Fruitful TouchStream Repair Tactics

[pythagoras]

With the recent interest in home TouchStream repair, you all need hints of fruitful approaches for TouchStreams that are completely dead, or showing bad sensors in diagnostics.

• * ST/Stealth Tented Frame Dissasembly: The first step is to separate the sensor boards and their plastic frame from the tented metal frame. The boards are attached to the metal frame with silicone rubber. To safely detach, slide a wide, thin piece of sheet metal or fiberboard between the plastic pieces and metal frame, gradually sawing through the silicone rubber with the wide sheet. WARNING: Using a narrow screwdriver, putty knife or pick will likely scratch the underside of the circuit boards in the process and cut the circuitry!
  
  * ST/LP Black Plastic Frames: The top and bottom frame pieces are snapped together (with a bit of acetone or superglue near the center cable and USB cable). A small screwdriver can unlock the bottom snaps, and can then inserted between top and bottom to GENTLY pry apart. A couple snaps usually break off even when super-careful, but as long as the top frame is not ripped it can be glued back down as necessary with acetone or superglue.
  
  * Parlex Center Cable on LPs: A ripped or worn center cable is one possible cause of left half sensor block failures (usually either 320 or 640 bad sensors) that is particularly easy to fix. The FingerWorks forum posts on long cable extensions had the exact part # and ordering instructions from DigiKey.
  
  * Circuit Boards: There were a couple generations of TouchStream circuit boards. In the first (2002?) generation, all four boards were the same size, sandwiched as pairs. These were much harder to repair 'cause the boards must be de-sandwiched before doing ANY chip/circuitry repairs. If NOT de-sandwiched by an experienced electronics technician with proper desoldering equipment, they probably won't work after soldering everything back together. In the second generation (2003-2004?), only the sensor boards were full size, and the processor boards were 1/3 width stubs (with plastic or aluminum plate filling out the sandwich). The second generation are considerably easier to repair ... sometimes you get lucky and the stuff needing repair is accessible without desoldering the sandwich.
  
  * FWSEN16A Chips: Though someone seems to have found some floating around China, these rarely fail in the field, and the ones in China might not be in the correct chip package. Sensor chips are more likely to come unsoldered than fail in the field. Each handles an 8x2 block of electrodes and daisy chains down a row pair. So if diagnostics show some bad sensors in one or two adjacent rows, it's worth resoldering the FWSEN16A chips in that row, or if that fails replacing the chip directly underneath or down-row of the ioslated failures. A BAD SENSOR CHIP CANNOT CAUSE FAILURE OF LARGE BLOCKS OF 4-16 SENSOR ROWS. Look elsewhere (Parlex cable, Xilinx, loose connections) in those cases.
  Probably if the community can round up a completely dead iGesture pad or two to sacrifice, someone with a desoldering gun could harvest the 40 chips off the dead pad to supply all the spares the community would ever need.
  
  * Xilinx FPGA Chips: According to the mystery manufacturer who used to do repairs, one of the two Xilinx FPGA chips located on the RIGHT half processor board is the next most likely cause of LEFT half sensor block (i.e. 320 or 640) failures (if not the Parlex cable). The Xilinx's are also the weak link most likely to fail during extremely strong static discharges. One of the Xilinx's feeds the left half sensors, the other the right half sensors. The one feeding the LEFT half sensor should have lots of direct connections to the Parlex center cable connector (discoverable with a continuity checker). Each Xilinx has its own EEPROM program. It should be possible for someone with access to a Xilinx universal programmer to desolder the Xilinx from a sacrificed TouchStream, read it's program off the EEPROM, and propagate this program to a set of spare Xilinx's procurable from Digikey or other chip supplier. Or for the truly clever reverse-engineer, there is a single row of (6?) thru-holes on the processor board that, when custom-wired to the proper pins of a Xilinx programmer, should be able to talk to (read/write) the Xilinx chip(s) serially without desoldering them?
  
