Agilent 4396B repair odds

[mark03]

I may have an opportunity to pick up one of these "combination" analyzers (spectrum, network, impedance) in non-working condition.  I'm curious what the general opinion is for this model, particularly VNA performance vs an 8753 C/D/E.  But mainly, how do you estimate a fair price for something which would sell for over $2k in working condition, when the repair odds are something like 1/3 easy fix, 1/3 tricky repair journey, and 1/3 not economically repairable?  Symptoms are "powers up, no display" and I don't have access to the unit to test the VGA output on the back.

An independent repair outfit, siliconinvestigations.com, claims "these analyzers were notorious for having display problems and failed power supplies" which I guess could be good news?  Hard to say.  I know of no documented LCD replacement, if that's the problem, but it was fairly straightforward on the 8753E which is similar (different LCD though).

Curious to hear how seasoned buyers of defective TE approach something like this.

[DaJMasta]

It's a pretty decent risk, while the fault could certainly be display or power supply related, they're also YIG oscillator based, and there are a lot of RF bits that can be challenging to troubleshoot without RF gear.  I don't mean to discourage too much, but especially if you're unfamiliar with that sort of gear or repairing it, I wouldn't put the odds of it being a simple failure and nothing more without any issues that manifest outside of what it can test on it's own (verifying output and input with external instruments).  I've got a 4395A as a project currently that was sold in what appeared to be working condition but was entirely untested aside from powering up - passes all self tests, but the output is way off frequency for a chunk of the sweep which manifests as noise that can't be normalized on the trace (not to mention no actual valid measurements at some frequencies)... it's currently low in the stack of projects.

I think the odds of your perspective unit being unrepairable are pretty low, but I would expect the repair to be somewhat in depth.  There are parts and assemblies from other units available, so it's not like they'd be impossible to find, but they could be expensive and there is some likelihood that you'll need to get one.  Testing the monitor output is extremely easy - a <$10 ebay special VGA to usb capture device can do it sometimes, but I got a nicer VGA input LCD at my local thrift store for $15 and it does a better job (wider signal compatibility).  There is a service manual available, but it's not schematic level, just assembly with block diagrams (at least, for the 4395A), and you'd have to see a lot of 4396As to make any real statement on reliability on their PSU or LCD... so I wouldn't put much stock into it.


So all in all, it's a gamble, as you probably already knew.  You've got to weigh that against how much you want it, how much experience/ability you have to repair something like it, and how much you can afford (in time and cost).  I think the general rule of thumb is to not gamble much and never assume it will be an easy repair, but if there's a chance of it and you're equipped to troubleshoot for the long haul, you're likely to get a working instrument out on the other side.

[mark03]

Indeed, a problem like the one with your 4395A would be well over my head.  I guess the real question I have (and maybe it is unanswerable) is, does P(difficult repair) = P(difficult repair | display is not working)?  In other words, does knowing about the display problem affect the underlying (prior) odds of something more serious, like your 4395A?  I can make an argument in my head for them being equal or unequal in either direction.

A few years ago I acquired four 53310A modulation-domain analyzers at auction.  One had no display, and the reason was not that anything was defective; someone had simply cannibalized the display board!  So I do worry that an analyzer with an unknown failure could have been used as a parts donor... among other things.

[DaJMasta]

If the cal seals are intact, you can be at least somewhat sure the parts are still inside it, the trouble is that even if the display was working, the self tests don't cover every little thing aside from calibration, though as with most equipment they are comprehensive.

If there's a way to verify that the instrument responds to input - i.e. you could change the active channel and see the lights change or you could hear relay clicks on startup or when changing reference/output level, then you can mostly rule out power supply and CPU board, at least, but if all you have to go on is pictures, that's still a gamble.  While a display out could be power supply, CPU board, display board, firmware corruption, backlight dead, or display itself dead.... it also prevents you from diagnosing almost anything else at a glance.


Sounds like it may be a bit risky for you.... but if you don't mind a challenge and are willing to bone up on your RF test equipment understanding (I'd point you towards TheSignalPath on youtube!), then factor the extra risk into your price.  If you can get it for cheap enough, maybe it's worth your investment.

Worth mentioning that you should consider a test set (S parameter or T/R and maybe even an impedance fixture) into the total price of using it... and at least that impedance analyzer fixture can be pretty pricey for whatever reason.

[mark03]

Repair journal here: https://www.eevblog.com/forum/repair/agilent-4396b-spectrumnetworkimpedance-analyzer-repair/