Hm.. personally I sorta dislike the schematics for having a number of symbols flipped upside down. The PCB's seems to have a few questionable designed parts, too. Aside from that there seems plenty of information to soak up. I don't think "porting" will work given the amount of miniaturisation we'll have to pull off.
In attempt to contribute to the thread, I thought to post my opinion on some major obstacles I see in implementing UTAH MEA this way, purely from engineering side:
- A solution has to be found regarding connecting wires for the array, which apparently demonstrated tendency to fatigue and break in Warvick's experiment. I can't really see this project progressing until this issue is solved. Being limited to off-the-shelf (COTS) components, integrating the whole implant on the top of the array is an unlikely option, so some sort of wired connection will be necessary. With only about 0.3mm pitch the copper wire thickness that can be used is rather limited - does anybody know how many strands did the wires in Warvick's experiment have? Were they single stranded? Multi stranded wires would be expected to be somewhat more durable.
Probably the next most manufacturable solution would be flat-flex cables, ball bonded to back of the array, likely in multiple layers due to large number of conductors required. These layers would need to be ball soldered together and to the array which may be too difficult without specialized equipment.
Or, more exotic solutions could be researched such as carbon fibre conductors bonded to array using conductive epoxy. Much more fatigue resistant and biocompatibile than copper, but may be too high resistance for stimulation.
Secondly, I see that there is no general consensus about what the circuit itself should look like, so I'm offering my opinion based on MEA research I've read so far. The circuit itself should obviously occupy as little volume as possible (a few cubic cm at most), preferably be in a thin form factor (1-3mm) as well as flexible and compliant. This calls for absolutely minimal number of tiniest SMT components available.
One other constraint is power dissipation in the implant issue which I don't often see addressed!. Considering human body only dissipates some 10mW/cm3 the implant has to be limited to similar power level to avoid thermal tissue damage. On the other hand, this power needs to be sufficient for data processing and transmission at required bandwidth, say 10ksps * 100 electrodes = 1Msps or so. Atxmega128b3 for example is already 21mW at 10 Mhz, very borderline indeed!
The system has to be able to set MEA electrodes to following states: logic high, logic low, high Z and record. Due to severe circuit constraints, biphasic voltage stimulation is the only practical method (has been well proven on in vitro arrays).
One proposed solution could involve 1 or more mcu's to provide first 3 requirements by interfacing every MEA pin to a port pin. For analog recording, 6 16-1 MUX chips followed by a low noise amplifier could be used in conduction with on-mcu ADC.
Some mcu's tend to have lots of ADC channels, up to 16, and even built in amplifiers such as AVR series. A good designer could be tempted to simply use several mcu's with MEA connected to their port pins, having internal analog mix and amplifier take care of everything else. If any filtering is required it could be done digitally; this design is probably the absolute minimum number of chips I can imagine with COTS.
I wouldn't venture into world of fpga's and external arc's; too many external components and too power hungry in opinion (you guys are free to prove me wrong though!)
I have worked a lot with wireless power for my thesis and it can indeed be solved with a few components; however communication is another story. Usually we want to keep processing on the implant side at minimum and send raw data, which in best case may be 1Msps * 10bits = 10Mbit/s, which is beyond typical radio standards (at allowed power level) and near field communication (even at Q = 10 (coils almost touching), bandwidth = 2.7Mhz for 27Mhz which is still not enough...) Indeed, it appears no practical method of intra-body communication at required bandwidth yet even exists! We have to either implement some very clever data compression or implement new methods. One of them could be what I call 'bio-port' concept, which in short means ditching all modulation and transmitting baseband digital signal by direct conduction through body, potentially doing hundreds of Mbps with very low power consumption. I will post more about this in future after I'm done with my thesis...
I think I have added enough brainstorming fuel so I should better end my rant here... Provided we can solve all issues I presented here, I would be happy to go ahead and design a pcb of my version of the solution.
