open-source electrophysiology

Open Ephys and Neuralynx: a head-to-head comparison

Added on by Open Ephys.

In our first blog post, we described the fantastic signal quality of the Open Ephys acquisition system. Today, we made our first direct comparison between signals recorded with Open Ephys and those recorded with a Neuralynx Digital Lynx system. The conclusion? The recordings are virtually indistinguishable. There's still more rigorous testing to be done, but to the trained eye of a seasoned electrophysiologist, the signal quality of the two systems is qualitatively identical. Read on to see for yourself.

We performed the test by recording from a mouse implanted with an eight tetrode flexDrive. The Omnetics connector on the electrode interface board was compatible with both the Open Ephys headstage and a special-ordered Neuralynx HS-36. 

Open Ephys headstage (left) and Neuralynx headstage (right) used for the test. The Open Ephys headstage is based on the Intan RHD2132 amplifier chip. It measures 25 x 13.5 mm and weighs 1.0 g. The Neuralynx headstage is based on the Analog Devices AD8643 chips (we think). It measures 27 x 21 mm and weighs 2.3 g.

Open Ephys headstage (left) and Neuralynx headstage (right) used for the test. The Open Ephys headstage is based on the Intan RHD2132 amplifier chip. It measures 25 x 13.5 mm and weighs 1.0 g. The Neuralynx headstage is based on the Analog Devices AD8643 chips (we think). It measures 27 x 21 mm and weighs 2.3 g.

The mouse was placed in a 1' x 1' arena, and attached to either recording system via a tether that was counter-balanced via a system of pulleys.

The tether used by the Open Ephys system (available for purchase from Intan Technologies) weighs 8.2 g/m, contains 12 conductors, and is 2.9 mm in diameter. The tether used by Neuralynx weighs 10.1 g/m, has 44 conductors, and is approximately 3 mm in diameter.

The tether used by the Open Ephys system (available for purchase from Intan Technologies) weighs 8.2 g/m, contains 12 conductors, and is 2.9 mm in diameter. The tether used by Neuralynx weighs 10.1 g/m, has 44 conductors, and is approximately 3 mm in diameter.

We made sure the settings on each system were the same (1 to 7500 Hz bandpass, ~30 kHz sampling rate), then recorded for 3 minutes with Open Ephys and 3 minutes with Neuralynx

The Open Ephys acquisition board (shown here in a special-edition yellow case) measures 16 x 16 x 4 cm, weighs 0.3 kg, and costs approximately $30 per channel. The Neuralynx Digital Lynx SX system measures 45 x 27 x 30 cm, weighs quite a bit more than 0.3 kg, and costs approximately $1000 per channel.

The Open Ephys acquisition board (shown here in a special-edition yellow case) measures 16 x 16 x 4 cm, weighs 0.3 kg, and costs approximately $30 per channel. The Neuralynx Digital Lynx SX system measures 45 x 27 x 30 cm, weighs quite a bit more than 0.3 kg, and costs approximately $1000 per channel.

The examples below all come from a single electrode, but the signals across all 32 channels are identical between the two systems, as far as we can tell. We hope to carry out a more detailed comparison in the near future, but we couldn't be happier with the initial results!

Several seconds of raw data from each system. If we can spot any difference, it's that the Neuralynx recording has slightly more high-frequency noise (but we haven't quantified this yet).

Several seconds of raw data from each system. If we can spot any difference, it's that the Neuralynx recording has slightly more high-frequency noise (but we haven't quantified this yet).

Zooming in, it's clear that the signal quality is basically identical.

Zooming in, it's clear that the signal quality is basically identical.

Spikes extracted from the continuous signals also look the same.

Spikes extracted from the continuous signals also look the same.

Since this was a hippocampal electrode, we could directly compare ripple oscillations recorded by the two systems. Again, we can't see any difference.

Since this was a hippocampal electrode, we could directly compare ripple oscillations recorded by the two systems. Again, we can't see any difference.

May 2013 Newsletter

Added on by Open Ephys.

In the last newsletter, we described the features of our new acquisition system based on the RHD2132 amplifier chips from Intan. Since then, we assembled that system and started using it in our experiments. We also set up a blog to document our progress.

