Agilent 54835A scope (4 channel 1GHz / 4Gs/s) repair & uphack

[Converter]

I think you should not worry. I, too, had it during the first boot, but I never saw any further problems during use.

[alpher]

You're right, I'm going to leave it as it is for now, just try to update to 4.5 tomorrow.
Bsods started after at the point windows asked me to choose drive for some intel ide controller,
there were 2 choices some intel number and a generic one.
Choose intel and bang bsod and hard lock , had to power cycle, after that had around 3 more bsods an had to powercycle twice or trice.
Wonder if I had chosen the generic controller things may have gone more smoothly 
I'll try it on the final board that I'm getting from a guy from Ukraine from all places, funny.
Thank you Converter for all you'r help.

[alpher]

Just a quick update,wasn't able to update software to ver4.5 yet, somehow second IDE chanel doesn't even show up when the OS starts, BiOS sees the CD drive no problem, but in windows device manager there is only the primary IDE.
I think I'm gonna wait for the 810 mobo before I try again, my idea now is to edit the GHO image, add a directory containing the 4.5 update and then use "subst" command in windows command prompt to point drive a: to the directory, in theory it should work, we'll see how it does in practice  .
In the meantime removed the acquisition board and the motherboard to some "bodywork" on the frame as mine came with a big dent in one of the back corners, luck has it didn't damage the acquisition board.
Since I have the board out , may try the 54846 uphack, we'll see.
When I removed the acq. board I noticed that the service manual mentions 9 torx screws to be removed, I counted 13  , overall removal is really easy compared to other scopes I worked on recently.
Here are some pics for the curious minds:



Here you can see how close the dent in the aluminum frame came to breaking the acq. board, luck was on my side here:









Wonder what material is the RF shield made of, is quite heavy could it be Lead ?

[alpher]

846A uphack didn't work for me as well, immediately fails calibration on all 4 channels.
Something else is at work here.
Maybe for 54846s they use a specially graded acq. boards or attenuators?

[alpher]

Just consulted service manual for these scopes, and the acq. boards have a different part numbers for 1.5 and 2.5GHz scopes, namely:
A13 54845-69527 1 4-CH ACQUISITION ASSEMBLY W/ATTENUATORS (Agilent Model 54845A)
A13 54846-69527 1 4-CH ACQUISITION ASSEMBLY W/ATTENUATORS (Agilent Model 54846A)

Even more telling is this:

A13A1
A13A2    54845-66502 2 2-CH ATTENUATOR ASSY (54835A and 54845A)

A13A1
A13A2    54846-66502 2 2-CH ATTENUATOR ASSY (54846A)

So even though the acq boards themselves may look the same, the attenuators are not.
What's significant here is that while 835A and 845A share the same attenuator assemblies, and people report successful uphack from 835 to 845.
The same is NOT true going from 845 to 846, since they have different attenuators, according to service manual at least.

[Converter]

I hope this was not unexpected for you? You have already seen the posts about the same unsuccessful attempts?

https://www.eevblog.com/forum/testgear/agilent-54835a-scope-(4-channel-1ghz-4gss)-repair-uphack/msg1347104/#msg1347104

https://www.eevblog.com/forum/testgear/agilent-54835a-scope-(4-channel-1ghz-4gss)-repair-uphack/msg1347123/#msg1347123

https://www.eevblog.com/forum/testgear/agilent-54835a-scope-(4-channel-1ghz-4gss)-repair-uphack/msg1348009/#msg1348009

Here you can see how close the dent in the aluminum frame came to breaking the acq. board, luck was on my side here:

And you can easily fix it. Simply use a hard, even base and some flat rectangles on the other side and hammer to align the dent. I often have to fix this.

[alpher]

Of course I've seen them, but the more data we have the better chances of successful hack,
Obviously now we know that there are some differences between 846 and 845 attenuators.
So someone with access to both types can find out the differences.

[alpher]

Here you can see how close the dent in the aluminum frame came to breaking the acq. board, luck was on my side here:

And you can easily fix it. Simply use a hard, even base and some flat rectangles on the other side and hammer to align the dent. I often have to fix this.

That's exactly what I did:

[Converter]

Believe me, this is all easily corrected. I came across more complex cases. Simply place another small flat bar at the height you want. I use ground metal parts. But you need to be very careful, because you can create new dents . Before this, do not forget to remove all boards from the chassis.

