14
\$\begingroup\$

this is my first post, and I need some help/advice finding the right integrated circuits. I'll start by describing the project background a bit. I've received a fellowship to develop an open source electroencephalograph, the finished hardware will provide a platform for people to develop various creative and therapeutic uses for an EEG system i.e music controllers, game controllers or brain training programs. I want to use an Atmel MCU for the analogue to digital conversion, I'd like it to connect to a computer via USB and I'd also like the unit to be powered by the USB connection. I need help finding a suitable op-amp IC to amplify the signals from the electrodes prior to the MCU. The finished device will be 16 channels, so I'd like to find an IC with multiple op-amps. The electrical activity picked up by the electrodes will be in the region of 200mV and less, so I'll need a lot of gain. Is it possible to adjust the gain of an op-amp circuit with an MCU by using a digital resistor program? It would be nice if the hardware could be reprogrammed to work with various audio and sensor inputs. Any help or advice would be fantastic.

Jim.

\$\endgroup\$
9
\$\begingroup\$

Your average instrumentation amp can easily do 1MHz bandwidth; and your EEG should be no more than 2kSPS. So a multiplexer / Sample and Hold ahead of the the instrumentation amp ought to save you there.

But consider that the amp should be only a few dollars. Is it worth the multiplexing? If you do Surface Mount, the size will be quite minimal.

The Arduino cannot digitise faster than about 10kSPS, so you would need a faster A/D to do 16 channels. Something that can do 12 bits at 100kSPS would be nice. They are also fairly cheap.

Note that for patient safety you need optical isolation on signals, and a good isolated supply (battery or similar). Don't mess with safety in this area - if you need to get a high speed data stream out, build your own isolators or use fibreoptics to transmit the signal.

\$\endgroup\$
  • \$\begingroup\$ Cheers, that sounds like the best approach, I think I'll probably look at designing and arduino clone that has a built in instrumentation amps and a hefty Atmel chip. If I power the unit via USB would I really need to use opto isolators? I thought the protective circuits on most peoples USB would be more than sufficient \$\endgroup\$ – Jim Dec 15 '09 at 13:13
  • \$\begingroup\$ I'm looking at using this Atmel chip -> search.digikey.com/scripts/DkSearch/… Not totally sure, but I think it will do the job, and it looks like it wont need a separate chip to communicate with USB \$\endgroup\$ – Jim Dec 15 '09 at 14:36
  • 2
    \$\begingroup\$ There are no "protective circuits" on USB ports. When you touch a USB port, you're touching Earth ground. There's zero isolation. The only time it's ok to run without isolation in the EEG circuit is if it's floating (if you're connected to a laptop and the laptop is not plugged into anything else). Just build in some real isolation so you don't have to worry about it. \$\endgroup\$ – endolith Dec 15 '09 at 20:42
  • 6
    \$\begingroup\$ Hello Jim, As endolith has commented, USB has no protection. The earth it is referenced to might be the local "ground" potential, or it might not. If your laptop is plugged in to a 2-pin adapter, then ground for the laptop will be around 120VAC (in Australia), giving you at least a lot of common mode noise to contend with, the likely possibility of your patient getting "tingles" from your leads connected to his head (!) or in the worst case (of a fault in the power supply), a dead body and a manslaughter charge. \$\endgroup\$ – Tony Barry Dec 19 '09 at 0:56
  • \$\begingroup\$ For your instrumentation amp, an INA129 from Burr Brown (Texas) might be good. They are available in surface mount (SMT, small, SOIC) as well as through hole (TH, big, DIL). You might end up just putting 16 of them down on the board to contend with local polarisation around your electrodes (which will give you your biggest DC offsets to deal with). 16 amps allows each amp to settle and you can then acquire through a multiplexer and A/D. Your A/D might be an AD7940 which Farnell sell for just 12 bucks-does 100kSPS at 14 bit, single ended 0-5V. Nice. \$\endgroup\$ – Tony Barry Dec 19 '09 at 1:16
6
\$\begingroup\$

For such situations I think you generally want to use an instrumentation amplifier type of op-amp. They're made for differential signals, so you can easily subtract out the noise, they have really high-gain, so you can amplify the weak signals, and they have really high impedances, so they can sense delicate signals.

