Tek TDS784A "ERR, in pixelator, pipeA, demux 200"

[Jwalling]

This is one I've never seen before. TDS784A that passes all tests and SPC except for this one:

:ERRLOG:FIRST "Tue 05-03-;6 12:56:31  ERROR: diagnostic test failure,  for DEMUX C, Bad pixelation, 718 errors, , ERR, in pixelator, pipeA, demux 200"

I've isolated the problem to the ACQ board; swapping in a known-good board corrects the failure. I just don't know what a "pixelator" is or which chip it might be. I have parts boards that I could get a replacement chip from, but not knowing which one makes the repair rather daunting...

Jay

[siggi]

Pipe A has to be referring to one of the two byte-wide sample channels between ADC and MUX. At a guess, pixelation would refer to the "broadsiding" in the MUX where it rearranges incoming samples for parallel writes to sample memory.
We reasoned out how this orders from ADC to memory recently in TekScopes, where someone had stuck bits in the ADC.
I'd try running a triangle wave at the offending channel and see whether there aren't acquisition speeds and/or channel combos where samples go wonky with a stride of 8. If so, it'd be the MUX.


Sent from my iPad using Tapatalk

[Jwalling]

Pipe A has to be referring to one of the two byte-wide sample channels between ADC and MUX. At a guess, pixelation would refer to the "broadsiding" in the MUX where it rearranges incoming samples for parallel writes to sample memory.
We reasoned out how this orders from ADC to memory recently in TekScopes, where someone had stuck bits in the ADC.
I'd try running a triangle wave at the offending channel and see whether there aren't acquisition speeds and/or channel combos where samples go wonky with a stride of 8. If so, it'd be the MUX.

Hi Siggi,

Not seeing anything unusual with a triangle wave signal from 1KHz to 100KHz with maximum memory depth. I attached a pic @ 100KHz. Not sure what you mean by "with a stride of 8"

Also not sure which is the offending channel...

Thanks

Jay

[siggi]

Not seeing anything unusual with a triangle wave signal from 1KHz to 100KHz with maximum memory depth. I attached a pic @ 100KHz. Not sure what you mean by "with a stride of 8"

Also not sure which is the offending channel...

Hey Jay,

this is mostly speculation, so take with a healthy grain of salt. This is also assuming the 784A is similar in architecture to the 544A, for which we have schematics on Håkan's site. There's also a decent description of the DEMUXes in the TDS520B component SM.

So, given these caveats, the acquisition system works as follows (apologies if this is old hat to you):
There are four ADCs, each of which I believe contains two separate flash converters. The ADCs have two byte-wide output channels. There's a DEMUX per ADC, these take the samples from the ADCs, decimate and buffer them until there's a "broadside" of eight samples ready to be written to acquisition memory.
Decimation modes in the DEMUX include normal (dropping samples) max/min/hires, so it could be a bit of an ordeal to test the DEMUX well. You'd effectively have to try all sampling rates in all acquisition modes :/.
I seem to remember that the normal decimation mode will favor one of the ADC "pipes" until the acquisition speed demands using both. The details are fuzzy and perhaps we didn't explore this thoroughly with the 520 with stuck ADC bits.

The bad samples in a stride eight would be where either the DEMUX internal logic or memory interface has problems, as the DEMUX internally buffers eight samples for a single write.

NB: if the 784A is shaped the same way the 544A, then demux "C" corresponds to channel 3. They go 1-A, 2-B, 3-C, 4-D.
Note, however that the acqusition board CAN acquire any of the channels with any of the ADCs, and at max acquisition speed with a single channel on, it interleaves all the ADCs to a single input.  Maybe that's the easiest way to see whether you can provoke some visible acquisition-time trouble?

Also, in the 520B schematics, U200 is DMUX C - maybe the error message is giving you the suspect component designator?

The 520B service manual also talks about a "pixel processor", which interacts in some way with the sample memory through the DMUXen. If the 784A has anything like that, then that's another possibility - e.g. some sort of post-acquisition problem.

As you see, this is all so much speculation from poring over service manuals and schematic fragments - but I'd be curious to know whether/how you resolve this.

Siggi

[Jwalling]

Thank you for all that Siggi! I played around with the scope yesterday especially running CH3 through the ringer and still was unable to see anything out of the ordinary all the way up to 1GHz at various sweep rates and memory depths. I think I will try replacing the A/D converter on CH3 and see what happens then. I'll wait for the weekend and have a few beers before I start to steady the the old hands as it's SMT 84 lead PLCC.

Jay

[ArcticGeek]

Jay,

How did you observe the error message you stated in your first post?  I assume that was the error message you got after you ran SPC, is that correct?

One thing you may want to try is to attach a serial console port while you run SPC.  The console port often times spews a bunch of information when running SPC if it finds something that isn't working or that it doesn't like.  Often times it will give you a clue as to where to look that might be causing the problem.

It might be a bit of a longshot, but it's simple enough to do without breaking out the soldering iron.

Good luck.
Mike

[siggi]

I played around with the scope yesterday especially running CH3 through the ringer and still was unable to see anything out of the ordinary all the way up to 1GHz at various sweep rates and memory depths.

Did you try hires as well?
Hires has to be doing addition in the DMUX, and perhaps something like division - bit shifting at least.

