highaltitude.log.20090226

[00:11] josepharmbruster (n=josephar@45.144.119.70.cfl.res.rr.com) joined #highaltitude.
[01:31] <shellevil> halo - atlantic evolution?
[01:41] Laurenceb (i=83e3dd23@gateway/web/ajax/mibbit.com/x-6c558314f9bd9cd1) left irc: "http://www.mibbit.com ajax IRC Client"
[02:10] kc0wys (n=kc0wys@75-130-209-194.dhcp.stls.mo.charter.com) joined #highaltitude.
[03:46] sYx66 (n=wooh@bas1-kitchener06-1096648054.dsl.bell.ca) joined #highaltitude.
[03:50] edmoore (n=edmoore@pomegranate.chu.cam.ac.uk) left irc:
[04:27] josepharmbruster (n=josephar@45.144.119.70.cfl.res.rr.com) left irc: Read error: 113 (No route to host)
[04:51] sYx66- (n=wooh@bas1-kitchener06-1096648054.dsl.bell.ca) joined #highaltitude.
[05:07] sYx66 (n=wooh@bas1-kitchener06-1096648054.dsl.bell.ca) left irc: Read error: 110 (Connection timed out)
[05:36] kc0wys (n=kc0wys@75-130-209-194.dhcp.stls.mo.charter.com) left irc:
[05:54] Simon-MPFH (n=simon@phantom.mpfh.co.uk) joined #highaltitude.
[08:58] edmoore (n=edmoore@pomegranate.chu.cam.ac.uk) joined #highaltitude.
[09:18] edmoore (n=edmoore@pomegranate.chu.cam.ac.uk) left irc: Read error: 113 (No route to host)
[09:54] rjharrison (n=rharriso@gateway.hgf.com) joined #highaltitude.
[11:20] jcoxon (i=zeusbot@lister.antycip.co.uk) joined #highaltitude.
[11:20] <jcoxon> morning all
[11:20] <jcoxon> just got a quote for some valves :-D
[11:23] <SpeedEvil> D= OMG! I assume.
[11:24] <jcoxon> the valves that laurence originally suggested are30 pounds each
[11:24] <SpeedEvil> That's not utterly horrible.
[11:24] <SpeedEvil> Are they specced to the conditions?
[11:24] <jcoxon> http://www.sensortechnics.com/index.php?fid=300&fpar=YToxOntzOjQ6InBjaWQiO3M6MzoiMTgzIjt9&isSSL=0&aps=0&blub=cc367a6648be72fe7ad8e2a6ea2eefd8
[11:25] <jcoxon> the quoted ones are 5v do 30psi and are pretty tiny
[11:26] <jcoxon> i'm quite tempted as they should be quite reliable
[11:26] <SpeedEvil> what's the flow rate at a couple of millibars?
[11:26] <jcoxon> hmmm minimum flow says 6l at 30psi
[11:26] <jcoxon> perhaps we should got for the 6psi version
[11:27] <SpeedEvil> the head'll be what - 10-20cm?
[11:27] Simon-MPFH (n=simon@phantom.mpfh.co.uk) left irc: "Leaving"
[11:28] <SpeedEvil> err
[11:28] <SpeedEvil> that's 6l of air
[11:28] <jcoxon> ooo good point
[11:29] <SpeedEvil> The operating temp is only down to 0.
[11:29] <SpeedEvil> I'd want to know how it's constructed to see if it might not jam at -50
[11:29] <SpeedEvil> actually
[11:29] <SpeedEvil> no
[11:29] <SpeedEvil> -15?
[11:29] <SpeedEvil> what's the minimum temp at altitude?
[11:29] <SpeedEvil> I keep forgetting it's not stratospheric.
[11:31] <jcoxon> it'll be the base of the stratosphere
[11:31] <jcoxon> so -50 is a good baseline
[11:33] <jcoxon> hmm i'm sure the one laurence found had a lower temp rating
[11:33] <jcoxon> let me have a browse
[11:36] <SpeedEvil> that one says -40C storage
[11:36] <SpeedEvil> isn't it going to be at ~10Km?
[11:36] <SpeedEvil> actually, on reflection, 'only 10Km' is quite high.
[11:36] rhjharrison (n=rharriso@gateway.hgf.com) joined #highaltitude.
[11:41] <jcoxon> they don't usually make valves for cold things
[11:41] <jcoxon> just warm things
[11:42] <SpeedEvil> Ok. alcohol is about 25* more viscous than air.
[11:42] <SpeedEvil> Very fuzzy memory says that the flow if it's turbulent might experience a drop in pressure of differences in viscosity squared times the original pressure drop
[11:43] <jcoxon> i should know this as a medic
[11:43] <SpeedEvil> so ~18000PSI to get 6l/s flow
[11:43] <SpeedEvil> so ~18000PSI to get 6l/min
[11:44] <SpeedEvil> Hey!
[11:44] <SpeedEvil> You've got helium - you've got carbon fibre, that might just work!
[11:44] <SpeedEvil> :)
[11:44] <SpeedEvil> found price for prepreg online the other day.
[11:44] <SpeedEvil> $150 for 60"*1 yard. ~200gsm
[11:45] <jcoxon> SpeedEvil: we don't need that much of a flow rate
[11:45] <SpeedEvil> jcoxon: of course not.
[11:46] <SpeedEvil> jcoxon: I'm just trying to remember how to work back for flow rate at 2000Pa or so with alcohol, not water.
[11:46] <SpeedEvil> Ah.
[11:47] <SpeedEvil> They give an equivalent orifice diameter for the 6psi one of 1.1mm dia
[11:49] <jcoxon> i think the 6 psi one might be best
[11:50] <SpeedEvil> with advances space-age simulation techniques I have the answer!
[11:50] edmoore (n=edmoore@pomegranate.chu.cam.ac.uk) joined #highaltitude.
[11:50] <SpeedEvil> Well - lighter, paper-clip, plastic cup
[11:50] <SpeedEvil> ~1.5ml/s for water
[11:50] <SpeedEvil> so maybe 2ml/s for air
[11:50] <SpeedEvil> alcohol rather
[11:51] <SpeedEvil> the hole is 1.3mm+-0.2mm or so
[11:51] <SpeedEvil> whyat's the total ballast mass?
[11:51] <jcoxon> depends on the payload weight
[11:51] <jcoxon> 5kg - payload weight = ballast weight
[11:51] <SpeedEvil> so 2Kg?
[11:51] <SpeedEvil> ish
[11:52] <jcoxon> i'm think 3 - 4
[11:52] <SpeedEvil> 0.5W*3000s isn't a really scary amount of power.
[11:52] <SpeedEvil> If it won't jam of course.
[11:54] <jcoxon> yeah
[11:54] <jcoxon> might email the engineer to ask if tehy have any low temp valevs
[11:54] rjharrison (n=rharriso@gateway.hgf.com) left irc: Read error: 110 (Connection timed out)
[11:55] <jcoxon> and whether i can wing any free samples (doubt it but worth a try)
[11:55] <jcoxon> right bbl
[11:55] jcoxon (i=zeusbot@lister.antycip.co.uk) left irc: "leaving"
[11:55] <rhjharrison> ok
[11:55] Nick change: rhjharrison -> rjharrison
[12:18] josepharmbruster (n=josephar@45.144.119.70.cfl.res.rr.com) joined #highaltitude.
[12:56] josepharmbruster (n=josephar@45.144.119.70.cfl.res.rr.com) left irc: Read error: 60 (Operation timed out)
[13:57] Laurenceb (i=83e34f19@gateway/web/ajax/mibbit.com/x-3c59971bed5dedf9) joined #highaltitude.
[14:03] borism_ (n=boris@195-50-212-67-dsl.krw.estpak.ee) joined #highaltitude.
[14:10] borism (n=boris@195-50-200-227-dsl.krw.estpak.ee) left irc: Read error: 145 (Connection timed out)
[14:47] edmoore_ (n=edmoore@chu-gw.churchillcambridge.co.uk) joined #highaltitude.
[15:00] edmoore (n=edmoore@pomegranate.chu.cam.ac.uk) left irc: Read error: 113 (No route to host)
[15:00] rjharrison (n=rharriso@gateway.hgf.com) left irc:
[15:46] edmoore_ (n=edmoore@chu-gw.churchillcambridge.co.uk) left irc:
[15:58] Simon-MPFH (n=simon@phantom.mpfh.co.uk) joined #highaltitude.
[16:55] jcoxon (n=jcoxon@host86-146-165-98.range86-146.btcentralplus.com) joined #highaltitude.
[16:55] <jcoxon> afternoon
[17:08] mc- (n=mfcastle@cpc3-glfd1-0-0-cust439.glfd.cable.ntl.com) joined #highaltitude.
[17:16] mc- (n=mfcastle@cpc3-glfd1-0-0-cust439.glfd.cable.ntl.com) left irc:
[17:32] <Laurenceb> http://www.katco.eu/mandm.htm?gclid=CNr28IC_-pgCFQQhnAodVF8DnA
[17:47] <jcoxon> Laurenceb, We have the X203SF0 @ ?29.97 each.
