SONY/Tektronix AWG520 2ch 1Gs/s with no output - repair attempt and teardown

[Martin Hodge]

Yes I got my 16MB SSD imaged using your ftp method. Then set my nic back to it's odd config at base 240. The Tek software couldn't see the nic otherwise, but there were no errors. I did however use my old PATA equipped PC with an adapter to image the 3.25GB hdd. I wrote the image to a CF card and verified it works in place of the spinning rust in the 610. We really need a way to generate vxworks partitions though.

How were you able to hack your kernel to read vxext? Did you use the jens-maus driver on github? I'm limited to 32 bit kernels on this old Pentium 4 and compiling a ko worked, but modprobe threw linker errors for an undefined division symbol and complained of "kernel taint" (using centos 7 minimal)

Could you PM or post here your driver hacks?

[DC1MC]

Yes I got my 16MB SSD imaged using your ftp method. Then set my nic back to it's odd config at base 240. The Tek software couldn't see the nic otherwise, but there were no errors. I did however use my old PATA equipped PC with an adapter to image the 3.25GB hdd. I wrote the image to a CF card and verified it works in place of the spinning rust in the 610. We really need a way to generate vxworks partitions though.

How were you able to hack your kernel to read vxext? Did you use the jens-maus driver on github? I'm limited to 32 bit kernels on this old Pentium 4 and compiling a ko worked, but modprobe threw linker errors for an undefined division symbol and complained of "kernel taint" (using centos 7 minimal)

Could you PM or post here your driver hacks?

Excellent news, I've also imagined the HDD via a PATA-USB adapter, the ftp method would have taken too long, but for somebody that is not yet ready to open the device and replace the drives, is an excellent method to be able to make backup copies of the drives content for future use. Because again, if the SSD dies, or gets corrupted, is game over.

I'll post tomorrow morning the source code of the modified driver, but most likely it will not be useful on older kernels, but for modern ones. The original driver should actually work on Centos 7, the "Tained" message is harmless and stems from the driver maker not including some GPL license marker in the source code, so the "ideological purity" of the kernel is tained, go figure  .

The linker error, that's a bit strange coming after the module compiled without errors, usually after "make install" I do a "depmod -a" to have all de post-compile dependencies fixed. I think a Live CD of Ubuntu 18.04 LTS or Mint Tricia and the modified driver will make a better solution to this.

The modified Linux device driver is able to read and write perfectly the loop mounted images AND the actual devices (I've used the actual HDD), as I don't have yet a method to create vxfat  partitions from scratch, I did a trick, I've deleted all the files on a drive image, the copied the first 4MB of the image and use it as a "partition image" for a new drive, worked like a charm, and if is stupid and it works, then is not stupid  .
Sometimes a proper "mkfs.vxfat" should be done, but as I don't believe that the OS and the partition type will go higher than 2GB, I think this humble solution should suffice for most of the people.

 Cheers,
 DC1MC

[DC1MC]

Hi Martin, for some reason your latest post that I was going to reply has disappeared, I think is the second time when this happened  .

In any case I've attached the archive with the modified driver from Jens Maus, it seems that unfortunately the creator has lost interest in it, my attempts to contact him have failed  , so I'm putting it here maybe it helps somebody with a modern distro (kernel 5.4+). On my Linux Mint Tricia (Ubuntu 18.04 based) it worked perfectly for reading and writing the files.

Regarding the strange macro that fails in your situation, I did a bit of googling and naturally, is a RedHat specific "optimization" that found it's way in your Centos7 machine. Unfortunately I only have access to 64-bit Centos machines, so I could not debug it, but I've found it refereed many times in Internet, at one time was even impossible to build the kernel itself because of this macro missing. The preferred solution I thin was to insert a short assembly sequence to convince the compiler to disable that optimization, kind of messy, IMHO.

Right now I have a fight with creating a programmer for the RTD2660 LCD boards because I have a feeling that I'll need it soon when the Chinese VGA screen will arrive, but mostly because one forum friend donated some 5" LCDs and their driver boards that may fit the bill for the screen replacement.

 Cheers,
 DC1MC

[Martin Hodge]

Hi Martin, for some reason your latest post that I was going to reply has disappeared, I think is the second time when this happened  .

Wrong_button_oh_noes = Attempt_edit(Fat_fingers, Tablet_keyboard, Old_age) ;

Thanks for the source, it'll at least be educational.

[Martin Hodge]

I cannot get my 610 to recognize anything other than the included 16MB SSD connected to the IDE master. I investigated the power issues you talked about on your 520, but on the 610 the two 5v pins are tied together both on the motherboard and on the SSD. So not realizing I've never actually looked at the SSD, I took it out of the PC cage and look what I found. 

The BIOS config has no problem seeing whatever I plug in to the master and reading correctly it's CHS and size, but every OS I've tried fails to work with it.

