- 1 Sensors
- 2 Controller and data transmission
- 3 Data logging and visualization
- 4 Development Budget
We need to decide whether to build them ourselves or whether to buy them ready-made and just worry about assembling the bits and making everything communicate. Important considerations would be affordability, accessibility and required precision.
Why thermocouples (TC) vs. another type of sensor? It depends on your application! TCs can be very robust, for nasty environments and super high temperatures, but their voltage is very small so they may require an amplifier with digital out (http://www.sparkfun.com/products/307, $12 - see also http://www.chinwah-engineering.com/USBThermocoupleProject.html). Other types of sensors (like DTS) may not be as rugged, but in the end they may be sufficient for bioprojects if the sensor is protected, and the total cost may be less ($4). If we can hack thermocouples well enough to make them cheap, there are some advantages, it just depends on the application whether they are worth it!
Commercial thermocouple "probes" are available from approx. $20 upwards
And we can use just the thermocouple wire to create our own, buying it by the foot from places like McMaster Carr both wires in a sheath for ~$1/foot, by the foot or Omega (bare wire is here http://www.omega.com/ppt/pptsc.asp?ref=SPIR&Nav=temh02). Omega is the ultimate source, but they seem to only sell it by the roll (25 foot minimum, buy both wires separately) or in the form of super nice manufactured probes ($).
Charlie has access to a good amount of Type K metal sheathed TC wire, plus assorted probes and a TC reader he can donate - as we go along our improving expertise will lead us to resources other people can use... like the relatively cheap McMaster Carr wire.
There is also a somewhat sketchy Instructable for how to build one - at approx. $15, it’s not going to be much cheaper, though, so choosing the DIY version would be mostly for the educational benefit of actually building it (almost) from scratch. This soldering method will work with the Omega wire, and better junctions can be made with a welder or capacitive discharge.
Presently have a prototype thermometer which uses a one-wire device: DS18B20 digital temperature sensor, DTS - $4, also from Sparkfun - it's not talking to my Arduino yet, but it will. Oh, yes...it will. (Rikke)
The digital temperature sensor (DTS) is indeed a good one to start with - inc. because one is already mounted in a probe. We'll need to test that probe to see what its sensitivity and accuracy is (rough probes can have a problem with response times, but in many cases the temperature does not change quickly in bio systems), but having a sensor that does not need auxillary electronics (like the TC does) ready is a big plus. http://www.danielandrade.net/2008/07/05/temperature-sensor-arduino/, http://www.instructables.com/id/Waterproof-a-LM35-Temperature-Sensor/
The thermistor has some advantages over other sensor types in that it is inexpensive to get started with (<$2 for bare sensors, even $.25 - almost disposible), the electronics and communications are simple, and it is easily available in a wide variety of sensor/probe designs. Here is an easy source (http://www.hacktronics.com/Sensors/Thermistor-Temperature-Sensor/flypage.tpl.html) but note they tend to be rated only up to the boiling point of water - reasonable for a bio based measurement system, but no further. They also offer advice on monitoring one with an Arduino (http://www.hacktronics.com/Tutorials/arduino-thermistor-tutorial.html), including software and the same is generally true for the DTS - we are building on the experience of others. Thermistors and thermistor probes are available everywhere (I got the one for my old distiller from Radio Shack, but they may not offer them anymore), including Jameco and Digikey (so many it can be confusing. It seems like the 10kOhm version is the one to standardize on, but we can discuss. We can calibrate/troubleshoot it with the thermocouple reader we now have. Thermistors are also common for the temperature compensation of pH sensors.
What shall our specifications be? Range of temperature, resolution, accuracy, response time, precision (repeatability), cost, durability, calibration potential, etc. We have the same need for all of our sensors, requiring us to somewhat anticipate what users will do with our board.
- Biologically relevant temperature range is approx. 0-100°C; accuracy should not be less than ±0.5°C at 25-35°C. pH range is (1-14), and required precision is approx. ±0.5, preferably better. dO probe should be able to measure % conc. with an accuracy of approx. ±2%, preferably better.