I found this table of hires bit depth in one of the user manuals:

100 ns and faster 8 bits
200 ns to 500 ns 9 bits
1
s to 2
s 10 bits
5
s to 10
s 11 bits
20
s to 50
s 12 bits
100
s to 200
s 13 bits
500
s 14 bits
1 ms and slower 15 bits

Good luck.

[Jwalling]

Jay,

How did you observe the error message you stated in your first post?  I assume that was the error message you got after you ran SPC, is that correct?

One thing you may want to try is to attach a serial console port while you run SPC.  The console port often times spews a bunch of information when running SPC if it finds something that isn't working or that it doesn't like.  Often times it will give you a clue as to where to look that might be causing the problem.

It might be a bit of a longshot, but it's simple enough to do without breaking out the soldering iron.

Good luck.
Mike

Hi Mike,

Oddly enough, the scope passes SPC. It's the full self test from the utility menu that generates the error. I've never connected to the console port - is that the serial port in the rear of the scope, or the box header on the CPU board? If it's the box header, is it the same pin-out that used on the TDS7000 series? I already have a cable for those...
Thanks.

Jay

[ArcticGeek]

Jay,

The console port attaches to the card edge style connector that is on the main processor board.  There is a post over on the Tektronix user forums where someone provided a way to take the serial/parallel option port and wire to the console port.  To do this you don't need any parts, its just a matter of wiring up a small adapter (I did mine using perfboard).  You can search for "console port" on the Tek forum and probably find it.

The nice part about the console port is that it spews out a ton of information - and it often gives a person a good clue where to look next.  I fixed a TDS754A that had a bad ASIC on the acquisition board by getting the info from the verbose error message on the console port.

Let me know how you make out...if you need help with the console port I can probably dig up the info I have on it.
Mike

[Jwalling]

Jay,

The console port attaches to the card edge style connector that is on the main processor board.  There is a post over on the Tektronix user forums where someone provided a way to take the serial/parallel option port and wire to the console port.  To do this you don't need any parts, its just a matter of wiring up a small adapter (I did mine using perfboard).  You can search for "console port" on the Tek forum and probably find it.

The nice part about the console port is that it spews out a ton of information - and it often gives a person a good clue where to look next.  I fixed a TDS754A that had a bad ASIC on the acquisition board by getting the info from the verbose error message on the console port.

Let me know how you make out...if you need help with the console port I can probably dig up the info I have on it.
Mike

Hi Mike,

Are you referring to this post by by "Garibeos" about 13 posts down?
https://forum.tek.com/viewtopic.php?t=137307&start=20
Does it appear to be accurate?
If you have the time, I'd like to see what info you have as well. A picture of it would be great!
I'm guessing you:
1. Take an old style floppy cable and connect this to the edge card connector on the CPU board.
2. Then you cut off the box header on the option 13 board leaving the ribbon cable.
3. Then splice the two ribbon cables together with the connection info in that link.

Last: Does the serial cable need to be a null-modem (crossover) or straight connect to the PC?
Thanks!

Jay

[Jwalling]

Did you try hires as well?
Hires has to be doing addition in the DMUX, and perhaps something like division - bit shifting at least.

I found this table of hires bit depth in one of the user manuals:

100 ns and faster 8 bits
200 ns to 500 ns 9 bits
1
s to 2
s 10 bits
5
s to 10
s 11 bits
20
s to 50
s 12 bits
100
s to 200
s 13 bits
500
s 14 bits
1 ms and slower 15 bits

Good luck.

Nothing of note there either Siggi. Everything still looks fine other than finding out that the Acquire menu button wasn't working (fixed) 

Jay

[ArcticGeek]

Jay,

Yes, that is the post.  Note that I followed the directions exactly as described in the very first post by "sschnelle".  I actually post over there in that forum from time to time, my userid over there is "rochoz".  I built an adapter pretty much as you described following the wiring plans outlines by sschnelle. 

You will need a null modem adapter to connect the serial/parallel port to your computer.  For my connection I use a USB-to-serial adapter on a laptop, and then a null modem adapter to connect to the serial/parallel option.  My home-built adapter has 2 connectors on it, which are dual-row 100-mil center berg strips.  I plug the serial/parallel option to one connector, and the other connector I plug in a cable I made up that has a berg-style connector on one end and a card edge on the other.

I'll try to post pictures of it later tonight when I get home (it might be a bit late as my daughter has a birthday party tonight).

Mike

[Jwalling]

Thank you for all that Siggi! I played around with the scope yesterday especially running CH3 through the ringer and still was unable to see anything out of the ordinary all the way up to 1GHz at various sweep rates and memory depths. I think I will try replacing the A/D converter on CH3 and see what happens then. I'll wait for the weekend and have a few beers before I start to steady the the old hands as it's SMT 84 lead PLCC.

Jay

Well, replacing the CH3 A/D converter U700 didn't change the fault. Note that CH3/4 and CH1/2 swap logical positions on the ACQ board through the demux chip U1250. With the scope upside down and facing you, the A/D order from left to right is CH3, CH4, CH1, CH2. I'm going to order a 20 pin IDC edge card connector and build a console port interface. In the meantime I'll perhaps try changing U1250 - another shot in the dark but it only costs some time.

Jay