  Also, the first and second generation circuit boards had different Xilinx's with different programming. The first generation Xilinx's probably aren't worth attempting to repair or stock, since not that many shipped.
  
  * Other Chips: The bullets above likely account for the majority of failures and common repairs. If some other random chip or solder connection is unlucky enough to fail, sometimes it'll be lucky enough to visibly pop out. On some processor boards there was a tiny inductor feeding from the USB Cable +5V wire to the voltage regulator that could get zapped by a power surge, and stick out like a burnt bloated flea. Replacing this inductor (or just short-circuiting across it) gets power flowing again. A good tech. should be able to fix the USB cable connections and replace the voltage regulator on units whose power busses are obviously dead, and be glad to do it for less than $1500!
  
  *Making TouchStreams out of 2 iGesture Pads: Sorry this will never work. The iGesture Pad processor boards don't have enough memory, or the right firmware.
  
  * iGesture repair: Same principles, but no left half to complicate things.

Despite appearances, some of us fought hard behind the scenes for a smoother FingerWorks ending, but, if anything, probably made it worse. Even with the benefit of hindsight, we might not manage to thread the chaos of cash, investors, legal and production constraints to shutdown smoothly in such a brief time frame. But FingerWorks took a fork in the road in the hopes of a greater good in the long term. So buck up, try to keep your TouchStreams working, and give up the gloomy talk about history being stopped. It's just on pause ... for a long while ...

[ivanw]

This is nice. This is understandable. Anyone could have guessed that something simple happened to explain the end of an episode. At last we know that this is not one of those damned profit motive again. It's just that it would have been comforting to know that sooner. I know, there has been so many good reasons why no one could talk by then.

Now for the repair advices. Thanks, whomever you are . Maybe we won't worship you for this but it will help the community in its contribution to make it happen again. As I said before, we had to work hard too. Keeping the spirit alive is our task in the process.


Ivan

[nomaded]

I think "Prometheus" would have been a better pseudo than "pythagoras".

Thank you very much for some of those insights are what to look for, for future repairs. I really hope you will be able to stay and participate in this little community - your experience is invaluable.

[The00Dustin]

Maybe we won't worship you for this

Won't we? hehe

[ivanw]

Worshiped? I would just say that some genuine experienced fellow would be appreciated to say the least. There must be some still alive I think. Is that truly not feasible? If so, could anyone explain why?

Any suggestion about coming up with someone that could build up the required skills and equipment? These are typical symptoms analysis that should be written down into some repair procedure. More information has to be collected as well, like schematics, test points and their typical readout: voltages, frequencies, wave shape. In fact, all this is the matter of a specialized organization. And I know that efficient servicing facility can only be achieved by someone close enough from the repair activity.

If one of us is ready to put his/her hand on a Xilinx programming device and manage to know how to operate with the technology, then I will be confident in our ability to repair a damaged device. All this talking comes from my previous life experience of the field. That's why I can tell that I am not able to do it myself.
- I know that I don't know what has to be known to do a good job.
- Does anyone knows/wants us to know he/she knows?

[pythagoras]

He he, we all have our deficiencies: I'm too clumsy with a soldering iron. Thus the need for a community effort, with experienced technicians.

Yes, hooking up the Xilinx programmer will be the hardest part, but it only has to be done once and propagated. But beyond tracing Xilinx pins and mapping out the power busses, lots of test point & waveform documentation may be overkill. Even someone who fully understands the system could easily get bogged down for hours scoping test points all over a misbehaving system. And to get to all test points, boards must be de-sandwiched, leaving one without connected sensor boards to examine ... kind of a Catch 22.

The point of the hints is that most repairs on second generation boards can be completed with a more 'superficial' procedure:

For a completely dead system (with no voltage on +3V power busses), priorities are:

• 1. Check whether connections from USB cable +5V & Gnd wires to processor board are still good. Resolder as necessary.
• 2. Check for bad inductor between USB cables and voltage regulator, or bad voltage regulator. Bypass burnt inductor with wire if necessary.
• 3. Replace voltage regulator.
• 4. It's possible (but unlikely) some other chip has gone bad and is shorting the power busses, but if so it should be hot to the touch!

