Advantest R3273 Spectrum analyzer stuck in LOADING screen

[baltersice]

Hi Guys, my ambitious repair project isn't progressing, so I wanted to ask for some hints:

Got a sick SA that doesn't seem to boot:

I have left it in the loading screen for 20 minutes, nothing changes.

Found an interesting serial interface, but it too gets stuck without putting out anything useful (to me at least):


Both previous owner and I tried swapping RAM chips with NOS parts, but that changed nothing or made it worse:

Changing the red marked Toshiba TC551001CF for what looked like brand new ones, caused complete deadness, not even self test was executed.
All the EPILDA Chips are new, solder joints were inspected with a microscope and look good. RAM self test passes too, but that could be testing the main processors RAM?

The Main processor is a Motorola MC68340, the flash card contains some interesting verbose files and some binaries: https://drive.google.com/file/d/1XZMoc3Zd86UlXWMjUhN6VH5KTtQ_qpTr/view?usp=sharing
But nothing in there has helped me so far.

I have changed some parts in the switching power supply and now I am very confident that it intact.


J12 was my serial interface, unpopulated J20 has some continuous high frequency digital signal going on, doesn't look like something meant to be read by human. J10 is silent permanently.
I have looked at the board with a thermal camera and found nothing suspicious.
I have unplugged everything (Coax connectors, floppy drive, front panel, all the RF boards, rear interface board) and that only caused more errors in the self test and the instrument didn't proceed to the loading screen any more.
I have temporarily replaced the backup battery, which fixed the appropriate self test fail, but didn't make the initial problem go away.


The previous owner has a working one of these and might be coaxed into making a measurement, if that helps me.
Apart from that the only option I see now is play around with the chips around the large Advantest labelled devices.
I wouldn't be surprised if the initial self test is only performed by and on devices around the Motorola processor in the top right corner.
As soon as it sends the boot message via serial it might hand over to the Advantest devices which may keep their errors to themselves.
In particular I could remove individual RAM chips and see if that changes anything visibly. If not, then that chip might be broken without a visible sign.

Thank you for reading, hope we can solve this puzzle, so that more interesting HF stuff can happen

[PA0PBZ]

I have temporarily replaced the backup battery, which fixed the appropriate self test fail, but didn't make the initial problem go away.

This seems strange: I guess the backup ram error is some kind of a checksum error or at least it is checking the data in the backup ram somehow, so how is replacing the battery going to fix that? Is there some kind of initialization/reset option, maybe powering up the device with a switch/jumper in another position? It could well be that the corrupt data in the backup ram is making the thing hang in bootup…

[baltersice]

with a switch/jumper in another position?

Thank you! Just verified that indeed the self test fail 'Backup RAM' changes from NG to OK if I equip a 3V backup battery and restart the unit once.
I have tried all the single DIP switch positions, 1=On tries to boot from floppy, 2=On was the way it was delivered to me, 5=On starts a display test. At some point I saw random date / time data in the top right corner of the display with 3=On, but I can't seem to reproduce it now.



All the ELPIDA HM538253BTT-7 in the bottom left are marketed as Video RAM and I think that is what they are used for here: If I remove one, the displayed image gets vertical lines, but is still recognizable. Is it safe to say that problems with the video ram would not cause the instrument to get stuck in the loading screen? Then I'd focus on throwing other random replacement parts at it for now.

The previous owner told me, that both of the LEDs (D15 and D14) eventually start flashing in a working unit. They are labelled 1 and 0 and are connected to the largest ADVANTEST chip next to them and GND. Does anyone know what they could mean? I have double-checked, it's numbers 1 and 0 but maybe they mean input / output? Maybe the large chip is receiving it's firmware but not sending out any data out after that?

[PA0PBZ]

Did you check the service manual?
http://www.ko4bb.com/manuals/92.70.248.105/Advantest_R3267_R3273_Spectrum_Analyzer_Service_Manual.pdf

[baltersice]

Yes, I have. It contains 95% performance verification and the remaining 5% show how to replace failed modules

[PA0PBZ]

Did you check the contents of the flash card with a known good one? I think I still have an image somewhere…

(from https://www.eevblog.com/forum/testgear/advantest-r3273-spectrum-analyzer-repair/ )

[baltersice]

Yes, I have. Tried both images, same result.