Testing the new hardware

We now have fully functional designs for headstages and acquisition boards based on Intan's Rhythm interface and digital SPI cable standards. With four headstages connected to one acquisition board, it's possible to stream 128 channels of neural data over a USB cable. We've been testing the system using our 16-tetrode flexDrive in behaving mice and the signal quality looks fantastic.

Images of our new headstages and acquisition board are now available on our website. We'll add details and pictures of the complete system shortly.

Manufacturing

Now that our designs have been validated, we're ready to start the manufacturing process. We'll be sending the first round of boards to the 24 labs that have signed up to be beta testers. In return, we hope that they'll help us add features to the software by creating new plugins for online analysis and by reporting and fixing issues with the software. If each lab contributes only one plugin, the capabilities of our software will increase tremendously.

If you're interested in trying out our system but didn't sign up for beta testing, we're working on an instruction manual that describes how to build it from scratch. We've already built two complete systems ourselves, and each took us the lesser part of an afternoon. And if all goes well with the beta boards, there's a good chance we'll manufacture another round of boards in a few months, so stay tuned for more information.

Open Ephys blog

For more frequent and more technical updates than we provide in our newsletters, you should check out the new Open Ephys blog. So far, we've written posts on using our system for tetrode recordings, making Intan-compatible fine wire tethers, and configuring our system for closed-loop control. We hope the blog will become a place where all users can share their unique applications.

As always, feel free to get in touch if you have additional questions or want to know how you can contribute.

Closed-loop feedback

Added on by Open Ephys.

One of the key advantages of the Open Ephys acquisition system is its prioritization of closed-loop feedback. Our software makes it easy to mix and match modules for acquiring data, detecting events, and sending triggers to external devices. The figure below depicts the configuration we've been using for our own experiments. In the middle, the histogram shows the amount of time elapsed between an event occurring and the feedback being delivered (measured for 1000 events). The delay is never more than 20 ms, and it's around 11 ms on average (white dotted line). This is short enough to facilitate a wide range of closed-loop experiments.

  1. Electrical potentials are generated by the brain and detected by an array of electrodes.
  2. Analog electrical signals are digitized by our headstage and sent to the acquisition board.
  3. The acquisition board packages samples into a buffer, synchronizes them with auxiliary digital and analog inputs, and sends them to a computer via USB.
  4. A dedicated module in our GUI unpacks the incoming data, converts it to floating-point values, and places it into a separate buffer for analysis.
  5. Another software module analyzes the incoming data stream and looks for specific types of events. These can be simple, first-order characteristics of a continuous signal, such as phase or amplitude, or more complex features that are matched to a template. Other modules could be created to analyze the attributes of spike activity, such as multiunit firing rate or decoded sequences. When the event is detected, this module generates a signal that's passed to all the modules farther down the signal chain.
  6. In the final step in the software processing, we have a module that communicates with a Pulse Pal, an open-source stimulator designed by Josh Sanders in the Kepecs Lab at CSHL. This module receives the signal sent by the event generator and triggers the Pulse Pal via USB.
  7. The Pulse Pal is pre-programmed to generate pulses of a precise frequency and duration. Any of its four output channels can be triggered independently via software.
  8. A Plexon PlexBright LED is activated by the Pulse Pal, generating a brief pulse of light that activates channelrhodopsin in the brain.

At every step of the way, it's possible to swap out the hardware or software modules for ones with different functionality. For example, someone proficient in C++ could write a module in Step 4 that communicates with National Instruments hardware. The module in Step 6 could be programmed to deliver feedback via an Arduino. Feedback can also come directly from the Open Ephys acquisition board. The hardware in Step 8 could be a laser, a current source, or a mechanical actuator. Right now, the range of available software modules is limited. As more labs start using our system, we hope they will help us extend its functionality by contributing modules with new capabilities.

Digital fine wire tether

Added on by Open Ephys.

For recordings in freely behaving mice, it is important to minimize the weight and torque applied by the cables. This is especially important for experiments that require natural behavior and becomes a real issue for channel counts over 32 where even light wire tethers become bulky.

The standardized interface cable for Intan RHD chips we use is ideal for this application. Thanks to the digital LVDS signal, only 12 conductors are needed for transmitting up to 64 channels of neural data. Cables that conform to this standard can be purchased from the Intan website in 3- or 6-foot lengths.

