Hébert Labs

2012: Ten Years Plus!

I began working on an R&D project to develop a mass spectrometer in February of 2001. The collapse of the dotcom bubble was still weeks away, the attacks of 9/11 wouldn't happen for seven months yet, and I was trying to get a new business off the ground.

Not quite a year earlier the DOD Explosives Safety Board held a conference, The UXO/Countermine Forum, in May of 2000. Children in war-ravaged regions were being disfigured by abandoned landmines and insurgent ingenuity was making them increasingly difficult to detect. In these proceedings I read that the most effective means of detecting buried landmines was still a trained dog.

I read that the canine olfactory system is sensitive enough to detect the chemical vapor signature of explosives present in the atmosphere above landmine sites. And in a moment of inspiration I believed it would be possible to develop a mass spectrometer with comparable sensitivity. Moreover, I believed it would be possible to make it man-portable.

It's been over a decade now, and it seems like I've spent an inordinate amount of time working on this project, this next-generation mass spectrometer. Normally I would just point out that it often takes longer than one expects to do what has never been done, and that is certainly true. But it also seems a bit dismissive of a legitimate question. Why has this taken so long? Am I chasing an uncatchable quarry?

Oh, to be sure, there has never been a guarantee of success. Yes, I've completed the proof-of-principle tests, but until and unless I build a demonstrable prototype I have nothing but a hypothetical possibility. And if, after over ten years, I am still unsuccessful, why have I not yet conceded failure?

The answer is simply because I have not yet failed. I have met with a number of setbacks, obstacles I've had to overcome, but so far I've successfully overcome them. I've yet to encounter any difficulty which might have suggested the objective could not be achieved.

I'm not going to address the problems I faced during the proof of principle testing phase because I'm not going to discuss the physics of how this mass spectrometer will work. I hope someday to commercialize this technology, or preferably to sell it for an obscenely huge pile of cash, so I protect the proprietary aspects of it jealously.

Still, I will tell you that I had effectively demonstrated every principle necessary for this technology to work within just a couple of years. The big problems, the lion's share of the delays, have been quite mundane. Indeed, I'm not at all quite certain just how much of the delay has been due to nothing more than exasperation.

My first obstacle was interfacing to the host computer. I knew I would have tremendous amounts of data to transfer, and a standard UART serial port would be a crippling bottleneck.

I thought to implement a USB interface, which would be more than fast enough, so I bought a book. In the process of finishing that book I learned that the only way to implement a USB interface would be to use an embedded µ-controller.

I also knew that I didn't have the expertise necessary to implement an embedded µ-controller system, plus I needed to maintain strict control over the timing of various control signals (a circumstance not amenable to the likes of interrupt service routines and such). So I thought to use a parallel I/O card in much the same way I had done for the ion trap I had built in graduate school.

I bought a PCMCIA 24 bit digital I/O card (a QUATECH IOP-241), and it worked fine for one or two of my proof-of-principle tests. But there were fundamental limitations of the card that prevented me from bringing all of the functions together in a single cohesive system. I had no choice but to develop an embedded µ-controller system to control the mass spectrometer functions and interface to the host computer.

Unfortunately that meant that I had to learn, not only embedded µ-controllers, but also high speed digital printed circuit design.

So, I undertook a simple embedded µ-controller project for no other reason than the learning experience. I never completed that project, but I did learn enough to move forward with the mass spectrometer.

At the same time I took training courses and became certified in PCB Design Principles and Advanced PCB Design Principles through the IPC Designers Council.

Still, even though I was using an embedded µ-controller, implementing a USB interface was a whole undertaking in and of itself. So I proceeded on the assumption that I would be able to interface my system via a standard parallel port, or at least try to use a UART serial port.

By the time I had the new system designed and ready to test I had no viable choice but to use a Windows operating system for my host computer. And while I could write programs to run under UNIX, or DOS, or even Windows 3.x, I was not prepared to write a program to run under Windows 98 or later. I simply didn't know how to write a program to run under a multitasking GUI operating system.

So I had yet another thing to learn to do just so I could do what I needed to do. I had to switch from sequence-driven programming to event-driven programming. I bought a copy of Visual C++ and set about learning. At the same time I was being confronted with the realization that I could no longer address hardware directly; I would have to learn to write my own device driver for Windows.

In the meantime I was forced to accept that portable computing devices (laptop and notebook computers) no longer included either parallel or UART serial ports. If I was going to interface to my control electronics it had to be via a USB connection.

But that, it turned out, was actually good news. Silicon Laboratories (SiLabs) had just introduced the CP210x UART to USB bridge chip. It included device drivers and a complete API library, all royalty free.

Unfortunately it only came in a package (MLP-28) that included 28 interconnect pads plus two additional GND pads on a 5 mm square chip, none of which were visible from above. This pushed my PWB fabrication capability to the very limit, and it flatly exceeded my assembly capability.

I burned through countless chips trying to perfect a technique that allowed me to assemble (place and solder) this part reliably. In time, a SiLabs competitor (FTDI) would introduce a competing chip that used a nearly as small footprint, but which had standard quad flat pack leads. Since I could easily assemble this part I tried switching to their chip, but other issues sent me back to the SiLabs CP2102.

