Author Topic: Wavecrest DTS-2077 Teardown  (Read 5142 times)

0 Members and 1 Guest are viewing this topic.

Online edpalmer42

  • Super Contributor
  • ***
  • Posts: 1723
  • Country: ca
Wavecrest DTS-2077 Teardown
« on: August 20, 2013, 05:14:56 am »
I recently purchased a Wavecrest DTS-2077 Digital Time Scope.  It's basically a Time Interval Counter that is primarily used to measure the time between events with extreme precision.  It has such high performance that even Agilent didn't try to build their own, they just OEM'd it from Wavecrest.  My unit has an Agilent part number that identifies it as part of a spares package (!) for one of their semiconductor manufacturing systems.  Although it was DOA, the problem turned out to be trivial.  Now that I've got it running, here's a teardown showing most of the internal boards - more on that later.   :)

Here are some key specifications:

Hardware resolution:  800 femtoseconds
Single Shot Accuracy:  +- 25ps
Internal Jitter:  < 6ps rms
Measurement Rate:  > 40,000 measurements/sec.
Maximum Frequency:  1.3 GHz

Size:  44 cm wide x 19cm high x 60 cm deep 
Weight:  18 kg
Power:  Less than 300W

Initial Cost:  If you have to ask, you can't afford it.  There are various DTS models.  I've seen prices of $50K to $90K.
Current Cost:  I've seen prices from < $1K to $20K.  I paid much less than that.


Front Panel



This picture shows the DTS measuring a signal derived from it's internal 100 MHz clock.  The main reason to do this is to measure the unit's internal noise and jitter.  In this case, the 1 Sigma jitter measured 1.771 picoseconds.  Notice the input and output connectors - all SMA.  No crappy BNC jacks here!  The main inputs even came with connector savers.


Inside View



Here's what you see after removing the top.  On the left is the card cage where most of the boards live.  On the right are the power supplies that run the system.  The small supply in the middle is the 24V supply that runs the 100 MHz OCXO (the square silver can below the power supply along the edge of the unit).  It's energized whenever the rear power switch is on.  The power switch on the front panel connects the 24V to a solid-state relay that's above the power supply.  This relay turns on power to the 4-part main power supply which takes up most of the section.  The main supply has ratings of +-15V@1A, 2 X 5V@10A, and 5V@35A.  I guess that explains why it needs two fans that each pump out almost 100 cfm!  Yes, it's noisy.  I SAID IT'S NOISY!!  ;)


Front Panel Processor



Sorry about the picture.  It's hard to get at this one.  This board is a stand-alone computer whose only purpose is to run the front panel.  When I saw the front panel working, I thought the unit was working.  Silly me!  On the right-hand edge of the board you can just see two small, grey, round cables.  These are the optical cables for the 19,200 bps serial link that connects the front panel to the main processor.


Card Cage



When you remove the rear panel, you see the card cage that holds three of the system's boards.  Also visible are the fans that provide cooling for the Main Board.


Processor Board



The Processor Board is a standard MultiBus I board with a 486DX2/66 processor, 8 MB of RAM and custom EPROMs to run the system.  Finding the manual for this board was a big help in my debugging.  I'm a little puzzled by the 'family history' of this board.  The manual that I found was written by Intel - I think they created Multibus, but the board appears to have been made by Radisys.  Not shown in this picture is the GPIB daughterboard that plugs into the left-most blue socket just below the EPROMs.  The daughterboard is visible in the Inside View picture above.


Main Processor

 



After a lot of reading and inspecting while I tried to revive this unit, I pulled the processor, just because I could and because I was running out of things to check.  Here's what I found.  That's one of the data leads that's bent over.  It's been like that since it left the factory in early 2000.  It must have passed factory testing because the type of socket used allowed the lead to touch the socket.  If you look at the detail view you can see a 'dent' in the bent pin that could be where it made contact with the socket.  When I got this unit, I could see that it had been dropped a couple of times.  Maybe that shifted the processor enough to break the connection.  Also notice the spot of green corrosion between the bent lead and the metal plate.  Nothing helps a computer run better than occasionally shorting one of the data leads to ground!


NVRAM Board



This board worried me for a couple of reasons.

First, look at all those NVRAM chips with 10 year data retention and 13 year old date codes!  They turned out to be okay, but I'd still like to check the batteries!  This was discussed in this thread:

https://www.eevblog.com/forum/testgear/dallas-nonvolatile-ram-chips-battery-surgery/

Second, notice the broken end on the right hand card extractor and the small piece broken out of the board where that missing end should be.  When the unit was dropped, it fell on the back panel and part of the panel was driven into this board, breaking the end of the extractor and driving the extractor into the board hard enough to break out a piece.  Luckily, there were no traces there, only a couple of ground and voltage planes.  I filed down the damage to make sure they weren't shorting together.

But then I thought about the force that caused the damage and realized that it would be transferred through the board to the backplane.  Did it damage the backplane?  Did it break a trace or a connector?  I had to tear the unit down completely to get to the backplane.  I found a shattered surface-mount resistor, but it was just part of some kind of protection circuit.  I replaced it, of course.  I resoldered the pins on the connectors that would have taken the brunt of the impact just because I was there.  A rather long session with a continuty tester didn't find any other problems with the backplane.


Main Board





Here's the good stuff.  To show the size of the board, the two rulers are 30cm long.  I know there are larger boards, but I've never worked on one.  The front panel is to the left.  The SMA connectors on the right side of the board connect to the IQM board (See below.).  The small board on the bottom of the picture is a signal generator used to calibrate the system to cancel out the delay caused by the cables used to connect the DTS to the devices being measured.  This unit is sensitive enough that you can measure the change in delay caused by unscrewing the SMA connector by one turn.