[17:51] <Laurenceb> cool
[17:51] <Laurenceb> not horrible
[17:51] <jcoxon> indeed
[17:51] <Laurenceb> hang on is that the same one... ?
[17:52] <Laurenceb> we need to be 100% sure its the one with extended temperature range ect
[17:52] <jcoxon> oh
[17:52] <jcoxon> its not, i needed to ask you that
[17:52] <jcoxon> SpeedEvil and I couldn't find the temp extended version
[17:53] <Laurenceb> X103SSB
[17:54] <jcoxon> where does that one say that its temperature extended?
[17:55] <Laurenceb> silicone IIRC
[17:55] <jcoxon> as thats just 6psi, 3v short pin connector, silicone, screw and gasket
[17:55] <jcoxon> oh right
[17:55] <jcoxon> thats what makes it extended
[17:55] <Laurenceb> yeah I think we want silicone
[17:56] <jcoxon> okay, doesn't mention anything about temp in the data sheet apart from 0-50degC
[17:56] <Laurenceb> yeah erm I did look into it...
[17:56] <jcoxon> don't worry i trust you!
[17:57] <Laurenceb> obviously its running outside of specs
[17:57] <jcoxon> yeah
[17:57] <Laurenceb> but its probably the o ring thats going to jam
[17:57] <jcoxon> i've got the email address for the salesperson for london now
[17:57] <Laurenceb> cool
[17:57] <jcoxon> so i'll email him now and ask for a quote for the X103SSB
[17:58] <jcoxon> i think we are at the point of if it jams it jams, there is literally nothing we can do to stop it
[17:58] <SpeedEvil> not quite nothing
[17:58] <SpeedEvil> you can insulate it a lot, and make it so it self-heats
[17:58] <jcoxon> hehe
[17:58] <jcoxon> yeah
[17:58] <jcoxon> true
[17:58] <SpeedEvil> also - pulse it
[17:59] <Laurenceb> a 3 way valve is really nice, as we can have a trailing hoise
[17:59] <SpeedEvil> if it's not going on/off, drive it at 5Hz 80% duty cycle
[17:59] <Laurenceb> http://wiki.ukhas.org.uk/_detail/projects:ballast.png?id=projects%3Aaldist&cache=cache
[17:59] <Laurenceb> described there
[17:59] <SpeedEvil> Laurenceb: err - why can't you have thart with only a 2-port?
[18:00] <Laurenceb> you can, but the flow rate will be decreased
[18:00] <Laurenceb> it takes a lot of current to keep that thing open
[18:00] <Laurenceb> so you want to increase flow
[18:00] <SpeedEvil> Oh - I see
[18:00] <jcoxon> Laurenceb, how many should i get?
[18:00] <SpeedEvil> Laurenceb: note however the air pressure
[18:01] <SpeedEvil> Laurenceb: also - the quick test I did with a plastic cup with a 1.2mm orifice and water indicated that alcohol might flow at ~1.5ml/s
[18:01] <SpeedEvil> Which isn't bad at all
[18:05] <SpeedEvil> Unless there is a problem with having the valve open for a minute or so to dump a reasonable amount.
[18:05] <SpeedEvil> 0.5W*4000 seconds isn't a lot of power
[18:07] <Laurenceb> sure, you can only get a small delta pressure
[18:08] <Laurenceb> theres flow specs somewhere for those valves
[18:08] <Laurenceb> sorry I've forgotten and dont have my design notes on me here
[18:08] <Laurenceb> but with something like 40cm of trailing pipe you get power down a lot
[18:08] <Laurenceb> jocoxon: 1 ?
[18:09] <Laurenceb> we only need 1, and they should be hard to fry :P
[18:09] <Laurenceb> but feel free to get as many as you want :D
[18:10] <jcoxon> hmmmm
[18:10] <jcoxon> should we have redundency valves?
[18:11] <Laurenceb> ooh god
[18:11] <Laurenceb> well what if it fails open
[18:11] <Laurenceb> guess thats less likely
[18:11] <Laurenceb> ok, yeah guess a second valve will be ok
[18:12] <jcoxon> that was why i originally suggested 2 ballast tanks
[18:12] <jcoxon> but what is worse - ballast or no ballast?
[18:12] <SpeedEvil> 0.5Wh isn't exactly an issue is it?
[18:13] <SpeedEvil> Just check that the pipe will siphon properly.
