Open-Source Spectrometer
Nathan McCorkle
I'm trying to lock down the requirements for an open-source
spectrometer, which I and colleagues will likely design in the coming
months if we can gauge that there is a decent market to make our
engineering time back with a few sales. At first glance we've
estimated that it will cost between $300-$700, . The price could be
flexible depending on customers desired configuration (e.g.
ethernet/no-ethernet, high-res version/low-res version, plastic
case/metal case, etc... basically anything that could be omitted from
the final design if it would lower cost and was desired)
Our general specs:
fiber-coupled
2048 or 3648 pixel array detector
USB and/or ethernet for data transfer
What we'd like to hear from you:
If you were going to spend $500 on a spectrometer,what would you want
it to do/have
Would kits be an option, or only assembled units? (or wired but not in
a box, for robot hobbyists looking to build a fire-fighting bot that
seeks out spectral signatures of burning materials!)
How often would you want a reading? e.g.... Would it be really fast
(1-10 times per second) for looking at fast chemical reactions? Or
more like one-shot/once-in-a-while, for things with a stable spectrum
(getting DNA/RNA concentrations, cell culture density, etc)
What kind of data connection would you want? USB and ethernet come to
my mind, but lots of connections are possible. The data link could be
a bottle neck if you want to stream readings quickly, so this ties in
with the previous question.
Let me know what you think.
-Nathan
--
Nathan McCorkle
Rochester Institute of Technology
College of Science, Biotechnology/Bioinformatics
Inventoriffic
Being able to handle elisa's would be a huge plus (i.e. microplate
reader) - do you think this would be possible?
USB would be fine for my usage. Please, please, make it arduino based
- it would be great to be able to construct it ourselves, and also
replace parts when necessary. That way people could create different
code for different uses, and a community could develop around it.
Great news!
Jay Woods
> Great! I'd love to buy a kit.
>
> Being able to handle elisa's would be a huge plus (i.e. microplate
> reader) - do you think this would be possible?
My usage would be for geological and biological thin sections preferably
under a petrographic microscope and reflection spectra from
polished/cleaved sections.
>
> USB would be fine for my usage. Please, please, make it arduino based
> - it would be great to be able to construct it ourselves, and also
> replace parts when necessary. That way people could create different
> code for different uses, and a community could develop around it.
+1 on arduino. The arduino can handle USB, bluetooth, or IP.
Marc Juul
> If you were going to spend $500 on a spectrometer,what would you want
> it to do/have
An included reference test-kit for calibration purposes would be nice.
Auto-calibration would be really nice.
I realize this is asking a lot, but it would be really useful if you
can figure out how to measure side-scatter, maybe by having the
ability to switch (manually rotating a mirror?) between having the
sensor measure the side-scattered and forward-scattered light. That
would give us some ability to filter based on particle size.
> Would kits be an option, or only assembled units? (or wired but not in
> a box, for robot hobbyists looking to build a fire-fighting bot that
> seeks out spectral signatures of burning materials!)
>
> How often would you want a reading? e.g.... Would it be really fast
> (1-10 times per second) for looking at fast chemical reactions? Or
> more like one-shot/once-in-a-while, for things with a stable spectrum
> (getting DNA/RNA concentrations, cell culture density, etc)
I would definitely prefer stuff like filtering, calibration and
accurate results over speed. If the device could read a 96-well plate,
then speed could become important since e.g. cell cultures would keep
changing which might result in a difference between first and last
well, but there are ways of dealing with that.
> What kind of data connection would you want?
USB would be fine. A Linux API would be a must. A compromise could be
to read quickly and buffer the data, sending it slowly to the
computer. I think this would work out well for anything that doesn't
require real-time availability of the results. Buffering might make it
difficult to integrate this with other tools like robots /
microfluidics, etc.
--
Marc Juul
Cathal Garvey
--
You received this message because you are subscribed to the Google Groups "DIYbio" group.
To post to this group, send email to diy...@googlegroups.com.
To unsubscribe from this group, send email to diybio+un...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/diybio?hl=en.
--
letters.cunningprojects.com
twitter.com/onetruecathal
http://www.indiebiotech.com
General Oya
I'm sure our local hackerspaces would be quick to put one together should you go w an arduino kit... I'm taking an instrumentation class this semester and would love to build some of these basic tools.
Ryan
Nathan McCorkle
diffraction (diffraction grating)
> --
> You received this message because you are subscribed to the Google Groups
> "DIYbio" group.
> To post to this group, send email to diy...@googlegroups.com.
> To unsubscribe from this group, send email to
> diybio+un...@googlegroups.com.
> For more options, visit this group at
> http://groups.google.com/group/diybio?hl=en.
>
--
Cathal Garvey
I am struck by the possibility of making a 3Dprinted or lasercut LED/Cuvette/CD holding case with a manual slider to move one part to select which band of light passes through the cuvette. With a photoresistor along the lightpath and at right angles to the cuvette, that'd give you a fair bit of spectro function, right?
The most expensive part, as always it seems, is the light source..
On 15 Feb 2011 08:01, "Nathan McCorkle" <nmz...@gmail.com> wrote:
The CDs just act like a grating, to spread the light out by
diffraction (diffraction grating)
On Tue, Feb 15, 2011 at 1:06 AM, General Oya <gener...@gmail.com> wrote:
> I understand UV/Vis LE...
--
Nathan McCorkle
Rochester Institute of Technology
College of Science, Biotechnology/Bioinformati...
You received this message because you are subscribed to the Google Groups "DIYbio" group.
To post to...
General Oya
http://www.instructables.com/id/A-simple-DIY-spectrophotometer/
http://www.creative-technology.net/MAKE.html
http://openwetware.org/wiki/Citizen_Science/Open_Spectrophotometer_Project/Application
http://blog.makezine.com/archive/2008/11/safety-spectrometer.html
Intrigued to see what we'll come up with. I'm all about compactness in the form of lab mobility and deployment.
Ryan
--
You received this message because you are subscribed to the Google Groups "DIYbio" group.
Inventoriffic
JonathanCline
> What kind of data connection would you want? USB and ethernet
regarding comm protocol is made based on bandwidth requirement first
(i.e. if you want to take many fast readings then it may be too much
data for a USB-serial comm protocol although that is one of the
easiest to interface).
There should be a couple GPIO pins reserved for interrupt output.
This would allow easier hardware chaining of other devices. For
example, a pin which pulses when any reading is done. This could be
tied to a separate controller for asynchronous query thru USB or
ethernet or etc. The thermocycler should behave the same way (i.e. a
standard set of pins to indicate device state). This would allow
interesting effects, such as the standard biologist wish list of "I
want it to send SMS to my cell phone when it's done" -- which is easy
with another generic hack, if the output signal is available from the
thermocycler or [insert any other piece of equipment here]. Think in
terms of hardware interoperability.
## Jonathan Cline
## jcl...@ieee.org
## Mobile: +1-805-617-0223
########################
Matt DiLeo
1) do whatever a NanoDrop does (to measure DNA/RNA purity and
concentration
2) act as a microtiter plate reader for ELISAs and to track cell
growth over time (turbidity, GFP, etc)
For both of these, some thought will be need to put into what cuvettes/
plates/tubes people are using so that there's compatibility
Nathan McCorkle
I'm making progress on optics design, with several Professors
interested and giving me their help, as much as their schedule allows.
I am insanely busy with classes, so this progress is slow. I am also
waiting to hear back from summer research programs, if I don't get
into one within the next month, I'll put the project up on
kickstarter, and if it gets funded that's all I'll work on over the
summer months.
As of now there are professors helping me that work in an optics lab,
are in chemistry and trying to develop cheap field assays that could
be helped with spectral data (who also has a background in laser
spectroscopy), and another professor that is in chemical engineering
but has built telescopes and is now building an IR spectrometer.
-Nathan
> --
> You received this message because you are subscribed to the Google Groups "DIYbio" group.
> To post to this group, send email to diy...@googlegroups.com.
> To unsubscribe from this group, send email to diybio+un...@googlegroups.com.
> For more options, visit this group at http://groups.google.com/group/diybio?hl=en.
>
>
--
Nathan McCorkle
Ok, this is a multi-list reply.
On Fri, Feb 11, 2011 at 8:56 AM, Jay Woods <wood...@cox.net> wrote:>
On Friday, February 11, 2011 06:45:04 am Inventoriffic wrote:>> Great!