From a quick perusal of the OpenEEG site as davr suggests, it looks like they use a TI INA114AP instrumentation amp as the main amp.

\$\endgroup\$
  • \$\begingroup\$ Cheers dude, they sound like the perfect little amp, is there an equivalent IC that can handle multiple channels? I'd like to make a 16 Channel EEG, but I don't want to have 16 separate IC's for the job. \$\endgroup\$ – Jim Dec 9 '09 at 13:10
  • \$\begingroup\$ I've just been looking at my electrodes, they are single connection 1pin mini DIN. If I use an instrumentation amp with + and - inputs, does that mean it will output the differential between two electrodes? I'm a tad confused as to how the electrodes wire up. I have been researching an electrode array called a montreal 10/20, where the input of all the electrodes is summed and used to cancel out noise, how would this fit in with the instrumentation amp circuit? Cheers again Todbot \$\endgroup\$ – Jim Dec 9 '09 at 18:28
  • \$\begingroup\$ Yeah no clue. I've not played with instrumentation amps in 15 years, and have never done EEG stuff. I'd pour over OpenEEG. It looks like there's multiple different electrode systems described there, maybe you can find some circuit bits that do what you want. \$\endgroup\$ – todbot Dec 9 '09 at 20:02
  • \$\begingroup\$ Cheers for the help todbot \$\endgroup\$ – Jim Dec 10 '09 at 2:08
  • \$\begingroup\$ Yes, it does the difference between two electrodes, with common-mode signals cancelled out by the DRL amplifier. \$\endgroup\$ – endolith Dec 15 '09 at 20:40
5
\$\begingroup\$

Any reason you aren't using OpenEEG, a low cost & open source EEG system? They've been around for a while, and have lots of useful information on their website.

\$\endgroup\$
  • \$\begingroup\$ I've looked it's ok, but it's quite a basic setup, low bit depth and limited channels, it's also quite a large piece of hardware. I want a much more compact and comprehensive system. A channel with a bit depth of 10 is like a bare minimum for getting a decent representation of a brain signal, I'll definitely want more than that \$\endgroup\$ – Jim Dec 9 '09 at 15:31
5
\$\begingroup\$

TI has the ADS1298 for EEG and ECG front ends. It is described as a 8-Channel, 24-bit analog-to-digital converter with integrated ECG front end.

\$\endgroup\$
  • \$\begingroup\$ cheers dude - I got an promo email last week from TI and this chip was on the front page - crazy! \$\endgroup\$ – Jim Mar 30 '10 at 10:58
  • \$\begingroup\$ I can't get this chip yet - darn, they said it would be available some time from May to mid summer! - talk about drumming up interest \$\endgroup\$ – Jim Mar 31 '10 at 8:29
  • \$\begingroup\$ Rather than using the raw chip directly, you might consider using the open-hardware ADS1298 Based Biopotential Sensor Platform based on that chip in your first prototype. \$\endgroup\$ – davidcary Oct 16 '11 at 17:39
4
\$\begingroup\$

AD620 op amp has a schematic in the datasheet for an ECG circuit. Here you can also find class sheets that use the AD620 to build an ECG (1, 2, 3, 4). Similar to the schematic in the datasheet just a lot more verbose. It is only single channel. Cant quickly answer the rest of your question but hope that helps.