[18:13] <SpeedEvil> the pipe needs to be under a critical diameter, so you can't get it running down the side
[18:14] <SpeedEvil> but wide enough to make it worthwhile siphoning
[18:14] <SpeedEvil> I don't know what the happy medium'd be
[18:15] <Laurenceb> as its a 3 port valve the pipe will be drained when the valve is off
[18:15] <Laurenceb> afk
[18:15] <SpeedEvil> If it runs down the side, you don't get any siphoning effect
[18:15] <SpeedEvil> any pressure addition from the height
[18:15] <jcoxon> i'm trying to think of a easy way to flight test this
[18:15] <SpeedEvil> normal launch with a timer?
[18:16] <jcoxon> then look at the increase in ascent rate
[18:16] <SpeedEvil> maintain the ballast till 15Km, let it go, and cut down at 20Km
[18:16] <jcoxon> i think a key thing will be turning it on and off
[18:16] <SpeedEvil> should get the temp to the right number.
[18:18] <jcoxon> Laurenceb, i'll get 2
[18:24] <Laurenceb> cool
[18:25] <jcoxon> just emailed for a quote
[18:25] <jcoxon> might even be able to get them sent out tomorrow
[18:26] <Laurenceb> I'm off, cya
[18:27] <jcoxon> cya
[18:28] Laurenceb (i=83e34f19@gateway/web/ajax/mibbit.com/x-3c59971bed5dedf9) left irc: "http://www.mibbit.com ajax IRC Client"
[18:33] RocketBoy (n=Steve@217.47.75.27) joined #highaltitude.
[18:35] <jcoxon> evening RocketBoy
[18:44] <RocketBoy> hey jcoxon - the MFSK-16 code is comming on well - the main bit should be finished by the w/e
[18:46] <RocketBoy> bbl
[18:46] <jcoxon> oh amazing
[18:52] <jcoxon> has the kit arrived yet?
[18:56] fergusnoble (n=fergusno@fn217.quns.cam.ac.uk) joined #highaltitude.
[19:09] <RocketBoy> jcoxon: no not sure when to expect it
[19:10] <RocketBoy> but ther is a bit to do yet - I can do a lot without it
[19:22] <jcoxon> oh cool
[19:23] <jcoxon> finally got some quotes for some miniture solenoid valves
[19:23] <jcoxon> around 30 pounds each
[19:23] <jcoxon> not bad really
[19:56] G8KHW (n=andy@217.47.75.27) joined #highaltitude.
[20:34] <natrium42> jcoxon, back from USA :)
[20:34] <natrium42> sYx66-, ping
[20:36] <jcoxon> hey
[20:36] <natrium42> got the balloons, went pretty smoth
[20:36] <natrium42> *smooth
[20:37] <jcoxon> excellent
[20:37] <jcoxon> got a quote for the x-valves
[20:37] <jcoxon> about 30 pounds each
[20:39] <natrium42> not too bad
[20:39] <natrium42> would probably cost more to make from scratch
[20:40] <jcoxon> yeah, i think they are a good move
[20:40] <jcoxon> the type i got quoted for itsn't the perfect one, so i've emailed for a quote on the ones that laurence originally researched into
[20:40] <natrium42> bbl, going to get some food
[20:40] <natrium42> literally just got back
[20:40] <jcoxon> might well be able to get htem sent to me tomorrow
[20:40] <jcoxon> np
[20:41] <natrium42> neat
[21:35] <natrium42> got the PCBs too
[21:35] <natrium42> they look great :)
[21:38] <shellevil> :)
[21:41] <natrium42> shellevil, they printed quite a few copies of each project
[21:41] <natrium42> so if you want, i can give you a few for your venus chips
[21:48] hallam (i=836fc8c8@gateway/web/ajax/mibbit.com/x-040602e87a1b5af5) joined #highaltitude.
[21:48] <hallam> hey dudes
[21:48] <hallam> How would I go about finding the rotation matrix that best maps one set of 3D unit vectors to another set?
[21:49] <hallam> I understand it's some kind of least squares deal
[21:51] Simon-MPFH (n=simon@phantom.mpfh.co.uk) left irc: "Leaving"
[21:57] <shellevil> umm. A set of unit vectors?
[21:58] <shellevil> you mean you have a set of vectors pointing in assorted directions, and you want to find some rotation that makes them similar to another set?
[21:58] <shellevil> ah
[21:58] <shellevil> would this be for star-tracker?
[21:58] <hallam> yup
[21:58] <shellevil> One very simple approach I've seen.
[21:59] <shellevil> Your camera has a field of view of 14 degrees, and a limiting magnitude of 5 say.
[21:59] Laurenceb (i=83e3dd23@gateway/web/ajax/mibbit.com/x-016dc3fde7237aff) joined #highaltitude.