I'd love to buy a kit.>>>> Being able to handle elisa's would be a
huge plus (i.e. microplate>> reader) - do you think this would be
possible?>
That would be great, but I think that would be a separate project,
Iwould just need to add some sort of interface to this unit so the
twocould talk. This might be via SPI bus.
> My usage would be for geological and biological thin sections preferably> under a petrographic microscope and reflection spectra from> polished/cleaved sections.>>
Ok so would you want a light source to be included in or with theunit,
if so the price will go up if its more than LED based (I.E. areal
lamp). If a lamp, what sort of bulbs are best, cheapest,
mostversatile/common? A multi-point LED calibration may actually be
OK,except you wouldn't get the UV without expensive UV LEDs (though
thespread of wavelengths should be inferable from the
opticalcalculations, so maybe LEDs would be ok, unless the LEDs
luminanceband is not tight enough )
>> USB would be fine for my usage. Please, please, make it arduino based>> - it would be great to be able to construct it ourselves, and also>> replace parts when necessary. That way people could create different>> code for different uses, and a community could develop around it.>> +1 on arduino. The arduino can handle USB, bluetooth, or IP.>>>> Great news!>>
It won't be an arduino, sorry, its just too slow for this.
On Fri, Feb 11, 2011 at 12:17 PM, Keith Mc <ac...@provide.net> wrote:>
Topic: Open-Source Spectrometer> Nathan McCorkle <nmz787@[...]> Feb 11
01:34AM -0500 wrote:>> I'm trying to lock down the requirements for an
open-source>> spectrometer, which I and colleagues will likely design
in the coming>> months if we can gauge that there is a decent market
[present]>> I am seeking a CHEAP, simple spectrometer to evaluate LED
emitted spectrums,> including non-visible light ranges. I am
especially interested in seeing the> intensity and freq distribution
of inexpensive IR LEDs.>> Freq Range: UV through IR. I am very
interested in IR wavelengths, to> evaluate IR emitters for robotic
projects.>
you mean near-IR right, say greater than 900nm wavelength, lets
referto it as NIR from now on in this conversation, just to avoid
confusionthat we're building an actual infrared spectrometer.
> Reading rate: relatively low, once for ever several seconds is fine.>> Data Connection: Anything that gets the result into a computer is fine.> Data rate is not as important as getting the spectrum curve.> Several seconds per transfer is sufficient...>> $500 is a bit high for me, as this is more hobby than for profit use.> Does anyone have a CHEAP way to do this task?>
how cheap? here's cheap and
quick:http://sci-toys.com/scitoys/scitoys/light/spectrograph/spectrograph.html
> (Note: I am on Digest here, so posted replies through the group take time to reach me...)>> Thanks!> - Keith Mc.>
On Fri, Feb 11, 2011 at 2:58 PM, r2k-in-the-vortex
<rkor...@gmail.com> wrote:> Hi>> i could use a proper spectrometer
for white led spectral distrobution> measurements(both rgb ones and uv
led and phosphoros based ones)> so the working range should cover
about 850nm to some 350nm(wishful> thinking)
ok thanks, I have to look into a different CCD as they have one with
aspecial window for better clarity below 400nm. Since they have
twotypes of CCDs that are electrically the same, below 400nm could be
anoption.
> 500-700$ would be an acceptable price(cheap compared to commercial> ones)>> i think data rates will not be a significant bottleneck, easiest> implementation would probably be serial link with an ftdi 232 chip -> results in an simple and cheap communications over USB> ethernet would be significantly more difficult from the programming> point of view>> sample rate of less than 1Sps, would be acceptable but i think its> simple enough to achieve higher sample rates, it would certainly be> benificial>> Br> Rainer>>
On Fri, Feb 11, 2011 at 4:51 PM, Marc Juul <mar...@gmail.com> wrote:>
On Thu, Feb 10, 2011 at 10:34 PM, Nathan McCorkle <nmz...@gmail.com>
wrote:>> If you were going to spend $500 on a spectrometer,what would
you want>> it to do/have>> An included reference test-kit for
calibration purposes would be nice.> Auto-calibration would be really
nice.
yes, very good idea, I mentioned this earlier in regards to types of bulbs, etc.
> I realize this is asking a lot, but it would be really useful if you> can figure out how to measure side-scatter, maybe by having the> ability to switch (manually rotating a mirror?) between having the> sensor measure the side-scattered and forward-scattered light. That> would give us some ability to filter based on particle size.
Ok, if you don't need it to be fast, you could use one CCD and have
anoptical multiplexer, or for faster/simultaneous operation you
wouldneed 3 spectrometers that were synchronized.
>>> Would kits be an option, or only assembled units? (or wired but not in>> a box, for robot hobbyists looking to build a fire-fighting bot that>> seeks out spectral signatures of burning materials!)>>>> How often would you want a reading? e.g.... Would it be really fast>> (1-10 times per second) for looking at fast chemical reactions? Or>> more like one-shot/once-in-a-while, for things with a stable spectrum>> (getting DNA/RNA concentrations, cell culture density, etc)>> I would definitely prefer stuff like filtering, calibration and> accurate results over speed. If the device could read a 96-well plate,> then speed could become important since e.g. cell cultures would keep> changing which might result in a difference between first and last> well, but there are ways of dealing with that.
Ok, thats mostly software, unless you meant optical filtering.
For plates if you're only doing growth curves, using a CCD is
probablya bit too much, it would probably be better to use a matrix
ofphotodiodes under the plate, and a matching matrix of LEDs above
it.96 wells could be multiplexed into a reasonable number of I/O ports
ona controller, but its a bit different project in that sense. If
youwant spectra from a plate, with a spectrometer you need some sort
ofoptical multiplexer, which could be an XY positioner for the plate
orinput fiber.
>>> What kind of data connection would you want?>> USB would be fine. A Linux API would be a must. A compromise could be> to read quickly and buffer the data, sending it slowly to the> computer. I think this would work out well for anything that doesn't> require real-time availability of the results. Buffering might make it> difficult to integrate this with other tools like robots /> microfluidics, etc.>> --> Marc Juul
> --
> You received this message because you are subscribed to the Google Groups "DIYbio" group.
> To post to this group, send email to diy...@googlegroups.com.
> To unsubscribe from this group, send email to diybio+un...@googlegroups.com.
> For more options, visit this group at http://groups.google.com/group/diybio?hl=en.
>
>
--
Nathan McCorkle
>
>> It won't be an arduino, sorry, its just too slow for this.
>
Not a DUE, but it would be tough to jam 7KB of spectral data into the
tiny amount of ram the arduino has. Not sure how good the Arduino's
PWM is, but the CCD needs several lines to be clocked, with the master
being 1-2mhz, and the data 1/4 the speed of the master.
>
> --
> You received this message because you are subscribed to the Google Groups "Open Manufacturing" group.
> To post to this group, send email to openmanu...@googlegroups.com.
> To unsubscribe from this group, send email to openmanufactur...@googlegroups.com.
> For more options, visit this group at http://groups.google.com/group/openmanufacturing?hl=en.
Nathan McCorkle
>
> On Oct 21, 2011, at 12:05 AM, Nathan McCorkle wrote:
>
>> On Thu, Oct 20, 2011 at 3:12 PM, Windell H. Oskay <win...@oskay.net> wrote:
>>>
>>>> It won't be an arduino, sorry, its just too slow for this.
>>>
>>> Really? Too slow for the Arduino DUE, even?
>>
>> Not a DUE, but it would be tough to jam 7KB of spectral data into the
>> tiny amount of ram the arduino has. Not sure how good the Arduino's
>> PWM is, but the CCD needs several lines to be clocked, with the master
>> being 1-2mhz, and the data 1/4 the speed of the master.
>
>
Gonna send it to the list, to try and avoid confusing others.
>
> Not sure how you're reaching these conclusions.
> 50 k of SRAM in that one, so 7 k is not going to be a challenge, nor is 1-2 MHz data transfer over SPI (or variants), when running at 96 MHz.
Hmm? Sorry, I meant a logical NOT to your "Too slow for the Arduino
DUE, even?". I mean that the DUE should be fast enough, but not the
1st gen Arduino.
>
> -Windell
Cathal Garvey
standard at the moment? That and/or the Mega.
