On Sun, Mar 24, 2019 at 6:15 PM Keith Monahan via Groups.Io
<keith@...> wrote:

On 3/24/2019 4:18 PM, Glen Slick wrote:

I don't remember having any problems when I tried using a Buslink
SPCI2P PCI-SATA card, which uses the Sil3112A. It might be that the
drivers are only preinstalled for that particular PCI device, and not
any other versions of Silicon Image PCI-SATA bridge chips. I'd have to
get set up to try the recovery installation process again to verify
that.

Yeah that's what you had said on the previous thread, which is why I
opted for a Sil3114A. Surely the drivers are the same or close for the
3112 vs the 3114. It's possible that the 2-port version is part of the
available upgrade kit, so they made to sure integrate those particular ones.

I took a look at the 16903A on my bench at the moment. The driver and
.INF file that are preinstalled only include PCI Device & Vendor ID
matches for the Sil3112 and the Sil3512, but not the Sil3114.

I just looked around on eBay, Sil3112 and Sil3512 cards don't appear
to be common now. I can't imagine there would be any real demand for
them, but that never stops people from listing obsolete parts. I
bought my Sil3112 cards 4 or 5 years ago, a name brand Buslink SPCI2P,
and a generic one that I can't find a match for anywhere on the net
now. The generic one was actually the one I installed, the Buslink
SPCI2P one is still sealed in its antistatic bag.

Glen,


On 3/24/2019 5:11 PM, Glen Slick wrote:

I finally found my recovery DVDs. The DVDs that I received as physical
copies from Agilent in the "Part Number: 16900-68722 16900A/16902A
Recovery DVD Kit" envelope are labeled "Logic Analyzer Recovery DVD
16900A and 16902A Only" / "Part Number: 16900-14121" on the DVD
itself.

Yeah, I was sent 16900-14121.ISO, and 16900-14122.ISO (for the BIOS upgrade). That's all I really needed for the 16900A.

I forgot that there is a different recovery DVD for the 16903A, part
number 16903-14107. I had to request an image for that one from
Keysight before it magically appeared (and quickly disappeared later)
on their FTP server.
Ah, now I remember even more. There used to be a handy text file on
their FTP server named "Decoder Ring (Which one do you need).txt"
which had the part numbers for the appropriate recovery media. On the
recovery media there is an XML file which has some hard coded values
to check in the BIOS for a match before the image restore will
proceed. While the same ADLINK M-815G motherboard is used in multiple
analyzers, it may have different analyzer dependent signatures in the
BIOS.

Thanks for that list. I saw you posted that on EEVBLOG too.

FWIW, those scripts are simply in some DOS batch files, which are easily modifiable in case there isn't a "hard match."

The recovery disks contain Norton .GHO and .GHS files that get combined to produce a single disk image. You can use those files directly without the need of any of their utilities. You can convert those to .ISO, or .VMDK for virtual machines, or really whatever you need.

I find the standardization and ease-to-decode files and file structure refreshing. I love being able to use the files beyond their original application.

Thanks,
Keith

Glen,

Thanks again for your comments.

Responses-inline.


On 3/24/2019 4:18 PM, Glen Slick wrote:

I don't remember having any problems when I tried using a Buslink
SPCI2P PCI-SATA card, which uses the Sil3112A. It might be that the
drivers are only preinstalled for that particular PCI device, and not
any other versions of Silicon Image PCI-SATA bridge chips. I'd have to
get set up to try the recovery installation process again to verify
that.

Yeah that's what you had said on the previous thread, which is why I opted for a Sil3114A. Surely the drivers are the same or close for the 3112 vs the 3114. It's possible that the 2-port version is part of the available upgrade kit, so they made to sure integrate those particular ones.

It wasn't the end of the world, just had to work around it. Like any of these related problems in IT.

Yes, you want to update to XP SP3, then install the .NET packages so
you can install the version 05.90.1104 analyzer application software.
The .NET package installation is one of the longest parts of the whole
system setup from scratch. A set up from scratch on a bare hard drive
can take a few hours to get everything installed.

It's just tedious because of the slow CPU. Even with SSD installed, the CPU was maxed at 100% for the most of the installation time. Definitely takes a couple hours in the best case circumstance. Installing XP from scratch is no picnic either.....in comparison to Win7/10 on a modern machine.