[21:59] <hallam> heh, very astude
[21:59] <hallam> astute*
[21:59] <Laurenceb> hello
[21:59] <shellevil> you make a list of all the angular seperations of stars within 14 degrees, and that magnitude
[22:00] <shellevil> you make a list of bright star-candidates in your field of view
[22:00] <shellevil> work out the angles
[22:00] <hallam> right, I've done all that, I've identified the stars
[22:00] <shellevil> you then compare the camera-angle list with the nice sorted list of the stars
[22:00] <shellevil> Umm.
[22:00] <shellevil> isn't that the problem done then?
[22:01] <hallam> that's assigned a star catalog number to every star in the image
[22:01] <shellevil> Oh - right
[22:01] <Laurenceb> oh wow
[22:01] <hallam> now I want to do the fairly simple job of mapping spacecraft axis to inertial axis
[22:01] <hallam> axes*
[22:01] <Laurenceb> how do you work it out?
[22:01] <hallam> the algorithm's pretty basic at the moment
[22:01] <Laurenceb> identifying stars
[22:01] <hallam> as shellevil says, with the addition of magnitude information
[22:01] <shellevil> hall: you along with the angle of the stars seperation have an orientation of the seperation
[22:02] <hallam> I just wanted something quick and dirty, can improve it loads later
[22:02] <shellevil> And you have the angle they appear in the camera
[22:02] <Laurenceb> for attitude, compare each stars pos to a reference attitude
[22:02] <hallam> shellevil: so I should use the planes formed by pairs of stars?
[22:03] <Laurenceb> yeah
[22:03] <shellevil> hall: yes, then you least-squares fit all of the errors
[22:03] <shellevil> if you care about that sort of accuracy
[22:03] <shellevil> you may get more accuracy by defocusing
[22:03] <Laurenceb> its a bit hard to describe on irc
[22:03] <shellevil> as you can pick a centroid out of a 4 pixel blob lots easier than a sharp sub-pixel spot
[22:03] <Laurenceb> yeah defocussing and sub pixel estimation is where you go for really high accuracy
[22:03] <hallam> shellevil: it's already defocused enough, I can barely detect the stars
[22:03] <shellevil> :)
[22:04] <hallam> they're spread over like 40 pixels
[22:04] <shellevil> Twiddle the little knob
[22:04] <Laurenceb> not that one
[22:04] <hallam> I'll make more of an effort to focus it, yeah, but it wasn't *too* far off - I blame crappy seeing
[22:04] <shellevil> Or bin it
[22:04] <hallam> should have another chance to take data tomorrow
[22:04] <hallam> bin it?
[22:04] <shellevil> bin as in take the pixels, and count them 2 by 2
[22:05] <shellevil> average over 2 pixel squares
[22:05] <hallam> yeah I'm convolving with an estimate of the PSF
[22:05] <shellevil> how are you detecting that 40 pixels?
[22:05] <shellevil> ah - ok, that works too :)
[22:05] <hallam> a pretty lousy estimate, but an estimate
[22:06] <hallam> http://www.mibbit.com/up/ErKXHwvF.png <-- results of the first attempt earlier this week
[22:06] <shellevil> But anyway - you pick up rotation around the camera axes from the difference between the stars 'reference' angle, and the apparant camera anal.
[22:07] <shellevil> and you pick up the exact camera pointing by looking at the star pairs actual directions.
[22:07] <hallam> I understand the principle, but not how to implement it
[22:07] <hallam> it's hard to explain over IRC I guess
[22:07] <shellevil> Arbitrary frame - looking out at the north horizon
[22:08] <shellevil> you have a pair of stars you 'know' are horizontally aligned at 10 degrees above the horizon, seperated by 10 degrees
[22:08] <shellevil> symmetrically about north
[22:09] <shellevil> If they are at 48 degrees inclination WRT each other, your camera is twisted 48 (or 48+180) from the reference horizon
[22:09] Laurenceb (i=83e3dd23@gateway/web/ajax/mibbit.com/x-016dc3fde7237aff) left irc: "http://www.mibbit.com ajax IRC Client"
[22:09] <hallam> so that gives you one Euler angle, and then you do the same thing again for the other two?
