Just been spending a bit of time looking around at column chromatography and it really doesn't look as scary or difficult as it's put out to be..
Trying to figure out roughly how big an apparatus would be needed to separate gram-quantities of mixed alkaloids, thinking a 500mm*50mm column would suffice as a rough estimate?
Lab-grade silica gel seems easy enough to source too but does it have to be such a fine grade? If so a diy could be ball-milling dessicant silica and washing it well with solvent first?
So then it's just separating out your compounds.
In lots of luck if you can see your compounds but being colourless alkaloids we're after seems you have to be more clever
For an easy mix of compounds that you know to be safe (enough) via bioassay then it could be as simple as either taking a ridiculous amount of fractions and evaporating and seeing what you get or continuously taking a drop or two on a pane of glass and switching receptacles when there's a gap between residue being shown.
A quick search shows this could be a good option, pretty much exactly what I was thinking of and seems to do the job.
Seems he had quite some issues with rf interference but with a darkened shielded case mounted to the bottom of the column and making a shielded usb cable should allow you to visualise on-the-go what's coming out of the bottom of your gravity fed column.
If done first with a small sample column and then repeated you could easily compare against the first spectra and start collecting when you see the required peaks..

Then it's onto actual analysis, which will be difficult to do against published spectra with diy equipment but if you can figure out a way to make some relatively well known standards (ie chromatograph some re-xed mhrb extract and guess the largest peak is dmt ;P) then shouldn't be at all difficult for separation of usable quantities..

Column chromatography and checking the fractions with TLC is the easiest/most logical way to go about it, otherwise you won`t have a way to really know what you are separating or not. You can probably make some DIY spectrometer, I think InMotion was working on one, but TLC will definitely be more practical and simpler to start I think.

Or was there any reason why you didn`t consider TLC?

You can probably also buy some pretty cheap spectrophotometer, I remember benz mentioning one he bought very cheap.

TLC somehow seems more complicated to me but then I like making circuits and having a nice graph instead of peering at little dots of liquid
And making a simple spectrometer would also allow for real time measurement as it's passing through, which if done right would give you more accurate results than just picking out fractions.
I'll have a wee look around the threads about it as I'm sure it's about but what do you use to visualise the spots with alkaloids?
I'm considering getting a wee column to play around with, not got any need for it at the moment but seems like it could be a really interesting technique.

I applaud the experimenting spirit, do let us know if you try it out

As for TLC, check my signature on `how analysis methods work` , and feel free to ask any further questions.

Well an arduino-compatible board and a uv detector cell is fuck all really, just looking around for deep uv led's, only able to find 390nm or so led's, afaik need something more 230-290 for photometry.
If I can find a few different wavelengths and programme it to sample one of each sequentially... :evil grin:
Shall this be my next project

hrm uv-c germicidal lamps are 254nm generally .. not really got that much of an idea what wavelengths to use - anybody got any experience?

wow now my heads swimming a little bit .. starting to make sense of it ..
can get:
254nm germicidal lamps
uvb reptile lamps
~390nm LED's
visible range LED's
Presumably as long as I get some variance even if it's not the peak of that compound's spectra it will be detected. One worry about intense light sources is the photodegradation of our lovely active compounds..
arduino nano for about £5, UV sensor £10, not many other components needed. Then it's just looking to find if there's a prewritten script to read analog input and forward it to the computer on a real-time plotting programme.. shall see.

Well very near to getting the necessary equipment ..
endlessness: in your sig. link there was a link to 3 columns that you said were for gram sizes etc but the link's dead, can you find another or give a rough size idea?

Believe it or not, column chromatography isn't simpler than TLC, actually it's the opposite.
Column chrom is a preparative technique. Like any art, it requires practice, particularly even packing of the stationary phase (I favor the old-school slurry method), and mobile phase composition.

A typical LC-UV setup can give information about the relative abundance of analytes present, because of the separation of said analytes on a stationary phase (packed column), prior to entering the flowcell on the absorbance detector. (my avatar is a capillary flowcell on an abs. detector, microflow LC)

A spectrophotometer involves no moving solvent. It may give you information about concentration of an analyte in solution, based on Lambert-Beer Law.

interesting project on that blog, horrible resolution though, because of the relatively low number of theoretical plates in the column.

benzyme; my purpose for looking into CC is for separation of a relatively well known group of alkaloids not necessarily analysis as such, but I would think if it's all set up like that then it'd be easier/smoother once it's all set up to have a simple uv photometer on the bottom of the column to visualise the compound separation as it's happening and being able to take fractions at just the right point, rather than taking loads of unknown fractions and then tlc 'ing them after.
It'll be interesting to see what comes out of it, going to try making a UV / visible spectrum detector with wide spectrum light sources, once that's working relatively ok then get a column and have a play with some known compounds and then move onto separation.
Total cost should be ~£30 for a simple detector just to roughly quantify the separation and column silica and solvent really isn't too much either.
Will keep you posted on how it goes!

Well so far so good..
Got a arduino nano and a uv sensor, and after a lot of faffing and head-scratching I've finally got it to work using python and matplotlib ..
Just been trying it out with a 'black light' fluorescent tube, to which the sensor is highly responsive, giving a reading of up to 200. A normal lamp will give a reading of up to 15 but with far less accuracy, confirming its selectivity for UV light. This means I can have a separate visible light detector alongside it to show different peaks. Next will be getting a UV-C lamp and boxing everything up and getting a column to actually use it all on.
Main issue on the software side is getting software that will plot real-time yet be able to show the whole graph with an accurate time not just a scrolling oscilloscope type chart with no reference. And it's getting the timing on the chip and the computer the same, it's a little out of sync so the longer I run the program the more laggy the response is from the chip.
Going good though, and will post a full tek once I'm done with it.

Check this out: http://store.publiclab.org/products/desktop-spectrometry-kit
Works for vis/NIR.

EDIT:
IR spectrophotometry gives useful information on the functional groups in a molecule and although, perhaps, a little outdated, is clearly MUCH cheaper to do than NMR.

Interesting! Could be worth having a play around with ..
Got to the point now where I have:
arduino+uv sensor mounted on breadboard
arduino programmed to forward information on to serial port
altered above program with matplotlib and python to export data to a text file of my choice and altered the axes for my data.
What I'd still like to do is modify the chart so instead of a scrolling chart over a small period it would be an autoscaling chart showing the entire range of data getting progressively smaller scale as time goes on. Unfortunately that seems to be beyond my capabilities at the moment but we shall see..
Anybody with python skills here?
Doesn't really matter too much though because I can just view it in openoffice later.
Next will be getting a column and running some tests to see how well it will detect my chosen compounds .. I may also cobble together a vis/IR sensor as well and chart all three next to each other

You said at the start you were planning on using a 500*50mm column - wouldn't a thinner one have a couple of advantages, namely using less silica gel and solvents?

Of course, sample size would necessarily be smaller and so the detection limit would be higher - i.e. not as good (which is only of relevance if you're looking at trace compounds, I suspect).

How will your detector be calibrated? Presumably you've already looked at stuff like this?