Repair: HP 8753C Network Analyzer

[hendorog]

The 8753C VNA I mentioned in TiN's repair thread has arrived today, and thankfully much better packed that his one was. 
I have been keen to get hold of one of these for quite a while.

http://www.ebay.com/itm/HP-Agilent-Network-Analyzer-No-Frame-No-Power-As-Is-/252411867849?nma=true&si=wEiNpJ6q0gpoS313Zzj%252BbBkVaqY%253D&orig_cvip=true&rt=nc&_trksid=p2047675.l2557



Delivery was very rapid, it shipped on Wednesday PM and arrived Friday PM - considering it came from the Philippines to New Zealand via Hong Kong. Shipping large items is quite costly to NZ, thankfully offset by the low price on this unit.



The rubber keypad has some missing keys, and the bezel is completely missing. I was worried that there would be some boards missing from inside, but fortunately everything looks to be present. Just lots of dust and dust bunnies everywhere.

This was always going to be a repair as the description stated no power and sure enough this was the case when it arrived so off with the covers.



There was a red LED showing on the back, which flicked off and on again about once every second. It looks like the power supply is cycling/resetting. According to the manual that LED should be off, and the green one below should be on.

I spent some time pulling boards and connectors trying to figure out if the fault was in the PSU (A15 pre-regulator) or in the boards connected to it. In the end I removed the A15 unit as it was still faulting with nothing connected.


Inside the A15 unit I checked the caps and discovered a short to ground. After lifting a few components which turned out to be fine, I managed to track that to a bad 60V/5A rectifier diode in -8V rail of the A15 pre-regulator. The short was via the transformer winding.



When the diodes were removed it was shorted in both directions, there was also a faint scorch mark on the PCB - CR6 and CR7 just left of the inductor (which looks like a transformer) in the middle of the pic



I'm going to replace the pair of those with these: http://www.diodes.com/_files/datasheets/ds28009.pdf
as I can get them locally from Jaycar tomorrow.

Of course I have no idea if that is the only fault, so we shall see what happens tomorrow.

To be continued...

Service manual for 8753A:
www.hparchive.com/Manuals/HP-8753A-Manual-Service.pdf

Service guid for 8753C:
http://www.siliconinvestigations.com/hprep/hp8752/08752-90136-C-serv.pdf

[TiN]

Yey, sister thread FTW! ebay listing shows me 375$ only for shipping, so it's more than was paid for piece of wreck on my table. Something tells me you will get yours fixed and working much faster than I do

I'd appreciate detailed photographs of all boards, if you plan disassembly and cleaning. Got mine disassembled tonight, it was fairly easy except few levers broke. Frame, shield cages and backplane PCB with connectors are riveted together though, so that's as much as you can go without special tools to install new rivets.

I'm wondering how hard could it be to upgrade 8753A to C-version, as they share same service manual and most of boards are seem to be same. I know source on C is 6GHz capable (opt.006) and receivers + different brains.

Good luck with repairs, looking forward for your success..

[hendorog]

Yes the shipping was a bitter pill, nearly twice the cost of the unit... I had some nightmares on that thinking I might have paid a fortune to ship a box of broken boards...

I'll take pics of the boards today, I have had them all out but not cleaned yet. Any tips on the best method of cleaning them?
I hope your confidence isn't misplaced, you have a lot more experience at these repairs than I do. At least the boards don't appear to have been worked on before so there is a chance there is just one fault.

My feeling is that upgrading is doable. Note that as I understand it the source in the B and C is not 6GHz capable, but the receivers are. The option 006 still leaves them with a 3GHz source. They get the 6GHz from the  85047 test set which contains a switched frequency doubler.
So that is good for you as the A model source won't need to be changed. 

Thanks and all the best on your repair too!

[veedub565]

Looks nice, we still use those a lot at work. Keep us posted

[TheSteve]

What do you intend to use it for? The 4 port models often go for less then the 2 port/s parameter units. I've only bought 2 port units as I'm most interested easy s parameters. Or will you be adding a test set?
Either way good luck with the repair - hopefully a simple power supply fault.

[hendorog]

What do you intend to use it for?

Thats a good question, but hard to answer - I need to practise my response as pretty soon SWMBO will be asking the same question.
There are a few reasons.
Mainly it is probably for learning, and just to see where it leads. Learning about this stuff is so much easier when you can see the results live on the screen.

I've been making my own VNA for up to 6GHz as there was nothing affordable available. I want to be able to compare what that says with this. Also since I have never used a real one I wasn't sure how the UI should work.