If an arduino *can* drive the spec, I'd be strongly in favour of using
Arduino. Not because of fanboyism, but because it's the most widely used
open source hardware chip, and there's already a huge userbase who know
how to code/recode arduinos. That, and it's the same platform used by
OpenPCR, meaning that the odds of someone creating a "master system"
that drives all sorts of sub-systems in a biotech lab are far higher.
Unless there are other reasons to avoid Arduino besides those
mentioned/discussed?
--
www.indiebiotech.com
twitter.com/onetruecathal
joindiaspora.com/u/cathalgarvey
PGP Public Key: http://bit.ly/CathalGKey
Lisa Thalheim
I'd put down the $500 for a spectrometer if its design was
hackable/extendable, meaning: modular and well documented. Personally,
I'd want a kit that makes tinkering easy.
Something I'd like to see in a spectrometer would be wavescans over a
range of, let's say, 200nm-1000nm, ideally even beyond that. Being
able to track reactions would be nice, but not a requirement.
Something else that's not on your list of requirements, but that i'd
like to add is: portability and low power consumption. I do have a
very decent spectrometer already, but a device that i could take on
hiking trips in remote areas would definitely be a selling point.
Again, this might come under the heading of "hackable", but designing
a device that can run off a 12V or 9V power source (which the solar
panel/battery combo I own outputs) would be rad.
USB would be an option, but with the transmitted data we're looking
at, I'd be even happier with an option of something simpler, like I2C
or a simple analog protocol that could be recorded/evaluated by an
Arduino or another small-scale development platform.
> --
> You received this message because you are subscribed to the Google Groups "DIYbio" group.
> To post to this group, send email to diy...@googlegroups.com.
> To unsubscribe from this group, send email to diybio+un...@googlegroups.com.
> For more options, visit this group at http://groups.google.com/group/diybio?hl=en.
>
>
Nathan McCorkle
beagleBone as a reference design for the open spectrometer, and
related future equipments (3D printer, laser/CNC controller,
integrated microfluidics Lab-On-Chips).
Pros:
500Mhz, lots of power so your hacked code doesn't have to be super-optimized
Runs debian-strain Angstrom linux, comes with gcc and python, etc...
familiar and easy to use development from GPIO signalling to web
servers
SPI for communication to other chips, USB to use cheap Wifi/Bluetooth
dongles via common linux drivers, etc...
Tons of GPIO, even crappy ADC, hardware PWM and timers
Cons:
Umm, we gotta design it, more complex since it requires external RAM, USB hub
Might not be able to get DIP or QFP packages, e.g. the processor would
likely be BGA... would be less kit-like on the soldering end of things
Ideas, comments? (Please)
John Griessen
> Cons:
> Umm, we gotta design it, more complex since it requires external RAM, USB hub
> Might not be able to get DIP or QFP packages, e.g. the processor would
> likely be BGA... would be less kit-like on the soldering end of things
External RAM is a lot of layout and design work to do since it clogs up power layout.
Some suggestions for processors easily able to handle your CCD timing.
==========
The MBED quick development board costs $50.80 qty one from Mouser...
96MHz, with 512KB FLASH, 64KB RAM, Ethernet, USB (host, device, and to-go), CAN bus, multiple serial, I2C and SPI buses, 12-bit
A/D and even a 10-bit D/A converter and real-time clock/calendar. Also on top is the USB connector (mini-B), some power regulation
circuitry (operating on 4.5 to 9 volts DC, or USB power), several indicator LEDs, and the reset button
develop code with
a closed source IDE made by ARM (hackaday article says the programming editor is odd,
and it's all website based, so could be slowww...), or use python-on-a-chip which is free-open
and sometimes needs some C code below the python, but it is the oldest platform of
python-on-a-chip, so has the most features in code already. Program in a python subset mostly.
=========
http://www.futurlec.com/ET-STM32_Stamp.shtml
Includes ST Microelectronics STM32F103 Microcontroller with 512kb Internal Flash Program Memory
32-Bit ARM Cortex M3 Processor Bit running at 72MHz or 90 MIPS
Direct In-Circuit Program Download with RS-232 Connection
Up to 48 I/O points
16 Channels 12-Bit A/D Converters
2 Channels 12-Bit D/A Converters
1 CAN Channel
5 USARTS and 2 I2C Ports, together with SPI
Two 32-Bit Timers with Four Capture and Compare Channels
On-Board Real-Time Clock with External Back-Up Battery Connection
And free-open tools!
http://code.google.com/p/python-on-a-chip/source/browse/src/platform/stm32/README?r=pymite-09
90 MIPs, has flat pack leads -- easy to solder...
======================
Why would you want 5X the speed for a little lab instrument?
Why can't you clock a CCD with only a 90 MIPS processor?
For that matter, why not a 16MHz arduino compatible for the ease of coding and free-openness
of the arduino IDE? http://www.pjrc.com/teensy/ shows ready-made boards with plenty of I/O
and you could also try python-on-a-chip on that one.
Another thing... if getting serial data clocked out of and into a CCD is trouble for you
and you think you need a complex 500 MHz system to do it, you don't. You can use 2
microcontrollers on one board -- 2 that are easy to solder. One handling
serdes (serialization-deserialization) for the CCD image sensor, and the other doing the usual stuff.
The two micros could be Teensy 2.0 Teensy++ 2.0 $16 $24 in qty one (complete devel boards with USB)
Teensy uses a similar Atmel AVR processor as the Arduino board, but there are differences. The most important is direct, built-in
USB on the Teensy compared to serial converted to USB by a separate chip on the Arduino. Teensy communications MUCH faster than
Arduino, and can implement many types of devices.
John
Nathan McCorkle
> On 11/16/11 12:14 PM, Nathan McCorkle wrote:
>>
>> I'm developing a line of scientific instrumentation that I want to be
>> 21st century compatible (option for dongle ethernet, wifi, bluetooth,
>> sd card, USB thumbstick)... but its hard to just throw linux at these
>> devices because I'm concerned about precise timing for sensor control
>> lines, as well as having around 30Mhz SPI functionality (fast ADC).
>>
>> Would an 500-700Mhz running Angstrom be able to function as a
>> real-time OS (or even predictably real, if under-utilized i.e.
>> processor has lots of idle time)? I'll probably run a web server to
>> take simple AJAX commands and then stream data out 0.6-1.0 MB/s
>>
>> What options do I have for scaling down cost (I don't need graphics
>> acceleration, for instance, and I don't need lots of RAM)
-----------------------------------------------
On Wed, Nov 16, 2011 at 12:43 PM, Mark Barton <ma...@atldes.com> wrote:
> Hi Nathan,
> I can't answer your question directly since I really don't have any real
> experience with near real-time Linux on the Beagleboard. You might be OK,
> but I am guessing that you would be pushing if you need precise timing for
> sensors and especially high speed SPI. What I have done in the past is to
> use a second smaller processor such as TI MSP/ARM family and ST's ARM
> offering to offload the real time grunt work and let Linux do what it does
> best, handling all of the associated peripherals that are supported, servers
> and so on. Granted that you end up with two development projects, but it may
> be actually less expensive in the long run if you don't have to tinker with
> performance issues.
>
> Just my 2 cents.
>
> Mark
----------------------------------------------------------------------
From: Nathan McCorkle <nmz...@gmail.com>
Date: Wed, Nov 16, 2011 at 12:57 PM
Subject: Re: [beagleboard] beagleBone clone, scaling down features to
decrease cost
To: beagl...@googlegroups.com
Thanks Mark, I'm actually developing now with a C2000 Picollo
controlStick, and have a Stellaris EKS-LM4F232 dev kit too.
I had considered your option of offloading, but how would I pipe my ~1
Megabyte/second of data over from one chip to linux?
That leads to then, what is the cheapest and smallest linux solution?
www.youtube.com/watch?v=ut9LfQmaYkU
Cathal Garvey
mean your "job" would then be to create a "shield" that could be
inserted into an off-the-shelf beaglebone to handle the Spectrometer stuff?
On 16/11/11 12:17, Nathan McCorkle wrote:
> So after a lot of thought, what does everyone think about using the
> beagleBone as a reference design for the open spectrometer, and
> related future equipments (3D printer, laser/CNC controller,
> integrated microfluidics Lab-On-Chips).
>
> Pros:
> 500Mhz, lots of power so your hacked code doesn't have to be super-optimized
> Runs debian-strain Angstrom linux, comes with gcc and python, etc...
