band C and D i could see the number change. I did this by varying a tone going into the input and pressing the peak search button to move the marker without closing the DAC adjustment menu. It seamed to prefer DAC setting 31 on the higher up band D while in band C it was choosing values around 40. It feels like it keeps some sort of running average on a band by band basis of successful preselect peaking commands. It seams to be one number for the entire band so it does not try to fix slopes.

Im guessing the way using preselect caused a slope on your 8566 is that the YTX tuning was already a bit off towards the end, but the error crossed zero around mid way. So as it went up in the sweep it went from for example(could be vice versa) being a tiny bit high at first, then spot on in the middle and then a tiny bit low towards the end. But it was not far out enough for the filter to start cutting it out. But when you pressed preselect on a tone towards the beginning of the band that made it offset it all so that that point was spot on. As a result the beginning of the band became spot on, midway it was a little bit low and the end was even more low. That could perhaps bring it out enough for the filter to start cutting.

Once i get home from work again i will certainly try the factory default settings and see what happens. But then again i did touch the IF offset pot for band C so i might have screwed that one up. I will also map out the best preselect DAC settings by manually adjusting it since i see the preselect button can't always be trusted. Also will take another look at the mixer diode bias in case there is something going on there.

I did do a full sweep from 2 to 20Ghz using my synthesizer a while ago and that is what made me discover this hole at 12GHz that lead me down this rabbit hole. I only recently got some proper Gore cables that are rated for these frequencies (Hard to find good ones on the cheap). So i used another high quality looking 1 meter cable that i don't know its ratings but it seamed to have reasonably low loss at 20GHz. I did indeed see a slope going down as i was expecting for cable loss. Forgot how much exactly it fell but i think it was somewhere in the neighborhood of -10 dB. Did see a little bit of few dB wiggle here or there(likely cable) but it was pretty flat overall with no jarring steps in the response apart from that dreaded 12.5GHz. The response of the Gigatronics 7100 was also given a quick spot check up to 20 GHz some time ago using that HP 8472B RF detector and it looked nice and flat.



By the way G0HZU i had my eye on a HP83752A but i could not find a good one for the right price. My Gigatronics can't do analog sweep and only has 1MHz frequency resolution cause i don't have those options installed.
 

Offline rastro

  • Frequent Contributor
  • **
  • Posts: 380
  • Country: 00
Re: HP 8566B Spectrum analyzer, trouble with band C (5.8 to 12.5 GHz)
« Reply #22 on: September 23, 2016, 01:39:15 pm »
Berni,

When you use 'PRESEL PEAK' the system remembers the optimal DAC value it found for that band.  There should be a specific value for each of the 4 upper frequency bands.  The appropriate DAC value for each band is loaded from memory as the SA sweeps between bands.  Also you can store 6 different system configurations in memory for convenient recall.  It appears these also retain the 'PRESEL PEAK' or DAC values at the time the set up was saved.  I recalled a couple of my old 'registers' (system configurations) and noticed different the DAC setting now had different values for the same band.  This could cause some confusion. :'(


-rastro
 

Offline rastro

  • Frequent Contributor
  • **
  • Posts: 380
  • Country: 00
Re: HP 8566B Spectrum analyzer, trouble with band C (5.8 to 12.5 GHz)
« Reply #23 on: September 23, 2016, 03:56:30 pm »
Latest finding...

Review of Pre-select filter DAC values:
1. A saved DAC value is loaded from memory when moving into each of the 4 upper bands.
2. 'PRESEL PEAK' routine adjusts the DAC until it finds the peak (best filter tuning) for the selected peak.  It then sets this as the DAC value for that band.
2a. Use peak search or adjust cursor to the peak of the signal you want to peak before running 'PRESEL PEAK' routine.
3. The DAC adjust <SHIFT><=GHz + dBm dB> manually (fine) tunes the pre-filter.  The DAC values are also remembered to last value in that band.  'INSTR PRESET' does not clear these values.
4. Factory reset puts all bands to DAC value 32 which is center of the 64-level DAC range.

This is my general understanding of DAC and pre-filter tuning:
A. In functional well adjusted system the pre-filter correctly follows the tuning of the LO across each of the bands.  There should be a flat level response within the specifications of the system datasheet.  It would seem the pre-filter should track the entire length of all upper bands set to DAC value 32. 

B. The 'PRESEL PEAK' routine is used to focus in on a specific signal in a particular band in order fine tune the pre-filter level to that specific frequency point so that the operator can get a more accurate power reading - one not affected by slight pre-filter misalignment.  In a functional/calibrated system it shouldn't make a significant difference in the readings before or after.  This new DAC setting will also affect every frequency along that band - hopefully only a small amount.

C. On a working/calibrated system how can the pre-filter DAC values in different bands start moving significantly away from 32??? 
1. Someone use the DAC adjust and leaves an arbitrary value there.
2. The 'PRESEL PEAK' routine is run without the frequency of interest properly identified or centered.  I did some experiments on a 12GHz signal to show how off-sets to the frequency of intrest from the cursor can create erroneous 'PRESEL PEAK' tunings.  The extent of error may also be affected by the BW setting that the mis-centering is done under.

Improper use of 'PRESEL PEAK' routine can induce SA performance errors (see pictures).
« Last Edit: September 23, 2016, 04:01:30 pm by rastro »
 

Offline rastro

  • Frequent Contributor
  • **
  • Posts: 380
  • Country: 00
Re: HP 8566B Spectrum analyzer, trouble with band C (5.8 to 12.5 GHz)
« Reply #24 on: September 23, 2016, 03:57:34 pm »
Remaining Pictures:
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf