bricked HP89410a with an attempted firmware update

[vcoguy]

I noticed that my (perfectly working and reliable) HP89410a Vector Signal Analyzer firmware had never been updated since it shipped in 1993.  Firmware for this model was updated as late as 2002, and is available online. 

Downloaded the latest FW from Keysight, followed the directions, created 3 floppies, booted with floppy #1, started the update, it wrote 1/3 of new version to internal flash, asked for floppy #2, I put it in, and.... totally locked up.  I let it run for several hours.  No response, had to cycle power.  Tons of bootup ROM checksum errors (because it didn't finish), then it halts, even before the normal HW checks.  The system can't do anything at that point.  Never finishes booting, no response to any buttons.

Went back and booted off floppy #1 again.  The system proceeded to restart the firmware update process.  Same story.  Floppy #1 completes successfully, system reads and unpacks file on floppy #2, begins writing flash, and locks up.

There are special notes about models with very early serial numbers needing a HW update before updating firmware, but my serial number is newer than the cutoff for the special procedure.

Any ideas?  I'm about to re-create a set of 3 new floppies and try again.  I see few alternatives beyond that, other than buying a new A42 memory module on Ebay.  But... then the required service procedure, immediately after install of A42, is to update firmware, which I would now be terrified of doing.

Aaarrrggghhh!  I know, I know... if it ain't broke, don't fix it!

[SilverSolder]

All floppy drives are old and creaky today... there is a good chance that a floppy issue is all there is to it.  The simplest answer that fits the facts is a marginal incompatibility between the disk drive used to write the data, vs. the one reading it (heads not perfectly aligned with each other).  I have seen that issue dozens of times back in the day when floppies were common, and always managed to find a source drive that worked... 

So, I would try formatting and writing the data on a different floppy drive (even a few different ones, if the first one doesn't work)  and see if you can find a "writer" that the drive in your analyzer likes a little better.

If that is not so easy in 2021, you could try taking the drive out of the analyzer and connect it to your PC, and format/write the data on the exact same drive that is going to be reading it back later.   That stands a very good chance of working.

[vcoguy]

I turned out that the floppy DRIVES were OK.  But floppy DISK #2 was also 30 years old, I guess.  The media was old and creaky, as you put it.  Time to throw that one out!

Found some newer-ish ones, formatted them, ran CHKDSK or SCANDISK or whatever it was called in the MSDOS days, to verify no issues.  Recreated the firmware disks, ran the update, and all is good.

Thanks for your help.  The simple solutions need to be checked out first! 

The old HP89410 VSA is back in business, with the last firmware update it will ever need.

Time for a beer.

 

[SilverSolder]

Nice work.  -  It is almost always something simple that causes the most problems! 

Interesting instrument, by the way, never "noticed" this one before.  What kinds of things is it good at?

[vcoguy]

The HP8910a (baseband) and HP89440 (with paired 2.65 GHz downconverter) were designed as very fast, FFT based spectrum analyzer / measuring receiver for complex I/Q digital modulation, like modern GSM / CDMA cell phone signals, digital TV, etc.  It's a big fast ADC digitizer front end, and lots of DSP math to analyze and process the signals and results.

As a scalar spectrum analyzer, it's quite good.  A little slower on broad sweeps, but very capable.  I've measured oscillator phase noise with it. But as a vector signal analyzer, it really shines.  It was quite ahead of it's time, back when it came out.  There was a whole writeup in the HP Journal Dec. 1993 about its design.

Any oscilloscope can show you amplitude vs. time.  For analog demodulation, this thing can show you AM, FM, or PM vs time.  One button measurements of PLL synthesizer lock time.  Or for digital modulation, it can display modulation symbols and decode the data.  It can display and measure the modulation impairments: transmit distortion, receiver error vector magnitude, etc.  I tracked down a lot of EVM problems back in the day.

Or you can use the two input channels and display correlation.  I've hunted down unusual ripple in an amplifier output by checking how much the ripple correlated to suspected "culprit signals" nearby on the PCB.

It has a source (which can be a swept sine wave, additive white Gaussian noise, or can be loaded with your own arbitrary signal).

One time I was stumped by how to measure the baseband IF filter selectivity for a RF receiver chipset that integrated the entire IF filter inside the chip.  Normally I would be able to sweep an RF input signal across the receiver frequency, and measure how signal strength was present at the output, as I swept across the band.  But this chip also had an AGC automatic gain control amplifier integrated, which I couldn't disable, that would vary the gain as the test signal swept through the desired frequency, thus messing up my measurement.  The VSA came to the rescue.  I would up driving the input with AWGN noise, so the AGC detector always saw a constant level, and didn't vary the gain.  Then probed the input with CH1 and the output with CH2 and set it up to calculate and display the CH2/CH1 transfer function.  Gain and phase!  A perfect solution.

(Something to think about:  How to measure gain and phase transfer function with a sinewave is easy.  Measure amplitude and phase shift at output on an oscope, right?  Look at zero crossings for phase, etc.  But... how to measure gain and phase when both signals are noise?)  What's the phase shift of one messy noisy waveform vs. another messy noisy waveform?  Not just one frequency, but a bazillion, all superimposed on each other.  Now do it again, with significant time delays due to propagation or filter group delay effects... Hard for my brain to visualize the math in that case.  But the 89410 had no problem.  There is a whole lot of sophisticated FFT and digital signal processing in that instrument, that is quite frankly, beyond me.

It's great a very low frequencies.  I even measured low frequency (sub 1 Hz) power supply R-C filter imperfections with this thing.  Like ESR and dielectric absorption effects on supply filter rejection.  It's a great DC coupled, millihertz vector network analyzer, where traditional RF VNA's fail because their direction couplers roll off, and directional bridges also fail to work.

I played with one of these at Motorola when they first came out in in the early1990s.  No other engineers in the lab were as fascinated with all the possibilities as I was.  Years later, I worked at Agilent, and worked a few cubes down from one of the HP program managers who worked on the design.  He enjoyed hearing me talk about all the ways I used it for tough measurement problems.

I have watched them on Ebay for all these years, and now have one of my own.  Fully loaded, it was $72K.  28 years later, you can buy one for a couple percent of the original list price.

I'm still learning new things to do with it.  Just a really cool instrument.

[SilverSolder]

It does indeed sound like a very cool instrument!  It seems to do much more than a spectrum analyzer, and going all the way down to DC is what peaked my interest in the first place.   It definitely looks like something that can provide a lot of entertainment and education just to understand what it does! 

Are there any "gotchas" when buying one, or are they generally pretty robust?

[vcoguy]

The part numbers and options were (These are from my notes and partially from recollection, I think they are correct)

HP89410A DC-10 MHz Vector Signal Analyzer - The base standalone unit that goes to 10 MHz.
HP89440A DC-1.8 GHz Vector Signal Analyzer - Adds a 1.8 GHz downconverter; in that case the HP89410A part is called the IF unit, and the downconverter is the RF unit
HP89441A DC-2.65 GHz Vector Signal Analyzer - Adds a 2.65 GHz downconverter.  I think this one replaced the HP89440A just a few years after it was introduced.
HP89430A HP89431A were the standalone RF downconverters, that could be purchased later on, to convert an existing 89410 DC-10 MHz unit into an equivalent 1.8/2.65 GHz 89440/89441 system.
HP89411A 21.4 MHz downconverter - This is just a smaller box that downmixes a 21.4 MHz IF output of an existing spectrum analyzer down to 5.6 MHz, so it lands in the 10 MHz band of the 89410.  This allows someone with an existing spectrum analyzer to have all the same VSA capabilities, by using the spec analyzer as a front end, and the VSA as the back end.

I never had one of the RF downconverters during my time at Motorola.  I just used a RF signal generator, and a Mini-Circuits mixer, to downmix any RF frequency, down to 5 MHz.  One extra manual step to do, but it saved money and bench space at the time.

Today, mine is one of those unusual retrofitted 89431A hybrid units, that is labeled DC-10 MHz on the top box,  but actually has the full 89441 capability to 2.65 GHz.

Option AY7 (Second 10 MHz Input Channel) was a must for me.  That one allows the CH2/CH1 network analysis measurements that I described.

Unless specifically disabled via the menus, the units do a self cal / self zero intermittently while being used.  Maybe every 30-60 minutes or so?  It's about 30 seconds of relays repeatedly clicking and clacking, while measurements stop.  When I first heard one do it, I thought that those poor relays would wear out over the years.  However, my 28 year old machine seems to work fine after thousands of self calibrations.  So they must be pretty robust.

The self-cal info is stored in battery backed NVRAM.  The lithium coin cells die over time, so the clock/calendar resets to May 1st 1986 or something.  The lithium coin cell is a somewhat oddball part number, BR2325.  Rare enough that you won't find it at most neighbor stores, but it can be ordered online.  You can slide out the NVRAM battery memory card from the back panel, to replace it, without taking the whole unit apart, so that's convenient.

The anti-glare screen over the front of the CRT tends to get blurred over time, near the lower corners.  I think the plastic in that location outgasses with heat, or maybe UV light affects it.  Many old units seem to show that.  It's just a cosmetic issue, the CRT remains readable, and functional.  I think that anti-glare screen also has a super micro fine wire mesh that serves as an EMI/RFI shield to keep stray signals out of the instrument.  So it's not as easy to replace as one might think. Some people have done a CRT to LCD replacement. 

Quite a few units that I see on Ebay have a black floppy drive, which has replaced the original beige one.  I don't know if those original beige ones had issues or not, but it's curious to see so many units with the wrong color floppy drive poking out.

[SilverSolder]

Do they have any kind of self check feature during boot or something, so you can quickly gauge if a unit is in serviceable condition or not?

I have little experience with downmixing RF.  What little I have tried (downmixing to look at a signal in a spectrum analyzer) seemed to introduce a lot of artifacts, to the point that it becomes difficult to tell which is real, and which is an illusion (to misquote the Moody Blues)!

[vcoguy]

Yes, a power up self test checks all the boards, memory, optional I/O, communication to downconverter (if there is one) etc.

Downmixing isn't so bad with one of these, because you have to mix down to somewhere DC-10 MHz, it samples at 25.6 MHz, and so it has a good antialiasing lowpass filter on the front end.

So if you wanted to see a signal at, for instance, 150 MHz, you would run that into a mixer RF port, and feed the LO port with a sig gen at 145 MHz.

The sum and difference frequencies would be created at 5 MHz and 295 MHz, and there would be LO feedthrough at 145 MHz, RF feedthrough at 150 MHz.  All of those would be well attenuated by the antialiasing filter, except for the desired 5 MHz term.  You could also run your LO at 155 MHz, with similar performance, except your downmixed spectrum would be inverted at 5 MHz.

It's tougher if you want to look at something just out of band, such as, for instance, a 10.7 MHz IF on a radio.  Now all of those signals are close enough that they might bleed through the filter, create even more mixing products, and cause some trouble.  Even that HP89411A downconverter for 21.4 MHz mixes it up to 45ish MHz as a first step, then takes it back down to 5.6 MHz , to avoid those sort of issues.

There is a nice convenient amber overdrive LED next to each of the input BNC connectors.  It quickly tells you if you're overdriving the input ADC, so you know to back off the drive level.  But if you don't see the adjacent green half scale LED lit up, then you know to increase the gain, because you aren't exercising all the bits in the ADC.  Hitting that front end at the right amplitude will keep it linear enough to avoid trouble, so long as you keep your mixing products out of band.

[SilverSolder]

What happens with a downconverter if you want to sweep a whole range of frequencies -  does the downconverter do the actual sweeping, or does it stay at a fixed frequency while the 89410a does the actual sweep?

[maxwell3e10]

This seems to be the closest to a general thread on the instrument.

I noticed on the back of 89410A there is a "Gate in" BNC input that is not populated. According to the service manual it is connected to A71 pass-through board, but otherwise I couldn't find any documentation. Does anyone know if it is associated with a certain option and what does it do? I would like to have blanking of the signal  but not associated with triggering and wondering if it may have that functionality.

[pquadrat]

I know, this is an old thread, but this is an update to the (unanswered) question in the first post.

I also got a HP 89410A with old firmware, got the firmware from Keysight, and started installing it.
After some trouble getting anything from the floppy drive, it started with the update, but after the restart,
it threw many checksum errors.

I was not able to get the Epson drive to work reliable.

One importatnt note: As You can see on the pic, pin 1 of the floppy cable is NOT on the side with the red marking.

The Epson SMD300 drive gets it 5V power on pins 7, 9 and 11 of the 34-pin header. Prices for these drives
on ebay are crazy, so I took a known good Teac FD-235HF drive, built an adapter cable, and it did work.

There seems to be a drive select problem, the drive LED light up all the time, but it did its job. This is probably
the last time to use the floppy, so I woill put the EPSON drive back into the unit, and keep the adapter cable.

Now the machine is working again, just gives a message about the firmware of the LAN module, which I bought seperately on ebay.
please see the picture.

Does anyone know how to update the firmware of the LAN adapter?

Thanks,
Peter

[pquadrat]

The LAN card firmware problem was easy: I had to put the card into the machine BEFORE starting the update.