> familiar and easy to use development from GPIO signalling to web
> servers
> SPI for communication to other chips, USB to use cheap Wifi/Bluetooth
> dongles via common linux drivers, etc...
> Tons of GPIO, even crappy ADC, hardware PWM and timers
>
> Cons:
> Umm, we gotta design it, more complex since it requires external RAM, USB hub
> Might not be able to get DIP or QFP packages, e.g. the processor would
> likely be BGA... would be less kit-like on the soldering end of things
>
> Ideas, comments? (Please)
Nathan McCorkle
> On 11/16/2011 06:17 AM, Nathan McCorkle wrote:
>>
>> Cons:
>> Umm, we gotta design it, more complex since it requires external RAM, USB
>> hub
>> Might not be able to get DIP or QFP packages, e.g. the processor would
>> likely be BGA... would be less kit-like on the soldering end of things
>
> External RAM is a lot of layout and design work to do since it clogs up
> power layout.
>
>
> Some suggestions for processors easily able to handle your CCD timing.
>
The timing isn't so much of a problem, as is both getting the data,
and getting the data to a user or storage, fast.
> ==========
>
> The MBED quick development board costs $50.80 qty one from Mouser...
> 96MHz, with 512KB FLASH, 64KB RAM, Ethernet, USB (host, device, and to-go),
> CAN bus, multiple serial, I2C and SPI buses, 12-bit A/D and even a 10-bit
> D/A converter and real-time clock/calendar. Also on top is the USB connector
> (mini-B), some power regulation circuitry (operating on 4.5 to 9 volts DC,
> or USB power), several indicator LEDs, and the reset button
Yeah but if I could, for $90 I can get a beagleBone @500Mhz, with
linux on it. Linux means cheap peripherals like $5 Wifi @ 54Mbps for
constant streaming of spectra over the net... python, c, java, octave
for program development
yeah but gcc is really the way to go for open source compilation of anything IMO
> ======================
>
> Why would you want 5X the speed for a little lab instrument?
why drive the Prius over the Cadillac? For the mA hit, its not that
much more. 5X the speed means there's room for complexity to grow into
new instrumentation or data processing schemes. higher-level
development, have you seen the node.js web browser IDE someone put
together for beagle platform?
http://www.youtube.com/watch?v=8qME7_Eza54
> Why can't you clock a CCD with only a 90 MIPS processor?
you can, but throw in a TCP/IP stack for a Wifi module, and you're
weeks into coding, and you haven't even touched the sd card or LCD/LED
matrix for cheap user input/output
>
> For that matter, why not a 16MHz arduino compatible for the ease of coding
> and free-openness
> of the arduino IDE? http://www.pjrc.com/teensy/ shows ready-made boards
> with plenty of I/O
> and you could also try python-on-a-chip on that one.
>
> Another thing... if getting serial data clocked out of and into a CCD is
> trouble for you
> and you think you need a complex 500 MHz system to do it, you don't. You
sure, ocean optics does this with a 8051 chip with high-speed USB bulk
transfer... I want to have ethernet, wifi, bluetooth, sd card, usb
stick storage options... USB and linux have all that already, free.
> can use 2
> microcontrollers on one board -- 2 that are easy to solder. One handling
> serdes (serialization-deserialization) for the CCD image sensor, and the
> other doing the usual stuff.
I've thought about this as well, see the last email about me talking
to the beagleBoard google group...
>
> The two micros could be Teensy 2.0 Teensy++ 2.0 $16 $24 in qty one
> (complete devel boards with USB)
>
> Teensy uses a similar Atmel AVR processor as the Arduino board, but there
> are differences. The most important is direct, built-in USB on the Teensy
> compared to serial converted to USB by a separate chip on the Arduino.
> Teensy communications MUCH faster than Arduino, and can implement many types
> of devices.
What's the cheapest embedded linux board that's out there that can
bring over the 1MB/s from a 2nd micro?
>
> John
Nathan McCorkle
> Surely you can buy BeagleBones pre-assembled with RAM etc? Wouldn't that
> mean your "job" would then be to create a "shield" that could be
> inserted into an off-the-shelf beaglebone to handle the Spectrometer stuff?
yep, but if they're selling it for $89, that means they're making
profit... so more integration could make things cheaper... but the
time investment would be a lot more if we were to clone/augment the
beagleBone.
I think with the beagleBone, and a second micro, ADC, and CCD... the
electronics could be under $150... it'd be about the size of two smart
phones, with a few $ invested in a USB dongle, you can interface the
world anyway you like.
>
> On 16/11/11 12:17, Nathan McCorkle wrote:
>> So after a lot of thought, what does everyone think about using the
>> beagleBone as a reference design for the open spectrometer, and
>> related future equipments (3D printer, laser/CNC controller,
>> integrated microfluidics Lab-On-Chips).
>>
>> Pros:
>> 500Mhz, lots of power so your hacked code doesn't have to be super-optimized
>> Runs debian-strain Angstrom linux, comes with gcc and python, etc...
>> familiar and easy to use development from GPIO signalling to web
>> servers
>> SPI for communication to other chips, USB to use cheap Wifi/Bluetooth
>> dongles via common linux drivers, etc...
>> Tons of GPIO, even crappy ADC, hardware PWM and timers
>>
>> Cons:
>> Umm, we gotta design it, more complex since it requires external RAM, USB hub
>> Might not be able to get DIP or QFP packages, e.g. the processor would
>> likely be BGA... would be less kit-like on the soldering end of things
>>
>> Ideas, comments? (Please)
>
>
> --
> www.indiebiotech.com
> twitter.com/onetruecathal
> joindiaspora.com/u/cathalgarvey
> PGP Public Key: http://bit.ly/CathalGKey
>
John Griessen
> Wouldn't that
> mean your "job" would then be to create a "shield"
Your job as product developer is to deliver more bang per buck, so the
$89 price of a Beagle Bone is hard to resell for cheap. And
Teensy ardino compatibles are better to design on top of than
straight "shield" versions with their large connectors and large
size and large price for newbies.
On 11/16/2011 12:17 PM, Nathan McCorkle wrote:
> What's the cheapest embedded linux board that's out there that can
> bring over the 1MB/s from a 2nd micro?
There's more to consider. Why not have the first micro reduce the data some
also? 1MB/s is raw, very raw. A little filtering by the first micro
would let you have a cheap flat pack leaded linux running micro instead
of the all out ones you like. Likewise with the Wifi.. 1MB/s is letting
your lab instrument do nothing but send raw data... If you reduce
it reasonably, you still have plenty of room for outside re-analysis.
Having the first processor be a python-on-a-chip one would let users
easily change the filtering routines and access raw data as well.
There's just not going to be much interest in the world for the CCD raw data though,
so why send it out? Making an open way to access it as it flies by inside the machine
is better.
The electronics bill of materials should be $20-$30 for two processors,
one running linux, and USB ports, not including the CCD.
John
John Griessen
> linux on it. Linux means cheap peripherals like $5 Wifi @ 54Mbps for
> constant streaming of spectra over the net... python, c, java, octave
> for program development
Here's another genuinely open source platform
that's probably able to do it all --
http://elinux.org/RaspberryPiBoard#Hardware_Details
It has three or more processors in one chip.
You can never have it all though, and this SOC with many many wires
coming out of it is in a BGA package. For $35, just use it like a chip
and don't re-engineer it until you can hire the original developers
for a re-spin.
The RaspberryPi folk have partners at Seneca College’s Centre for the Development of
Open Technology in Canada that have optimized a lot of the linux drivers
to work well with the mere 256MB of RAM in it.
It's got a HDMI connector for video out for a display, and
based on Broadcomm's system on a chip BCM2835 containing
a ARM CPU, GPU, DSP with stacked on package RAM of 128MB or 256MB.
The model B version (256MB and ethernet) will sell for $35.
http://www.raspberrypi.org/faqs
You could use the thing for info processing, ignoring the video out for a while,
then productize further with a 7 inch LCD touch screen outputting graphic spectral response representations,
and taking finger commands. There are touch screens that size for $225 right now.
John
Nathan McCorkle
complaints about the processor's documentation and something about the
project not releasing designs and plans:
(search raspberry)
http://hackaday.com/2011/11/01/say-hello-to-our-little-friend-the-beaglebone/
I've contacted broadcom, we'll see what they say about docs.