- Concerning sensor validation: We have a pretty accurate digital body thermometer and commercial thermocouples for the thermometer. Should we choose to build a pH meter, we can easily get fairly accurate pH strips, and also borrow a 'proper' pH meter through Sean. Same applies for the dO probe, and getting a sample run through a spectrophotometer shouldn't be impossible either. No. of living cells in a suspension (Colony Forming Units pr. volume, CFU/mL) can be measured by plating a diluted sample on a petri dish and counting the no. of colonies formed, then calculating backwards to the original cell density. Can we get petri dishes, some pre-hopped DME and some agar from some of our brewing contacts? /Rikke
Here's a basic tutorial on how common probes work - http://www.sensorland.com/HowPage037.html and this article is titled "building the simplest possible pH meter" - http://www.66pacific.com/ph/simplest_ph.aspx - at least once you have found a sensor; as with the thermocouple, you need an amplifier because of the millivolt signal. This led me to his source for sensors (http://www.americanmarineusa.com/ - they supply to the aquarium industry so also sell conductivity and dissolved oxygen meters) and their cost of $50 is comparable to others we have seen. $30 for a similar one at another aquarium supply joint http://www.bulkreefsupply.com/store/brs-brand-ph-probe.html, with BNC connector (http://www.practicalmaker.com/tutorials/bnc-sensor-shield-documentation).
Maybe we can buy one of these instruments (or similar), break them open and figure out how to read the output from an arduino:
- HANNA Instruments HI 98103 $55
- Milwaukee pH600 $20 - doesn't look like it needs specific buffers for calibration, but the accuracy is probably not great. Maybe it's enough, though.
- Google shopping results approx. $40 upwards
- SOTA pH Electrode $100 - expensive, but so so sweet: designed for continuous measurement, and comes with any kind of connector.
You can also get little tester units, such as this Jenco 610 pH tester for $30 - perhaps it could be hacked?
There’s also the option of attempting to build one ourselves using this (or a similar) schematic with the Arduino instead of a voltmeter - not necessarily cheaper, although it’d certainly be both fun and informative. This site mentions the Ion Selective Field Effect Transistor (ISFET - see http://www.colorado.edu/MCEN/micronanobio/Homework/Homework_Nano-ScaleEngineering_3_2008_Solutions.pdf for more exotic applications) as a solid state alternative to traditional electrodes, but I can't find a source for the right/inexpensive one yet; they are said to be less accurate/sensitive, but that is something we are willing to consider. As soon as some one packages a sensor (e.g. http://www.microsens.ch/products/pdf/MSFET_datasheet%20.pdf) into a full blown "probe" the price goes way up.
Bilding from scratch a pH meter: http://damien.douxchamps.net/elec/ph_meter/, and googling on "DIY pH sensor" will find you more meter schematics. Many of them are referenced here at the pHduino project site: http://code.google.com/p/phduino/
Dissolved oxygen (DO) probes
- DO-BTA Dissolved Oxygen Sensor $209 - cheapest commercial product I could find. In this case, it seems relatively safe to assume that bulding one ourselves would be cheaper.
Potential DIY designs and progress
- Mod an automotive O2 sensor to make it a membrane electrode - New sensors for out of date cars are available on eBay for $10. Although these sensors typically operate at ~300C (won't work for us), they do have the required platinum, anodes, and teflon membrane. I'm thinking I can knock out the zirconium matrix and add a KCl electrolyte and see if we can get a reaction started at room temp (fingers crossed).
- Progress thus far: Ordered 3 $6-$10 probes on ebay to futz with
- Build an intensity or time based optode from scratch - Recently, people have been using a ruthenium complex as a visual (fluorescent) indicator of oxygen concentration. This complex is excited by a blue LED, then its transmission is measured by a filtered photoresistor (more details here in pdf). There could be some serious tecnical hurdles to overcome on this one, but if it works, this would be a way better sensor in the long run - no calibration needed, all solid state (super low maintenance). The Ru molecule is expensive (~$70/mg), but could probably be used for quite a few electrodes.