[22:10] <shellevil> Now you know the twist, if they are both off to the left of your camera frame, you know that your camera is tilted +-48 degrees, and panned either above or below the horizon
[22:10] <shellevil> you now have two possible pointings - and you disambiguate between them by using other objects
[22:11] <hallam> okay
[22:12] <hallam> I guess you can do that for all pairs, and then average the resulting euler angles
[22:12] <hallam> I'll give it a try
[22:12] <hallam> I should probably take a look at these spacecraft attitude determination books that I got on inter-library loan, as well
[22:13] <shellevil> For each pair of stars, you get two roll/pitch/yaw values, make those into a table. Sort by pitch
[22:13] <shellevil> err
[22:13] <shellevil> roll
[22:13] <shellevil> you should get two distinct sets, one scattered everywhere, one coming up with almost identical yaw/pitch values
[22:14] <hallam> yeah
[22:14] <hallam> I kind of want to do a linear algebra approach though
[22:14] <shellevil> Fair enough
[22:14] <hallam> leave everything as cartesian unit vectors
[22:14] <shellevil> I tend to avoid the nasty maths if you can.
[22:14] <hallam> then it should just be LQ factorisation or something (basically I want to do it in order to learn the nasty maths because I'm not very good at it)
[22:14] <shellevil> I'm sure laurencb would suggest using a FFT.
[22:14] <hallam> it also avoids the problem of gimbal lock
[22:14] <hallam> haha
[22:15] <shellevil> I'm not sure how.
[22:15] <shellevil> Yeah.
[22:15] <shellevil> I used to have a much better grasp of the maths than I do now.
[22:16] <shellevil> Is this camera sensitive to IR?
[22:16] <shellevil> or is it a colour one?
[22:16] <hallam> it's colour, it's sensitive to IR, I don't know if the lens has a filter
[22:16] <shellevil> pity
[22:16] <hallam> guess it's easy enough to check
[22:16] Laurenceb (i=83e3dd23@gateway/web/ajax/mibbit.com/x-c80aac4086b9bb79) joined #highaltitude.
[22:16] <shellevil> as the colour knocks off a lot of sensitivity
[22:16] <hallam> yeah I really want to move to a better sensor, it's kind of sucky
[22:16] <Laurenceb> hallam: use an extended kalman filter
[22:17] <hallam> thanks Laurenceb
[22:17] <hallam> that's your solution to everything
[22:17] <Laurenceb> convert each star detected to a pointing vector
[22:17] <shellevil> No it's not.
[22:17] Action: shellevil is surprised Laurencebs solution doesn't involve a fft.
[22:17] <Laurenceb> then that vector is a measurement
[22:17] <Laurenceb> so you have a large number of measurements
[22:17] <Laurenceb> the measurement model is a rotation matrix
[22:18] <hallam> Laurenceb: in all seriousness I intend to use some kind of EKF for frame-to-frame propagation, but I want a least-squares solution to get a result from a single frame
[22:18] <Laurenceb> rotating the star from some reference point to its observed
[22:18] <hallam> and I think least-squares is optimal for the single-frame case
[22:18] <Laurenceb> hmm
[22:18] <Laurenceb> IMO start off with EKF
[22:18] <hallam> and yeah, I'm trying to find out how to get that rotation matrix
[22:18] <Laurenceb> less work
[22:18] <hallam> which I acutally just did
[22:18] <Laurenceb> you can c&p my code off the wiki :P
[22:18] <hallam> woohoo
[22:19] Action: shellevil wonders if cypress'll sample 'b' grade imagers.
[22:19] <Laurenceb> http://wiki.ukhas.org.uk/code:4_state_extended_kalman_filter_in_matlab
[22:19] <Laurenceb> basically use that then have each star as a measurement
[22:19] <shellevil> I know fillfactory would sell 'b' grade ones for 1/5th the price of 'a' grade.
[22:20] <Laurenceb> in a similar fasion to the way the accel is used
[22:20] <shellevil> Lens model I suppose is also an issue.
[22:21] <Laurenceb> yeah
[22:21] <Laurenceb> you need to convert 2D pos on image plane to vector
[22:21] <Laurenceb> hopefully its not a fish eye or anything :P
[22:21] <shellevil> Well...
[22:22] <shellevil> you actually need either a very sensitive camera, or quite a wide field of view to have no 'holes'
[22:22] Action: shellevil tries to remember the planetary probe sat that had that problem.
[22:22] <shellevil> I think it was a comet sample probe.
[22:23] <Laurenceb> so to identify stars... you assume some random star in your image is star x then see if it has a matching star at some radius that would be expected?