[PA0PBZ]

Does the behaviour change when you take the flash card out?

[baltersice]

Yes, it says PLEASE INSERT SYSTEM DISK in that case.

But that gives me a good idea: It displays the files it is loading on the bottom of the screen.
I could try and modify a file that has been read fine before to see how it behaves then.
Does it complain with a proper error message or will the display just freeze in the LOADING screen maybe?



Nope in that case it puts out a weird and misleading, but clearly recognizable error message

[SaabFAN]

Does the unit have a fan? My R3131 has a quite noticable fan in terms of noise created

Also I had to press a button on the processor-board to clear the NV-RAM to get rid of any errors during the self-test. Maybe your NV-RAM has some random data in it that screws up the boot-process.

If that doesn't help and the software seems to be ok, my next guess would be either that error-message concerning "Block 10 not detected!!" or a problem with RAM at higher addresses.
The CPU seems to be OK, otherwise it wouldn't say "ok" after unzipping the data from the Flash-Card.
Check for CS-Signals of the RAMs. They are most likely generated by some 74-Logic chips near the RAMs. And also check all connections between the CPU-Board and the other components.

Do you have a logic analyzer?

[baltersice]

I does have an audible fan, but not unbearable
The only button on the MB seems to be a reset kind of deal.

I started probing around some of RAM ICs finding some odd 'rounded' signal all over the board:

Pretty sure that in a mostly digital section nothing should look like that??

I also got something entirely new by probing a specific pin while booting:


Maybe it would have booted if the coax cables had been plugged in then?

[PA0PBZ]

Pretty sure that in a mostly digital section nothing should look like that??

That's normal

Why the vertical red line on your display that shows in a lot of your pictures/movies?

Also, the question marks on top, any idea what that should be, date/time maybe?

[SMdude]

What were the errors in your earlier vid?
Looked like back up ram
and the menu.x Not enough space?

Just curious if you have tried warming things up a bit?
I have had bad flash roms in embedded systems that would not work while cool, add a bit of heat and you would never know that there had been a problem! Till it cools down again..
Worth a shot if you haven't tried already.

[baltersice]

Not sure about the red line, I'd count it the many signs of wear and tear on the device.
Yes, the questionmarks will be replaced with date / time when that is set.

Self test error Backup RAM is reported when no backup battery is equipped.
menu.x not enough space occurred when I deleted half of the menu.x file on the CF card, to see how it would react.

add a bit of heat

Haha, what a good idea!! I have actually tried Daves old inverted air duster trick, but heat not yet!
It would make sense, because an hour ago I got further than ever before in the boot process after probing around a lot and consequently letting everything warm up.
Well, what do you know ... It's not fully ready to analyze spectrums, but it's the most promising sign of life so far:

[SMdude]

Awesome!
Now, when it cooled down again, did it still boot?
The tedious bit is to narrow down which part exactly...

[nctnico]

Also I had to press a button on the processor-board to clear the NV-RAM to get rid of any errors during the self-test. Maybe your NV-RAM has some random data in it that screws up the boot-process.

I'd start here. If it boots and display an error message then it is very unlikely the CPU part is bad. With a different Advantest SA I had to keep a button pressed on the CPU board during startup to clear the NV-RAM and get it to work again. It probably is something simple with this SA as well.

Don't swap components or probe stuff when you aren't 100% it is broken. You likely ruin more than you fix.

[baltersice]

https://www.youtube.com/watch?v=9L_hi7juvmE

It is almost all good now thanks to SMdudes idea! I've 'resoldered' the big Advantest chip next to the video RAM. (Would it be crazy to call that a GPU?)
Now only warning windows and pop up messages are broken ... ?






And I don't have a step attenuator installed yet ... 400$ replacement part

I also tried fixing the the red stripe on the display, but nothing helps or even changes, so I bought a reasonably priced replacement display.