If you want something even more lightweight and flexible, it's possible to build your own cables. We did this by soldering wires to two 12-pin Omnetics PZN-12 polarized nano connectors. Here, we've used Cooner CZ 1187 wire, FEP Insulation 38AWG with 0.012" diameter and 0.720Ω/foot. This is the standard wire for analog tethers because it is very flexible and light, but also durable. The cables sold by Intan are 0.423Ω/ft for the LVDS and 0.172Ω/ft for ground and power, so we're at the upper end of the possible resistance values, but it seems possible that the 40AWG version of the wire could work for the LVDS pairs. For the GND and VCC traces using two 38AWG wires or going to a thicker wire with <0.2Ω/foot is recommended unless the tether is pretty short.

The wiring diagram of the cable is simple: There are two rows, each with 6 conductors. Each pair consists of a 'top' and 'bottom' conductor which must be wired straight to the same pair, except with the top and bottom cables switched at the opposite end. If one connector is laid facing the other back-to-back and one connector is upside-down, each pin needs to be connected with its opposite pin (see illustration below).

Each tether will need 12 conductors total: 5 LVDS pairs, plus power and ground. To begin, securely clamp one of the connectors so there's enough space to lay out wires in front of it. Measure out 12 equal lengths of wire.

Now for each LVDS pair, de-insulate ~1 mm on one end of a wire (sharp forceps work well for the cooner wire), tin the wire, and solder it to one pin on the connector. Use plenty of flux when dealing with small wires like these and solder quickly to avoid heating up the plastic body of the connector.  Attach a label to one of the wires in each pair indicating the number (1-6) and which wire is top or bottom.

For each LVDS pair, twist the wires (top and bottom) until they maintain contact even when the tether is bent. Don't simply twist both wires together so they they remain under tension - instead move one wire around the other, without twisting each of the individual wires. Otherwise, the tether will twist around itself later. The ground and power wires don't need to be twisted.

Once all wires on one connector are soldered, fix the tether to the table with standard lab tape or Kapton tape about 1 cm from the connector. This way you can gently pull on the wires to ensure they are the same length, without the risk of breaking the solder joints. Next, lay out the tether so that the free end with the labelled wire ends can be soldered to the 2nd connector.

Make sure the tether is straight and that all wires are tightly twisted with no open loops. Fix the free end to the table with another piece of tape, so that its easy to cut individual wires to the same length, and solder them to the 2nd connector. Make sure that all LVDS pairs remain well-twisted, and add a few more twists on the free end where needed.

After soldering, carefully connect the tether to the acquisition system and headstage and verify that it's working. Use a 64 channel headstage or a 2-to-1 adapter to test both of the 32 channel data lines.

Remove the flux from the connectors with ethanol, and secure the solder joints with a thin coat of epoxy. Tie the wires together at the connectors, and at regular intervals throughout the cable. Add some more epoxy to the knot at each connector and to the sides of the connector to make a solid connection that can withstand handling.

For added strength, it might be useful to add a thin string in parallel with the wires.

Et voilà, 64 channels of neural data on a tether even lighter than those used for conventional 16-channel analog recordings.

March 2013 Newsletter

Added on by Open Ephys.

Since the last newsletter, we've made significant progress updating our hardware to incorporate the new digital chips from Intan. The design for our new acquisition board can be viewed on our GitHub site, with a full bill of materials on Google Docs. Read on to find out about our plans for distributing these boards.

Open Ephys RHD2132-based acquisition system

If you've been following the latest updates from Intan Technologies, you know that they've recently released an evaluation system for their RHD2000-series chips. We worked with Intan to come up with some of the specifications for these boards, and will also adopt those standards for our own hardware. Both the Intan RHD2000 Evaluation System and our new acquisition systems will feature:

  • compatibility between acquisition boards, headstages, and software
  • 32-channel headstages with 36-pin Omnetics connectors (available from Intan for $745 each)
  • 4 headstage connectors (for a total of 128 channels of data acquisition when using 32-channel headstages)
  • 12-wire cables that can be daisy-changed up to 10 meters (available from Intan for $175 per 3-foot length)
  • 8 analog-to-digital converters (separate from the converters on the headstages)
  • 8 digital-to-analog converters (for real-time audio monitoring and signal generation)

Both the Intan evaluation boards and our acquisition boards are powered by the Opal Kelly XEM6010 FPGA development board. This means they can both use the same firmware (Intan's "Rhythm" USB/FPGA interface), and that the same FPGA can be used with either system.