I picked up a few new bulk package Adaptec AUA-5100B PCI USB 2.0 cards
cheap on eBay. Those use the NEC uPD720101 (now Renesas), which is one
of the better USB 2.0 controllers from what I remember from my USB
development days. No problem getting that to work in the 16900A with
the preinstalled drivers. The Adaptec AUA-2000C appears to use the
same uPD720101 and those are also cheap in new bulk package condition
on eBay.

I think this would be worthwhile upgrade. I'll check out similar cards.

I believe GBit Ethernet card is just a standard Intel PRO/1000 XT
PCI-X card, plugged into a 32-bit slot, with a low profile bracket.

You're right. FWIW, when I used to specialize in Networking, Intel's cards were really only second to 3Com for drivers and support. These particular cards used to be really expensive back in the day!

One thing I did do is install an add-in AGP graphics adapter. Nothing
fancy, just wanted DVI output. I think it was an older passively
cooled VisionTek Radeon 7K. That's somewhat period appropriate for a
Windows XP system.

The default resolution works fine on VGA. I have a 4x3 monitor on the LA cart that I built by hand, attached to an adjustable monitor arm.

Thanks,
Keith

On Sun, 24 Mar 2019 14:45:44 -0400, you wrote:

I purchased a fully loaded 16900A for quite a bargain recently. It has
(1) 16950B, 16M, 667MHZ state mode, and 4ghz timing zoom module. It also
has (5) 16910A's, 16M option, 500MHZ state mode option. The 16950B is
68-channels, and the 16910A's are 102-channels. I wasn't a math major,
but that's almost 600-channels of simultaneous capture.

A couple friends asked, what are you going to do with all those
channels? While truthfully I don't know, I can certainly do whatever the
heck I'd want to do with this beast!

Debugging I2C communications between microprocessors:

8 channels for data,
roughly 8 channels for status (start/stop/first byte, ack/nak, etc per
byte)
8 channels (roughly) for message byte count, resetting after a stop:

24 channels

Then you monitor anything that the I2C message was supposed to
trigger....


Now the 24 channels worth of data is parallel, state machine, and you
have to figure out how to take it apart.

Harvey


<snip>

That's great ... for you. Not so great for me. Now I have no reason not to work on my taxes.

V.

Test 48 now passes.? The analyzer seems to be operating correctly! Wow.

I'm going to need to do a writeup on what I found; what was significant and what wasn't, so this can be useful for others in the future.

Peter

I looked and looked and turned up a K&L 2 GHz LPF.

The procedure is not the clearest but I identified what seemed like the right spur, the screen looked somewhat like the one shown, and with that one selection it said it was done.

I immediately noticed a huge difference:? the PLL looks nicely locked all the way down to 300 KHz and there is no error message on preset.

I am going to see if test 48 now passes.

Doing the cavity oscillator spur correction turns out to be very critical.

Peter

The manual says this:

?You should perform this procedure with the recommended filter, or a filter
with at least 50 dB of rejection at 2.9 GHz, and a passband which includes
800 MHz. The filter makes spur identification substantially faster and more
reliable.

If you have a 1 GHz low pass, try it. I think I once did it with a 1.8 GHz low pass. Note, there is a filter-less Test 54 procedure in the manual too, but it looks a lot less pleasant.

Vladan

Ok, I will do this test.

Where can I find the actual specs for this filter?? Online I see reference to it being a 1.9 GHz LPF, in the manual it shows in in a diagram as a 4.1 GHz LPF, and the other post mentioned a Minicircuits part which is a 1.2 GHz LPF.

Peter

You do need to check that before you do more esoteric troubleshooting. If the 3.8 GHz oscillator is off, then for a given YTO frequency the IF will be off too. Too much IF error, and the YTO loop will not lock.

I have to mention that the 3.8 GHz oscillator has a varactor inside. This varactor is used to move the cavity oscillator's frequency slightly, depending on the programmed source frequency. As I mentioned in an earlier post, this is done to avoid some spurs. Every time the firmware shifts the cavity oscillator's frequency, the YTO gets moved by the same amount and that keeps the IF at the right place. For this reason, you can't just attach a counter to the 3.8 GHz oscillator and adjust it. It's frequency is dithered, depending on the desired source frequency. So, you need to do test 54, either with a filter, or using the more tedious method which I think is described in the manual. The filter is not critical in the sense that you don't need the exact HP/Agilent part number. It just needs to have a certain stop band. The manual probably says something about it.