The GHz frequency VNA motivation originally came from testing my home made antenna's for FPV model aircraft at 5.8GHz. However since then I've learned more and have more ideas of things to use it for. Building LNA's, phased antennas, filters, SDR etc, etc.

The other consideration is financial - If I ever have to sell it I should be able to get my money back as these are rare in NZ but still desired.

The 4 port models often go for less then the 2 port/s parameter units. I've only bought 2 port units as I'm most interested easy s parameters. Or will you be adding a test set?

Yes that is what I have found too - the 2 port units are very nice but very hard to get hold of cheap. The downside is that the 006 option is more than just a software key, and the ports are not available externally (I think thats another option) so bodging something using RF parts is more difficult.

The other consideration for me is the freight cost - the 2 port units are larger and heavier. I did work out that I could possibly just squeeze a 4 port unit with the test set packaged separately into the limits of the NZ Post cheap freight forwarding service. The 2 port units are too big for that. That was the plan until this one came up anyway.

I want to get a test set. Ideally I'd like the 85047A with built in doubler to 6GHz, but those go for crazy money. So I'm thinking I will get a T/R set for up to 3GHz and bodge up something with a Hittite doubler, splitter and a directional coupler for the rest of the range.
I have some HP coaxial switches that I could use.

Either way good luck with the repair - hopefully a simple power supply fault.

Thanks, I hope so too.

[hendorog]

Looks nice, we still use those a lot at work. Keep us posted

Thanks, I might need to ask you guys who have these already how to use it if/when its working

[TheSteve]

I've bought an 8714C and 8711B, both had issues related to the power supply. The RF portions of HP network analyzers are impressively reliable. There is still an 8711B sitting on ebay I know of that should only have a power supply fault(only good to 1.3 GHz though).
I have assembled some 5.8 GHz antennas as well and am still looking for a deal on a cheap/broken 6 GHz VNA. I really only look for ones with power supply or display type faults. For now I use a 6 GHz spectrum analyzer that has a builtin return loss bridge - so I can at least see if they are "tuned".

[hendorog]

I've bought an 8714C and 8711B, both had issues related to the power supply. The RF portions of HP network analyzers are impressively reliable. There is still an 8711B sitting on ebay I know of that should only have a power supply fault(only good to 1.3 GHz though).
I have assembled some 5.8 GHz antennas as well and am still looking for a deal on a cheap/broken 6 GHz VNA. I really only look for ones with power supply or display type faults. For now I use a 6 GHz spectrum analyzer that has a builtin return loss bridge - so I can at least see if they are "tuned".

They look very nice. You could make a transverter to extend the frequency range perhaps?

I've also been using the SA approach too with the Signalhound SA+TG. Actually I might not have taken the plunge on the VNA if I didn't have an SA in case I needed it for troubleshooting.

[hendorog]

Well there was a twist in the tail of this repair.

Sorry this is a bit of a ramble, I had intentions of posting updates regularly but it was a fight against time to get as far as I could before the end of the weekend.

Well I replaced both of the diodes in the -8v rectifier in A15 - and it fired up happy as larry with just the post regulator and cpu boards in. The A15 PSU red LED went off and the green one was on which is good. The A8 postreg board was showing all but one green LED lit for the power rails. The LED which was off was just the front panel power for external probe/bias, and had a blown fuse, so I could ignore that for now.

So looking good and I started adding boards the back in again.

Unfortunately when I nearly had all of the boards back in, the A15 pre regulator reset and then blew one of the new diodes.
Tried again and I got it to the point where it would run, but as soon as I added more boards it would die. Blew another couple of diodes trying to figure out which board was faulty. Lucky I brought 10 of them... 

I started using a multimeter with a temperature probe to keep an eye on the replaced diodes, and was able to nurse it for long enough to see the screen was good and the front panel was working - apart from the missing buttons of course and glitches caused by the missing boards.

I started looking for what was causing the problem assuming that something was overloading that rail, but couldn't get anywhere. The problem rail was -5.2V, and was used in several boards. The confusing thing was that the rail was fused on the A8 post reg board, and that fuse did not blow. So either the problem was prior to the fuse, or there was still an issue in the A15 PSU which feeds it with -8v. I don't have board extenders so there wasn't much I could probe with the boards in place.

Several dead ends later I was still confused by one thing. The -5.2v rail was fused for 4A and the fuse wasn't blowing - so perhaps there really was nothing else wrong. I thought I better find out what the current actually was. Easy enough to do by desoldering the -8v line from A15 to A8. And sure enough it was only drawing 1.6A with all of the boards installed.