[22:23] <shellevil> lau: and then compare with multiple ohter pairs
[22:23] <Laurenceb> if not then move on
[22:23] <hallam> pretty much
[22:23] <Laurenceb> ok, fairly easy
[22:23] <hallam> yeah it's not hard to come up with a basic algorithm
[22:23] <Laurenceb> star trackers are piss
[22:23] <hallam> there's a lot of scope for improvement though
[22:24] <hallam> thanks Laurenceb, it's my thesis project
[22:24] <shellevil> :)
[22:24] <Laurenceb> yeah, all the work here goes into sub pixel accuracy
[22:24] <Laurenceb> hallam: you can do lots of fun stuff with extended kalman
[22:24] <hallam> http://www.mibbit.com/up/8pliGjPH.png <-- sky coverage analysis
[22:24] <Laurenceb> e.g. use the magnitude to derive an error estimate
[22:24] <hallam> I think I'll run out of time before I run out of fun stuff
[22:24] <hallam> EKF is on the list though
[22:24] <Laurenceb> combine with mems gyro
[22:24] <Laurenceb> ect
[22:25] <hallam> already doing that for image stabilisation
[22:25] <Laurenceb> hallam: I've got the basic code on the wiki
[22:25] RocketBoy (n=Steve@217.47.75.27) left irc: Connection timed out
[22:25] <Laurenceb> you have mems gyros rigged up?
[22:25] <hallam> I saw, I'll have a read later
[22:25] <hallam> yeah
[22:25] <shellevil> hall: that's only valid if you have no pairs of stars at irresolvable seperations
[22:26] <hallam> how do you mean?
[22:26] <Laurenceb> hallam: what gyros/imu?
[22:26] <hallam> IDG300, some other one for the other axis
[22:26] <shellevil> hallam: I'd personally be worried about relying too much on brightness. One supernova, ... :)
[22:26] <Laurenceb> nice
[22:26] <Laurenceb> you can try out some IMU code XD
[22:27] <hallam> the gyros aren't sensitive enough really, they're just on the edge of being useful, but I already had the board made up, so
[22:27] <shellevil> hallam: is that assuming the stars are randomly oriented?
[22:27] <Laurenceb> sparkfun or your own design?
[22:27] <hallam> shellevil: no, it's the actual sky
[22:27] <hallam> my own
[22:27] <shellevil> hallam: k
[22:27] <Laurenceb> nice
[22:27] <Laurenceb> accel on there as well?
[22:28] <hallam> yes, but I'm not using them for this
[22:28] <Laurenceb> any chance of sharing the pcb design?
[22:28] <shellevil> magnetometers?
[22:28] <hallam> yup
[22:28] <Laurenceb> I was thinking of using the sparkfun 5dof imu, but ideally you want a 3rd gyro
[22:28] <Laurenceb> thanks
[22:29] <Laurenceb> I got +-1 degree or so accuracy with the sparkfun board
[22:29] <Laurenceb> drifts about yaw obviously
[22:30] <shellevil> hallam: what's this going on?
[22:30] <shellevil> hallam: suborbital rocket balloon launched?
[22:30] Action: shellevil forgot.
[22:31] <hallam> maybe on the rocket, for now just the balloon to test it in a space-ish optical environment
[22:31] <hallam> kind of its own project really
[22:31] <shellevil> Planetarium.
[22:33] <shellevil> I've got some optical fibre bundle coming from dealextreme.com
[22:33] <shellevil> going to try for a 'realistic' sky patch
[22:33] <shellevil> drill holes in bit of plasterboard
[22:34] <hallam> nice
[22:34] <hallam> a lot more reliable than the British weather
[22:34] <shellevil> apply one or more fibres (for brightness variation) with a little drop of coloured enamel on, and a power LED on the end.
[22:36] <shellevil> I also want a virtual sky, but that's more complex.
[22:36] <shellevil> (take RI=2 beads (sold for road marking) photoprocess a teeny 10*10 pixel image on the back, illuminate with UV)
[22:52] <hallam> http://www.mibbit.com/up/egwyGVXM.png <-- this is so much better than the bodge method I was using before
[22:55] <SpeedEvil> neat
[22:56] <hallam> now the tough part
[22:56] <hallam> make it work for exposures shorter than 2 seconds
[22:56] <SpeedEvil> :)
[22:58] <Laurenceb> how do you locate the exact poition?
[22:58] <Laurenceb> best fit to a guassian?
[22:58] <Laurenceb> fft :P
[22:59] josepharmbruster (n=josephar@45.144.119.70.cfl.res.rr.com) joined #highaltitude.
[23:01] <SpeedEvil> hallam: what are the residuals like?
[23:01] <SpeedEvil> I suppose you should really correct the databse for nearby stars for optimal accuracy
[23:02] <SpeedEvil> say the top star has a star 1/5th of the brightness just to the left of its blob, distorting its apparant position
[23:05] <hallam> I *think* that kind of thing is at the arc-second level rather than arc-minute
[23:06] <hallam> I'm aiming for about 30 arc-sec
[23:06] <hallam> will get back to you on the residuals, just tidying some of the code up
[23:06] <SpeedEvil> maybe
[23:06] <jcoxon> evening all
[23:06] <SpeedEvil> e
[23:07] <Laurenceb> hi
[23:07] <Laurenceb> hallam: how are you locating the stars?