Vladan

开云体育

Dear All,

I checked the FM coil drive and it is right about at zero so that seems like it's not the issue.

Nobody seems to know how the firmware works so we are left to guess and try things.? There is some range which the firmware can correct for and I heard from someone who needed to nudge his phase lock board to get his system into that range.? I tried the same thing and can get some measure of success as well, although apparently I have another problem causing me grief.? At this point I have this temporary fix in which permits the unit to run on default pretune values but won't pass test 48 which sets more precise pretune values.

Below about 50 MHz I have a PLL instability which gets worse as the frequency decreases.? The PLL begins to lose lock below about 25 MHz.? I took two photos of the unit output when set to 25 MHz in CW mode showing the instability, then I put it into source tune mode which opens the phase lock loop.? See attached.? Notice how the output is nice and stable as soon as the loop is opened.

Not sure what to make of this.? The YTO main coil drive doesn't look any noisier at 25 MHz than it does at higher frequencies.

On Sun, Mar 24, 2019 at 11:45 AM Keith Monahan via Groups.Io
<keith@...> wrote:

* The keysight-provided recovery DVD includes scripts that use Norton
Ghost to restore the hard drive to factory condition.

I finally found my recovery DVDs. The DVDs that I received as physical
copies from Agilent in the "Part Number: 16900-68722 16900A/16902A
Recovery DVD Kit" envelope are labeled "Logic Analyzer Recovery DVD
16900A and 16902A Only" / "Part Number: 16900-14121" on the DVD
itself.

I forgot that there is a different recovery DVD for the 16903A, part
number 16903-14107. I had to request an image for that one from
Keysight before it magically appeared (and quickly disappeared later)
on their FTP server.

Ah, now I remember even more. There used to be a handy text file on
their FTP server named "Decoder Ring (Which one do you need).txt"
which had the part numbers for the appropriate recovery media. On the
recovery media there is an XML file which has some hard coded values
to check in the BIOS for a match before the image restore will
proceed. While the same ADLINK M-815G motherboard is used in multiple
analyzers, it may have different analyzer dependent signatures in the
BIOS.

16900-14121 DVD \Recovery\ag169xx.xml

<!-- Recovery Set for Agilent 16900A/16902A -->

<FrameDescription>
<Device VendorID="0x8086" DeviceID="0x244B"/>
<FrameString Segment="F000" Offset="0892" Value="16900"/>
</FrameDescription>

16903-14107 DVD \Recovery\ag169xx.xml

<!-- Recovery Set for Agilent 16903A -->

<FrameDescription>
<Device VendorID="0x8086" DeviceID="0x244B"/>
<FrameString Segment="F000" Offset="0892" Value="16903"/>
<FrameString Segment="F000" Offset="E0C1" Value="M-815G"/>
</FrameDescription>


"Decoder Ring (Which one do you need).txt":

168xA/AD Win XP MY41000801-MY41000899 Radisys SC815E 01680-14100
MY41000901-MY41001099 Motorola VP22 01680-14104
MY41001101-later ADLINK M-815G 01680-14106

168xxA MY46000101-MY46000999 Intel D915GUX 16800-14102
MY46001000-later ADLINK M-880 16800-14103

16900A, 16902A MY43000001-MY43001999 Motorola VP22 16900-14109
MY43002000-later ADLINK M-815G 16900-14121

16901A MY46000101-MY46000399 Intel D915GUX 16901-14101
MY46000400-later ADLINK M-880 16901-14102

16902B All ADLINK M-880 16902-14100

16903A MY43000001-MY43000199 Radisys SC815E 16903-14100
MY43000200-MY43001999 Motorola VP22 16903-14104
MY43002000-later ADLINK M-815G 16903-14107

If your serial number starts with SG, then the same number ranges apply.
For example, SG43000001 is the same as MY43000001

Hello,

OK, so all caps and diodes turn out to be functional - this means the transformer is arcing :-/

