Re: [AUDITORY] EEG purchase

---- External Email: Use caution with attachments, links, or sharing data ----

Dear all,

I wanted to publicly thank everybody for all the responses I received regarding my upcoming EEG system purchasing. I received very useful feedback that will help me make an inform decision on which system will be most optimal for my lab. It’s quite obvious from our exchange that every system has its pros and cons, and it all comes to finding the system that better suits your needs. I’m still doing my math but I’ll make a decision soon. Nevertheless, I learned a lot from this exchange. Kudos to the Auditory List for being such a useful tool!

Finally, I was contacted by Biosemi regarding the issue that Amos brought up regarding the dithering process with their system. They provided a very detail response that they encouraged me to share with all of you. For those of you interested on the topic, I’m including their response below.

Thank you again everybody. Best of luck for the new academic year!

Sincerely,
Carlos

RESPONSE FROM BIOSEMI

- The spike problem doesn’t play a role with the special ABR channels because these are amplified an extra factor of 64 before the ADC. So the spikes are far below the noise level over the full bandwidth. 100Hz to 3kHz.

- The spurious tones being discussed in this thread are inherent to Sigma-Delta ADC converters.
- The phenomenon of sharp spikes at 8 Hz intervals in spectra of long periods are caused by spurious tones generated in the Sigma-Delta ADCs.

- All the ADCs in the system show the same behavior. However, the amplitude of the spikes differ slightly between the channels due to (internal) ADC tolerances and different electrode offset levels at the inputs, Therefore, referencing decreases the spikes, but does not eliminate them completely (and the effectiveness of referencing differs for different channels).
- The spikes are perfectly locked to the sampling frequency, and therefore of nearly infinitely narrow width (see NoiseSpectrumShortedADCzoomed.jpg attached and at https://imgur.com/a/Y9KVhK0). The spikes are always fixed at the same frequencies (multiples of 8 Hz). The noise power in the spikes is very small. Also note that only FFTs over windows in the order of hundreds of seconds at the high sample rates show the effect.
- The spurs (spurious tones) have an amplitude of approx. 20 nVrms in the passband (see NoiseSpectrumShortedADCLin.jpg attached and at the imgur link). The maximum input signal is 0.5 V peak-to-peak (approx 0.2 Vrms). This means a SFDR (Spurious Free Dynamic Range) of 140 dB, which is according to the specs of the ADC. So, our circuitry is able to use the ADC to its specs (we do not add or increase spurs). The Sigma-Delta ADCs used in competing systems have a similar SFDR specs and are likely to show similar spurious tones. Note that nearly all contemporary EEG system are based on Sigma-Delta ADCs.
- Besides the spurs at 8 Hz intervals, ActiveTwo shows a perfectly clean spectrum, see plots of measurements with two EX electrodes in a bowl of water (NoiseSpectrumElectrodesLin.jpg and NoiseSpectrumElectrodesLog.jpg -- FFT of the difference between EX1 and EX2, 1 second FFTs, 0.5 second overlap, average of 1999 FFTs over a period of 1000 seconds). The small peaks at 50 and 450 Hz are due to mains interference.
- Competing systems may have a HIGHER noise density at high frequencies than ActiveTwo (consistent with the choice of a chopper preamp which has constant noise density, whereas we use a linear preamp with decreasing noise density at high frequencies). This higher noise may cover the spurs in those systems. In addition, competing systems may use very high-order (very steep) FIR decimation filters, whereas we use a 5th order CIC filter. High-order FIR filters may be a bit less susceptible to spurs, but generate a ripple in the passband (we have a flat frequency response) and ringing on step input signals (we have no ringing and overshoot).
- For high dynamic range in the time domain, a Sigma-Delta ADC (which we use in the ActiveTwo) is the best choice. However, for high dynamic range in the frequency domain it is preferable to use a SAR (Successive Approximation Register) type ADC, which we use in the hypersampling version of ActiveTwo, see https://www.neurospec.com/Products/Details/1076/activetwo---hyper-sampling-edition
- The whole spurious tones issue only plays a role with input signals on standard active electrode channels (not ABR channels) above 1kHz, because below 1 kHz the spurious tones are below the amplifier noise.