The 6A1 diodes were overrated in both voltage and current - 100V vs 60v and 6A vs 5A and were physically much larger so they must be better - yeah right. I even tried using two in parallel, but that didn't help.

One of the original diodes was still working, I'd removed it along with the dead one. I measured it and noticed that its forward voltage read quite a bit lower than the new ones.... Maybe HP had used some flash harry diodes and my replacements were not up to the task? So google got a bit of a thrashing and I schooled up on how to choose diodes in a bridge rectifier.
This page was helpful http://electronics.stackexchange.com/questions/48688/selecting-the-right-bridge-rectifier

The part number was readable so I found the datasheet of the original part [SB560] - (doh, should have done that sooner, instead I was working from the schematic for the older 8753A/B and I couldn't find those part numbers on the web):
https://www.fairchildsemi.com/datasheets/SB/SB560.pdf

Here is the replacement I used [6A1]
http://www.diodes.com/_files/datasheets/ds28009.pdf

Vf = 0.67 vs 0.9
Thermal resistance = 25degC/W vs about 10-20degC/W depending upon lead length.

Then I saw that the thermal resistance of the new diode seems to require a ground plane of 25mm^2. This board didn't have that, so maybe the diode just couldn't dump enough heat out the leads into the PCB? And the physical package being rather large (9mm diameter) maybe not be as efficient at dissipating heat into the air.
The other difference was the original diodes were Schottky. I don't know if that makes any other  difference in power dissipation apart from the lower forward voltage?

So off I went to the old mates at Jaycar again and came back with pretty much every Schottky diode they had over 3A.

First cab of the rank was MBR735 - 35V 7.5A
https://www.fairchildsemi.com/datasheets/MB/MBR735.pdf

Aaaand we have a winner.

I plugged all of the RF cables back in and it has been solid ever since.

Well apart from one loud bang complete with smoke from the CRT enclosure.
The CRT - which is colour and quite dim - turned bright green.
I quickly turned it off, thinking that was good while it lasted....

Crossed my fingers and turned it back on and happy days it still works. I expect was due to dust on the tube.
I've been able to test the source and it looks OK - unlocked as I need to feed the R input to lock it and I've only got one N-SMA adapter.










Here area some board screenshots TiN, sorry just using a phone camera and I haven't yet taken the RF shields off.
I will try and do better ones, and I will do the RF modules as I need to clean them anyway - they are full of dust.

[TerraHertz]

The other difference was the original diodes were Schottky. I don't know if that makes any other  difference in power dissipation apart from the lower forward voltage?

It does make a difference, most importantly in this instance because the lower Vf results in lower dissipation at the same current. This is probably why the first diodes you tried were failing.
In switchmode supplies there are other important parameters too, such as turn-on and turn-off times, and stored charge - all of which can cause excessive thermal dissipation at high switching frequencies.

[hendorog]

The other difference was the original diodes were Schottky. I don't know if that makes any other  difference in power dissipation apart from the lower forward voltage?

It does make a difference, most importantly in this instance because the lower Vf results in lower dissipation at the same current. This is probably why the first diodes you tried were failing.
In switchmode supplies there are other important parameters too, such as turn-on and turn-off times, and stored charge - all of which can cause excessive thermal dissipation at high switching frequencies.

Thanks - I've got a bit of learning to do on switching supplies. I really fell for the numbers on the original diodes.

[Kleinstein]

Turn on time of the diodes is usually not a problem unless you use high voltage PIN diodes and really fast slopes. Even a "slow" diode like the 6A1 or an 1N4001 will be rather fast on turn on - mainly limited by lead inductance.

The turn off time can be important. The stored charge is just a different way to express the same property.

Chances are the lower V_f solved the problem. Power ratings on a tightly packed board can be tricky - this is especially bad with SMT parts.

[hendorog]

The other difference was the original diodes were Schottky. I don't know if that makes any other  difference in power dissipation apart from the lower forward voltage?

It does make a difference, most importantly in this instance because the lower Vf results in lower dissipation at the same current. This is probably why the first diodes you tried were failing.

The turn off time can be important. The stored charge is just a different way to express the same property.

Chances are the lower V_f solved the problem. Power ratings on a tightly packed board can be tricky - this is especially bad with SMT parts.