Update: Now I see on your new photos that you have already successfully corrected everything. Congratulations.

[Krzys]

Just a quick update,wasn't able to update software to ver4.5 yet, somehow second IDE chanel doesn't even show up when the OS starts, BiOS sees the CD drive no problem, but in windows device manager there is only the primary IDE.

What do you mean by "BiOS sees the CD drive no problem"? Do you mean that autodetection detects CD drive?

By default, only first IDE channel is enabled in WinBios setup - you need to enable both in Peripherals. (And it is being reset to defaults on each change that prompts you to enter the BIOS - for example on memory size change. Quite annoying.)

Even though secondary IDE channel is disabled in WinBios setup, BIOS could autodetect devices on that channel - but it wouldn't work anyway.

Edit: I'm probably wrong about autodetection always working, as today it detected no drive - it turned out that the setting for IDE channels was set to "None". Autodetection started working just after changing it to "Both", saving settings and reboot. It seems inconsistent to what I've observed while trying to connect CD drive previously. (OTOH Windows 95 recognized both channels just fine.)

I don't know if enabling secondary IDE channel is enough to see it in Windows.

Also please note that some CD drives might not work.

In my case I've tried some PATA laptop DVD drives. First one ("modern", DVD-RW, Sony NEC Optiarc AD-7563A from 2008, rated 1,5A) , was detected as secondary slave (my desktop IDE <-> laptop CD adapter has no jumpers), but it wasn't working (at least it was not possible to boot from it).

Second one - old 20x/2x DVD-ROM (Panasonic SR-8171C, from 1998 (yay!), specc'd at 1,8A) was detected as primary master and worked just fine - I've went through full Windows XP repair install with it.

As for +5V power supply, I've used Vcc and ground from USB connector. Somehow it was enough.

[Gemikro]

Hi folks !

i recently could acquire a four channel 54815A for free which would have been thrown away by the company otherwise 
The only defect part i found was the Dallas CMOS battery so far which was easily replaced.

The unit came with 8 probes, original keyboard and track ball.
All channels are working and can be calibrated.
I already replaced some parts as the system fans, CPU fan and the hard drive and increased system memory to 1042MB to be able to upgrade to Win98.
I also replaced the network car to a newer 3com Pci card which is working quite well and upgraded the software to version 4.5

I also would like to upgrade the scope to 54825A but did not dare to solder on the main board yet.

One bad thing is, i could not get the onboard usb controller to work yet.
I could install the 98FE usb driver and the onboard controller is recognized, but it does not detect any attached usb devices so far.

I know that 98SE has better support for usb and tried to fresh install (from Cd-Rom attached to the second IDE) or upgrade to 98SE but in both cases the installation always hangs at the hardware detection process 
Has anyone tried to install 98SE on the original Atlas III mainboard ?

Max

[Krzys]

I tried to install XP but the problem is that the drivers for the acquisition card are not available. The Windows98 drivers won't work on XP because XP is completely different when it comes to drivers.

Do you know what the VIDs/PIDs are for the acq card? Maybe in the range 15BC:0500 to 15BC:0507?

No, according to the list from the BIOS I think it is 103C:1020 (that is the only unknown PCI device listed)

I don't know if you are aware of this, but on webarchive there's an old dump of Keysight's FTP site containing something that looks like N5383A installation ISOs. Four files named M815G-T3A-Disk1.iso, M815G-T3A-Disk2.iso, M815G-T3A-Disk3.iso and M815G-T3A-Disk4.iso contains ghost image of Windows XP installation, together with drivers for graphics card and acq board - probably all revisions, including the oldest one, I see the following VEN/DEVs (acq boards olny):

Code: [Select]

VEN_15BC&DEV_0508
VEN_15BC&DEV_0507
VEN_15BC&DEV_0503
VEN_15BC&DEV_0501
VEN_15BC&DEV_0500
VEN_103C&DEV_1020xml file found in the root dir of the first CD states:

<!-- All 548xx Infiniium Models-->
<FrameDescription>
   <Device VendorID="103C" DeviceID="1020"/>
</FrameDescription>

so it explicitly lists first version of acq board (the one shipped with first Infiniiums, W95-based). That comment: "<!-- All 548xx Infiniium Models-->" is even more promising.

Anyway, I think it should work on 54832B with either acq board. I cannot test this myself, as I don't own 54832B.