From: AUDITORY - Research in Auditory Perception <AUDITORY@xxxxxxxxxxxxxxx> on behalf of Ravinderjit Singh <singh415@xxxxxxxxxx>
Reply-To: Ravinderjit Singh <singh415@xxxxxxxxxx>
Date: Monday, August 22, 2022 at 11:21 PM
To: "AUDITORY@xxxxxxxxxxxxxxx" <AUDITORY@xxxxxxxxxxxxxxx>
Subject: Re: EEG purchase

Hello all,

Curious if anyone has any experience in comparing the OpenBCI products (Like the Mark IV) compared to the Biosemi Active Two system. Curious what the difference between the two products would be for cortical measures. Currently the OpenBCI products cannot sample as fast as Biosemi, but OpenBCI is fast enough for cortical measures. I prefer the price point and openness of OpenBCI 🙂

Best,

Rav

-----------------------------------------------------------------------------------

Ravinderjit Singh, PhD

MD/PhD Student

Purdue University - Weldon School of Biomedical EngMark IVineering

Indiana University - School of Medicine

From: AUDITORY - Research in Auditory Perception <AUDITORY@xxxxxxxxxxxxxxx> on behalf of Kaganovich, Natalya B <kaganovi@xxxxxxxxxx>
Sent: Friday, August 19, 2022 10:38 AM
To: AUDITORY@xxxxxxxxxxxxxxx <AUDITORY@xxxxxxxxxxxxxxx>
Subject: Re: [AUDITORY] EEG purchase

Hello Carlos,

I have been using the Biosemi Active Two system with children ages 7-14 or so and with adults for about 10 years. I love it. I only have 32 electrodes, which was an intentional choice. I find the application quite fast and easy – to speed up the process, we use 2 people to apply gel into electrode holders and to plug in electrodes. It takes about 12-15 minutes start to finish. I worked with Neuroscan during the post-doc. It may have changed by now, but the data quality with Biosemi is amazing compared to Neuroscan. Lloyd Smith and his team at CortechSolutions (the Biosemi distributor in the US) have been extremely supportive and respond promptly. Whenever we need to repair electrodes, we get them back within a week or so. Lloyd is extremely knowledgeable and was able to guide us through unexpected issues when the did arise.

I will mention a couple of things to consider:

The wires are rather delicate. I believe the later version of electrode sets have the connection between the wire and the electrode strengthened. We do have both graduate and undergraduate students using the electrodes but provide extensive training on handling. We only had a few breakages.
Biosemi does not have in-house data processing software. This may or may not be an issue for you – many labs now use Matlab-based programs for data analysis. I use EMSE, which is now also owned by CortechSolutions.

Good luck with your lab set up!

Natalya

Natalya Kaganovich, PhD

Associate Professor

Department of Speech, Language, and Hearing Sciences

Department of Psychological Sciences

Purdue University

kaganovi@xxxxxxxxxx

(765)494-4233

From: AUDITORY - Research in Auditory Perception <AUDITORY@xxxxxxxxxxxxxxx> On Behalf Of Carlos Benitez-Barrera
Sent: Friday, August 12, 2022 11:21 AM
To: AUDITORY@xxxxxxxxxxxxxxx
Subject: EEG purchase

---- External Email: Use caution with attachments, links, or sharing data ----

Dear Auditory List,

I am planning on purchasing an EEG system for my new lab at UW-Madison, USA. I have been in contact with several EEG providers, and I’m still undecisive on which system would be best for my lab. I’m familiar with Neuroscan and EGI, but I’m actually leaning towards Biosemi or Brain Products (Brain and Vision in the US). I run simple auditory paradigms with children (ages 3 to 14) and adults. The main characteristics that I’m looking for are (not necessarily in order):

Ease of use
Prep cap time (important to minimize with children)
Customer support
Signal quality

I’m shooting for a 64-channel system. Also, I’m still hesitating between saline caps or gel caps. I heard that the Biosemi caps despite of being gel based are very fast to get going.

Anyway, I’m just looking for some advice from any of you working with these systems. Any experiences or recommendations will help!

Thank you in advance for your taking your time to reply!

Sincerely,

Carlos