For a completely dead system (no USB enumeration but good voltage on power busses):

• 1. Check whether 2 data wires from the USB cable to processor board are still good. Resolder as necessary.
• 2. Visually inspect all boards for loose parts or solder connections. Resolder as necessary.
• 3. Replace Xilinx(s)?
• 4. Possible but unlikely some other chip has failed, like the DSP. But chances of the system still working after blindly replacing several components are slim, so probably just SOL.

For a system that enumerates USB but reports bad sensors on left side:

• 1. Check (replace?) Parlex center cable.
• 2. Inspect and resolder sensor chips underneath bad sensor region.
• 3. Inspect and resolder supporting circuitry at ends of sensor rows.
• 4. If half or all left side sensors affected, replace Xilinx on right side processor board that feeds left side.
• 5. If isolated sensors affected, replace sensor chips in affected area.
• 6. If entire row pair affected, replace support circuitry at ends of sensor rows.

[ivanw]

Sure, we won't get all the appropriate material we could dream of but any of them may prove to be missing at times. These are things to be considered at this stage where some of us may be ready to get involved.

I also agree on the illusion of being able to face any situation. Although understanding the theory is a good start point, some intricate malfunctions may still defeat diagnostic, but I've also seen some soldering iron experts lost for hours...

Those situation related procedures you're describing here may well cover many likely situations. It seems reasonable to get prepared for those first and get ready for those who already have a malfunctioning device.
--
The Xilinx's are probably the first issue for devices that have suffered from an electrostatic shock. Being able to replace and reprogram one is sadly the less easy part. This means that one has to show up here with the ability to handle this aspect.

For the other scenarios, it is only a question of having some real-life experience in the matter. I am pretty sure there may be some qualified and ready to help forum member around here.

My expertise is not on the IC domain. I know a good deal about oscilloscopes and measurement but I have no logic circuit experience. I'm ready to hold the iron and dispense good advices, that's about all. ...and I am in France!

[aat0995]

i have not yet got a TS but actively looking for one. I am also looking at this troubleshooting place as I know sooner or later I will need it. Unfortuantely, my skills in electronics is zoro. I am good at programming and I can learn any language quite easily. I even looked at xlinx documentation but I don't think I can learn it that easily. If any electronics engineer comes forward it will be very helpful.

I thought I would post as I think there are two many people looking and no one replying. I would as well give my support

[ibrandt]

Thanks much for the post!

Timely for me as my ribbon cable seems to be acting up, and I just noticed a split in the USB cable shielding just as it leaves the frame that fully exposes wire. Looks like I'll be opening 'er up.

Anyone know what's involved in repairing the USB cable? Is it a cut and re-solder job, or is there a connector where I'd just buy a matching replacement cable from somewhere?

[Michael.Massing]

I just noticed a split in the USB cable shielding just as it leaves the frame that fully exposes wire. Looks like I'll be opening 'er up.

Anyone know what's involved in repairing the USB cable? Is it a cut and re-solder job, or is there a connector where I'd just buy a matching replacement cable from somewhere?

I have this problem, too. I did talk to FingerWorks about it - they had never heard of such a thing, but offered to repair/replace the board. Silly me for not acting on their offer before it was too late! Armed with the information above, I'm going to Fry's Electronics today to see whether they have any suggestions for either adding external shielding and strain relief, or repairing/replacing the USB cable and re-anchoring it from inside the case. The board still works fine, but I can't really travel with it, because the additional strain of moving the unshielded cable around usually knocks it out of commission in various unpredictable ways.

I'll post again if I can come up with a satisfactory solution.

Thanks for letting me know that I was not the only one in the entire history of the TouchStream LP to see this problem.

Best to all,
M.

[ivanw]

I'll post again if I can come up with a satisfactory solution.