John Griessen
> I've heard
> complaints about the processor's documentation and something about the
> project not releasing designs and plans:
> (search raspberry)
> http://hackaday.com/2011/11/01/say-hello-to-our-little-friend-the-beaglebone/
>
> I've contacted broadcom, we'll see what they say about docs
The raspberry developers are a non-profit, they say they will release plans
and open hardware details after a while -- the code is all FOSS licensed...
They're probably just worried about their launch and making enough to sell at
their promised low prices of $25 and $35 before anyone else uses the software
development by Seneca College. They said they had to plead with
Broadcomm and explain their whole strategy just to get them to sell in such small
quantities as 10K parts at a time. So Broadcomm may not tell you much -- it's not their
mission to supply teeny open hardware launches, but to sell chips in 100K or 1M
lots. From what I read, the Broadcomm
chip is under development and isn't quite done yet as far as what internal code
is accessible how. Things like "It has a DSP, but no software access to it
in the current model of chip". So, the DSP was for some special batch of chips
for one customer, and they need to do more work before putting it out to
the rest. Broadcomm said they are planning to though, according to r-Pi.
JG
John Griessen
> Yeah the raspberry pi has come up before on my end, but
A big reason to use it is it solves your want for display, GUI, and
standard linux app running quite well without you changing it a bit,
and at a fabulous price. Don't even care how open or hackable it becomes, just use
it, and connect a Teensy to it by SPI and not even use up one of it's
fancy ports. Selling a lab instrument with ethernet or not is just ordering
an R-Pi model A or B. Let the serdes functions for the CCD image chip get done by
a Teensy or an Econotag and add your hardware hacks there.
Limit yourself to linux development, scripting and the like
on the credit card sized R-Pi board and be way ahead.
In case you switch to a camera image sensor instead of a line sensor,
R-Pi is planning a camera interface version next year. (That's not a daughter card,
it's a re-spin of the board design)
Designing and fabbing a board with BGA chips and linux bring up would probably
sink you or me before enough money came in to pay for it. The R-Pi folks
understand they need to sell cheap -- they're selling to education buyers.
DIYbio is similar -- very cheap oriented, so your budget to sell a spectrometer
is much better if you don't burn $150 on electronics bill of materials
and have a decent case for your machine. Many sales depend more on the
look of the box than what's in it. Many people are not comfortable
with a box that is wood with spectrometer burned on it and rectangular --
they want more industrial design like Apple sells. The case should be 3D
printed curvy parts -- could be done on a makerbot, which makes rough output,
if smoothed up with paint. Colorful sells also. The kickstarter is about expired,
but you can probably get approved for another one.
JG
Nathan McCorkle
> On 11/16/2011 04:58 PM, Nathan McCorkle wrote:
>>
>> Yeah the raspberry pi has come up before on my end, but
>
> A big reason to use it is it solves your want for display, GUI, and
> standard linux app running quite well without you changing it a bit,
it has no display, so not sure what you're talking about there...
linux is standard, and building GUIs is quite simple... what standard
linux app are you referring to?
> and at a fabulous price. Don't even care how open or hackable it becomes,
I and others definitely care about it being hackable... I have a lot
of faith in TI products, can get all the info I want/need on them...
they're a huge company that also sells millions upon millions of
chips, so that Broadcom isn't open with their materials is very
discouraging (for me to invest time in their product)
> just use
> it, and connect a Teensy to it by SPI and not even use up one of it's
> fancy ports. Selling a lab instrument with ethernet or not is just ordering
> an R-Pi model A or B. Let the serdes functions for the CCD image chip get
> done by
> a Teensy or an Econotag and add your hardware hacks there.
> Limit yourself to linux development, scripting and the like
> on the credit card sized R-Pi board and be way ahead.
>
> In case you switch to a camera image sensor instead of a line sensor,
> R-Pi is planning a camera interface version next year. (That's not a
> daughter card,
> it's a re-spin of the board design)
>
> Designing and fabbing a board with BGA chips and linux bring up would
> probably
> sink you or me before enough money came in to pay for it. The R-Pi folks
I and the main EE on this project both work with a company that does
this all the time, we know the costs involved. That's specifically why
I'd rather buy a beagleBone and design a daughter board for it.
> understand they need to sell cheap -- they're selling to education buyers.
> DIYbio is similar -- very cheap oriented, so your budget to sell a
> spectrometer
> is much better if you don't burn $150 on electronics bill of materials
but the competiton is still more than an order of magnitude greater
cost... still winning whether electronics are $50 or $150
> and have a decent case for your machine. Many sales depend more on the
> look of the box than what's in it. Many people are not comfortable
> with a box that is wood with spectrometer burned on it and rectangular --
openPCR is in wood and it looks great! Our box was a test that we
threw together in 15 minutes, its in no way representative of a final
enclosure.
> they want more industrial design like Apple sells. The case should be 3D
> printed curvy parts -- could be done on a makerbot, which makes rough
> output,
> if smoothed up with paint. Colorful sells also. The kickstarter is about
> expired,
> but you can probably get approved for another one.
>
I'm all for looking nice, but a case is a case... its not part of the
function critical to the instrument function
> JG
Nathan McCorkle
> On 11/16/2011 12:10 PM, Cathal Garvey wrote:
>>
>> Wouldn't that
>> mean your "job" would then be to create a "shield"
>
> Your job as product developer is to deliver more bang per buck, so the
> $89 price of a Beagle Bone is hard to resell for cheap. And
> Teensy ardino compatibles are better to design on top of than
> straight "shield" versions with their large connectors and large
> size and large price for newbies.
>
> On 11/16/2011 12:17 PM, Nathan McCorkle wrote:
>> What's the cheapest embedded linux board that's out there that can
>> bring over the 1MB/s from a 2nd micro?
>
> There's more to consider. Why not have the first micro reduce the data some
> also? 1MB/s is raw, very raw. A little filtering by the first micro
> would let you have a cheap flat pack leaded linux running micro instead
> of the all out ones you like. Likewise with the Wifi.. 1MB/s is letting
> your lab instrument do nothing but send raw data... If you reduce
> it reasonably, you still have plenty of room for outside re-analysis.
In my experience no science compresses their data, except for CERN
maybe... I'm not sure a slow micro could handle a compression scheme
though. Something linux-based could though.
>
> Having the first processor be a python-on-a-chip one would let users
> easily change the filtering routines and access raw data as well.
>
I don't like the sound of python on a chip for more than a learning
tool, but my friend said there was a microfluidic controller that had
lots of MUXed pins for valving that ran with a python-on-a-chip
chip... he did say it didn't go anywhere, can't say it was because of
the chip though :)
> There's just not going to be much interest in the world for the CCD raw data
> though,
> so why send it out? Making an open way to access it as it flies by inside
> the machine
> is better.
>
> The electronics bill of materials should be $20-$30 for two processors,
> one running linux, and USB ports, not including the CCD.
>
Ok, TI launchpad is $4.30, has SPI, master clock at 16Mhz with
62.5ns/instruction. Arduino 16Mhz (not sure on timings, probably
similar)... both compile with GCC (this is a feature that I'd like to
preserve no matter what platform) and are coded in C, can optimize
with assembly if needed... Launchpad is cheaper really because its
subsidized by TI
Link one of those uCs to the CCD, link the CCD out to the nice ADC
with SPI out, link the SPI out to this FTDI SPI-USB virtual com port
chip (FT232H, USB 2.0, win, linux, mac drivers, $4.3), link that to
whatever system you want.
http://www.mouser.com/search/refine.aspx?Ntk=P_MarCom&Ntt=123533426
Just found this via the Arduino wikipedia page, near the bottom...
this ($35 100Mhz ARM) or the $5-cheaper 60Mhz might run linux (uClinux
comes to mind) and a web server, then attach the CCD subsystem and
comms dongle via USB.