Hmm.
Even at the quoted worst case Vf of 0.9v and 1.6A the 6A1 diode only had to dissipate around 1.5W.
According to the datasheet the worst case is 20degC/W so only 30 degC temperature rise above ambient.
They were at least 5 times hotter than that even when the current was reduced, so I'm pretty sure Vf isn't the primary driver.

So there must be some other factor at play. Either turn off time,  or the 6A1 simply cannot operate without a ground plane to dump heat into.


 

[orin]

The other difference was the original diodes were Schottky. I don't know if that makes any other  difference in power dissipation apart from the lower forward voltage?

It does make a difference, most importantly in this instance because the lower Vf results in lower dissipation at the same current. This is probably why the first diodes you tried were failing.

The turn off time can be important. The stored charge is just a different way to express the same property.

Chances are the lower V_f solved the problem. Power ratings on a tightly packed board can be tricky - this is especially bad with SMT parts.

Hmm.
Even at the quoted worst case Vf of 0.9v and 1.6A the 6A1 diode only had to dissipate around 1.5W.
According to the datasheet the worst case is 20degC/W so only 30 degC temperature rise above ambient.
They were at least 5 times hotter than that even when the current was reduced, so I'm pretty sure Vf isn't the primary driver.

So there must be some other factor at play. Either turn off time,  or the 6A1 simply cannot operate without a ground plane to dump heat into.

It's switching loss in the diode.  See the first part of this:

http://www.allaboutcircuits.com/technical-articles/switching-losses-effects-on-semiconductors/

The loss is usually negligible in a linear supply at 50/60 Hz, but the switching supply in the 8753C runs at 60kHz, so the switching losses are 1000 times greater since you get the loss every time the diode switches off.  Switching losses in Schottky diodes are much less, hence their use in switching supplies.

Has Dave done a video on switching losses yet?

[hendorog]

It's switching loss in the diode.  See the first part of this:

http://www.allaboutcircuits.com/technical-articles/switching-losses-effects-on-semiconductors/

The loss is usually negligible in a linear supply at 50/60 Hz, but the switching supply in the 8753C runs at 60kHz, so the switching losses are 1000 times greater since you get the loss every time the diode switches off.  Switching losses in Schottky diodes are much less, hence their use in switching supplies.

Thanks that makes perfect sense now. It was quite amazing how cool the new diodes run.

In light of that I will try some other Schottky's with the same axial package as the original. The current ones work well but are not physically robust in that spot.

[hendorog]

Some more progress. I installed these axial diodes instead of the TO220 ones.

Pulled the A3 source unit out and gave it a quick dusting. Got some photos of the board part numbers, but I didn't disassemble completely.

New keypad rubber arrived this morning, installed and that makes it much more pleasant to use - no need for a bent paper clip to press buttons anymore!

I joined the HP Agilent equipment group on yahoo, and quickly found that is a goldmine of information about these units.
Some interesting finds:

* Software options can be enabled by removing an eeprom and modifying the correct locations
http://www.ko4bb.com/doku2015/doku.php?id=test_equipment:restore_eeprom_of_hp_8753
* Options can be enabled by emailing the right guy with your serial and model # [email sent, will see if this is still the case]
* External monitor can be connected via one of these [easy but expensive option]:
http://www.aliexpress.com/item/Industrial-video-converter-GBS8219-XVGA-BOX-CGA-EGA-MDA-RGB-RGBSOG-RGBSync-RGBHV-TO-VGA-CONVERTER/32631766393.html?spm=2114.01010208.3.9.3Gcsw0&ws_ab_test=searchweb201556_8,searchweb201602_2_10037_10017_405_404_407_406_10032_10040,searchweb201603_8&btsid=3114812f-5851-4da5-b29d-b9ec26c469b4
or, [this requires simple hardware mod but is cheaper]
https://groups.yahoo.com/neo/groups/hp_agilent_equipment/conversations/messages/73654

http://www.aliexpress.com/item/Games-video-converter-CGA-EGA-YUV-TO-VGA-signal-2VGA-OUTPUT-GBS8220/541420074.html?spm=2114.01010208.3.1.SLP6F7&ws_ab_test=searchweb201556_8,searchweb201602_2_10037_10017_405_404_407_10033_406_10032_10040,searchweb201603_8&btsid=b558e5e2-ad7d-4f95-a22a-f1240243664b





















Post keyboard replacement:

[hendorog]

A few updates:

I've made a new front panel using the 3d printer, this makes the rubber soft keys much easier to use.
This is used in place of both the bezel/glass part and the menu button cover which were missing.
I've found some genuine parts which are on the way.





