Building speed measurement hardware

[LinusLarsson]

I can confirm the rumor, the PI 3 do overheat to crash condition when stressed. Heatsinks needed..

[PhenixFord]

Just a thought here. Have you thought about using something other than a video reference. Maybe something like this. \/\/

http://www.automationdirect.com/adc/...DC_Rectangular

Works on the same principle as a driveway alarm. For "point reference" you would need a long pole with a reflector on one side. You could use a PVC conduit to keep the weight down. Maybe even make it sectional for easy transport. This device would deliver a pulse output when triggered. Only drawback would be that your flyby's would have to pass between the pole and the photo eye. And you would need a power source to operate the eye. Just a thought.

[LinusLarsson]

It's not really realistic, is it. If I understand you correctly you would need to pass at the same height every time? The device sends out a directed light with a receiver at the other end?

What I really like about cameras is that the speed is "within the limits of the system" verifiable. It's all on film.

[LinusLarsson]

I've got a realtime videofeed up playing on the base station now (while it's recording). It's pretty much zero lag. perhaps 100ms or so, really cool.

[LinusLarsson]

I didn't mount the heatsink, it just stands there... And the crashes are gone..

[PhenixFord]

Originally Posted by LinusLarsson

It's not really realistic, is it. If I understand you correctly you would need to pass at the same height every time? The device sends out a directed light with a receiver at the other end?

What I really like about cameras is that the speed is "within the limits of the system" verifiable. It's all on film.

Device(s) work differently depending on the type ( expense ) you choose. The device I would "test" would be the type that sends out a light source and looks for the return from a reflector. I believe that these are noted as "Light Operated Devices". The device is triggered when it sees a dark spot on the reflector. The reflector doesn't necessarily have to be a single point. The reflector can de a strip of reflective tape. I'm not really sure how long the tape could possibly be without looking at the specs of the device. But I've seen the reflector strip be as long as 8 feet. The range between the reflector and the device can be rather far. I've worked with these devices used as light curtains to safeguard conveyor system that were 40' to 50' long. If precision is what you are after, you would need software to look for the leading edge in the pulse output when it transitions from (-) to (+). It is very realistic. These devices are used on high speed production lines quite a bit. I don't know if they have ever been used in this type of arrangement before. But I think it could work.

[LinusLarsson]

Ok, nice to know, will read up on them. But no, I'm not changing input devices at this stage. But options for the future or is a good thing.

Another thing about the hardware I'm using is that it's available for cheap all over the world. It's important if others in the want to build there own timing system.

[Vinger]

Thks for all the hard work, sure would be nice if you can share eventually so that we can build our own system. What is keeping speed flying from flourishing is the lack of a cheap and accurate timing system.

Looking forward to your results every day.....

[LinusLarsson]

Originally Posted by Vinger

Thks for all the hard work, sure would be nice if you can share eventually so that we can build our own system. What is keeping speed flying from flourishing is the lack of a cheap and accurate timing system.

Looking forward to your results every day.....

Thank you! I'm mildly optimistic this will pan out to something useful. One of the problems I have is to measure 100M, down to centimeters accuracy. Measuring tape this length probably stretch with temperature (and the ones I have access to are like 30 years old. I think I can source a 50M laser meter (perhaps low cost, not fluke), how accurate are these at full distances?

My field is pretty much 100M so 200M is out of the question for me. Of course you should be able to set distance in the program.

[Vinger]

We use a laser measure, just get a white board of about 1mx1m as a target for the laser beam, easier to see. Luckily for us we can get the 200m track into our club area.

A white foam board also works well and cheaper.

[Ace Dude]

I'd think a good quality metal or fiberglass 100m tape measure would have minimal stretch. You can also buy measuring rope. You can measure it out once and put stakes in the ground so you'll never have to measure it again.

I'd put 4 stakes in the ground, two at each end to ensure it's square as the camera needs to be in perfect alignment with both stakes. You can use a triangle to ensure two sides are square and then flip it over for the other two sides.

[LinusLarsson]

I'm going to start with a laser I think, and use some simple trigonometry to get a box, should be doable, I really want this perfect.

However the good thing is that having it perfectly 100M is important for internet bragging. On competition, as long as everybody is running the same course 30cm off isn't that big of a deal. But I still want it to be perfect.

There are so many small sub-projects to this thing. For example I will be mounting the camera-stations on very short camera stands. And those things uses 1/4 inch bolts (standard camera)... Finding the mating nut where I live is a challenge... Nobody use imperial system here

[Ace Dude]

I suggested perfection a while back. I'd still be concerned about measuring time independently on both Pi clocks. Would be interesting to see how much the clock drift there is even during short periods. Ideally, you'd want to have one accurate clock measuring elapsed time for both start and finish.

Check the accuracy on the laser measure as well as accuracy will vary with length and price. They have very nice accurate ones, but they're also very expensive. It would be best to borrow one if you know anyone who is a surveyor or works in construction. Of course, if you know a surveyor they could lay out the whole thing for you very accurately.

A friend of mine has one of these for his job, but they are not cheap.

https://www.us.hilti.com/measuring-s..._METERS?page=0

[AlastairC]

There is a good few pages here on setting up a NTP server on the PI with a GPS seed: http://www.satsignal.eu/ntp/Raspberry-Pi-NTP.html
Also gives the accuracy in time expected. Should be enough for what is needed.

[LinusLarsson]

Yeah, NTP should be no problem. Within 1ms should apparently be quite easy. It doesn't have to be the correct calendar time, but they need to be the same. Perhaps one Pi being the master and the other slave. Preferably I would use the base station as server, but I don't want to have too much crazy requirements on that. Like an NTP server.

[Ace Dude]

Yup, it's all about elapsed time.

[LinusLarsson]

About clock drift, I don't think there is any measurable amount. As long as they stay within 1ms I'm happy. It's also easy to verify using a camera. And in the unlikely case NTP doesn't work out, I come up with something else.

But of course, if I don't use any kind of external clocking I need to check what kind of clock skew there is on the PI hardware. But I really can't see this being a problem on a 5 second run. It has to be off by a lot...

[LinusLarsson]

I checked this up. And there actually is a reasonable amount of clock skew on the PI (Don't know which version). As much as 0.1 - 0.2ms per second... That's borderline... Not the biggest fish in the water but still..

[Ace Dude]

Originally Posted by LinusLarsson

About clock drift, I don't think there is any measurable amount. As long as they stay within 1ms I'm happy. It's also easy to verify using a camera. And in the unlikely case NTP doesn't work out, I come up with something else.

But of course, if I don't use any kind of external clocking I need to check what kind of clock skew there is on the PI hardware. But I really can't see this being a problem on a 5 second run. It has to be off by a lot...

There's always some clock drift. That's why we have NTP in the first place. I'd think 1ms should be able to be easily achievable, however, it would be cool just to validate that from a systems integration testing perspective.

[AlastairC]

Originally Posted by LinusLarsson

I checked this up. And there actually is a reasonable amount of clock skew on the PI (Don't know which version). As much as 0.1 - 0.2ms per second... That's borderline... Not the biggest fish in the water but still..

This is a blog post on the subject and how to test (as this is a couple of years old you may want to run the test with your new V3): http://blog.remibergsma.com/2013/05/...-pi-keep-time/

You can always for the NTP update at the beginning of each logged run so thereby minimising the drift.

Cheers,
Alastair