However, in reality, the "All 548xx Infiniium Models" promise might be false, as I couldn't make it to work.

I had to use Atlas III based 54810 scope with lowly K6 200 MHz, upgraded to 96 MiB of RAM (I had a hard time getting my two pairs of 32 MB EDO sticks to work with each other - on each (re)boot I had a different size of ram. Of course the two pairs on their own worked just fine separately. So in the end I've used one pair of two 32MB sticks and another pair of two 16 MB sticks totaling 96 MiB). I've also used 6,5 GiB 2,5 IDE HDD from some old laptop, that was competing in terms of noise generated with the scope itself (prior to this I've used 40G hdd, but I had issues as it was recognized as 8 GB by old Atlas III mobo.).

This configuration is even below XP minimum requirements (when it comes to CPU speed). 96 MiB of RAM is a bit low for even bare XP from 2001, lest for XP SP2 loaded with tons of Agilent bloatware. However I had no choice, as my K6-2 400 MHz based 54815 that takes SDRAM DIMM sticks decided to die, when I was trying to power it up (after not using it for some two years - but neither I used that lowly 54810 during this time).

So I restored the ghost image to the HDD, put XP setup bootstrap files on it (WinMBoardMig.zip that I got from some chasms of the internet did not help here - only turned bootlooping XP installation into hanging one - just after it loaded everything and switched to 32-bit protected mode), then I did repair install of provided WinXP SP2 (it is in WinXPFiles directory or something like that).

The good thing is that I had no unrecognised/unknown devices in device manager. Now the bad.

After reboot, I've installed latest version (5.71) of scope software ... which crashed on some SSE instruction (well, it was intended for those PIII based boards it seems...). a4sse, a sse emulator, BSOD'ed right after executing provided exe file.

As I imagined that rewritting scope app to not use SSE might be a major task on its own, I've tried the oldest scope app from Keysights website (4.21) which only barked that "Low physical memory detected.  Check physical memory size in the control panel and reboot", then caused some driver to allocate 1 GiB of memory (mostly in pagefile...) and hung afterwards.

After noping out that memory error it had shown another one: "The scope vxd is not registered or is out of date." - lol, they didn't updated that string for XP version. And it looked like it is not working anyway (all self tests failed, unable to calibrate, togle LEDs, read knobs etc).

I've decided to try Win'95 scope app instead, since it was destined for my scope and certainly it shouldn't expect it to be 5483x instead, hoping that interface between the driver and scope app didn't changed in the meantime.

(Both scope apps, after a quick look, seems to share their codebase.)

Win'95 app used file "\\.\hp548xx.vxd" to communicate with the acq board driver, whereas the XP one used "\\.\Scope0" instead. So I hexedited w'95 HP5481X.EXE accordingly, only to discover that it is working just like XP one (albeit with no message about wrong vxd and not eating 1 GiB of RAM). It couldn't read my scope s/n, was detecting it as 54855, couldn't find calibration data (despite it being where it should be) and of course no communication with front panel controls. Should I also say it didn't measured anything?

After a reboot to w'95 it turned out that on w'95 directly launching HP5481X.EXE produces the exact same result as directly launching it under XP. It turned out that on Windows 95 another exe is launched first - hp548Ldr.exe - which apparently tried to establish scope type and then launch either HP5481X.EXE or HP5484X.EXE with some magic parameters.

hp548Ldr.exe had at least one bug in it - if it failed to open file exposed by acq board driver, it didn't send IOCTL and didn't initialise some variable, but afterwards it tries to access memory pointed by that variable, like that not placed IOCTL call was successfull. So it crashed on XP, since it apparently failed to open that \\.\Scope0.

I've downloaded a little utility by the name of "DosDeviceInspector" to see if there is really \\.\Scope0 device (or maybe other devices of interest). Of course it also crashed on SSE instruction (file is dated 2016-01-12, so probably they used compiler recent enough to enable SSE by default). The instruction seemed to be irrelevant to the function of the program, as noping it out was enough to make it working and make sure that indeed, there is device called \\.\Scope0. Also there's another device of interest - \\.\Tombstone0 ("Tombstone" seems to be code name/nickname of first version of acq board - VEN_103C&DEV_1020 - thus tstone.inf/tstone.sys pair). However this device couldn't be opened on Windows XP by hp548Ldr.exe as well.

