MYOsc - A submission to the hive mind

LesHall
Posts: 36
Joined: Tue Feb 17, 2015 2:45 am

MYOsc - A submission to the hive mind

Postby LesHall » Tue Feb 24, 2015 5:37 am

Dear Hive Mind,

I had plenty-o-fun distilling down a Myoelectric control circuit to it's bare essentials. What I am presenting here is some general and untested ideas for electronic circuit implementations of myoelectric sensors / drivers. These are just "thought experiments" and are completely untested. Unfortunately I'm not set up to do any testing, so I'm hoping someone can step up to the plate and give these or similarly themed circuits a try.

Let's begin with the following schematic from advancer Technologies:

Muscle Sensor v3 Schematic.jpg
Muscle Sensor v3 Schematic.jpg (140.11 KiB) Viewed 2044 times


Things to note are: the AD8226 differential amplifier has a noise spec of 8nV/sqrt(Hz), which kind of sets the bar for the noise spec. 8 or lower seems to be a safe bet, higher is taking a risk since this circuit is known to work properly. The 240 Ohm resistor R1 is selected for a gain of approximately 200, and the circuit works for power supply voltages of +/-3.5V so we can expect a signal output in the neighborhood of +/2V that was boosted from +/-2V / 200, or in other words the input signal is roundabout +/- 10mV.

The rest of the circuit is very straightforward and well designed from a conventional engineering sub circuit standpoint. First there is the diffamp mentioned above, then a full wave rectifier, a first order low-pass filter, and an adjustable gain stage of unspecified gain range. The circuit is created by Brian E. Kaminski. So it's obvious what the circuit does, and it appers to be very simple, straightforward, and elegantly designed.

Enter MYOsc, my idea for reducing the circuit to it's bare essentials. Please note again that unlike the tested and working circuit above, my circuits drawn below are crude and early stage experimental in nature. After discussign how they work or might work, I will add a word on carbon fiber implementation. OK< have a gaze for a moment or two at the image below (quote courtesy of http://www.adafruit.com).

MYOsc.jpg
MYOsc.jpg (1.39 MiB) Viewed 2044 times


All circuits are based on the NE5532 dual low noise opamp with a noise spec of 5nV/sqrt(Hz), so about the same noise spec and slightly better than the sensing diffamp on the Advancer circuit. All three of these circuits are also oscillators based on the relaxation oscillator which produces a square wave output on the opamp's Vout pin, a square wave divided down from the output on the non inverting pin, and an exponential rising and falling wave on the inverting pin.

The idea here is to slightly skew the steady-state oscillation with the MyoElectric stimulus. For example, if we take the relaxation oscillator and add a resistor to the inverting input and a voltage source to the other side of that resistor, then we can modulate the frequency of the oscillator. We know also that frequency is incredibly sensitive and can be measured or used in circuits with very high precision, so if we vary the frequency of the oscillation (which is large signal) with the Myoelectric sensor signals (which are small signal), then most likely the variation in frequency will be small but measurable to a reasonable level of accuracy.

For example, let's say we are running our oscillator at 1kHz or so and we monitor it with a micro controller such as an ATtiny85 running at 8MHz. Our processor is running 8,000 times faster than our oscillator, suggesting that if all other factors such as counter/timer implementation as in the Arduino pulseIn() measurement function, there should be plenty of accuracy to do things like control the opening and closing of an e-NABLE hand. That's my story and I'm sticking to it!

Of course, if the accuracy is insufficient, a differential preamp my be required. I'm hoping, however that these circuits will be sensitive enough. One way to make them more sensitive is to vary the voltage division in the divider side of the oscillator. If we use, for example, 200k Ohms in the top resistor and 1k Ohms in the lower resistor, we will get approximately 200:1 division, effectively bringing the sense signal down to the level of the input signal. With tat in mind, our sense signal should be sufficient to modify the frequency significantly so that accuracy will be greatly enhanced.

To me the most interesting circuit is the circuit labeled as 2 with a circle around the 2. Here we see a relaxation oscillator with the two ground pins connected not to ground but to two muscle electrodes. A third contact, not shown, would connect ground to the body. Then when muscular flexion occurs, the oscillator is imbalanced and will change frequency, especially with large feedback resistors and a small divider resistor and a small capacitor.

OK, so let's say one of these circuits or a similar circuit actually works. We can apply some electrodes to our forearm and get a large signal square wave oscillation with frequency that is a measurably strong function of the muscle flexion. What to do with that? Well of course there is always the micro controller solution, however simpler may be to drive a solenoid directly or with a simple interface circuit such as a 555 circuit to convert the frequency to a pulse width modulated (PWM) signal.