Any outcome will be instructive anyway. I simply have this weakness in mind every time I have to move this device. Along with the static threat, it could turn us all into maniacs with their rituals.

Take care and good luck.

[Rqyteqto]

Michael - You might take a look at some of the liquid plastic products. Or liquid electrical tape. Its a goo that can be painted on whatever that cures on contact with air and forms a very good seal with excellent flexibility. It will bond to nearly anything. Just make sure you clean it well first with alcohol or something similar to eliminate any grease or oils.

I use this stuff all the time in lieu of electrical tape, to seal stuff, repair wires that have had their insulation compromised and so on.

[Michael.Massing]

Michael - You might take a look at some of the liquid plastic products. Or liquid electrical tape. Its a goo that can be painted on whatever that cures on contact with air and forms a very good seal with excellent flexibility. It will bond to nearly anything. Just make sure you clean it well first with alcohol or something similar to eliminate any grease or oils.

I use this stuff all the time in lieu of electrical tape, to seal stuff, repair wires that have had their insulation compromised and so on.

Thanks, I did decide after the trip to Fry's last night to go with a less invasive fix, and this sounds as if it might be what I'm looking for. Could I use liquid electrical tape over an entire half-inch/5cm of exposed USB cable, or would I want to maybe splice some insulation into the gap first? (I may have just answered my own question, but I'd appreciate hearing what you think.)

Thanks,
M.

[Rqyteqto]

No need for anything else. Just apply successive coats and build it up gradually. Let each coat cure fully (maybe 15 to 20 minutes). I've covered large areas of metal with the stuff. Its really almost like duct tape and vice grips and should be part of every general tool kit.

Hint: The same stuff is available in most boat stores as liquid rope whipping. Its got a zillion uses.

Just be sure whatever you apply it to is clean and free of oils, grease and dust. For wiring, I'd separate the individual wires as much as is possible and coat them several times before squeezing them together and coating them a bundle. Again, make sure the existing insulation is clean, might even help to roughen it a bit with some fine sand paper but thoroughly clean it afterwards.

[-jeffB]

Wow, looks like I missed some good stuff in the past couple of months...

* Xilinx FPGA Chips: According to the mystery manufacturer who used to do repairs, one of the two Xilinx FPGA chips located on the RIGHT half processor board is the next most likely cause of LEFT half sensor block (i.e. 320 or 640) failures (if not the Parlex cable). The Xilinx's are also the weak link most likely to fail during extremely strong static discharges. One of the Xilinx's feeds the left half sensors, the other the right half sensors. The one feeding the LEFT half sensor should have lots of direct connections to the Parlex center cable connector (discoverable with a continuity checker). Each Xilinx has its own EEPROM program. It should be possible for someone with access to a Xilinx universal programmer to desolder the Xilinx from a sacrificed TouchStream, read it's program off the EEPROM, and propagate this program to a set of spare Xilinx's procurable from Digikey or other chip supplier. Or for the truly clever reverse-engineer, there is a single row of (6?) thru-holes on the processor board that, when custom-wired to the proper pins of a Xilinx programmer, should be able to talk to (read/write) the Xilinx chip(s) serially without desoldering them?

Also, the first and second generation circuit boards had different Xilinx's with different programming. The first generation Xilinx's probably aren't worth attempting to repair or stock, since not that many shipped.

As I recall, some (all?) Xilinx arrays use external serial EEPROMs to store their programs. This might make the task easier. Most programmable logic parts include some type of "read-enable" fuse that, when blown, prevents programmers or diagnostic tools from reading out program info; this is to prevent, or at least impede, reverse engineering by competitors. I'd assume it's not possible to do this with external EEPROMs. Do you know, or can someone with a disassembled TouchStream find out, how the Xilinx chips are set up?

[Michael.Massing]

Michael - You might take a look at some of the liquid plastic products. Or liquid electrical tape....

Thanks, I did decide after the trip to Fry's last night to go with a less invasive fix, and this sounds as if it might be what I'm looking for.