In the end, with the addition of a stereo microscope mounted to a fine pitch placement table, some trial and error and a couple of newly implemented assembly techniques, I was finally able to assemble the MLP-28 chip reliably. But it had taken thousands of dollars and more time than I'd have believed to accomplish.

Still, I had the basis for my control electronics and had even developed an impressively elegant circuit to handle the most difficult part of my system's function. Truly, this circuit was the last significant obstacle I had to overcome. It was the hump over which success was more probable than failure. It involved some incredibly timed interdependent signals operating at very high digital speeds, but it worked beautifully.

Once that circuit was working I improved the firmware in the embedded µ-controller, and the software in the host computer, dramatically increasing the data transfer efficiency. For the first time in a long time things were going well.

But not for too long. As I was working on one of the other control subsystems I came to a point where I needed to build a new set of electronics, to incorporate what I had done, and discovered that a particular manufacturer had discontinued a critical chip that my system was designed around. All of that critical timing, all of that elegant design, was for naught. I had no choice but to redesign the entire control system from scratch.

I suppose it could have been much worse. I could have been all the way into production when that particular manufacturer decided to pull the rug out from under me. And if that sounds like just a taste of sour grapes, there's a reason. If that chip manufacturer wasn't prepared to support such a specialized and proprietary chip long term, they shouldn't have offered it to begin with. I often wonder how many man-years of productivity were flushed away because of their decision in this matter.

But I don't have time to dwell on the past, and I didn't have time to waste on yet another empty promise of some specialty chip, no matter how convenient. So I redesigned the entire control circuit, but only using the most standard chips available.

I also decided to modularize the new design. In the course of all this I had worked up the hardware for the complete mass spectrometer. The physical prototype had vacuum, atmospheric sampling, and battery power to run for hours. It was ready to go, but for want of control electronics. I had made real progress and it was time to start finalizing things. By modularizing the electronics I could start bringing subsystems online one at a time until I had a complete and working prototype.

But, as usual, getting further down the trail meant discovering a new stream to ford. Impedance discontinuities were rendering the modularized control electronics dysfunctional. Keeping the traces short was no longer sufficient. I had to take measures to control the impedance of the traces, minimizing any discontinuities, especially at board-to-board transitions. That meant I had to fabricate 4-layer circuit boards.

You can read about my efforts to develop a lamination press and expand my PWB fabrication capability to include multilayer panels in another project report. It took some time, but was a successful effort. Even so, that doesn't mean I'm home free yet.

It turns out that making 4-layer boards means I've exceeded the reliability of the proprietary conductive epoxy that LPKF sells for "plating" through holes and vias. A board with dozens of vias might have a few issues, but a board with hundreds of vias had an overwhelming number of problems to find.

And as if that weren't enough, despite making sure that my latest design used only standard electronic parts, one of them has been made obsolete since my last parts order.

So I had to modify the new design to accommodate a replacement for that part before putting in my latest parts order. Even still, by the time I got the new parts list together and an order placed, three of the parts were put on backorder. So now I'm waiting for mid-August (a two and a half month delay) when my new parts kit should arrive.

I'm using this time to develop electroplating capabilities. I've designed and built a reverse pulse plating controller and I'm experimenting with plating techniques and chemistries. I hope to have a new set of PWBs fabricated, with electroplated through holes and vias, by the time the new parts arrive.

It's been a long time since I began this project. And there have been other delays too, things that had nothing to do with my research, things that make up the fabric of life. Nonetheless, I have pressed onward and upward in this effort, meeting each challenge along the way without yet encountering any difficulty that was irresolvable.

At no time in all of these 10 plus years have I ever believed I was more than 18 to 24 months away from being finished, usually believing I could be done in a year or less. And but for these delays I might have been. So I suppose it really is the case that sometimes, doing what's never been done just takes longer than one expects.

Applications for a Man-Portable Hypersensitive Mass Spectrometer

 

When Michael Faraday demonstrated his newly invented dynamo (an early type of electric generator), Sir Robert Peel, the British Prime Minister, famously asked, "Of what use is it?" Michael Faraday equally famously replied, "I don't know, but I'll wager that some day you'll tax it."

I can't help but feel a bit of a kindred spirit. Whenever I tell anyone what I'm working on, inevitably they ask the same question. "Of what use is it?"

Simply put, I can't even begin to imagine the full potential of this or any new technology. But what I can imagine is pretty impressive.

So where to begin imagining, if not the full then at least the immediate potential of this technology? Well, the impetus for the inspiration of this project was the extreme sensitivity of the canine olfactory system. So, as a first estimate of its potential, imagine anything a dog can be trained to sniff for.

  • Detecting buried landmines and unexploded ordnance
  • Detecting contraband
  • Detecting cancer
  • Locating disaster survivors
  • Locating disaster victims
  • Locating/tracking individuals and escapees

Then there are also traditional mass spectrometer applications that will be enhanced by the new capabilities of this technology.

  • Mobile atmospheric studies
  • On-site analysis of potential toxins, contaminants or contraband
  • Monitoring manufacturing processes for extreme purity against contamination

And of course, these are only a few that I can think of spontaneously. Who can imagine what other applications will grow from these? Will detecting cancer lead to other medical diagnostic techniques? Will tracking individuals lead to new identification techniques?

Hopefully we'll all find out in the not-too-distant future.