Having issues getting hp548Ldr.exe to work, I've tried to figure out whether it is really important, looking like just a launcher for either HP5481X.EXE or HP5484X.EXE.

The magic done by hp548Ldr.exe takes a form of some magic spellparameters which turned out to be:

Code: [Select]

HP5481X.EXE /scope /hpibWhen launched this way, scope software just works - on win'95 of course.

Equiped with magic parameters, I've swapped HDDs to boot Win XP and try starting scope app new way.

It turned out to be a mess - now it started crashing with unprivileged instruction. W'95 app was writing directly to registers of (I suppose) graphic card! I've tried noping the entire function that did ins and outs but then it started to crash in other places.

This time I gave up. Maybe I will try something more in the future, preferably after fixing my other scope (MOBO not working - no sign of life, I've even tried running it off external power supply; but that a good sign anyway as it is easier and probably cheaper to get equivalent AT MOBO, if I couldn't fix this one). Doing anything on that low-end XP machine is an exercise in patience (if not in futility)  - even turning it on (or off) takes several minutes. Imagine running scope app under the debugger...

Hypothesis: in that particular configuration XP driver does not work. Maybe there were different VEN_103C&DEV_1020 (with different interface to the PC I mean, as there were PCB revisions anyway, as cards from 54810 & 54815 looks different) and the XP driver was tailored only to the acq card that was shipped in XP scopes, or maybe the drivers are checking for certain characteristic of VP22 motherboard (unlikely, since folks have changed their MOBOs - but maybe tstone.sys checks for that, where drivers for newer acq boards don't?). I can also imagine different revisions of acq board firmware. Or maybe something is misconfigured (for example some service occupies the driver and latest scope knows how to deal with it, while older ones do not). Or my repair reinstallation had broken something important.

There are plenty of debug strings in at least Mesa.sys. Printing them via DbgPrint seems to depend on a value of some variable. I've tried changing it to expected 2 in driver binary, but no luck, maybe I misread something.

Or maybe I'm missing something major. Just before writing this post I believed that tstone/tombstone is a codename for acq board. But in tstone.inf file it is described as "Infiniium Display" which does not seem to make much sense for me (also I vaguely remember something about some PCI-PCI bridges? maybe "tombstone" acq card is just an one piece needed in case of later scopes, whereas on previous generations it was the only one?). Maybe there were some major changes between Win9x based Infiniiums and XP based one, despite retaining the same VEN/DEV of acq board?

Update: For explanation, why it failed, see my latest post here.

[alpher]

Well, very interesting stuff, could you send us the link to these images ?
I'll be willing to try them on a i810 board with  P3 cpu that most certainly has the required sse and I have more than enough of sdram sticks for xp to run comfortably.
If only the damn board ever decides to show up.


p.s. Dobra robota panie Krzysztofie.

[Krzys]

Well, very interesting stuff, could you send us the link to these images ?

Here's the link to said ftp dump:

https://archive.org/download/ftp.keysight.com/2015.03.ftp.keysight.com.tar/

recovery .GHO is divided into 4 CDs, so we need the following 4 ISO files: M815G-T3A-Disk1.iso, M815G-T3A-Disk2.iso, M815G-T3A-Disk3.iso and M815G-T3A-Disk4.iso.

Some files from the same directory: 54830-97013.pdf, Mesa-BIOS-HowTO.DOC, VP22.pdf, VP22AG09.zip.

ISO of CD that makes recovery HDD image from fully installed scope: makeimage11.iso

Here's main archive.org page for that download: https://archive.org/details/ftp.keysight.com (in case somebody wanted to download the whole 183 GiB tar file).

If booting recovery CD wouldn't let you install it, you could just use ghost to restore these .gho files. In my case I've used ghost32.exe (Norton Ghost 8.0.0.984 from 2003) under WinPE, after collecting all .gho files in one directory.

I'll be willing to try them on a i810 board with  P3 cpu that most certainly has the required sse and I have more than enough of sdram sticks for xp to run comfortably.

Sure it has SSE, but I guess that the issue lies elsewhere anyway. Though I wish you luck with it :)

If only the damn board ever decides to show up.


p.s. Dobra robota panie Krzysztofie.

Well, thanks, though I don't think I achieved anything - quite the contrary. But maybe someone else will eventually do it.