If you're eager to test out the new Intan chips as soon as possible, we encourage you to order the RHD2000 evaluation system, which is available now from Intan. Depending on your needs, however, it might be worth waiting for our boards to be released. Our acquisition system will include some changes that make it more user-friendly for neuroscientists:

  • 5V logic on the digital input and output channels (vs. 3.3V for the Intan evaluation system)
  •  +/-5V input range for the analog-to-digital converters (vs. 0 to 3.3V)
  • +/-5V output range for the digital-to-analog converters (vs. +/- 3.3V)
  • HDMI connectors to interface with peripheral BNCs (vs. screw terminals)
  • an injection-molded case
  • tricolor indicator LEDs

These changes make our boards slightly more complex, but, in return, they will be more convenient for experimenters.

Plan for distribution

When new tools come along in neuroscience, it's typical that the first and last step in propagating them is to publish a methods paper. It's up to the readers to figure out how to put the tools together (or to hassle the original authors for more information). We want to do things a little differently. In addition to posting the design files and complete assembly instructions online prior to publication, we'd also like to manufacture a small number of boards and give them away to interested labs for free. In return, we hope that those labs will contribute to improving our open-source data acquisition software.

This is our first time manufacturing anything at this scale, so we expect there to be some hiccups along the way. Our current plan is to test a few of the new boards by the end of this month, then place an order with Screaming Circuits to produce a larger run. If all goes according to plan, we should have the boards ready by the end of April. But there's always a chance that things will be pushed back if our prototypes don't work as expected.

This effort will be supported by generous donations from a few principal investigators. We only have enough funds to purchase the populated circuit boards and cases, not complete systems. If you receive one of our acquisition boards, you'll still have to buy headstages and cables from Intan, and an FPGA from Opal Kelly (a minimum investment of $1319). Considering that 32-channel systems from commercial vendors start at around $20k, this is not a bad deal.

How to sign up

If you'd like to test out our acquisition boards in your lab, please fill out this request form.

We will have a limited number of boards to give away, so signing up doesn't guarantee you'll receive one. If we've already talked about sending you a system or two, your chances are much better. But please fill out the form anyway, so we can keep track of everyone who's interested.

Keep in mind that we're a team of graduate students—not professional engineers—so there's a chance that the first round of boards will be buggy or won't work at all. We also can't guarantee there will be any technical support (beyond the online documentation) if things go wrong. We do know that, at the very least, this effort will teach us useful lessons about the best ways to distribute open-source hardware. At best, though, it will be an important step toward making multichannel electrophysiology more flexible and more affordable.

 

January 2013 Newsletter

Added on by Open Ephys.

This is the third installment of the Open Ephys newsletter, which we plan to send out on a monthly basis. Since last time, we took a trip to Janelia Farm to discuss open-source standards for multichannel electrophysiology, updated our headstage designs for the new Intan chips, and fleshed out the documentation for both users and developers of the Open Ephys GUI.

Developing open-source standards

On December 11, Josh and Jakob traveled to Janelia Farm to talk with members of the Applied Physics and Instrumentation Group, other group leaders, and representatives from Blackrock Microsystems. Dr. Tim Harris initiated the meeting in order to reduce some of the redundancy in open-source tool development, a goal we could not be more enthusiastic about. We discussed the need to develop open-source standards for extracellular electrophysiology, which have been conspicuously absent until now. In communities such as the audio recording industry, standardized interfaces make it possible for hardware and software from different vendors to work together seamlessly.

In our discussions, we identified a few interfaces where standards would be especially useful to have:

  • Connections between implanted electrodes and a detachable headstage
  • Connector and cable type between headstage and acquisition board
  • Communication protocol between Intan chips and an FPGA-based acquisition board
  • Data transmission between an FPGA-based acquisition board and a computer
  • Writing data to disk, including file types for raw and pre-processed data

We've been working closely with Reid Harrison of Intan to develop standards for each of these interfaces. As a result, our headstages and acquisition boards will be fully compatible with the next generation of Intan evaluation hardware. Whether or not these catch on, we hope our efforts will draw attention to the importance of standards for moving our field forward.