- A better optode might be based on Erythrosine B (FD&C Red No. 3) (details here: http://nathan.instras.com/documentDB/paper-429.pdf). A fairly complete list of dyes that would work is in this paper http://www.jbc.org/content/262/12/5476.full.pdf but erythrosine is probably the easiest/cheapest/most readily available of these. It even has better sensitivity (phosphorescence lifetime change with change in oxygen) and produces a stronger signal. The drawbacks are that it is somewhat non-linear and it does photodegrade (slowly). Erythrosine dispersed in any kind of oxygen permeable and optically clear medium would work, including in silicone or aerogel/xerogel. Clear silicone is probably easiest. Silicone caulk from the hardware store, preferably aquarium grade, may work (ie: would not react with the dye during curing). If not, then platinum cured silicone (LSR) such as Smooth On  will almost certainly work. A green LED (530nm) is ideal for inducing phosphorescence. The silicone/erythrosine sensor can be completely sealed in black silicone to prevent interference from ambient light, and allowing an unfiltered light sensor to be used. Exact probe dimensions are not critical: thin probes (1-2mm) would respond faster to changes in oxygen (< a minute), thicker probes (5-10mm, loaded with enough dye that they are almost opaque) would have a stronger signal (phosphorescence up to 0.5-1% of the brightness of the illuminating LED). (AI)
- Optode signal processing might have the highest sensitivity if it is based on phase detection: a sine-wave input signal at 10-100kHz (with dc offset) is fed to the illuminating LED, then the output from the photodiode is filtered to remove dc offset and high-frequency noise, and both the input and output are fed through zero-crossing detection (only from above to below zero). The duration of pulses during which the input signal is below zero but the processed output signal is above zero is a function of the phase angle, and directly related to the phosphorescence decay time. Lowpass filtering of those pulses will produce a DC voltage which directly corresponds to oxygen levels. This measurement should be very sensitive, and immune to most types of noise/interference. This does sound complicated, but it may be easier to get good results with this method than with gating+integration, because it relies a lot less on specific part tolerances. (AI)
- Currently lots of DIY spectroscopy projects under development
- Relatively easy build, can be made using a LED and an old cell phone CCD
- Can be used for chemical analysis as well
- Verification of results with known absorbance values should be easy
- Will likely need re-calibration for every use
- Could be very hard to pack into a probe
Calibrated capacitance + conductivity sensor
Industry has commercial probes available which measure living biomass; we think we may be able to retroengineer such a thing. With enough calibration, it might be possible to do this by measuring capacitance alone.
[The basic principle behind these probes is the different electrical properties of living and dead cells; both are conductive - being essential very long and folded chains of carbon molecules - but living cells also act as capacitors (batteries); active transport across the cell membrane of electrically charged ions/molecules establishes a negative potential/charge on the order of -70mV in resting mammalian neurons.]
More accurately: living cells have intact membranes, which can be polarized by an external electrical field; the negative potential of the membranes is correct but has no effect on the measurement. Biomass measurement relies on the difference in capacitance at different frequencies: low frequencies polarize cell membranes, high frequencies do not, the difference provides an estimate of the total surface area of membranes in the sample.
- .pdf with technical notes about a commercial OD probe
- Industrial application of NIR spectroscopy in fermentation and cell growth monitoring
- Cell phone spectrophotometer
- Article on how to build your own spectrophtometer
- DIY Spectrometer
- DIY Spectrometer FAQ - lots of useful links to other DIY spectro projects
Controller and data transmission
Is Arduino going to be our platform?
Here's a possible design: An arduino hooked up to a bunch of different probes and in turn hooked up to either an ether/wifi shield or a full linux box (I've previously used the excellent and tiny 1 watt biffer board in this manner). As data comes in it should be timestamped, categorized (pH, temperature, etc) and sent (via shield or pc) to a server somewhere on the internet using e.g. JSON over HTTP post. The server would run a custom web app (e.g. Rails) that receives data, logs it to a database and generates graphs on demand. Add a Comet server and the graphs could be live-updated as the data comes in. We could add features that lets new users sign up and get a unique key which they use when transmitting their own data to the JSON web service on our server. The server then uses the key to associate the data with the user, and the user can look at their graphs and share them with others. We can implement the "export to CSV" on the server side, allowing users to analyze the data using the tool of their choice. If we put all of the code on github, then others can easily fork the code and add their own features. To me, the hardest part of this project will be finding/building cheap measurement probes that are accurate enough (how accurate is enough?) and require little or no calibration. - (Juul)
Sensor data collection
Which board would be optimal for our purposes? Does this depend on the sensors?
Wireless data transmission
How? Immediate suggestion would be ethernet shield, but others options are available and perhaps preferable.
Data logging and visualization
Data should be recorded at regular intervals; ‘smoothing’ by logging average values over each interval rather than point values may also be an option. Export to a simple spread sheet arranged to make later data analysis and visualization easy - how to do this?
- I suggest that we use a simple data serialization format like JSON for logging the data, and then include an "export to CSV" function where you can select the data you want exported (pH, temperature, etc). This should allow people to use a variety of programming languages and data analysis tools without a lot of work on their part or ours - (Juul).
It would be awesome to have the visuals built in from the beginning, as good graphics will greatly increase people's understanding of the correlations between the factors we measure.
A virtual space for the collaboration, organization and publication of the project. So far, this wiki seems to work well for our collaboration and organization purposes, but it may not be the best platform for presenting, sharing and comparing data?
- I could do a quick Rails site with some live graphing and throw the code on github. We could keep it really simple for the first version, but encourage people to add functionality for collaboration. - (Juul)
- Ru catalyst for DO design - $190: Sean C
- Ethernet shield for arduino - $56.01: Sean C
- Mylar/Vinyl films - $4.55: Sean C
Amount Remaining: $749.44