> John
Nathan McCorkle
Nathan McCorkle
Seems the FTDI USB driver is supported in rooted Android...
http://www.ftdichip.com/Android.htm
iPhone has a serial port cable, so pulling the data into the uC and
putting it out the UART could work... @57.6Kbps you could sustain
about 1 read per second (1 read is 59104 bits), or multi-shot into
buffer, then long readout
http://blog.makezine.com/archive/2011/07/59-cable-lets-you-connect-iphone-to-arduino-no-jailbreaking.html
http://www.redpark.com/c2db9.html
either uC would need some extra RAM for buffering, but this is solved
and looks easy:
http://www.arduino.cc/playground/Main/SpiRAM
So that would instantly give you access to a huge market of Android
and iDevices, new and used:
display, storage, comms (cell, wifi, GPS)
John Griessen
> constant streaming of spectra over the net... python, c, java, octave
> for program development
Here's something to think of:
Assumptions:
1. CCD serdes processor is not churning all the time, even when asked
for streaming of spectra over the net. It goes, then stops, idles.
2. wireless comm is something that can be to a gateway, then
to Wifi with TCPIP, since the gateway can be fairly generic and
optional. (For instance an NSLU2 with a Wifi attached via USB).
So then you could use an Econotag
http://redwirellc.com/store/node/1
to do the serdes function,
plus a cheapest linux running micro -- maybe the STM32 or STellaris chips --
telling it when to serdes and data compression, and when to send some wireless serial data to the
gateway via 802.15.4 250kbps 2.4GHz wireless. The linux running
micro now is not having to be real time responsive, just a help for
development, maybe ease of putting GUIs on the system.
The Econotag could be running pymite for easy direct user programming
of the CCD serdes function in python for ease of adapting to
a newer better image sensor du jour.
Now that's performance for price, AND development longevity, AND
user adaptable function. And it forces you to create a simple to describe
data format to send serially and stick to it. That will save you time.
Linux bring up on a prototype might not be a time saver, so Beagle Bone
seems good for that, being all done already. You could just resell
the $89 and $55 boards to start...their BOMs cost maybe $45 for when you want
to cost reduce it later...
But then, there's this write up..
http://dev.frozeneskimo.com/notes/getting_started_with_cortex_m3_cmsis_and_gnu_tools
I'm emailing him for cheap linux running suggestions.
John
Nathan McCorkle
https://github.com/nmz787/open-spectrometer/blob/master/pcb-design/openSpec_Block_Diagram_1.png
With this design it should be
A) programmable with Arduino IDE
B) be adaptable for Ardunio shield or standalone PCB
C) ~60 fps of RAW uncompressed pixel data sent off-board via either
USB 2.0 or to the uSD card
Cost for standalone board should be (roughly) something like this:
$2.82 * 2 - atmega328
$2.01 - WM8253
$4.25 - FT232H
$3 - uSD slot (push/push)
$20 - CCD (TCD1304AP)
$1.45 - uUSB B connector (through hole, DX4R005J91R1500)
$0.26 - EEPROM for FT232H (93LC56B)
$10 for power management and various capacitors, resistors, and crystals
$20 for PCB, potentially 1/10th that if made at home
Total: $66.61
Plus $150 grating
Plus $20 aluminum/steel plate for optics (probably a high estimate)
Add 3D printed case and some 3D printed optics mounts, practically
free if you have a 3D printer
Total $236.61
Depending on the type of fiber you get (TOSlink vs Thor labs or Wards)
add $4-$80
Paint/evap on some phosphor/fluorophore to the CCD to allow UV
detection - maybe about $20
(this stuff might work, though a different Lumogen is what all the CCD
UV papers talk about, this may be equivalent:
http://www.kremer-pigments.com/shopus/index.php?cat=010703&lang=ENG&product=94730)
Add a light source - $20 UV germicidal lamp, $1 CFL, $0.1 LED, etc...
<$350 UV-Vis spectrometer that should be good for DNA
I just got the ADC and Hi-Speed FTDI chips in the mail today, I should
be able to breadboard them within the next few weeks... then I'll
probably try to make some PCBs with copper-clad and the laser cutter
etch method (coat copper-clad with spray paint then etch with laser,
then with copper etcher)
I've also got a sample of the $150 grating, so maybe I can make some
progress on a 3D printed tip-tilt mount (hard for me to get started
because the point of rotation should ideally be at the center of the
convex grating's surface) for that during my Christmas/New Years
school break.
We also got in some cheap holographic transmission grating last week,
the kind that comes in 8.5x11 inch sheets.
Lots to do and play with!
Jonathan Cline
> Arduino IDE
It is important in an OSS community to learn from previous projects.
OpenPCR for example chose Arduino (I suggested PIC or etc)
and I believe they learned the hard way that the Atmel parts
are not a good fit and Arduino costs more with fewer peripherals.
Get their opinion on the project to see if they might have chosen
differently.
Recently I got an STM32 board from ST which is $15 and incredibly
functional; it includes all the firmware drivers needed and
hardware support for built in peripherals. No external
FTDI USB bridge chip would be needed. Check out
STM32-comStick or STM32F4DISCOVERY. Probably this
choice boils down to resolution on the A/D. STM32 is 12 bit.
STM32 uses FreeRTOS for it's framework. It's sufficient
though not very professional.
Another option is the PIC32 UBW32 board which has excellent
USB support as well and all onboard peripherals. Hands down,
Microchip wins for the best free C compiler and IDE. Also
choice boils down to resolution on the A/D.
You may not need 16 bit. Do the calculation.
Pick the hardware with the best fit and function. The days
of needing Arduino because only those boards are available
are over - there are now very usable boards from each
microcontroller vendor - and usually cheaper than Arduino
considering the full system cost.
Cost is everything. Trim cost out of every corner of the
design. Running linux means increasing the cost by at least
30% - typically it requires twice the amount of RAM and
code space compared to RTOS (yes, GNU libs are fat cows).
External SPI RAM will be slow and perhaps noisy, if you
need to add that. This could interfere with validity of the
data. Something to consider. Better to get a chip with
proper internal memory size.
Nathan McCorkle
> On Dec 14, 1:59 am, Nathan McCorkle <nmz...@gmail.com> wrote:
>
>> Arduino IDE
>
> It is important in an OSS community to learn from previous projects.
> OpenPCR for example chose Arduino (I suggested PIC or etc)
> and I believe they learned the hard way that the Atmel parts
> are not a good fit and Arduino costs more with fewer peripherals.
Josh and Tito, Cathal... do you guys really think the Atmel chips
aren't a good fit?
the openSpectrometer team had thought that Arduino IDE would make it
painless to begin tinkering with... or even just doing firmware
upgrades.
> Get their opinion on the project to see if they might have chosen
> differently.
> Recently I got an STM32 board from ST which is $15 and incredibly
> functional; it includes all the firmware drivers needed and
Sounds good, actually, but how easy is it to start programming it (I
mean from the time you get it in the mail, setting up the software,
and finally flashing it)?
I think I can get away with using Atmel chips... but if the
STM32F4DISCOVERY board is here to stay at $16 USD, it would probably
be worth using if its easy enough for other hackers to get into.
> hardware support for built in peripherals. No external
> FTDI USB bridge chip would be needed. Check out
> STM32-comStick or STM32F4DISCOVERY. Probably this
> choice boils down to resolution on the A/D. STM32 is 12 bit.
> STM32 uses FreeRTOS for it's framework. It's sufficient
> though not very professional.
What does FreeRTOS bring over using a C-style single threaded program
ala Arduino or less complex micros?
> Another option is the PIC32 UBW32 board which has excellent
> USB support as well and all onboard peripherals. Hands down,
> Microchip wins for the best free C compiler and IDE. Also
> choice boils down to resolution on the A/D.
> You may not need 16 bit. Do the calculation.
> Pick the hardware with the best fit and function. The days
> of needing Arduino because only those boards are available
> are over - there are now very usable boards from each
> microcontroller vendor - and usually cheaper than Arduino
> considering the full system cost.
> Cost is everything. Trim cost out of every corner of the
> design. Running linux means increasing the cost by at least
> 30% - typically it requires twice the amount of RAM and
> code space compared to RTOS (yes, GNU libs are fat cows).
>
> External SPI RAM will be slow and perhaps noisy, if you
> need to add that. This could interfere with validity of the
> data. Something to consider. Better to get a chip with
> proper internal memory size.
>
>
> ## Jonathan Cline
> ## jcl...@ieee.org
> ## Mobile: +1-805-617-0223
> ########################
>
> --
> You received this message because you are subscribed to the Google Groups "DIYbio" group.
> To post to this group, send email to diy...@googlegroups.com.
> To unsubscribe from this group, send email to diybio+un...@googlegroups.com.
> For more options, visit this group at http://groups.google.com/group/diybio?hl=en.
