Beta Testers Update
If you’re a beta tester I know you’ve been waiting forever, but believe me when I say we’re making great progress. We now have the capacity (human and machine, for details keep reading) to make the 80-odd required devices for the beta. We also have the quality specs necessary to make those units useful for you (calibration and comparison to known devices, for details keep reading). We also have some great software tools for analyzing and sharing data quickly, easily, and (dare I say) beautifully. Now we just have to crank them out. Stay patient, we’ll get there soon.
Detroit Maker Faire
PhotosynQ was on display at the Detroit Maker Faire! We took about a hundred measurements all around the faire, and had our very own tee shirts (see picture). We met all kinds of folks, including educators like Melissa from Tinker Tailor, the guys putting together a new makerspace in Lansing called Innovation 5, and we were located right next to Josh and the fabulous folks at All Hands Active where he had is Soil Cam 9000 on display – I absolutely love that thing. We are just itching to get our first real beta units out into the world to folks like this to see what we can discover together!
3D printer up and running!
We recently bought a SeeMeCNC Rostock Max V2, and it is awesome! It took some adjusting and Jake’s been working on it for about 3 weeks now, but I’m very pleased with the quality and relative simplicity. That beings aid, let me share some lessons with you:
- First things first – adjust the z thickness to it’s as big as possible but still sufficient for your project, and make sure the first layer is fairly thick to help it stick to the bed.
- Make sure you include the big circle of abs around your part first, so the abs has a chance to stick to the plate (we smear glue on the plate as suggested by SeeMeCNC to get it to stick, that works well).
- For anything more than the smallest, simplest parts, you ABSOLUTELY MUST enclose the machine to prevent heat loss. The bed will heat up to only about 90C, which just isn’t hot enough to prevent warping of parts. It’s not shown in the picture, but we’re building a plexiglass case to completely enclose the whole unit – this helps the bed get up to 105 – 110 AND it keeps the printed piece itself warm so that warping isn’t an issue. We struggled with this for a while. We tested it by covering it in cardboard, so if you don’t want to get plexiglass don’t worry, there are cheaper options:)
We’ll be using this to print the cases for the beta units. We’ve tested making all of the different case components and have successfully achieved the necessary tolerances, so now it’s off to the races!
Pick and Place machine up and running!
FINALLY. The pick and place saga has ended. We initially attempted to order a Madell Technologies DP2006-3 way back in DECEMBER OF 2013! It has room for up to 100 parts, separate machine vision on the pick and the place, runs at ~2000 parts per hour and is under 20k with all the bells and whistles. MSU purchasing took forever to approve the purchase, then once we got it over a month ago our great undergrads Chris and Jake have been pulling their hair out to get it installed and working. It hasn’t been easy. We’ve had numerous issues setting up the machine, some our fault and some their fault, and while it’s taken a long time to get this thing working I can’t say Madell hasn’t been responsive – in fact their customer service is amazing. However, the number of bugs and issues with the device has definitely required that support, especially as it relates to the software and computer vision system. For anyone thinking about buying this machine, two things: 1) just get everything preinstalled including computer and vacuum pump (unless your time is worth $1 per hour, just let them do it, trust me here), 2) lights, lights, lights, get lights everywhere, pretend your creating a photobox to take product photos, no shadows and super bright lights, the machine vision system is finicky and the most lights the better! and 3) expect at least 3 weeks for install and testing.
ANYWAY, it works and now we can produce at a minimum 10 boards per day, so that is no longer the limiting step in our production chain.
Calibration, calibration, calibration!
Note: For your calibration nerds out there, there’s no graphs below – sorry about that. We’re just now at the point that it’s worthwhile to start really graphing our calibrations and comparisons to commercial instruments. Expect to see that data in the near future.
One of the big reasons for delay in releasing the MultispeQs is the process of properly calibrating the outgoing units. Initially in our zeal to hit self-imposed deadlines, we sent about 15 units of varying quality stages of development to beta testers. While we did learned some important lessons, we felt they were very time consuming both for us and the users. So the next 15 units that go into the world are going to be calibrated to commercial equipment AND calibratable (ie users can recalibrate as needed). We’re focused on the two most demanded measurements – chlorophyll fluorescence (that includes calculating fv/fm, fs, fm’, Phi2, qE, qI, NPQ, and all the other related acronyms), and SPAD (a measure of greenness similar to NDVI originally created by Minolta).
Chlorophyll Fluorescence Phi2
In our initial comparisons with LiCOR 64 (a top of the line fluorometer costing about 50k), LiCOR would say .85 and we would say .74 or so… this was too large of an offset to be acceptable. We made some adjustments and added a new offset due to electronic noise and addressed some background IR light which was bumping our signal higher than it should have been. After that, we were within about .02 of the LiCOR… I wanted to improve more, until one day during testing the LiCOR itself decided to show values about .05 higher than it had the day before for no apparent reason. Jesse Traub, my partner in calibrating, couldn’t figure out the issue, but said after looking back at his data that this kind of variation wasn’t uncommon. After that, I decided that .02 is easily within the variation of any given instrument.