Along these lines, it may be preferable to work with a 555 or similar PWM oscillator, make it sensitive to small inputs, and then apply the myoelectric signals, buffered or maybe even unbuffered, to an appropriate set of input terminals. The resulting PWM signal could be applied directly to a servo actuator, eliminating the need for any microprocessor at all. Also the frequency modulation signal, perhaps with a diode added to the MYOsc circuit to make it have a narrow fixed pulse width on one side of ground and a variable pulse width on the other side of ground, could be applied directly to the servo. The low pass filter function would be provided by the mechanical inertial of the system.

So you see there are many possibilities here and the ideas are wide open to exploration. i'll leave you with one more thought, which is a word on implementation. Hmmm, don't wan to lose this long edit, will speak to implementation in the next post.

Les

LesHall
Posts: 36
Joined: Tue Feb 17, 2015 2:45 am

Re: MYOsc - A submission to the hive mind

Postby LesHall » Tue Feb 24, 2015 5:52 am

Now a word on Implementation. The current setup of using electrodes with snap-on wire leads in a harness and a TRS connector with jack may possibly be reduced with the following arrangement.

The idea is to use carbon fiber PLA for the circuit and electrodes. This filament consists of 85% PLA and 15% chopped carbon fibers. Because the fibers are conductive, the material may be used as a resistor or a high resistance skin contact. With a dual extrusion printer or some imaginative use of a single extrusion printer, a circuit complete with spring-loaded chip-gripping contacts (utilizing the springiness of the chip pins), SMT or thru-hole capacitor mounts (friction fit SMT or lead-wrapping thru-hole for solderless alternatives), and epoxy potting to hold it all together.

The result could be a single unit that receives power and ground and outputs signal on a total of three contacts, and simply rests on the skin surface to pick up the myoelecctric signal. Simple yet effective, it eliminates the need for a circuit board altogether, eliminates the electrodes, the wires, etc, and should be low enough in cost to be disposable or cleaned easily many times before disposal.

Yes, tonight I am doing the happy dance of 1,000 Snoopy Dogs! I love imagineering stuff, don't you? What would really rock would be if these concepts actually worked. Either way, I'll enjoy my happy dance and stop typing. bye now...

Les

Vince-Heidelberg
Posts: 4
Joined: Sun Mar 22, 2015 3:16 am

Re: MYOsc - A submission to the hive mind

Postby Vince-Heidelberg » Sun Mar 22, 2015 1:55 pm

The Question is why should we do this ? I would prefer to stick to normal EMG Circuits.
If you want to skip the need of an uC you can add some analog circuit do make a basic threshold control for open/close the Hand.
The Circuit from AT is a good one , only point i would add is a Bandpassfilter for the normal Freq of EMG that we are focus on.
If you want to test arround go to TI and order some FreeSamples and start testing around for example the INA128P as Inst.Amp.

LesHall
Posts: 36
Joined: Tue Feb 17, 2015 2:45 am

Re: MYOsc - A submission to the hive mind

Postby LesHall » Sun Mar 22, 2015 2:30 pm

That's a good question and I have an even more important question: Why SHOULDN"T we do this? for liability reasons. If we do some fancy fan dangled BS is our circuits and then an accident happens, what do you think the lawyers are going to say? right.

So then why did i post this here? For research purposes. I had the ideas, I just LOVE to share my ideas and maybe they could help someone out. Like for example, if we were to make an oscillator that varied it's actual pulse width but not frequency as a function of the body's signals, then we could directly drive a servo with it. No processor, no programming, no power consumption, fewer parts, higher reliability, and a simpler more elegant system overall. So yeah, let's do the practical where practical is necessary and the research where research is necessary. 'nuff said!

Les :)

Vince-Heidelberg
Posts: 4
Joined: Sun Mar 22, 2015 3:16 am

Re: MYOsc - A submission to the hive mind

Postby Vince-Heidelberg » Sun Mar 22, 2015 3:24 pm

Sorry i dont mean that in a bad way when i ask why should we do this !
Dont get me wrong i like the idea mybe my english is not that good and you understand me wrong :-)
Im not a Pro but i think if you just look for the Freq you need a much higher NoiseControl so at the END you will need more Parts and better qualitity.
The carbon fibers PLA sounds interesting do you have a link or informations about it for me ?
When you start do think about the desing beware that this will not akt like an antenna for Noise.

The Control itself its easier to setup with a uC then catch the Freq and drive the Servo direkt.
What do you do if a NoiseFreq come up and drive the Servo without you doing anything.
I have some PDF about the FreqDomain on EMG Signals if you want i can upload it when it helps you.

When you looking for a System without uC and Programming have a look at the analog circuits for threshold control.
Looking at the Freq will not be very practical at the end. Too much things you have to look at.
But as you sad its because of the research itself so if you need help in anyway , have questions ... just let me know i will try to help you.

If we do some fancy fan dangled BS is our circuits and then an accident happens, what do you think the lawyers are going to say?



From that point of view there is so much that we have to look at.
Too sell a medical Dev you have stick to very strict rules !


Return to “Myoelectric”

Who is online

Users browsing this forum: No registered users and 1 guest