>
--
Derek
not too hard to get into as far as real time operating systems go.
An RTOS comes into it's own when you have a lot of very time-sensitive
tasks to take care of. For instance, keeping a servo motor at a given
position requires a very short pulse at very precisely defined
intervals. When you are trying to control a half dozen of these, while
also reading a bunch of sensors, you can no longer easily do this in a
single control loop - you need an RTOS.
But it's an added layer of complexity that may not be warranted in the
spectrophotometer project.
You can do more for less money with some other alternatives, but I
like the ATMEL chips and the Arduino IDE for ease of use for novices.
To make a good user-hackable device it still seems a decent choice to
me. In the end, I suspect your lighting and optics drive your overall
price more than does a few dollars in the processor choice.
Just a random opinion...
--Derek
On Jan 1, 1:10 am, Nathan McCorkle <nmz...@gmail.com> wrote:
Nathan McCorkle
After days of trying to get the Atmega timer to not only do PWM (which
was easy) but also fire an interrupt in a manner synced to the PWM
high level, I need a break and got the Parallax Propeller chip up and
running. For an easy example project I was able to use two of the
cores to independently and simultaneously toggle GPIO pins. Tomorrow I
should be able to get the CCD working on the Propeller, then I'll move
on to playing with the high-speed ADC chip (First I'll try making a
breakout board with the laser cutter, wax paper or paraffin, and
copper etch).
Once I get the ADC running then I'll interface that with either the
Propeller or another Atmega (not sure if the Propeller will have free
cogs (the processing cores) left, depending on if I bit-bang the ADC
or use hardware timers)... and then add on the FTDI High-Speed USB
interface, and finally the SD card.
I see that all taking another month, since school is back on this
week. At that point I can finally concentrate my time on setting up
the optics.
> For more options, visit this group at http://groups.google.com/group/diybio?hl=en.
Nathan McCorkle
Atmega/Arduino... more to come.
Nathan McCorkle
whatsoever, on 60 lines of Assembly code. The Propeller has a pretty
powerful set of Assembly instructions, though each takes about 4 clock
cycles to complete, so even though its running at 96Mhz, it's not as
fast as it sounds, though with some timer trickery you can make signal
changes on the GPIO pins at 96Mhz.
Last night I finished soldering a breakout board for the ADC, which I
made using copper clad PCB sprayed with black spray paint and cut on a
laser cutter, then etched with HCL and H2O2. In the next week or so I
should have digitized pixels coming over to the Propeller, and then
I'll start getting them sent out via USB.
Check out the breakout board here:
http://www.youtube.com/watch?v=nidjp-jl7Fk
Last project update was: On Sat, Jan 7, 2012 at 7:47 AM, Nathan
McCorkle <nmz...@gmail.com>
Nathan McCorkle
https://github.com/nmz787/open-spectrometer/blob/master/pcb-design/propeller_square/trace_layers.png
After tons of reading about electronics, it still isn't as good as I
can do, but I think it's a good start to at least make sure the
general circuit works as expected/intended... then I can think about
noise reduction a little more. For example I'm only really using 3 of
the 4 layers of the PCB right now, routing power traces through the
3rd layer or changing the layer stack with GND on both exteriors is
also an decent idea, but having signals inside makes debugging hard...
but I'm pretty happy with this for a 'first' pass.
There seems to be a lot of room on the PCB, but I wanted to keep as
much digital out of the ground plane around the analog area. After
debugging I'll probably want to split the board into two, so I think
the room will get diced away and lead to slightly lower board cost
altogether.
Right now, for this 4-layer board, 3 copies of the PCB from OSHPark
was $67.... so <$33 per PCB. Previously I estimated about $46 in
parts, so 46+33 = $79 with no optics (or SD card or USB cable or
housing or upgraded/coated CCD window).
I've added the current files to github, but I'm not sure all the
libraries are there, but that shouldn't stop a determined person from
googling for the missing modules/footprints! I'll figure that out soon
enough!
-Nathan
Nathan McCorkle
it into the http://www.freerouting.net/ autorouter for the rest.
I didn't want to go 2-layer as all the low-noise analog design stuff I
read seemed like enough reason to go 4-layer. I think I might break
the board into two, which would let me choose different number of
layers for each, if 4 are really needed.
I'm also not sure if I need the datasheet-recommended audio
transistor, since I didn't seem to noticed any loading effect when I
connected my o-scope with and without the transistor. The way I've
arranged the analog section, if I remove the transistor there's about
2 or 3mm between the Analog output and the ADC input... while I
believe the datasheet may be covering cases where the CCD is in a
hand-held barcode reader and several feet from the rest of the
circuitry. I could be wrong though, so its in there for now.
Like I said earlier, I'm not even using the second inner layer, just
top and bottom, with an unbroken ground plane (this is what I really
was going for, I didn't want to cut into the plane, as a lot of analog
design stuff recommends that over the tying two plane sections at the
ADC).
On Wed, Jun 4, 2014 at 1:52 PM, Jerry Biehler <jerry....@gmail.com> wrote:
> That board does not look nearly complex enough to need 4 layers. How are you routing it?
>
> -Jerry
>> dorkbotpdx-blabber mailing list
>> dorkbotpd...@dorkbot.org
>> http://music.columbia.edu/mailman/listinfo/dorkbotpdx-blabber
> _______________________________________________
> dorkbotpdx-blabber mailing list
> dorkbotpd...@dorkbot.org
> http://music.columbia.edu/mailman/listinfo/dorkbotpdx-blabber
--
-Nathan
Nathan McCorkle
https://github.com/nmz787/open-spectrometer/
and the (untested) board is on OSHPark:
https://oshpark.com/shared_projects/4HVNNy11
I still need to figure out the BOM
-Nathan
Nathan McCorkle
I hadn't thought much this time around regarding trace widths... the
analog section has its own LDO tied to the USB + after it passes by
some caps and an inductor at the microUSB port. The FTDI has a
regulator built-in, and the processor, level shifter (which I might
not need, I can't remember but it was on a Propeller demo board that
also had an FTDI), EEPROM and ADC digital lines are all running off of
another LDO (same type as Analog section uses). The ADC and processor
and FTDI all have their caps super close (as much as I thought I
could) and have the smallest-value cap closest to the IC, and am
making sure the power line passes under the cap mount on its way to
the pin (I don't have the caps on a side-leg of the power trace, like
a T with the caps on one side and the right side going to the pin, the
bottom coming from LDO).
I also didn't do any signal length matching, I tried to do this for
the USB data + and -... but from the FTDI to the processor was
autorouted (since I'm not actually needing USB high-speed, I think,
right now at least). I'd like to get the ADC control lines and CCD
control lines all length-matched... my thoughts are it will reduce
sampling error/noise and maybe remove clocking/control-line jitter. I
know the clocks/controls are relatively slow compared to radio or
other matched-pair interfaces... but I'm trying to watch all angles
for a good/robust overall design.
Thanks again!
P.S. here's a link to the project I posted on hackaday, it has some
schematic/PCB pics (they're in the github repo too) for easy viewing:
http://hackaday.io/project/1342-open-Spectrometer
On Wed, Jun 4, 2014 at 3:12 PM, <dst...@gmail.com> wrote:
> Nathan, Congrats on the design and build -- that's a lot of work.
> On the noise/grounding/4-layer -- The only need to go 4 layer would be if
> you (later on) wanted to physically shrink the layout and put parts on both
> sides of the board to make a really compact module. I don't have the layout
> software for your design to look in details, but 2-layers is cheaper and
> quite sufficient in terms of noise. The more important issue is impedance.
> (I've done a lot of RF design.)
> You want very, very low impedance for bypassing (DC) so short distances for
> SM caps and avoid long ground lines and 'loops' -- which it looks like you
> are doing. Mohm impedance signal lines (ADC input?) can be isolated by
> ground area but avoid adding a lot of extra capacitance to ground by
> trace-length over/next to ground -- it looks like you've probably done ok.
> The default 8-mil trace widths of many layout programs is an inherent
> pitfall for RF design so use wide traces for power, grounds, etc.and say
> 15-mil or so for important high-impedance signal lines -- though it sounds
> like you've kept them short which is good.
> Finally, use a super-clean DC supply or add secondary on-board regulation
> for an added 30+ dB supply noise immunity. Wall-warts can be rather noisy
> and infect your design but local bypassed regulation can do wonders.
