Looking for PCIE LVDS card for Agilent MSO Infiniium scope

[nexus]

Hi all,

Working on a few pieces of Agilent gear. One is a DSO81304B scope that I took a gamble on. Fortunately, it was not obliterated in overseas shipping and it seems to be almost fully functional.

The principal issue it has is with the display. I believe the PCI-E based LVDS display card is dead. I attached a pic of the card for reference.

Here is a link to the part on the keysight website:
https://www.keysight.com/my/partDetail/0960-2530 . It appears to be out of stock since there is no "Add to cart" button as with other parts.

The unit does not want to boot with this card installed. However, without this card it boots just fine and works with an external VGA monitor.

This card was used in a lot of DSO infiniium scopes. According to the KS website, the list is as follows:

DSO80204B
DSO80304B
DSO80404B
DSO80604B
DSO8064A
DSO80804B
DSO81004B
DSO8104A
DSO81204B
DSO81304B
MSO8064A
MSO8104A

Would anyone happen to have this part laying around or in a parts unit? Or know where to acquire a replacement board?

I've also found the main LVDS chipset IC on eBay (Chrontel CH7308) and ordered it, but no guarantees that it will fix the problem.

Many thanks,
Kevin

[nctnico]

There are a whole bunch of fuses on the board. I'd check if nothing is shorted on the display. Maybe it simply draws too much current. Does the scope boot with the card installed but with the display NOT connected?

There are also a whole bunch of tantalum capacitors installed on the card. Are none of these shorted?

BTW: is U4 an I2C Eeprom or something else? If it is something else, then it could be U6 has been pulled out. The Chrontel chip likely needs configuration data in order to work properly and that is typically stored in a I2C Eeprom.

[gslick]

The principal issue it has is with the display. I believe the PCI-E based LVDS display card is dead. I attached a pic of the card for reference.

FWIW, while that display card plugs into a PCI-E slot, it is not really a PCI-E device. I believe that display card is what is called an ADD2 card (Advanced Digital Display Card – 2nd Generation), which connects to the SDVO (Serial Digital Video Out) interface channel, which is multiplexed on a portion of the x16 graphics PCI Express interface. The SDVO display channel is an Intel chipset specific feature present on the Intel 82915G GMCH (Graphics and Memory Controller Hub), and some other Intel chipsets.

[gslick]

BTW: is U4 an I2C Eeprom or something else? If it is something else, then it could be U6 has been pulled out. The Chrontel chip likely needs configuration data in order to work properly and that is typically stored in a I2C Eeprom.

That part is not populated in the image from the Keysight parts page:

https://www.keysight.com/my/partDetail/0960-2530

[nctnico]

It is. U6 and U4 are on the board twice. Once as DIP8 and once as SOIC8. U4 in SOIC8 is populated but I can't see what it is.

[nexus]

Yup, I did probe through all of the fuses on the board. None of them were blown.

I also checked the tants in circuit fwiw, the capacitance was pretty close on my trusty fluke 179, so I think the tants are also fine.

I did some thermography when it was plugged in; only the chipset (Chrontel CH7308) gets warm on the board.

I also probed the 7 pin header (upper right hand corner in pic - J3 "BKL PWR"). No voltages on there, which are all switched through power mosfets directly by enable/disable lines on the CH7308 chipset. Everything so far points to the chipset being dead.

[nexus]

U4 is marked ATMEL632. I am guessing it is a I2C EEPROM but I cannot find any datasheets that indicate this package marking directly. See attached close up pic.

[nexus]

I see. Would there be any alternative display cards that can be used, if this PCIE based display protocol is an actual industry standard used in other embedded applications?

[nctnico]

It is a 24C16 Eeprom so that is clear.

The fuses on the board are self-resetting so these will measure fine; you won't get any information from here. Did you try to boot without the display and backlight attached?

Another measurement to make it whether the crystal oscillator for the Chrontel chip works. I have heard stories about crystals going bad. You'll need another oscilloscope to check. Alternatively, you can plug the board into another PC (preferably one you are no longer using for anything serious) and measure it in there while using an external monitor.

And is the resolution set correcty? If there is a service manual for this oscilloscope, there should be a section on how the BIOS needs to be configured. Maybe the display resolution is configured wrong. Do you get any beeps from the motherboard speaker/buzzer? Has the oscilloscope ever worked since you got it? Maybe the BIOS data is scrambled and needs a reset + configuration.

BTW: The Chrontel chip isn't a graphics chip / GPU / videocard. The actual video card is on the motherboard and the chipset allows to have the video output onto the AGP connector instead of the regular AGP signals to transfer data between video card and CPU. If it is a PCIe card (which I doubt), it will need to be inserted into the right slot. Likely this info is in the service manual as well.

[nexus]

Yup, the unit boots and is fully functional with an external display. I did try to boot without the display and backlight connectors attached to the LVDS card. Same exact symptoms, so I do believe some part of the card itself is faulty.

I will check the crystal. It is a 14.318 MHz osc; I'm not sure if I have that particular flavor laying around in my parts bin.

[gslick]

I see. Would there be any alternative display cards that can be used, if this PCIE based display protocol is an actual industry standard used in other embedded applications?