-edit-
I don't know what I was doing wrong, but after trying again this is no longer an issue. 

[DC1MC]

Hi Martin, welcome in $current year  !!!

So IBM or whoever was producing this enormously overpriced SSD run out of SSDs and used my solution, but with a crappy PCMCIA adapter  .

OK, so let's see what did you put there, as you may remember, I've only got the OS to work with 32MB CF card, in that double adapter, nothing else was working. I do plan to try with higher capacities CF cards (I'm tracing some 64, 128 and 256MB on  >:Dbay, let's see if I'll be able to get some).

IMO, there are two ways to proceed:

- Step one: check if a "normal" DOS works correctly, boot from a DOS floppy and try to fdisk/format the master device that you intend to use and try to boot form it, assuming of course that is recognized in the BIOS, initially try with single device, also check the master/slave/single configuration pins. The DOS should boot.

- Step two: once you're sure that the you can actually boot something, dd the image to the new device and IMPORTANT, don't forget to run the command sync before ejecting the media that you intend to use.

- Step three (optional): you can make an archive of the image of the OS disk, post it here or privately, and I will try to run it, both in simulator mode and on my actual device that is stil open and expects the LCD replacement of the CRT.

 Good luck and don't despair, it will work, and is nice that you've take a look, the PCMCIA card doesn't look too future proof to me, is a smart idea to be replaced, of the IBM Deathstar, I refrain from commenting  .

 Cheers,
 DC1MC

[Martin Hodge]

Definitely use an external monitor if you plan to do any OS hacking. Otherwise you will mistakenly think things have crashed when actually the system is just busy. 

[DC1MC]

Definitely use an external monitor if you plan to do any OS hacking. Otherwise you will mistakenly think things have crashed when actually the system is just busy. 

And, did it work ?

[Martin Hodge]

So, yeah. Not very useful without a mouse, but we know other OS is possible. Going to reach for the stars and try win98 next.

(This is on a 1GB IBM Microdrive through a cheap 1 socket IDE-CF adapter..)

[Martin Hodge]

Windows 98 down. What's next? Really need a linux CLI though. The biggest hurdle is getting the install media connected/transferred/copied.

[DC1MC]

OK, this is perfect  , that means hardwarewise you're set, now you need to choose a media for the OS, as I've said, I was only able to correctly boot the OS on a 32MB card (I will try with other sizes, but 2GB and 4GB were definitely not suitable for the OS, the data storage wasn't having any problems), and one for the data storage (most likely a 4GB card will be most happy to serve).
And then get rid of the Deathstar  and the strange OS contraption and that was it  .
 
I will keep mine open a bit longer, to replace the CRT as well, because I don't  want to have a museum piece, but a useful device in the long run and that was it.

Cheers,
DC1MC

[Martin Hodge]

I have been successful replacing the spinning rust with a CF card, which is all I really care about. I used linux dd over PATA on a separate machine, the FTP method only succeeded in producing a workable empty vxfat partition, no files. The PATA method also copied the files, though that's not essential if one keeps backups of their files which is easy to do via the built-in FTP server or floppies.

I don't want to replace the stock SSD with a CF card. I don't see the point. It's only used to boot the Tek software and occasionally store cal data. With no moving parts, and seldom being written to, it should last as long as the rest of the machine. (I tried, but only had a 512 MB card which didn't work.)

Apologies for the Wild-Windows-Installing tangent. I should have made a separate thread. My motivation for all that was to see if Windows' Device Manager would reveal anything more about the hardware. Specifically that EISA card labeled "GPIB", which also contains the big CPLD that is the gateway to the AWG subsystem and the front panel.

[DC1MC]

Hi Martin, congratulations for a job well done  , now the generator will live a bit longer and maybe some others will find this thread and save their device. Is a pithy to scrap them because of these bloody disks.
Regarding the OS disk, indeed the chance to have it defective is small, but better to make sure that this one could be replaced as well and have a safety copy of the software, then not having any, I see that you've already tried with a 512MB CF card for the OS, I've got locally a 64 and 256MB, just to use the time mine is open and test the highest capacity that can be used, expect news in a couple of days.

Regarding the VXFAT partition image, the Linux driver really works OK on Ubuntu 18.04 and probably 20, if you're interested in the files in the partition, get a live CD and run it to get the files, the command to mount the image is in this thread as well.

So now, all that remains is for my LCD to arrive and some tests with different sizes of CF.

 Cheers,
 DC1MC

[DC1MC]

As promised, a quick update regarding the size of the OS SSD (originally 10MB), I was able to test it with 16, 32 and 64MB, the 128MB boots and loads the first part of the OS (vxworks.sys, the one printing dots on the screen) and then reboots immediately and does this in a loop.
I will not do any OS disk size testing anymore, I think 16, 32, 64MB CF, while not common, are and will still be available for a reasonable long period of time.
Do take a backup of your OS and data disk as described !!!