> As you likely suspect, noise will probably not be your biggest issue first
> time around. Good luck!
> Cheers,
> Dave
>
>
> On Wednesday, June 4, 2014 2:40:56 PM UTC-7, Nathan McCorkle wrote:
>>
>> I used KiCad, did the analog stuff and some more by hand, then pulled
>> it into the http://www.freerouting.net/ autorouter for the rest.
>>
>> I didn't want to go 2-layer as all the low-noise analog design stuff I
>> read seemed like enough reason to go 4-layer. I think I might break
>> the board into two, which would let me choose different number of
>> layers for each, if 4 are really needed.
>>
>> I'm also not sure if I need the datasheet-recommended audio
>> transistor, since I didn't seem to noticed any loading effect when I
>> connected my o-scope with and without the transistor. The way I've
>> arranged the analog section, if I remove the transistor there's about
>> 2 or 3mm between the Analog output and the ADC input... while I
>> believe the datasheet may be covering cases where the CCD is in a
>> hand-held barcode reader and several feet from the rest of the
>> circuitry. I could be wrong though, so its in there for now.
>>
>> Like I said earlier, I'm not even using the second inner layer, just
>> top and bottom, with an unbroken ground plane (this is what I really
>> was going for, I didn't want to cut into the plane, as a lot of analog
>> design stuff recommends that over the tying two plane sections at the
>> ADC).
>>
>> -Nathan
>
> --
> Post to this group at plots-sp...@googlegroups.com
>
> Public Lab mailing lists (http://publiclab.org/lists) are great for
> discussion, but to get attribution, open source your work, and make it easy
> for others to find and cite your contributions, please publish your work at
> http://publiclab.org
> ---
> To unsubscribe from this group and stop receiving emails from it, send an
> email to plots-spectrome...@googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.
--
-Nathan
Jebus Jones
Nathan McCorkle
bought one as its got too many caveats that lower quality and
usability (for example it is a video camera so is limited to 30 FPS,
if your computer can keep up that fast... while I wanted to do up to
60 FPS to your computer direct to an SD card... the bayer filter
reduces signal then adds noise when you convert from color to black
and white... it is a CMOS device, so you can watch a live video and
literally see significant digital noise in dark/black frames (the
raspberry pi camera is pretty bad as far as digital noise goes too)).
There are also driver problems, I believe the auto-exposure is always
on, so you can't do HDR type stuff without a physical neutral-density
filter... or be sure your current measurement was taken with the same
settings as your last measurement. The grating is also a piece of DVD,
which means it isn't optimal for spreading the spectrum out (the lines
are curved not straight as in a normal grating), and I believe its
used as a transmission mode grating, and being plastic it is going to
limit UV throughput significantly (so even if the sensor could be made
sensitive to UV, that would block it quite a bit).
That said, if you're new to all the science and technology, you
probably didn't spend too much to get started so it is probably OK.
There's a lot to learn, you've gotta start somewhere!
On Wed, Jun 4, 2014 at 3:30 PM, Jebus Jones <rai...@hotmail.com> wrote:
> I bought this: http://store.publiclab.org/ Did I do the right thing?
>
> To unsubscribe from this group, send email to
> diybio+un...@googlegroups.com. For more options, visit this group at
> Learn more at www.diybio.org
> You received this message because you are subscribed to the Google Groups
> Visit this group at http://groups.google.com/group/diybio.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/diybio/20bce774-5f80-47f1-8521-2bce40d18a46%40googlegroups.com.
Jebus Jones
On Friday, 11 February 2011 06:34:39 UTC, Nathan McCorkle wrote:
Dear DIY folk,I'm trying to lock down the requirements for an open-source
spectrometer, which I and colleagues will likely design in the coming
months if we can gauge that there is a decent market to make our
engineering time back with a few sales. At first glance we've
estimated that it will cost between $300-$700, . The price could be
flexible depending on customers desired configuration (e.g.
ethernet/no-ethernet, high-res version/low-res version, plastic
case/metal case, etc... basically anything that could be omitted from
the final design if it would lower cost and was desired)Our general specs:
fiber-coupled
2048 or 3648 pixel array detector
USB and/or ethernet for data transferWhat we'd like to hear from you:
If you were going to spend $500 on a spectrometer,what would you want
it to do/haveWould kits be an option, or only assembled units? (or wired but not in
a box, for robot hobbyists looking to build a fire-fighting bot that
seeks out spectral signatures of burning materials!)How often would you want a reading? e.g.... Would it be really fast
(1-10 times per second) for looking at fast chemical reactions? Or
more like one-shot/once-in-a-while, for things with a stable spectrum
(getting DNA/RNA concentrations, cell culture density, etc)What kind of data connection would you want? USB and ethernet come to
my mind, but lots of connections are possible. The data link could be
a bottle neck if you want to stream readings quickly, so this ties in
with the previous question.Let me know what you think.
-Nathan
--
Nathan McCorkle
Rochester Institute of Technology
College of Science, Biotechnology/Bioinformatics
It will be fun to learn the basic principles, which hopefully I can build on. If what you are saying is accurate (and you make it sound pretty shitty, lol) it's a shame, because there simply isn't a decent cheap (and good) spectrometer out there for the DIYbio community. Your project sounds OK, but at $300+ (I will have to read back to get the exact number) it's still not chicken feed, particularly when you add up the costs of putting together a half decent home based lab. I guess you are trying to bring the barrier to entry down, which can only be a good thing.
Nathan McCorkle
swap (when its confirmed to work) my board in-place of the webcam on
the PLOTS 'spectrometer mount' (the rest of the spectrometer that
includes the optics) and have a much better instrument.
Realistically though, I kind of doubt spectrophotometers will become
cheap through design only. The demand has to be present to make things
like optics cheaper. This will only happen if someone can figure out
how to make a really good device, for really cheap, and appeal to
large markets (i.e. home consumers, non-laboratory people).
The $76 for electronics right now will likely deflate a bit, if I
convert to a 2-layer board the PCB price is cut in half immediately,
not counting further board optimization that could reduce the real
estate.
I actually got the price wrong, it was late and I guess I am bad with
math when I'm tired :P
With $46 of parts, and 3 PCBs for $66, that comes to $22 per PCB... so
$68 for a low-noise, fast speed, PC-less single-line camera.
The price inflation to $300 includes an concave flat-field
aberration-corrected grating... from China its $150... from the USA
it's about $800... but the upside is this type of grating ensures a
very linear frequency transform (the separation of mixed light into
monochromatic along the sensor) and has low higher-order effects... so
better SNR basically. This is VERY useful for something like Raman
spectroscopy. Also a UV-compatible fiber optic is not cheap, ~$80.
Adding onto the $300 price tag for a great device, would be ~$200 in
filters, and a cheap $5 laser or a nicer few hundred $ laser
(wavelength stable, good coherence, cheap lasers can do this but is
hit-or-miss according to Sam's laser FAQ)... and you make it to $500
or $700 total cost... but you've got a Raman spectrometer with
reflectance probe capability (for dipping into a stream or some
unknown liquid container, for example) that usually start at around
$7k or so. The $20k Raman spectrometers (the 'gun' looking variety)
come with pattern matching and a database of existing chemicals... the
PLOTS group has made some progress and gained experience with pattern
recognition/matching... but it still has a long way to go to be useful
for very fine analysis (i.e. is there alcohol in my mixed drink, if so
how much).
> -- You received this message because you are subscribed to the Google Groups
> DIYbio group. To post to this group, send email to diy...@googlegroups.com.
> To unsubscribe from this group, send email to
> diybio+un...@googlegroups.com. For more options, visit this group at
> https://groups.google.com/d/forum/diybio?hl=en
> Learn more at www.diybio.org
> ---
> You received this message because you are subscribed to the Google Groups
> "DIYbio" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to diybio+un...@googlegroups.com.
> To post to this group, send email to diy...@googlegroups.com.
> Visit this group at http://groups.google.com/group/diybio.
> To view this discussion on the web visit
Jebus Jones
John Griessen
> then you would just be another vendor selling equipment to those labs best equipped to afford it, lol.
at. Every variant product can have a different price and different buyers. If you make
a version that connects via ethernet to get web page data, you 'll have one customer, vs.
USB and an app for Windows and mobile phones, you'll have another, and that depends zero on
the "instrument" per se.