What is the motherboard? Is it an Intel D915GUX? The Intel D915GUX motherboard GMA900 Graphics Controller contained within the Intel 82915G GMCH (Graphics and Memory Controller Hub) supports an ADD2 card (Advanced Digital Display Card – 2nd Generation) in the PCI Express x16 slot. There is only one PCI Express x16 slot on the Intel D915GUX motherboard, so there is no question about installing the card in the correct slot.

The ADD2 card isn't really an industry standard, it is an Intel chipset specific interface.

https://en.wikipedia.org/wiki/Serial_Digital_Video_Out

[nexus]

The motherboard is indeed an Intel D915GUX. Thank you for that additional clarification regarding ADD2. I guess I'm in luck that Agilent decided to use a fairly uncommon and (now antiquated) Intel proprietary interface to drive the display!

I have ordered both the Chrontel CH7308 chipset IC and new 14.318 MHz crystals. I will update once those parts have arrived and I have a chance to place them on the board and retest.

I noticed that the system only has 1G of ram and uses a 2.9 GHz celeron. I ordered up 4x 1G sticks of ECC 533mhz ram and a Pentium 4 640 which should be compatible with the chipset and bios on this board. Also a 64G intel SLC SSD. Nothing like a little retro PC build.

The first machine I ever built was also an LGA775 board, but that supported the awesome q6600 which I had cranked up to 3GHz.
Cooler master 690, big typhoon cooler, evga 8800 GTS's in SLI... when life was simple.

[gslick]

I have an Agilent 16901A logic analyzer which also uses an Intel D915GUX motherboard. Without double checking, I'm pretty sure it also uses the same 0960-2530 LVDS display adapter card. (There are also later versions of the 16901A which use an Adlink M-880 or M-890 motherboard. Those Adlink motherboards have the LVDS display connector integrated on the motherboard).

The standard CPU on the Intel D915GUX motherboard version of the Agilent 16901A logic analyzer is an Intel Celeron 352 - Celeron D CPU 3.20GHz. I haven't tried to upgrade to a faster processor. I did upgrade the standard 1GB of system RAM to 4GB. Without checking I'm not sure how much system RAM was actually available to the OS, I think it was closer to 3GB than 4GB after the BIOS subtracts reserved sections of the motherboard address space.

[Pitrsek]

Beware, the display settings seems to be saved in your motherboard bios. I've decided to upgrade the internals and now my LVDS output is not working as VBIOS of new board is not configured properly.

[coromonadalix]

see  PITRSEK thread for some tools ??

https://www.eevblog.com/forum/testgear/agilent-mso-infiniium-scope-change-of-motherboard-issue-with-vbios/msg4411234/#msg4411234

[nexus]

So after finding a drop in 7x5mm 14.318MHz on mouser, I ordered that with some new 18pF caps.

Replaced the crystal and caps, and guess what, its working! Touchscreen is also fully functional.

While I had it open I also did some upgrades...

CPU: From Celeron D 341 >>> Pentium 4 660 (also upgraded cpu cooler from intel all-aluminum to copper slug version).
RAM: 512 MB DDR2 533 >>> 4 GB DDR2 533
HD: 40gb 7200 rpm drive >>> 64gb Intel X25-E SLC SSD

System is much faster now. As fast as it will get with the stock mobo anyways.

Time to play with GHz signals!

[gslick]

While I had it open I also did some upgrades...

CPU: From Celeron D 341 >>> Pentium 4 660 (also upgraded cpu cooler from intel all-aluminum to copper slug version).

After you replaced the Celeron D 341 with the Pentium 4 660 on the Intel D915GUX motherboard, did the BIOS have an option to enable or disable Hyper-Threading on the Pentium 4 660, or did the BIOS hard code Hyper-Threading to be enabled or disabled? I wonder if there would be any noticeable performance difference at all with Hyper-Threading enabled vs disabled.

My 16901A logic analyzer with the Intel D915GUX motherboard came standard with a slightly faster Celeron D 352 (3.2GHz Cedar Mill processor with 512KB L2 cache, vs 2.93GHz Prescott processor with 256KB L2 cache for the Celeron D 341). There would be less of a performance difference replacing a Celeron D 352 with a Pentium 4 660 than replacing a Celeron D 341. Still, I might give it a try just for the heck of it (or maybe a Pentium 4 650 which has a lower TDP)

https://www.cpu-world.com/Compare/383/Intel_Celeron_D_341_vs_Intel_Pentium_4_660.html
https://www.cpu-world.com/Compare/384/Intel_Celeron_D_352_vs_Intel_Pentium_4_650.html

[nexus]

Hyper threading was turned off in bios when I popped the P4 in.

Out of curiosity i turned it on. System booted just fine, but the Infiniium scope app would not open and kept crashing. Something about HT is not compatable with that. So I just turned it back off in bios again and everything is happy.

The 2MB cache in P4 might be a nice addition. FSB might also run a little faster. It was a pain to remove the cpu in my unit though; the heatsink had hot snot on mounting points from the bottom of the board. I had to unscrew and lift the motherboard up to get the glue off. The original ram sticks were also hot-snotted on the latches, so had to carefully peel that off too. Not sure if your logic analyzer has hot snot on the ram, cooler as well.

After you replaced the Celeron D 341 with the Pentium 4 660 on the Intel D915GUX motherboard, did the BIOS have an option to enable or disable Hyper-Threading on the Pentium 4 660, or did the BIOS hard code Hyper-Threading to be enabled or disabled? I wonder if there would be any noticeable performance difference at all with Hyper-Threading enabled vs disabled.