 Cheers,
 DC1MC

[DC1MC]

Some forum member replaced the CRT on his Tek 540 scope, the mechanical part of the front panel is identical, so have a look here, while my similar conversion kit is slowly crawling to my place:

https://www.eevblog.com/forum/testgear/tek-tds540-crt-to-lcd-conversion-experience/

 Cheers,
 DC1MC

[Martin Hodge]

(Months pass)

Are you still planning to do the LCD replacement? Are you still waiting for it to arrive?

[DC1MC]

Hello Martin,
the display arrived and was installed, it fits perfectly  , as the procedure is 100% similar as in this thread that I've mentioned,

https://www.eevblog.com/forum/testgear/tek-tds540-crt-to-lcd-conversion-experience/

I didn't want to abuse the eevblog forum server resources with more similar pictures, in the end it worked very nicely and I even put the gray contrast enhancing filter as a LCD protector, even if it took out a bit of the brightness.
If you consider that it has some value I'll dig the pictures and post them.

 Cheers,
 DC1MC

[tchiwam]

And I am in the same boat here with a dead , white screen AWG520

I dismantled it quite a bit to find a bastard PICMG /PC104+ ish type of CPU board with a ISA cage.

Internal HDD being 32MB SSD ADTX AX-SSD-20032A , it shows some data with HexDump, but no known partition found
Removable HDD  20GB Hitatchi IC25N020ATMR04-0  with 4GB Partition so I guess the OS supports 4GB fat partitions

I bypassed the CRT and connected directly to the VGA from the CPU and I get a
Memory count 64MB
BIOS screen
then:
"Operating system not found"

My guess is my MBR / partition is borked... 

Then made a dd image, then read in the Tek forum that I probably killed my chances to get it booting again....
Now I did manage to recover my cal.dat and setup.inf files by mounting it fat16 with an offset using linux.
parted and fdisk didn't want to find the partitions,
testdisk worked and even found the partition bootable...

I have an old HD where I'll dd the fixed image and give it a try....

Oh well, will get there at some point.

[tchiwam]

One step forward, now I got to on the AWG520
VxLd 1.2 No BOOTROM SYS

If I do the Qemu thing I get:
qemu-system-i386 -drive format=raw,file=AWG520HD0.img -drive format=raw,file=AWG520HD1.img
SeaBIOS (version ?-20190711_202441-buildvm-armv7-10.arm.fedoraproject.org-2.fc31)
iPXE (http://ipxe.org) 00:03.0 C980 PCI2.10 PnP PMM+07F922A0+07EF22A0 C980

Booting from Hard Disk...
VXLd 1.2 No BOOTROMSYS

[tchiwam]

FYI My CPU card is a Pentium 133
Transcend 64MB 144P

Manufacturer: HAGIWARA
PN: HPU5810ADA133 TJ

I found some for sale at sub 100USD on far east sites

They use a Xilinx for the PCI to ISA bridge on the adapter card. Would be fun to compare these cards.

[tchiwam]

And I got it booted.... Now I have an error 5448 in Wave memory bank 1

  Thanks for the thread.

Making sure the BOOTROM.SYS was at the right place with the attribute SYS seems to work... Well one or both was needed...

[DC1MC]

@tchiwam - congratulations for revival  , the Tek forum thread is depressing, (and incorrect as we have seen), only here you can find absolution  .
Don't worry about calibration data, they are rebuild every time you run the autocal, is not like other instruments where if a battery or memory dies, you're done.
As I was lucky to have the rest of my hw in a good condition, I didn't have the opportunity to look further in the technical details, in your case I will start with inspecting the board and measuring the power supply for values and ripple.
Don't forget the pictures and good luck !!!

 Cheers,
 DC1MC

[tchiwam]

I was looking at the CRT to LCD upgrade. If you check on the LCD connector of the CPU card you should find all the signals you need for the LCD. Usually the PICMG manufacturers put a default Video timing.

Try to use those, if you are lucky you may even be able to run full digital from the CPU card. Or at least get good sync.

Which kit did you finally go for ?  It looks like the pixel rate is more than 640x480 .Something like 720x...or 800x600 < my guess since the output is VGA

Something else I have on my board looks like a Ethernet 100B-T. Would be fun to use that PCI bus instead of the ISA card.

I already checked the power rails and they are very good. When I was doing my investigation I removed the ACQ section and the 2 power connectors. Giving me more than 30 minutes before over heating.

My guess is someone removed the tray and got the SSD corrupted.

I ordered a Kingspec 64GB PATA 2.5" SSD. I already use that in my TDS694C and the max Partition size it 4GB there too.

Just for giggles I will try 128MB on the CPU.

[tchiwam]

Working !

[tchiwam]

Cpu