[baltersice]

Gentlemen, little bit of an update:
The spectrum analyzer almost worked for a while! As well as one can expect, given that it still doesn't have its step attenuator:


It did have a dead red column of pixels and all pop up messages / windows / overlays were distorted.
I threw replacement parts at it until these problems were no more - it was the Elipida VRAM chips.
Now there's an all new problem



Any qualified engineers around? 
Join the Discord: https://discord.gg/vyBnAq

[SMdude]

Errr, had you been fiddling with anything just before you got that error?
Connector not seated(though you have probably checked them all before posting )
Dropped a power rail??

Heat, cool and wriggle test?

I'm glad I didn't buy that one!!

[Pat_G]

I was wondering if you made any progress with the analyzer since your last post.  I have an R3267 which looks to have an identical mother board to yours.  My unit worked when I first got it but now it gets stuck at the blue loading screen and all the front panel green LED's start blinking.  I am going to try applying some heat later to see if that changes anything.  It will occasionally start up properly.  I have not tried replacing any parts until I narrow down the bad components.  I don't want to cause any new problems.

[baltersice]

Really? Wow, what a coincidence! So I fixed the getting stuck in L O A D I N G . . . screen by resoldering all the pins on the large ADVANTEST chip next to the VRAM (lets call it GPU?). I think that is what did it in the end. But before that I also replaced all the RAM chips I could find reasonably priced replacement parts for, that is all except for the OKI ones.

Now I am occasionally getting the Cannot detect Mod. DSP Board error and that too seemed like a thermal issue, but not consistently. I've tried warming the main board and the DSP board without finding a consistent behavior. I have also checked the vertical connector board for bad solder joints but found none. Yet.

Am a bit stressed out at the moment, but I'll surely post here when I find something new.

[baltersice]

Still stressed, but I've got a little bit of progress to report:



I've fixed all the startup problems for good as described above. In dozens of boots I haven't seen a single error. I've also paid dearly for a new step attenuator.
Now when selecting Full Span the trace goes all the way to the ceiling. I'm being told that that is normal, can anyone confirm?
It's not really a problem, since it really only occurs in full span, as demonstrated here:

https://www.youtube.com/watch?v=NadX6FRe6Jw

Now 2 more, possibly related problems:

• Self tests are failing
• There's a DC bias on the external mixer input

In most other devices I've seen, there's a way to see the actual self-test results vs. values that are required for 'PASS'. What R3273 is showing me is not very helpful.
Maybe fails are guaranteed because of new step attenuator and missing calibration? Any ideas?

The previous owner has disconnected something internally and added these loose cables instead. It is apparently hard to troubleshoot that latter DC bias problem without a riser card.
Maybe I can take the board out and check some components for unwanted continuity?

[hafrse]

Still stressed, but I've got a little bit of progress to report:



I've fixed all the startup problems for good as described above. In dozens of boots I haven't seen a single error. I've also paid dearly for a new step attenuator.
Now when selecting Full Span the trace goes all the way to the ceiling. I'm being told that that is normal, can anyone confirm?
It's not really a problem, since it really only occurs in full span, as demonstrated here:

https://www.youtube.com/watch?v=NadX6FRe6Jw

Now 2 more, possibly related problems:

• Self tests are failing
• There's a DC bias on the external mixer input

In most other devices I've seen, there's a way to see the actual self-test results vs. values that are required for 'PASS'. What R3273 is showing me is not very helpful.
Maybe fails are guaranteed because of new step attenuator and missing calibration? Any ideas?

The previous owner has disconnected something internally and added these loose cables instead. It is apparently hard to troubleshoot that latter DC bias problem without a riser card.
Maybe I can take the board out and check some components for unwanted continuity?

Hi ,
Regadring "...the trace goes all the way to the ceiling...."
I have 2 R3273, when the CMOS battery dies and after replacement,  the trace goes all the way to the ceilling on full span. It seems that there is a bug in the software,  some parameter is lost with the original battery! very weird .
Regarding the self calibration, at which step it fails?

[mvanocht]

Hello, I also have an R3273 and the NVRAM Battery went bad.  Now..."the trace goes all the way to the ceiling...."
,  the trace goes all the way to the ceiling on full span. I agree it seems that there is a bug in the software,  some parameter is lost.  Anyone figure out how to solve this?

[hafrse]

Unfortunatly, this happens when the CMOS battery dies... I had 4 R3273, same thing happened when I replaced the CMOS battery!!!