[alpher]

Well, my 810 motherboard just showed up, been fighting with it since very morning.
As usual when a few faults combine, it takes some serious time.
Had a flaky ide cable, flaky CF card, bad cf to ide adapter, don't ask.
Now it's a quarter to 10pm and I finally got it to boot reliably from the CD drive, got the proper CF card etc, just to find out that the copy of 5484xB recovery image from Tonys drive is corrupt.
Downloaded it twice, burned trice tried a couple of different drives and, nada:


Any help here guys?

[Tony_G]

Sorry to hear about that image being corrupt. Unfortunately I never used that as I had a 54845A and not the B. I used the images in the 3_5, 4_3 & 4_5 folders. Did you try them?

TonyG

[alpher]

I wanted to try the B version because I'm hoping that the "B" hardware may be closer to my new "upgraded to socket 370 P3" than the original socket 7 system was.
Problem solved itself this morning though, I've found a good version here:
https://archive.org/download/ftp.keysight.com/2015.03.ftp.keysight.com.tar/ftp.keysight.com%2Fcos%2Foutbound%2F54845B%2F54846B%20Recovery%20CD%204.50%2FPholyRecovery.iso

Now trying the install, we'll see how it goes.
So far I'm getting windows protection errors after second reboot.
BTW. same thing happened with the image that Converter send me.

[alpher]

Still no go, after almost 2 full days I wasn't able to successfully install any of the scope images on the new motherboard.
Tried almost every combination in the BIOS settings, disabled cpu cache, forced IDE to PIO only, etc.
No matter what I do it always errors out at the same spot.
Now, to be sure that the hardware works with windows 98 I've installed both the first and the second edition on the very hardware that the scope must work with, no major problems.
Here is a screenshot of the device manager of the W98FE (that's the one that scope is based on, I think?), showing what needs the driver installed.
Despite that windows installs and runs fine.:
 

Now getting to the scope recovery images, I've tried both, mostly the 54845B image downloaded from archive.org, tried also the original 4.3 version from Tony's drive that worked for me before (on original socket 7 motherboard).
They fail everytime  I tried.
The sequence is as follows:















After that the scope locks up, and I have to power cycle.
Unfortunately after restart all I get is windows protection error.
Desperately need a second opinion here.
 

[tautech]

Desperately need a second opinion here.
 

NEVER let Windows decide what it thinks is the right driver for anything !

Get the correct drivers from the HW manufacturer always.
For some old HW you will have to dive into some archives probably.
Good luck.

[nctnico]

You'll need to get the Win95 drivers from the original disk. Letting Windows select the acquisition hardware drivers isn't going to work.

[alpher]

The thing is that I'm using the original HP/Agilent recovery images that have to have the proper drivers etc.
Would be really nice if someone with a real 54845B could look inside and tell what is the actual chipset that the motherboard  is based on.
I know that the original 54845A is a socket 7  system , can be VIA , Intel HX and possibly others.

[tautech]

The thing is that I'm using the original HP/Agilent recovery images that have to have the proper drivers etc.
Would be really nice if someone with a real 54845B could look inside and tell what is the actual chipset that the motherboard  is based on.
I know that the original 54845A is a socket 7  system , can be VIA , Intel HX and possibly others.

Doesn't matter what the HW is it still needs the right drivers.
Treat this stage like you would for building a custom PC and get the correct driver for each piece of HW and NOT from the first source you find online but from the HW maker.
Only resort to other online sources if the manufacturer doesn't have the driver (or driver package) available for download. (unlikely)

[alpher]

I don't think that you grasp the problem at hand here, there is not a driver or software package available for these scopes.
All you have are a couple of recovery images that include the OS (windows 98 first edition most likely on the 54845B ,for sure on 54845A) plus the necessary drivers plus custom scope application, that is. There's no more AFAIK ?.

[tautech]

I don't think that you grasp the problem at hand here, there is not a driver or software package available for these scopes.

Oh yes I certainly do.
It's a PC first running scope SW. Any scope running Windows should be thought of like this.

Get the PC bit right and the rest will fall into line.
Have a squiz through this thread and it may open your eyes:
https://www.eevblog.com/forum/testgear/upgrading-mainboard-in-lecroy-dda-3000-(aka-wavepro-7300a)/

Just trying to help. 

[alpher]

I've read the first page, will keep reading for the sake of it, but have a distinct feeling of wasting my time there.
At first glance , apples and very old oranges.
We're talking windows 98 (first edition) versus upgrading P4 to a dual core or more.