New headstages

We finished the first revision of our updated headstage design in December. If you have access to Eagle PCB software, you can check out the design files on GitHub: https://github.com/open-ephys/headstage. By incorporating the new RHD2132 Intan chips (as opposed to the older RHA2132 chips), these headstages will feature:

  • 32 channels
  • on-chip analog-to-digital conversion
  • low-voltage differential signaling, which allows significantly longer cable lengths
  • reduced size and weight
  • 36-pin Omnetics connectors compatible with many current implants and probes (mates with Omnetics part #A79026-001)
  • a built-in 3-axis accelerometer (for detecting head movements)

The printed circuit boards just arrived, and we're excited to test them out as soon as possible. We ordered a bunch of extras, so if you're interested in building your own headstages (instead of purchasing them through Intan), we might be able to send some bare boards your way.

Updated documentation

In the last month, we added a page to the GitHub wiki to orient new users to our cross-platform data acquisition package: https://github.com/open-ephys/GUI/wiki/User-documentation. If you've been using the Open Ephys GUI for data acquisition and any aspects of the software have been confusing, please let us know. We will gladly use your input to revise the documentation (which is still very much a work in progress).

We also posted some much-needed developer documentation: http://htmlpreview.github.com/?https://raw.github.com/open-ephys/GUI/master/Source/Docs/html/classes.html. We're using Doxygen to extract the documentation directly from the source code, which makes it trivial to keep everything up to date. As of last week, all of the major classes and methods are described. If you plan on changing the software yourself, we highly recommend browsing through these pages.

If you have any questions, or would like to get involved in our efforts, please get in touch with us through the contact page (http://open-ephys.org/contact) or by replying to this email.

December 2012 Newsletter

Added on by Open Ephys.

This is the second installment of the Open Ephys newsletter, which we plan to send out on a monthly basis. We reached an important milestone in November by carrying out the first neural recordings with our custom acquisition system. We've also made progress on updating our hardware and software.

Neural recordings

Over the past few weeks, we've been using the Open Ephys acquisition system to record data from awake, behaving mice. Previously, we'd only tested our acquisition board and headstages with signal generators and pre-recorded data. While our measurements indicated that the data quality would be excellent, we were obviously eager to see how it performed with actual neural signals. We're happy to report that everything looks great!

We implanted an array of low-impedance tungsten electrodes in mouse hippocampus and recorded with a skull screw reference. We see beautiful theta, gamma, and ripple oscillations, with barely detectable 60 Hz line noise. Since we're digitizing the signals directly on the headstage, movement artifacts are greatly attenuated. Saturating events are very infrequent, even in an animal actively running on a track. The next step is to implant an array of independently movable tetrodes. We'll keep you posted on our progress.

Upcoming hardware revisions

We're on schedule to have our hardware updated by the time the new Intan chips are available in early 2013. As described in the last email, the new chips will feature an onboard analog-to-digital converter, low-voltage differential signaling, and a number of other improvements. The new headstages will be a major upgrade in terms of size, weight, and performance. We already redesigned the circuit board, which now measures a mere 21 x 13 mm. We also added a 3-axis accelerometer to use as an alternate indicator of animal activity.

Another thing we're changing is the connector. Our current headstage design uses 0.4 mm pitch Molex connectors because of their compact size and low price point. Although these connectors can yield stable recordings in awake animals (as described above), they require additional reinforcement. We're also concerned about their long-term durability. Since Omnetics is the standard for the field (and many people have already requested Omnetics adapters for our hardware), our new headstages will use their 34-pin connectors.

Software improvements

The software interface we use for experimental control, visualization, and recording has received some significant upgrades. In order to collect real data, we made the recording capabilities more robust and updated the file format. We also added modules for real-time event detection, TTL-triggered recording, and digital referencing, all of which can be used in combination with data from any input source. We're making progress on fleshing out the documentation, with new wiki entries on data format (https://github.com/open-ephys/GUI/wiki/Data-format) and creating custom data-processing modules (https://github.com/open-ephys/GUI/wiki/Custom-processors).