The board was rebuilt nevertheless to do some experimentation. I set up a driver similar to the one in the analyzer and fed it pulses using a signal generator (HP8116A). It was fairly easy to get an output voltage above 1 kV, but there were clearly sparking sounds... Rather than playing with fire, I made a few measurements only and then shut down. I append a couple of pictures showing rebuild, setup and scope pictures. The measured voltage is the output from the +9 kV line.?
Ch1 measures the Uctrl signal directly from the signal generator. Ch2 measures the voltage over the 0.5 Ohms resistor, i.e., 1 A/division. Ch3 measures the voltage at the A6A1 primary winding. The DC voltage is set at 18 V.
?
It was very clear that pulses in phase with the transformer resonant frequency gave a drastically higher response than pulses being out of phase. See for example Capt 6, Capt 7 and Capt 8. It appears as if the resonance is around 70 kHz - quite a bit higher than the driver in the analyzer that runs around 30 kHz? Capt 2 runs at 30 kHz and does manage to generate a reasonable output voltage with a slightly higher Uctrl (input pulses) - this opens the MOSFET to 2 A pulses rather than the 0.25 A pulses earlier.
The pulse width is the whole time just 3 us in order to hold the power down - not sure it was sufficient or the right way to do it.

The high voltage output seemed never to be quite stable - neither at only a hundred Volts nor up in the kV levels. Possibly there was arcing all of the time although not audible? Once voltage was around 1 kV, it could also jump up to 2.5 kV for a short time, but then drop down again. A clear cracking sound was then heard from the supply...

One thought regarding early trouble shooting. Rather than taking the A6A1 apart immediately, I should perhaps have tried to alter the feedback loop it was sitting in so that the controller would reduce the voltage to safe levels (shorter pulses - possible)? One could then see at what voltage level the unit would fail - perhaps a way of finding more things out in a controlled way. The feedback is from the 110 line through a simple voltage divider. Just feed a DC voltage through a 10 kOhm resistor at the divider point?

Anyhow. Now I'm very much looking forward to the HV unit that has been sent to me. Hope the thread has been of use. Many thanks for all comments and suggestions!

Regards,
?Staffan

PS. If anyone has experience with winding high voltage transformers I think the forum would be _very_ interested to learn - I would at least :-)

I don't have the required low pass filter so can't do that step. What will happen if that's messed up?

On Sun, Mar 24, 2019 at 11:45 AM Keith Monahan via Groups.Io
<keith@...> wrote:

* The Adlink 815G industrial motherboard(Note earlier versions used a
different mobo) that the system is based on doesn't support SATA
natively. I added a QNINE(some off-brand Chinese) Sil3114-based 4-port
card from Amazon. The card specifically supports Windows XP, and I
believed the drivers to be already installed. They definitely were not!

* Getting XP to boot off the SSD was a nightmare. Windows kept blue
screening a STOP 0x0000007B error, which means Inaccessible Boot Device.
The Award BIOS 6.00PG present had no real relevant options to help, and
does not see the HDD present, even after getting it work.

I don't remember having any problems when I tried using a Buslink
SPCI2P PCI-SATA card, which uses the Sil3112A. It might be that the
drivers are only preinstalled for that particular PCI device, and not
any other versions of Silicon Image PCI-SATA bridge chips. I'd have to
get set up to try the recovery installation process again to verify
that.

* After getting the machine to boot, I updated to Service Pack 3 using
the Microsoft downloadable EXE, and then installed the Keysight Logic
Analyzer 5.90 application. That installer needs to install .NET 3.51 and
.NET 4.00 and this takes A LOOOOOOONG time. It also installs the Windows
XP drivers(both of them, there's two) to support the PCI card which
talks to the modules.

Yes, you want to update to XP SP3, then install the .NET packages so
you can install the version 05.90.1104 analyzer application software.
The .NET package installation is one of the longest parts of the whole
system setup from scratch. A set up from scratch on a bare hard drive
can take a few hours to get everything installed.

* The USB ports are version 1.1 by the best I can tell. Using flash
drives are painfully the slow. My chassis had the gigabit ethernet card
which is MUCH MUCH faster for doing samba transfers. Note that some
modern samba installations (I'm looking at you Ubuntu) turn of NTLM and
lanman hashes, which means that you need to re-enable support to be able
to connect --- otherwise your passwords fail when trying to connect!