[23:08] <Laurenceb> best fit to a circle? guassian?
[23:08] <hallam> be a bit more specific?
[23:08] <hallam> oh, detection
[23:08] <hallam> hi j
[23:08] <hallam> first convolution with a gaussian kernel to remove hot pixels and lower noise floor
[23:08] <jcoxon> natrium42, ping
[23:08] <hallam> then threshold
[23:09] <hallam> then discard all pixels that aren't local maxima (higher than their 8 neighbours)
[23:09] <Laurenceb> ah cool
[23:09] <Laurenceb> that would be how I'd do it :P
[23:09] <hallam> then centroid based on original pixels
[23:10] <Laurenceb> whats a centroid?
[23:10] <hallam> I haven't yet tested it on an image that doesn't have horrible JPEG artifacting (don't ask), it tends to be off by a pixel or so on the one I've got
[23:10] <hallam> centroid = center of mass
[23:10] <Laurenceb> or... brightness?
[23:10] <hallam> right
[23:10] <Laurenceb> ok
[23:10] <hallam> should get to about 0.1 pixels
[23:10] Action: Laurenceb sloww
[23:12] <jcoxon> Laurenceb, we've got our balloons :-)
[23:13] <hallam> sweet!!
[23:13] <Laurenceb> cool
[23:14] <jcoxon> also natrium42 has got his PCBs
[23:14] <jcoxon> natrium42, popped over the border to the US to get them from teh Fedex place
[23:15] <Laurenceb> just my AD7746s to arrive
[23:17] <jcoxon> and get this valve
[23:36] <natrium42> hi
[23:36] <natrium42> Laurenceb, got quite a few copies of the venus pcbs
[23:37] <natrium42> 20 pcs i think
[23:38] <jcoxon> hey natrium42
[23:38] <natrium42> hi jcoxon
[23:39] <jcoxon> i think we should push ahead with a name
[23:39] <hallam> what's Venus?
[23:39] <jcoxon> ridiculous gps chipset
[23:40] <natrium42> hallam, http://www.natrium42.com/gallery2/v/balloon/transatlantic/venus-gps.jpg.html
[23:40] <natrium42> everything but antenna is in that chip
[23:41] <natrium42> jcoxon, transatlantic halo?
[23:41] <jcoxon> i sort of prefer atlantic halo
[23:41] <jcoxon> oooo i'm torn
[23:42] <hallam> transathalontic?
[23:42] <natrium42> haha
[23:42] <hallam> cute board
[23:42] <jcoxon> bbiab
[23:42] <natrium42> i also though about "aurora"
[23:42] <natrium42> maybe atlantic aurora or something like that
[23:42] <hallam> natrium42: do you have a source for the sarantel antennas? are they the active ones?
[23:43] <natrium42> hallam, i got mine here --> http://www.jdgastore.com/
[23:43] <Laurenceb> natrium42: what ant is that using?
[23:43] <natrium42> got 3 passives and 2 actives, i think
[23:43] <hallam> CU Spaceflight has a ZP program called Aurora (of course we wouldn't object to you using it but it might cause confusion)
[23:43] <natrium42> that pcb is for a passive
[23:44] G8KHW (n=andy@217.47.75.27) left irc: Remote closed the connection
[23:44] <natrium42> hallam, ah, bastards! :D
[23:44] <hallam> cool
[23:44] <natrium42> let's use atlantic halo then
[23:44] <natrium42> unless there are objections
[23:45] <hallam> Sounds like a fine name to me
[23:45] <Laurenceb> natrium42: sarantel antennas?
[23:45] <natrium42> Laurenceb, yep
[23:45] <Laurenceb> hmm smd?
[23:46] <natrium42> yes, i have the passive ones which are SMD
[23:46] <natrium42> they are discontinued now for some reason
[23:46] <Laurenceb> ok
[23:46] <natrium42> but i also made PCBs for the bare ones
[23:47] <natrium42> http://www.natrium42.com/gallery2/v/balloon/transatlantic/venus-gps-b.jpg.html
[23:48] <natrium42> jcoxon, got the datasheet for STX2 :)
[23:52] <hallam> bbl folks
[23:53] <natrium42> later
[23:53] hallam (i=836fc8c8@gateway/web/ajax/mibbit.com/x-040602e87a1b5af5) left irc: "http://www.mibbit.com ajax IRC Client"
[00:00] --- Fri Feb 27 2009