That said, there are still a variety of ways in which the software could be made more reliable and user friendly. If you or anyone you know is interested in helping with C++ software development, please get in touch with us through our contact page (http://open-ephys.org/contact/) or by replying to this email. We'll make sure your message reaches the most appropriate recipient.

November 2012 Newsletter

Added on by Open Ephys.

Open Ephys is still in its infancy, but we've already made great strides toward our first major goal: a complete hardware and software solution for recording, visualization, and closed-loop feedback. At the recent Society for Neuroscience conference in New Orleans, we showed off a prototype system that could acquire 128 channels of data and respond to neural events in 20 milliseconds or less. Given that the plan for this system was hatched a little more than a year ago, we're optimistic that the production of a more polished system will happen in a matter of months, rather than years.

To learn more about the tools we're developing, check out our website (http://open-ephys.org) or download the poster we presented at SfN (bit.ly/RpxRim).

The road ahead

What's not currently on our website is our plan for moving forward, which became much clearer at the conference. As you may know, our hardware incorporates amplifier chips from Intan Technologies (http://www.intantech.com), which have allowed us to dramatically simplify and miniaturize our recording system. According to Reid Harrison, the founder of Intan, an updated version of the amplifier chips will be produced in December. The new chips will feature an analog-to-digital converter, low-voltage differential signaling, and a number of other improvements. If these chips pass the necessary tests, we will start using them in our system. We are already updating our designs so we can accommodate the new chips as soon as they're ready.

Besides revising our hardware, we also have clear-cut plans for manufacturing and distributing our recording systems. If our updated designs are sound, we will have our headstages and compatible cables available for purchase from the Intan website. Sometime in early 2013, we will mass-produce a round of about 50 acquisition boards and cases, which we will send to interested labs. The total cost of a 64-channel system will be around $1500, and there will be minimal assembly required. Of course, since the design is completely open-source, you could build your own system right away using the designs on GitHub (https://github.com/open-ephys/). But we recommend waiting for the next round of revisions to become available, as there are still a few updates we want to add.

How you can help

While our plans for improving the Open Ephys hardware are mainly settled, there's a lot of work to be done to make our software as flexible and user-friendly as possible. In parallel with our hardware, we've been developing a cross-platform application (called "the GUI") for experimental control (https://github.com/open-ephys/GUI). The major strength of the GUI is its modularity, which makes it easy to modify and extend. The software was designed from the start to be compatible with almost any input source, so it can also be used with hardware other than the Open Ephys acquisition system. It currently includes basic modules for filtering, spike detection, audio monitoring, recording, and LFP and spike visualization. But its functionality is very stripped down, and it still has more than a few bugs.

If you or anyone you know is interested in volunteering their time to improve our software, please get in touch with us. Possible projects range from the simple (adding a buffer to the audio monitor) to the complex (creating an OpenGL plotting library). Knowledge of C++ is required, but if you already have basic skills in another language, it's not very difficult to learn!

We'd also be thrilled to have full-time programmers helping with software development. If anyone has funding available for such a position, please let us know. We realize this is a major investment, but we feel strongly that easy-to-use, open-source data acquisition software would be an invaluable asset to the entire community. In the scheme of things, the cost of hiring a programmer for a year is much less than the price of a new commercial ephys system. Our software reached its current state through the efforts of two graduate students working in their spare time for one year. A more concerted effort to polish the software could generate substantial returns in even less time.

Channels of communication

We plan on using this mailing list to deliver updates on a monthly basis. If that's too much for you, feel free to unsubscribe using the link at the bottom of this message. But if you're craving up-to-the-minute information on our progress, we plan on using GitHub as our main channel of communication. Once you create a GitHub account (a very easy process), you'll be able to:

- subscribe to feeds that broadcast updates to all our repositories
- edit the wiki pages for each repository, which will eventually evolve into comprehensive online documentation (most likely hosted somewhere other than GitHub)
- use the "issues" system as a basic forum for solving problems and getting feedback

It's also fine to get in touch with the Open Ephys team directly through our contact page (http://open-ephys.org/contact/) or by replying to this email. We'll make sure your message reaches the most appropriate recipient.