On a side note, have you guys thought in actually getting involved with some real research, rather than just keeping this a lucrative hobby project? Ideally you would find interested research institute or university team who could provide you funding, and perhaps even a surgery for free (as I presume Warvick did?)
@Axon in my experiences mcu's have their issues with input buffers, most eat current for breakfast when the input voltage is floating halfway the supply voltage. As for communication i suggest optical systems, IrDA physical layer with custom protocol running on top of it can give high bandwidth with low power and a tiny footprint, only drawback it's usualy halfduplex. I'd say FPGA have a decently low power demand for the amount of signal processing you could run right in place, some even come with DAC and DSPs so they are very worth to have a look at and experiment with. About heat.. while the human body generates very little heat, it's very effective at transporting heat away and distributing it (i'd love to see a coated resistor injected under the skin with thermal imaging recording it to see the heat distribution and temperature rise *hint hint* ).
The issue you are mentioning seems interesting indeed - on which family of microcontrollers have you noticed it particularly? Cross conduction seems to become less of a problem with decrease in supply voltage; however I must also admit I never tried measuring increase in current draw in 5V mcu's yet. This may explain why ADC reference voltage is usually chosen more like 1V or so which puts the pin well into logic zero region.
5V microcontrollers are more interesting for this application since they ofer higher drive strength; note that Warvick's implant used as much as 20V (though current limited so we don't know what exactly the voltage drop was) which suggests that even 5V maz not be enough for successful stimulation. And it's rather hard to test this without doing actual implantation. In vitro MEA's like I posted above seem to be OK with about 5V; but if more is required COTS implementation becomes dramatically more complicated.
I have actually investigated optical communication before, and it seems to be somewhat on par with baseband conduction methods (which I still think would be lower power than LED drive; a comparative study between these would be useful indeed).
Are you aware of any existing IRDA devices in sufficiently small form factor to be used in implants?
Regarding FPGA's, Lattice seems to offer some ultra low power devices with inbuilt configuration memories and simplified power supplies which may be borderline usable for implant application; however the voltage is lower than 5V again, and external ADC may not be worth it again.
There might exist biomonitoring IC's that could potentially integrate multiplexing, amplifier and ADC into single device which would offer superior performance to crappy mcu ADC's. In that case FPGA might become a more sound option in my opinion. EEG IC solutions may be something to look for...
Atmel's atmega and attiny series are affected by the increased current at vcc/2 input voltage. From the datasheet:“If the input buffer is enabled and the input signal is left floating or has an analog signal level close to VCC/2, the input buffer will use excessive power”. For my tests that resulted in 300μA extra supply current for each input pin hovering at half the supply voltage. Impossible to turn of the input buffer which causes the extra current. If you know a mcu family that does not suffer from this lemme know (I have some older circuits I'd love to test again without the increased current)
There are many companies producing analog frontends for reading bioelectric signals. Texas instrument has some nice chips ( ADS129x series). Extremly low power (less than 1mW per channel) and up to 8 channels.
For IrDA I had an eye on the MAX3120 for a very low power infrared port (tiny footprint , only makes about 115kbaud but that's fine for many applications, shutdown current in the nA). There are probably newer chips around. The fastest non-IrDA communication I know would be 10Mbit manchester encoded (bit more power hungry but fast). Got a couple of more optical com circuits around. With an FPGA or CPLD they can work anywhere from 0 to about 2 , maybe 4MBit/s.
Using an FPGA to do the time-sensitive operations such as generating the control signals for the multiplexers and providing outputs to build a bunch of discreet DAC might be a pretty good way to handle things. except for the massive number of transistors we pull into this.
Different operating voltages within the system should be no problem. You'd need to separate analog and digital circuitry anyway. Nothing prevents one from having multiple power rails around.
About form factors , those aren't that much of an issue. Throwing enough money at chipfab they can get your circuit size down. There are also services offering 3d circuits with bare die's arranged in a volume with interconnecting wires. crazy stuff, not cheap but very awesome. And even without, there's plenty of space inside a body.