\$\endgroup\$
  • \$\begingroup\$ Unfortunately the signals an ECG handles are considerably stronger than the signals handled by an EEG, thanks anyway but I'd imagine they would be way too weak to pick up a signal through the skull. \$\endgroup\$ – Jim Dec 9 '09 at 12:39
  • 1
    \$\begingroup\$ Actually Jim, the hardware Cyphunk has suggested should be just about right for your purposes. The skin potentials an ECG picks up are comparable to those of an EEG (both measure nerve impulse). The AD620, AD624 or AD625 should suit your purposes admirably, depending on what functionality and circuit setup features you prefer. \$\endgroup\$ – Sketchy Fletchy Dec 10 '09 at 0:29
  • \$\begingroup\$ Cheers Sketchy, I was hoping to make something with 16 channels, so I've been looking for a chip that has multiple instrumentation amplifiers. Do you know of any, you can recommend? \$\endgroup\$ – Jim Dec 10 '09 at 2:19
  • 1
    \$\begingroup\$ There is a reason they only put one instrumentation amp in a package. The electrical layout is much better with a single amp per package, and there is no interaction between the channels. \$\endgroup\$ – Connor Wolf Jul 29 '10 at 7:26
2
\$\begingroup\$

In the 70's we developed 8 and 16 chan eeg telemetry units for hospitals. Need to keep input leads twisted, shielded and away from the transmitter antenna. Foor safety, We used the very first lithium cells to supply isolated power. Used L113 micro-power, flat-pack opamps for signal amplification. The outputs were multiplexed to the transmitter input. The fun part of this design was the method of de-multiplexing required to separating the 8/16 eeg signals. Have fun - this is a neat project!

\$\endgroup\$
2
\$\begingroup\$

I built an EMG amp as my master's thesis. It uses mostly standard parts (no expensive INAs) and has the safety features required for medical electronics. The requirements are similar to EEG amps, I guess. The low-pass filter and the gain stage can be programmed via 2-bit interfaces (00,01,10,11), which is cool if you want to hook it up to a microcontroller.

With an ADC, it might be better to do the isolation on the digital side, but you might be able to use some ideas from the amp anyway. One nice feature is the active shield at the differential input wires which allows wire lengths of < 10 ft (< 3 m) between the electrodes and the preamp, i.e. no little preamp box outside of the the amp's main housing.

The thesis itself is not available online, but you can find the key chapter in a PhD thesis that is partly based on my work. Feel free to check here (cf. chapter 8). Sorry the documentation is in German, but the circuit diagrams are quite international, I guess.

Also, I am not aware of multi-channel InAmps.

Related: Noise reduction strategies in electrophysiology

\$\endgroup\$
  • \$\begingroup\$ Cool - thanks for the PDF....I'm having a good read through it now, it looks real extensive! I've been using the ASD1298 by Texas instruments (has 8 instrument amps) it looks like the best candidate for my project. Just so busy with other projects right now! Will have to get cracking with the EEG soon. Thanks for the info and help :) \$\endgroup\$ – Jim Jul 28 '10 at 11:24
1
\$\begingroup\$

ModularEEG uses INA114.

Soundcard EEG uses AD8221 instrumentation amp. Yes, you can adjust the gain with a digital pot, but why would you want to?

alt text http://www.analog.com/static/imported-files/images/verified_circuits/CN0114_00_0415.gif

Also, a search: http://www.google.com/search?q=site%3Ati.com+PHYSIOLOGICAL+AMPLIFIERS%3A+EEG

\$\endgroup\$
  • \$\begingroup\$ I want to make a multi purpose board, that can be reprogrammed to work with various audio and sensor inputs. It's kinda like an Arduino but for people who want to do more serious DSP. Cheers for the help \$\endgroup\$ – Jim Dec 21 '09 at 11:05
  • \$\begingroup\$ I've gone and bought some of those texas instrument instrumentation amps. They look like the business, I'm just trying to work out a way to do multi channel. It would be good if all the signals could pass through the MCU ADCs and I could cross reference them afterward with the computer. I was thinking it would be good if the user could just select (in the software user interface) an arbitrary channel that the computer would then treat as the common mode, for example \$\endgroup\$ – Jim Dec 21 '09 at 11:15
  • \$\begingroup\$ The software end should be easy enough for me to work out, it'll probably take much longer for me to work out how the hardware will go! \$\endgroup\$ – Jim Dec 21 '09 at 11:19

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.