- This topic has 14 replies, 6 voices, and was last updated 6 years, 9 months ago by
.
-
Posts
-
Hi.
I am about to embark on making a VTI using Arduino components as set out by Piotr Smolarz.
See http://smopi.news.nstrefa.pl/index.php?pages/Video-Time-Inserter
This will be used to time stamp a video signal from a Watec camera when timing lunar impacts or occultations.
Has any one done any thing similar for any tips and advice.
Kind regards.
Tom Moran.
I’d be interested to hear how you get on with this.
I spent 18 months playing around with Watec cameras, using Raspberry Pis and USB video capture dongles to hunt for meteors. The project in an abeyance at the moment, and to be honest I’m not sure the USB video capture hardware was up to the job. Essentially, your video signal is mostly black, and as I understood it, the cheap capture dongles were turning the gain up way too high so we were mostly recording noise. Any real structure was totally saturated. As I understand it, better video capture hardware has brightness controls that let you adjust these things.
Having had that experience, I would be wary of putting anything in my signal chain between the camera and my digitiser. I’d want to know exactly what the video time inserter is doing. If it’s passing the analogue signal through, maybe it’ll be fine. If it’s digitising the signal and then converting back to analogue, I suspect it’ll be a disaster! 🙂
The time signal should be very accurate if it’s done competently. 🙂
There’s a caveat here that if you rely on the GPS unit spitting out NMEA data, that comes in ASCII format over a slow serial connection. By the time it’s made it down the wire and through your serial buffers, you’ll probably only get ~ 100ms timing precision.
The GPS chip will also produce a PPS signal, which is pin which gives you pips once a second, on the second. Using the GPIO lines on a RPi or Arduino, you can sample that at high frequency to get a very good time standard.
My Polish isn’t very good, so I don’t understand much of the attached webpage, but it seems to mention PPS towards the bottom, which implies this particular box ought to have much better than millisecond precision.
It’s also worth calibrating the camera since, depending on the model, you may have a few fields of buffering depending on the sensor readout before the analogue video gets out of the camera. I’m sure that the UK asteroid occultation guys (Tim, Alex etc) will have a good handle on that.
Yes, the slow speed (4800 or 9600) of the serial comms links many receivers use worried me and I know that several that are ostensibly USB actually use a USB/Serial interface internally and so are just as bad. I noticed that to get round this the NMEATIME2 shareware uses a Kalman filter to filter NMEA arrival times and manages to get an accuracy down into the single milliseconds.
I’m told – would have to check old emails for detail – that the 1pps is very accurate long term (so count 100 pulses and the average interpulse time will average very close to 1000ms) but the individual arrival time of a pulse is still somewhat dithered – on some hardware.
Can dig out the emails if you are interested.
Hi Tom,
Wojciech Burzyński (PL) gave a talk on this at last year’s ESOP in Freiberg. The PDF of his presentation can be found here.
Using a 1PPS GPS receiver module they found that their Arduino-based VTI gave timing accuracies as good as the off-the-shelf IOTA-VTI and GPSBOXSPRITE2 models used by many occultation observers around the world.
Happy assembly!
Alex.
Wow.
Thanks for the replies guys. Some of it is way past my current knowledge. However I agree with Alex about the timing accuracy of this device being as good if not better than some commercially available. The total cost of the three main boards came to £35 so worth the risk.
My Polish is non existant also. However if you open his web page (http://smopi.news.nstrefa.pl/index.php?pages/Video-Time-Inserter) in Google Chrome it will offer to translate the page.
Thanks again.
Tom Moran
I forgot to say that the three boards might take a month or so to arrive. Then I will begin my learning curve! I will post and update my progress as I go.
Regards.
Tom
Hi Tom,
If you are timing lunar impact flashes I guess a timing accuracy of only 0.1s is required to confirm a simultaneous event but researchers may prefer to 0.01s for light curve analysis.
Analogue video at 50 fields per second gives exposures of 0.02s per field, which is the fastest rate currently used by most occultation observers. All these GPS-linked VTIs have an accuracy of 1 – 100 microseconds.
For more information on VTIs and video camera timing delays (internal delays and integration effects) have a look at the work of Gerhard Dangl (AT) here.
Cheers,
Alex.
Alex: once the video digitiser has de-interlaced the video, don’t you effectively get 25 frames per second — i.e. 0.04s resolution?
This is what my USB digitiser was delivering.
Thats an interesting presentation. Am I reading that right? The 1pps pulse goes straight into what appears to be an Arduino GPIO – should work!
Hi Dominic,
The video camera will output 25 interlaced frames per second (PAL) and the VTI timestamps each video field, 50 per second. Many occultation (and meteor observers) record to AVI files using a USB video grabber. Limovie and Tangra can read and analyse these at the deinterlaced field level. We use this to get good time resolution (0.02s) on lunar occultations and bright asteroidal events…
https://www.britastro.org/node/8992
The pairs of video fields from an integrating video camera will give (almost) identical images.
UFO Analyser processes meteor AVIs at field level to estimate the meteor’s velocity.
Cheers,
Alex.
Any home brew project is susceptible to misunderstandings in how the time stamp is applied, and you need specialised timing equipment to unravel what is going on at frame and field level. For this reason i would not necessarily go down a DIY route. I would buy a VTI that is fully characterised. There are many, but the two in common use by occultation observers is the GPSBOXSPRITE2 by Blackboxcamera (london) and the IOTA VTI V3 by Videotimers (US). Both these derive the time from GPS and produce time stamps at field level. Ideal for our proposes.
I recently used a WAT-910 HX and recorded a Gemind lunar impact (30cm F/4). It had already been picked up by a 1 m scope but I was able to find the exact frame to 0.02s, and it was there (just). A good example of the importance of coordinated time.
Nick said that frame buffering can effect timeing. It does and a full description for various cameras in general use is published by G Dangl http://www.dangl.at/ausruest/vid_tim/vid_tim1.htm
The project described at ESOP36 (thank Alex) probably has been tested now, and verified, and similar components should work in the same way ( i guess) – best of luck with the project, and please report any occultations you time to the Lunar Section. They are very useful
- You must be logged in to reply to this topic.