I'm no use as an electrical engineer, but I came across some work from this guy's lab where they were implanting chips in mice, and using inductive coupling to power the device. Here's a link to his lab's webpage:
The aim was to run long-term data collection experiments from implants in rodents. From what I remember, they had the size down to about a cubic centimeter, and they were putting them in mice, not rats. This is pretty vague information I realise, but it might be of some use.
If you got a schematics for what you want I could do that PCB and assembly if you like I have access to an entire electronics assembly plant here at work I will let you know that not having a battery could cause you some problems because you are going to be beaming your circuit with RF the entire time you use it serious shielding will be needed to keep the noise out. I assume you want to used 0201 components for the project to keep the size as small as possible.
my first thought would be to take an RDIF tag and modify the design to run the I/O.
I have literally no idea what a MEA is currently but ridiculously small electronics are what I do for a living so let me know if you want any help with it.
I will look in to wtf an MEA is to get a better grasp on the project in the meantime. Is this project to produce a single implant proto or to produce them in a decent quantity? if its for quantity your budget most likely won't get you there. I just had to drop 10k on a test fixture for a board were building. producing electronics can get expensive fast. if its just for a single unit you might be ok to hack and tack to get it done.
A few days ago i stumbled over some links about raster microscopy, one included tip preparation. Those tips are basically Pt-Ir wires etched into very fine tips. That's not a MEA yet, but it could be a start towards constructing a MEA at home on a tight budget. It still needs isolation layer, good ways to mount them etc. but at least the tips could be done. related link: http://physics.nd.edu/assets/118523/czerepak_annareduced.pdf
Comments
The issue you are mentioning seems interesting indeed - on which family of microcontrollers have you noticed it particularly? Cross conduction seems to become less of a problem with decrease in supply voltage; however I must also admit I never tried measuring increase in current draw in 5V mcu's yet. This may explain why ADC reference voltage is usually chosen more like 1V or so which puts the pin well into logic zero region.
There are many companies producing analog frontends for reading bioelectric signals. Texas instrument has some nice chips ( ADS129x series). Extremly low power (less than 1mW per channel) and up to 8 channels.
For IrDA I had an eye on the MAX3120 for a very low power infrared port (tiny footprint , only makes about 115kbaud but that's fine for many applications, shutdown current in the nA). There are probably newer chips around. The fastest non-IrDA communication I know would be 10Mbit manchester encoded (bit more power hungry but fast). Got a couple of more optical com circuits around. With an FPGA or CPLD they can work anywhere from 0 to about 2 , maybe 4MBit/s.
Using an FPGA to do the time-sensitive operations such as generating the control signals for the multiplexers and providing outputs to build a bunch of discreet DAC might be a pretty good way to handle things. except for the massive number of transistors we pull into this.
Different operating voltages within the system should be no problem. You'd need to separate analog and digital circuitry anyway. Nothing prevents one from having multiple power rails around.
About form factors , those aren't that much of an issue. Throwing enough money at chipfab they can get your circuit size down. There are also services offering 3d circuits with bare die's arranged in a volume with interconnecting wires. crazy stuff, not cheap but very awesome. And even without, there's plenty of space inside a body.
https://engineering.purdue.edu/BME/People/viewPersonById?resource_id=20328
The aim was to run long-term data collection experiments from implants in rodents. From what I remember, they had the size down to about a cubic centimeter, and they were putting them in mice, not rats. This is pretty vague information I realise, but it might be of some use.
Currently we are waiting for glims and cassox to wrap up their current projects before we really get started on this one.
There are a few links to MEAs in this thread you might want to read.
A few days ago i stumbled over some links about raster microscopy, one included tip preparation. Those tips are basically Pt-Ir wires etched into very fine tips. That's not a MEA yet, but it could be a start towards constructing a MEA at home on a tight budget. It still needs isolation layer, good ways to mount them etc. but at least the tips could be done. related link: http://physics.nd.edu/assets/118523/czerepak_annareduced.pdf