We’ve been measuring SPAD successfully for 8 months, but we have this persistent issues of showing a little more than double the ‘normal’ SPAD meter value (SPAD shows 45, we show 110, for example). The correlation with Minolta SPAD values is excellent, but the raw values are always about double. So last week, we tried to get to the bottom of this. First, what is SPAD? It measures absorbance through a leaf at two wavelengths: 650 (red) and 940 (infrared) (see here for Minolta meter and description and here for manual). To calculate SPAD, you have to take a blank value (no leaf) as a baseline. If we call the blank absorbance 650 and 940 and the leaf absorbance 650′ and 940′ respectively, then you calculate SPAD in the following way:
SPAD = Log ( 940′ / 940) / (650′ / 650 )
So, we tried getting these values using a normal spectrometer – surely that would give us Minolta-esque SPAD values… but no! It was consistent with our values (in the 130 range)… we tried many other things I won’t go into here (using Ln instead of Log, reducing lost light in the MultispeQ, etc etc) but kept getting the same answer. Then Sebastian looked into the original SPAD manual from Minolta to confirm how they are calculating SPAD, and here’s the description:
The values obtained in steps 1 and 2 (940′, 940, 650′, and 650) are processed to calculate the SPAD value, which corresponds to the amount of chlorophyll present in the sample leaf.
In other words, they don’t say!! Argh! So who knows what they do or how they calculate it, but it sure seems that for some reason they are multiplying the actual value by some factor to get it to land in the 20 – 60 range. So, we’ll do the same thing, and we’ll have great, consistent SPAD values!
Finally, there’s actually immensely more to the calibration thing than I’ve mentioned here. Once things have settled and we’re sure we’re not going to be changing things a lot more, I’ll document exactly what gets calibrated when and how in more detail.
Light Intensity PAR (Photosynthetically Active Radiation)
In order to measure Phi2 (photosynthetic efficiency), it’s important to have an accurate light sensor. PAR is only the portion of the spectrum which plants can absorb – about 400nm – 700nm. This is surprisingly tricky to measure! There are many very good and accurate light meters available, like the TCS light sensor available as a breakout board from Adafruit. However, their either measure too light (from 420 – 650) or too much in the case of the other TCS sensors (from 400 – 1200). So decided to go with the TCS 34715, which has a broad range from 400 – 1200nm, then adding a hot mirror to eliminate any light from 700 – 1200nm. This, combined with a tighter case design, resulted in a pretty accurate PAR sensor, which if you compare to LiCOR is currently within about 10% accuracy (LiCORs PAR sensor alone costs $500!). Users can recalibrate the sensors if they change over time.
Using the MultispeQ with the Chrome app
Our initial idea for using a MultispeQ fluoromoter was an Android app, so that people can use the device in the field. However, we quickly realized that we also needed a desktop solution so we began working on a Chrome app. Now, 76 commits later, we almost have a version, we feel comfortable with, so you can use it. The developer version is on Github (https://github.com/Photosynq/PhotosynQ-ChromeApp). We spent a lot of time on the UI, to make it as easy as possible to work with the app. It not only includes measurements, but also the tools to give you the ability to design your own measurement and programming little code snippets called macros to evaluate your measurements. Right now we are trying to finish the help section, the script generator and the missing communication pieces with the database. And of cause, we are trying to catch as many bugs as possible before the first release. Almost a 1000 test measurements have been done with the very early units and saved to our website, using the chrome and the android app. We’d like to thank everyone testing the software at this very early stage and helping us improving it. — Sebastian
Here’s some key features in the Chrome app (also see images below):
- Taking measurements for projects, including user specific answers.
- Selecting measurement protocols available from photosynq and run them.
- Using the console to communicate with the device if you’re getting an expert
- Manual control for instant measurements, recording environmental parameters with just one click.
- Drag ‘n drop tool to build measurement scripts
- Programming tool to build macros to evaluate measurements
- Data viewer to look at locally saved data.
- Measurement and image caching for offline work
- Project preview
- User sign in Notification log
- Location service Device information
- We’re redesigning the Android App to be more user friendly and faster, especially for breeders and those performing hundreds of measurements per day
- Expect to see a complete redesign of the website, and the merging of our database site and landing page. Finally, we’ll have a coherent face to present to the world – thanks Venturit!
- We’re hiring Geoff Rhodes to manage the assembly and shipping of MultispeQs to beta testers and collaborators. He’s also going to be working on calibrating devices and developing new methods. Expect to hear more from him in the future. Also, we’ve had Jake Nuremberg and Chris Zatske working on the project, and we’ve added all these fells to our collaborators page if you want details on their awesomeness.