OK, I did try this, and it seems to have helped somewhat - I did get several days of stability, especially once I provided a sturdier lap-level support platform. Now I find, though, that the kb does still go out from time to time (the lower half and/or the mouse functions become inoperative). The "fix" seems to be just running diagnostics.

Since the USB cable fluctuation seems stabilized by the goo, and the strain on the middle cable is presumably stabilized by the new support platform, I now read the kb behavior as acting a bit as if it is overheating. The USB connection was merely sealed, though, not trimmed, reconnected, or repaired in any way. Any further thoughts?

Thanks to all,
M.

p.s. The liquid electrical tape is indeed amazing stuff. I built up the exposed area slowly over the course of a couple of days to the diameter of the original cable sheathing, with some additional painting at the juncture into the kb to provide some strain relief. It looks reasonably unobtrusive, and seems to have helped narrow down the variables on the failures.

[jwr]

Hey, some news on the repair front. The right half of my failed TouchStream has been opened and boards desoldered. Several components have been replaced, but it seems that indeed the problem is with one of the Xilinx FPGAs.

Many thanks to pythagoras for the tips, they are extremely helpful.

Next things to try will be to read the programming off the JTAG vias for the FPGA, so that it can be replaced and reprogrammed. Hopefully Fingerworks did not protect the software from being read. If they did, I hope some good soul has the binary lying around on the hard drive somewhere.

I'll be posting new pictures of the processor board soon (probably tomorrow).

[goldfish]

Wow! I missed this thread. Good stuff, many thanks to Pythagoras for the excellent info.

Great to know in case my iGesture breaks down. Makes me wish I was an electrical engineer.

The FingerWorks technology really was innovative and superior--which of course was why the company was bought out. Regardless, the market for products which minimize RSI will only grow, and grow immensely.

Were "history" to restart, I think the market would definitely still be there. With a big marketing machine raising sales, and economies of scale lowering costs, the future for this technology would be very bright, and would benefit many more thousands--in addition to the thousands already helped.

Regardless, thanks again to Pythagoras for the repair tips, and as well to the creators of FW for the superb products.

[Michael.Massing]

As a quick update, I think I have narrowed my particular problem down to the ribbon cable. With this (apparent) knowledge, I've been taking steps to minimize cable strain and shock, both for home use and traveling.

A neoprene pad - custom cut from a spent wetsuit - was devised to both cushion and stabilize the ribbon connectors when I fold the kb for travel: keying surface out, since it is relatively indestructable. To my pleasant surprise, I discovered that deploying the pad under the keyboard during use at home made a significant improvement in "deadening" the previously imperceptible bounce of the metal tent, and made the keyboard as a whole much closer to the "zero-force" ideal.

I still have a bit of a potential issue with increasing the angle between the halves of the kb (I use the right half more perpendicular to my body's center line and closer to the body, to compensate for trackpad use), since that torques the connectors more than they were designed for, but so far so good.

Best to all,
M.

[jwr]

For all those interested, I've posted pictures of the right half of the keyboard, at http://jan.rychter.com/articles/2006/to ... pening.ucw. You can see how the board with the CPU looks like

It seems that indeed one of the Xilinx FPGAs has failed in my keyboard. We'll have a try at it with a programmer using the JTAG interface. I hope to high heaven that the code in the Xilinx does not have the protect flag set! If it does, there is little else I can do unless a kind soul sends me the code or makes it available on P2P somewhere.

More to come.

[jwr]

Luckily the FPGAs did not have the read protect bit set and we were able to read the codes off them. I have them stored for future reference.

[ahales]

Luckily the FPGAs did not have the read protect bit set and we were able to read the codes off them. I have them stored for future reference.

I have some very skilled colleagues that could help me with replacing the FPGA, but would need the code for reprogramming the new one.

Does anyone has the FPGA codes ?

Thanks !

[TorbenGB]

Does anyone has the FPGA codes ?

I have sent you a personal message.

[ahales]

Thanks a lot man !!!

[jwr]

…and let us all know if the repairs were successful!