There's very little info available on the inner workings of the DTS series so it would be a major project to figure out what these boards are doing.  The user manual includes a very simplified block diagram of the system and one reference said that US Patent #6226231 describes part of the system of a related unit.  There is no service manual available.  As a result, I really can't provide anything more than the pictures themselves.

Unfortunately, I was so excited when I found the bent pin on the processor that the only thing in my mind was putting everything back together so I could test it.  I forgot to take a picture of the other side of this board.   :palm:  :-[ Of course, it's a royal pain to remove and install.  Here's why:

If you look at the detail view above, to remove the main board, I had to disconnect the flexible, continuous ground plane that connects the board to the front panel with 8 bolts.  If I removed the bolts on the board, metal clamps on the other side would fall off.  Repositioning those clamps didn't seem like it would be fun.  But my Allen wrenches wouldn't fit in the bolts in the front panel due to interference with the bolts on the board.  I had to remove one bolt from the board, remove the corresponding bolt from the front panel, replace the bolt on the board, remove the other bolt from the board, and remove the other bolt from the front panel.  Reminds me of the 'Towers of Hanoi' game!  Repeat all that three more times, then simultaneously unscrew the four SMA connectors and you're done.

So there won't be a picture of the top side of the main board.  It's too bad because the board was *really* interesting.   >:D  Actually, this unit is based mostly on ECL chips rather than analog circuits so there was very little to see.  There wasn't any of the RF voodoo circuitry that you'd see in a spectrum analyzer.


IQM - Interval Quantizer Mux(?) Board & Daughterboards









The IQM board receives the outputs from the Main Board and the input from the 100 MHz OCXO.  The 100 MHz signal routes through the central daughterboard and is then sent to the daughterboards on the left and right.  The SMA connector at the bottom center connects to a 100 MHz monitor jack on the rear panel.  The output from the IQM board is data, rather than RF, that appears to be processed  through the NVRAM board.

If you look along the bottom of the board, on the left and right daughterboards, you'll see three black 10 turn pots and, where the fourth pot should be there are two resistors.  When the unit was dropped on it's rear panel, a part of the panel ripped the fourth pot off the board.  I found it floating around inside the unit.  I was able to measure the value and temporarily replace it with two resistors.  It will cost about $14 to replace because it's a Vishay pot that has a temperature coefficient of 15 ppm/degree C.  I'd never even heard of pots with temperature coefficients.


That completes the teardown.  I now have a working unit.  All I have to do is figure out how to use it and how to interpret the results.  When you're dealing with picoseconds that can be a challenge.  So far I've found that it really doesn't like low frequency sine waves.  Below about 100 MHz, the slew rate is too low and the result is a significant increase in measured jitter due to trigger noise rather than jitter in the DUT.  I'm still investigating ways to square up those low frequencies so that they can be properly measured.

Actually, a bigger challenge is where am I going to put this beast??  :-\

Ed

References:

DTS-2077 User Manual:  http://www.wavecrestdts.com/manuals/2079.pdf
US Patent #6226231:  http://www.google.com/patents/US6226231
 
« Last Edit: August 20, 2013, 05:36:03 am by edpalmer42 »
 

Online free_electron

  • Super Contributor
  • ***
  • Posts: 7335
  • Country: us
    • SiliconValleyGarage
Re: Wavecrest DTS-2077 Teardown
« Reply #1 on: August 20, 2013, 06:32:27 am »
Made by the church of weirdshitology....

I dismantled one a coup,e of years ago. That multibus was indeed invented by intel. Radisys took over that productline as well as the iRMX rtos.
Professional Electron Wrangler.
Any comments, or points of view expressed, are my own and not endorsed , induced or compensated by my employer(s).
 

Offline ddavidebor

  • Super Contributor
  • ***
  • Posts: 1162
  • Country: gb
    • Fermium
Wavecrest DTS-2077 Teardown
« Reply #2 on: August 20, 2013, 06:46:50 am »
Should be a goos idea to save the content of those nvram if you can. But they last 20 year usually.
Davide Bortolami,
Fermium LTD
 

Offline Vgkid

  • Super Contributor
  • ***
  • Posts: 2565
  • Country: us
Re: Wavecrest DTS-2077 Teardown
« Reply #3 on: August 20, 2013, 02:25:31 pm »
Neat teardown, I'm surprised to see fiber optics being used.(do you mean for the front end data transfer to the nvram?)
If you own any North Hills Electronics gear, message me. L&N Fan
 

Online edpalmer42

  • Super Contributor
  • ***
  • Posts: 1723
  • Country: ca
Re: Wavecrest DTS-2077 Teardown
« Reply #4 on: August 20, 2013, 02:51:05 pm »
Should be a goos idea to save the content of those nvram if you can. But they last 20 year usually.

I wish I could.  The front panel board has one NVRAM in a socket so I was able to save it.  The others are all soldered in.  The edge connectors on the NVRAM aren't Multibus connectors so I can't plug that board into another Multibus computer and read it.  I might be able to plug another CPU board into this chassis, but that would be scary.  There's even an NVRAM chip on the main board - soldered, of course.

Ed
 

Online edpalmer42

  • Super Contributor
  • ***
  • Posts: 1723
  • Country: ca
Re: Wavecrest DTS-2077 Teardown
« Reply #5 on: August 20, 2013, 02:57:59 pm »
Neat teardown, I'm surprised to see fiber optics being used.(do you mean for the front end data transfer to the nvram?)

I don't understand why they used fiber instead of a couple of optoisolators.  As far as I can tell, the only thing going over that link is user info - commands from the user to make this measurement and then measurement results and status info back to the user to be displayed on the screen.  The low speed of the link (19,200 bps) tells you that there won't be time for much else.

Ed
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf