Seems that the talk I gave at the BSN2009 converence is now online. After watching the video, I think I need to spend more time in the gym. The video is in some windows streaming format, but I’m trying to get a more sane version if it post. For the time being, this link should work..
So I’ve not posted in awhile, and I’m not going to try to play catch-up, but I’ll apprise whoever is reading this of some stuff that’s on my mind. First off, why do people fear using real tools to get a job done. Like what’s the point of a little duster when you can use a leaf blower. Cleaned out my computer case in 2 seconds and didn’t destroy the ozone or cause pollution.
Quick note that I have a talk at the Body Sensor Network conference last week. It went really well, and I met some cool people.
I’m going to be putting together some demos and new material in regards to my research shortly. I’m going to be creating a fork of the development effort to look into potential spin-off possibilities. I’m getting a license from the university for my code, and part of my royaltee payments will go back to me — funny how that works. I also migrated to Altium Designer, and now, Cadsoft Eagle looks like microsoft Paint. Bought year’s license for $130 given that I’m a student, normal price is $14,000. And I hate to say it, but after using it for some time, I think that’s a reasonable price given what it offers. The lab is coming together, and I hope to have a dedicated office space in the next 6 months. Anyhow, stay tuned for some updates.
I’ve just finished the second week of human trials, and I’ve already learned a lot about device usability and some weaknesses which I want to incorporate into a newer version of the chest strap. The data is quite fun to go through, and it looks better than I expected for an ambulatory device. Here’s an excerpt from one of our subjects..
It was a few years in the works, but the time is finally here — time to collect some real data from random test subjects. We’re going to be collecting SpO2, ECG, EMG, BioImpedence (respiration), and 3-axis accelerometry at 256Hz. The firmware seems fairly solid, and the signals all look really good.
By friday, we should have 24 hours worth of nice, clean, data. Else, I’m going to have a busy haloween weekend.
After dealing with a few ugly hurdles (mostly having to do with a horrible new serial module in the MSP430F2618), life is starting to look really good. The filesystem is working very nicley with a shared serial bus with the accelerometer (SPI) and the bioImpedence circuit (AD5933 impedence measurement IC; I2C). See how nicely they play together..
And even the SpO2 signal is finally looking really /really/ nice..
Though that data is from a transmissive sensor. We’re still investigating the efficacy of an upper arm reflective sensor, but I’m quite pleased given the complexity of the new device.
Stephan from VG Bio came over for a visit and we spent the day debugging the BioImpedence circuit based on the AD5933 IC. It consists of a voltage to current converter, a differential amplifier, and the AD5933 IC (as well as a number of passives). Using the development software, we were able to graphically visualize Stephan’s breathing. Pretty cool.
Look here for more photos.
A long time in the works, but this version adds full pulse oximetry support, includes a mini-B usb connector for interfacing and battery charging, and a lot of RF filtering to help make the ECG signal more stable in noisy environments.
EDIT: Forgot to mention, we’re also switching to the LIS302DL accelerometer, same one as in the iPhone, and we’ve switched our reference voltage amp to the AD8644 to help with the SP02 stuff. The new design frees up some A2Ds which we’re using to monitor battery voltage and bringing out the other unused channels to the big connector on the bottom.
I actually had a great time and talked to lots of interesting people at the ESC conference. I did notice two trends that seemed prevalent — not that they have much to do with my project, but that I still found interesting.
First, the massive interest in capacitative touch sensors. It seems as if a number of big players from Silicon Laboratories, Atmel, Cypress, etc. are really promoting their touch sensor products. Atmel bought Qmatrix which was the original player in this space, and it seems as if other vendors developed their own tech. I didn’t see much of the guts of the Silicon Labs product, but I did get a demo board of the Ateml product. It doesn’t work very well. The Cypress guys seemed to have a neat product — it seems as if they’ve incorporated FPGA-like logical blocks that you can configure on the fly — as well as assign any pin any function (or so it seemed). Though I can’t really recommend them as they made me sit through a 10 minute long speal which was really 20 minutes long – was boring and conveyed no information for the promise of a dev board to play with — only to walk away with a tshirt instead. *Grumble*
The second big trend is in using 2.4Ghz transiever ICs for zigbee, zigbee variants, or for propriatary devices. All the products seemed quite low power, small form facture, required a minimal of extneral components and promised quite a range. Atmel’s new Raven product demos software developed by some company up in SF which lets you put an IP stack on one of Atmel’s transievers. They were able to ‘ping’ a remote device which was really nice. I like the idea of being able to telnet to my wireless peripherals. The guy I talked to was an eecs major from Cal.
On other notes, they had some other interesting stuff going on. They took apart a $2000 10″ OLED display to show the innards, but it was fairly boring. Supposedly they took apart a space-suite the day before but I missed that. They supposedly had beer, but it was all gone by the time I got there. I played with some real oscilliscopes that made me drool with envy — seems doing real time decode of various serial protocols is the new big thing in that catagory. Of honorable mention is a company that makes hardened memory that looks like physical keys, and a company playing this video on one of it’s demo boxes of an elephant painting. I have no idea what they were selling, but was blown away by the video. It make me seriously respect elephants.
Finally, I had a nice talk with the people at TI and they mentioned how Harvard, and Rice University came out with a nearly identicall device as the Tricorder we are working on. Here’s a photo of the rice device. It’s the same size if you exclude the antenna at the top of our device, has the same microcontroller, and uses the same instrumenation amplifier. We started our project almost 2 years before we posted anything online — I’m not sure what the harvard groups device looks like, but I would imagine it to also be simmilar. It makes me think of the quote by the inventor of PCR (who is a surf bum from my understanding). He said something to the likes of his discovery was a natural consequence of several discoveries that preceded it. I wonder if this device is just a function of the technology becoming mature. They did introudce a new processor that’s 2x faster, has more memory, and uses less power when it’s in a suspend mode. Given that it’s pin for pin compatible with the processor we’re using, I hope to implement them on the next version (which I hope to get out this weekend).
I also went to a tasty Thai restaurant afterwards with Corpse and Taner. Super tasty food. I’m not telling you were as I dont like places getting popular; I find that popular restaurants get complacent and the food starts to suck. Happened to Thep Phamon.
So my thought is that we’ll have to try a number of configuration for the pulse oximeter, and it will take too long to iterate through a PCB house, so I decided to make the boards in-house using our Sherline 2000 CNC mill. One comment : IT SUCKS. The damn OS crashes every 5 minutes, it takes 20 minutes to reboot because it wants a network connection, and the application run so slowly that it’s hard to tell if it’s frozen or it’s still drawing the next dialog screen. And when it does work, it throws random errors midway and forces you to start over again. Not to mention the device itself. It’s unstable and the whole platform will rock messing up the z-axis.
Ok, enough of the ranting. After 3 broken end mills, 5 hours over 2 days, and dozens of PC boards, I have successfully milled out a transimpedence amplification circuit based on the OPA380 which is supposed to be the dog’s bollox of tranimpedence amps. The white thing is the backside of the PIN photodiode. I didn’t have a 10MOhm 0805 SMD resistor on hand (I only stock up to 2MOhm, so I went with a through hole resistor).
I might test it out tonight after a few beers, else I’ll test it out in the morning.