I picked up a few new bulk package Adaptec AUA-5100B PCI USB 2.0 cards
cheap on eBay. Those use the NEC uPD720101 (now Renesas), which is one
of the better USB 2.0 controllers from what I remember from my USB
development days. No problem getting that to work in the 16900A with
the preinstalled drivers. The Adaptec AUA-2000C appears to use the
same uPD720101 and those are also cheap in new bulk package condition
on eBay.

I believe GBit Ethernet card is just a standard Intel PRO/1000 XT
PCI-X card, plugged into a 32-bit slot, with a low profile bracket.

* I am surprised at just how much this thing is just a regular of-era
Windows XP PC. All the normal PC-like things apply, and I'm convinced
that you could use a more modern PC, more modern (but still 32-bit!!!)
operating system, and really get this thing humming. I'm not sure
there's an obvious benefit, because this LA is plenty usable in its
current form. But for fun, who knows? I think there's threads on eevblog
about this!

I wasn't really into trying to hot rod my 16900A. A motherboard swap
would be a mechanical challenge due to the motherboard I/O cutouts
being integrated into the motherboard tray instead of using a
replaceable pop-in/out I/O shield you could change to match a
different motherboard. As I already mentioned previously you can use
the 16900A mainframe CPU as just a server for the acquisition modules
and run the analyzer application remotely on a much faster system over
the GBit Ethernet.

One thing I did do is install an add-in AGP graphics adapter. Nothing
fancy, just wanted DVI output. I think it was an older passively
cooled VisionTek Radeon 7K. That's somewhat period appropriate for a
Windows XP system. There might have already been drivers preinstalled
so I didn't have to install any vendor driver crapware.

One limitation with the 16900A is only 3 PCI slots (1 shared with the
AGP slot). So you can have only 3 of a GBit Ethernet card, a USB 2.0
card, a PCI-SATA card, and a PCI/AGP graphics card, if you are into
trying to max out the system.

I don't know if this will help, but here goes.

First, don't worry about the apparent offset in coil tuning current. YTOs are only specified to operate over a certain range, like an octave or two, but can run a ways beyond at either end. It doesn't help (other than mathematically to get the tuning chart numbers) to extrapolate down to zero coil current - the YTO will stall out somewhere a little below its lowest spec frequency and coil current. Any slope variations or offsets extrapolated at zero just represent the character of the magnetic structure in the YTO.

The proper range, slope, and offset needed for tuning a particular YTO are typically handled by the YTO driver circuit, which can make it so the net result is a nice convenient GHz/V type system, at its input. There are usually sufficient adjustments (pots and/or DAC settings) in the driver circuitry to accommodate the tolerances of the YTO, the 3.8 GHz oscillator in this case, and the rest of the parts, to make everything come out right, during calibration.

There may be circuitry included to improve the tuning linearity (due to magnetic saturation), and compensate for the dynamic response (magnetic delay and hysteresis). Ideally, the YTO frequency is perfectly and instantly proportional to the magnetic field, but in reality the field depends on an iron core, coil inductance, AC core losses, and so forth.

The FM coil may be used in a PLL, for fine tuning. Its effect is very small (but much faster), compared to the main coil. It too has to be properly driven, of course.

In a down-converting generator made for reaching low frequency, it would normally be set up to avoid spectral inversion - that's why there's a minimum output frequency that can be expected. If the 3.8 GHz LO is free-running, its drift may be enough to limit how far down you can go at the low end, regardless of the YTO's tuning precision. For sources that can get way down toward zero (DC even), both the LO and YTO need to be precisely controlled (synthesized).

The coarse and fine tuning, compensation, and calibration all contribute to successful operation, so there may be quite a few possible suspects.

If you have schematics, you should be able to find all of these elements, and track down the proper conditions for diagnostics and repair.

Good luck.

Ed

Did you do the cavity oscillator correction, test 54, using the low pass filter? I know we talked about it, but I want to make sure this setting is correct. This is normally a firmware correction, you shouldn't have to touch the tuning screw unless someone played with it before you got the analyzer.

Vladan