Software options are installed and working for 002 (Harmonic analysis), 006 (6GHz), and 010 (Time domain) - thanks to a top bloke on the HP Agilent group.

Have purchased this 85044A T/R Test set for $200 which was a great price. (There was an even cheaper one which looked a bit scruffy, but they ignored my messages about shipping and ignored my offers so I thought stuff them)
http://www.ebay.com/itm/222084316012?_trksid=p2057872.m2749.l2649&ssPageName=STRK%3AMEBIDX%3AIT
and this 7mm to SMA adapter:
http://www.ebay.com/itm/7mm-to-SMA-f-Adaptor-OMNI-SPECTRA-OSM-2082-2700-00-/172255016882?hash=item281b32f7b2:g:3UAAAOSwiYFXHyXl

On the electronic front I've discovered one of the tests is failing: Test 17 - ADC Linearity. The instrument seems to be measuring normally so any error this is causing is reasonably subtle. Noticing that the test passes when run straight after power on from cold, I also got it to pass the test when warm by pulling the A10 ADC board, cooling it (gently) with freezer spray and then reinserting, powering up and testing before it got warm.

Tried to narrow it down by only cooling only one half of the board, but had a wee hiccup when the PSU suddenly shutdown. Fortunately it came back up after a few minutes, probably I overdid it on the cooling spray and caused some condensation. I've got another board on the way as a backup before I try again...

[TurboTom]

Hey, well done repair job, congrats on that! On my 8753C, I had an analog multiplexer (U8) on the A10 digital IF board fail. It was quite difficult to spot since the instrument initially only reported the well-known "No IF found..." message as if there wasn't any RF fed back to the reference input. Funnily, the message disappeared in the upper frequency regions if a reference signal was fed back. But still the analog bus readings were all over the place. Fortunately, there are many test points at the top edge of the PCBs so step by step I could narrow the problem down. The schematic available on the web (for the 8753B) fortunately is similar enough to be of considerable help.

I replaced the HI1-518-5 multiplexer with its "commercial" counterpart HI3-518-5 since an original spare wasn't available. The instrument performs well after the rapair and all readings are correct again (as far as I can tell). It seems the multiplexers are somewhat prone to degradation/failure so yo may check yours as well.

Good luck and all the best,

Thomas

[dr.diesel]

WooHoo, I just bought one of these busted on eBay as well.  Anxiously awaiting for it to arrive.

[TiN]

VNA spree beware .

hendorog
Are you considering to get LCD kit for your box?

[hendorog]

Hey, well done repair job, congrats on that! On my 8753C, I had an analog multiplexer (U8) on the A10 digital IF board fail. It was quite difficult to spot since the instrument initially only reported the well-known "No IF found..." message as if there wasn't any RF fed back to the reference input. Funnily, the message disappeared in the upper frequency regions if a reference signal was fed back. But still the analog bus readings were all over the place. Fortunately, there are many test points at the top edge of the PCBs so step by step I could narrow the problem down. The schematic available on the web (for the 8753B) fortunately is similar enough to be of considerable help.

I replaced the HI1-518-5 multiplexer with its "commercial" counterpart HI3-518-5 since an original spare wasn't available. The instrument performs well after the rapair and all readings are correct again (as far as I can tell). It seems the multiplexers are somewhat prone to degradation/failure so yo may check yours as well.

Good luck and all the best,

Thomas

Golden info on the multiplexers, thanks! I was looking at the items in the signal path wondering what the most likely culprit could be.

That gives me somewhere to start as tracing a linearity fault through the board didn't sound like much fun

[Vgkid]

Nice job on the new front display.

[hendorog]

WooHoo, I just bought one of these busted on eBay as well.  Anxiously awaiting for it to arrive.

Nice one! Looking forward to hearing more about the repair.
Which one did you get?

hendorog
Are you considering to get LCD kit for your box?

I thought about it but decided probably not -  I've run the CRT adjustments and got it to the point where its OK to use in daylight and perfectly fine in the evening.

The LCD kits didn't seem to be marked as compatible with the C model either. Other seem to have installed them though so probably not an issue there.

I'm considering getting one of the cheap modules mentioned above to plug in an external monitor - which would be easier on the eyes than the small LCD.

How are you guys getting on with your one?

[hendorog]

Nice job on the new front display.

Thanks - forgot to mention I posted the files on thingiverse in case anyone needs it.
http://www.thingiverse.com/thing:1675565