Perpetual Thrill

It’s the first step of their grant process and is due tomorrow. The submission required a brief overview of the design and intent of this project, nothing complicated and all stuff I have thought though pretty well. You can find an excerpt on the new About page.

It also required an image, so I whipped this up in Sketchup. It’s a pretty close approximation, though I didn’t have time to rebuild that anatomical heart as a steel latice, which is what the actual sculpture will look like. Also, Sketchup does not have a lighting model – given that As One is a nighttime-only experience, the full-bright rendering is not doing us any favors.

(Not shown: the other operator behind that genny baffle / fuel depot there, or the probable crowd of passers-by)

Ok, now we are in business. Here is the current prototype, with the Olimex MOD-EKG under battery power on one side of the isolator and the microcontroller and computer on the other.

Aaaand, here’s the output from that run through a processing sketch that shows a totally reasonable ECG. Success!

Lastly, here is that same output, captured to a CSV and high-pass filtered with a 21 sample mean. Even with heavy filtering, the QRS complex is super clear and easy to spot. Excellent.

Here is the current setup, wherein I am attempting to isolate the sensor and microcontroller from the computer with a midi breakout board and a midi box on the other end. The theory is that since midi is optically isolated by contract the computer’s noisy ground plane will no longer be a problem.

Unfortunately, this did not work. Apparently, the length of the midi cable itself acts as enough of an antenna to introduce noise into the system (!). When the cable is plugged in, the ECG board will not read, and when unplugged it picks up my heartbeat just fine. Maddening, yes, but at this point also borderline hilarious.

I would try optically isolating the Olimex board’s serial port, but it is hardwired to 115200 baud, which is faster than any optical isolators support. Instead, I am going to try this weird-ass digital isolator chip. The datasheet says it is capable of 2mbps data transmission, which is more than enough.

In an effort to make sure the microprocessor and ECG boards do not receive noisy USB power, I made this terrible thing – a USB cable with the power wire physically cut. Unfortunately, while data transfer is successful, the setup was still too noisy, which surprised me.

I take this to mean that the common ground shared between USB and the batteries is where the noise is coming from. It’s frustrating, particularly given that USB isolators start at like fifty bucks. Also, given that the noise is low-frequency (otherwise the op-amps on the Olimex board would catch it) an RF choke is not useful here.

Actually, I tried three different power supplies, and all of them seem to be too noisy for the Olimex MOD-EKG board. Which is too bad, because it outputs a one-byte ECG reading over UART, which fits our processing needs perfectly. For now, I am running on batteries for testing per this pic:

This is using the Sparkfun AD8232 sensor board along with their Arduino clone. It matches the commercial sensor I am testing with, but requires a lot of pressure. Of note, only three of the pads are connected, and the right-leg drive is feeding into the right pinky area.