Looking for a designer (UK)

Thread Starter

Mike Brickman

Joined Mar 12, 2019
18
So, I'm a programmer but know nothing about electronic engineering. I have developed scoring and results software for the sport of Dog Agility but need to take this a step further and integrate a timing solution. This would involve a simple light curtain design. I have a 'theoretical' design but have no idea how to turn this into an actual working system.

Any suggestions about how I find a 'volunteer' to help achieve this. I have a small budget, but this is a hobby sport so funds are very limited. Ideally I would like to find a retired or hobby designer who is looking for a fun project. All materials and costs would be paid plus a little beer money. If they are based in the South West of England and have a dog then so much the better.

Apologies if this is a bit off topic, but I had no idea where to start.
 

hexreader

Joined Apr 16, 2011
581
There are a lot of very helpful people on this forum. I recommend that you give concise details of the project here and you might get a good range of ideas that could lead to better solutions.

Too much to hope that all of the sensors could be fitted within tunnels I guess? Much simpler than detecting dog position in the wide open
 

Uilnaydar

Joined Jan 30, 2008
118
Agree with Hex. Get a concise set of specs together and toss it up here and unleash the AAC hounds.

As a programmer, YOU could do this project given enough time and research. This is a perfect project for those wireless modules that speak Arduino. From watching (I'm a bird dog trial guy, not agility) agility sensors at each obstacle would "fire" on proximity and wirelessly send to the main timer host. Then it's just about what HMI/crowd display you want.

Sensor options
  • Is this a light curtain?
  • Collar device for proximity?
  • wires/wireless?

Scorer's table options
  • Penalties for "ugly" dogs
  • go/no-go for valid run (maybe a sensor failed)
  • Any judge related scoring modifers (are there penalties etc)
  • so on and so forth

Wireless scoreboard options
  • Lights for each obstacle cleared
  • Time each obstacle cleared
  • Total elapsed time
  • Best time (Time to beat)
  • etc etc etc
Fun project!
 

Thread Starter

Mike Brickman

Joined Mar 12, 2019
18
OK more details...

For those that don't know - Dog Agility is a sport for dogs and handlers where dogs have to complete a course of obstacles as quickly as possible without picking up too many faults. Obstacles comprise jumps, tunnels, weaves, see-saws etc. Courses normally start and end with a jump. We need to accurately measure the time between the first and last jump. This is currently done using a light gate.

I have designed a scoring system that uses tablets to mark faults and enter times. However this currently requires the scorer to copy the time from a display which wastes time and introduces transcription errors. Whilst the existing timing systems could theoretically be modified to communicate with the tablets, the basic design is around 20 years old and involves a lot of cables (and lead acid batteries!). Not a good idea as these are a potential tripping hazard and we get too much down time due to damaged cables and connectors.

My plan is to bring the system up to date using modern electronics and make everything wireless. I have looked at using ToF sensors but there are too many issues at this time (black dogs don't reflect much light, accuracy is another issue, as is operating in direct sunlight).

So I plan to stick with the light gate concept with the specification as follows:

Each timing gate will comprise two uprights: a transmitter and a receiver. These will be approximately 1000mm high. The transmitter will have 10 Ir emitters (eg Vishay VSMB2000X01) spaced at 100mm intervals. These will pulse so that only one transmitter is on at a time - but for 3 decimal place accuracy the cycle will complete every 1/1000 sec. The receiver will have have 10 identically positioned sensors (eg VEMD2000X01). These will be poled during each pulse to determine if any 'beams' have been broken. Each receiver will be able to sense signals from its matching partner as well as those above and below. This will create a sensing pattern similar to the diagram below.



The receiver will need to 'tune' itself into the transmitter from time to time to ensure that it knows the when the emitters should be sending, but also what a normal 'open' gate looks like. Competitions are mostly outdoors on rough ground so there will be some variation in signals - especially on windy days. Also sunlight needs to be taken into account (even in England!).

As a dog passes through the gate - the receiver will 'intelligently' determine that this has happened (as opposed to a butterfly passing through) and will signal a control using that will start or stop the clock as appropriate. The nature of the signalling (WiFi/Bluetooth/IEEE 802.15.4) will be determined later.

Both gates will be battery powered (probably replaceable 18650 units).

As I say - I am a programmer not an electronics engineer so forgive the imprecise terminology.

I need help with turning this into a hardware design. I can probably do the micro programming work and plan to order a Arduino kit to start brushing up my skills.
 

djsfantasi

Joined Apr 11, 2010
9,156
Your cross-crossing beam picture seems a bit over the top. Touch screens have been built using just parallel beams. By scanning all beams and counting the quantity interrupted, you can differentiate between dog and butterfly.

The first question is how far apart the beams need to be. I’d think a generous spacing would be fine. A dog’s a bit bigger than a butterfly. But the delta in jump differences may be large. Let N be the number of sensors.

Given sufficiently sturdy mounting, I don’t see wind being a factor. The issue is more of filtering out sunlight. I’m thinking of a monochromatic laser and corresponding phototransistor or detector. I.e., green laser and photodetector that only responds to light in the spectrum of the laser. Low tech, one could use colored filters.

And to make it further bulletproof, you could borrow a technique from IR sensor pairs. Pulse the lasers at a pre-defined carrier frequency and your receiver only responds to a signal at that frequency.

This last scheme reduces the need for “tuning”. The receiver either does or doesn’t recognize a signal of the appropriate color and frequency.

Depending on how accurate the timing is required, you can code appropriately. The driving of the lasers can be done with one pin! To be fast, I’d read 8 sensors at a time. The Arduino supports port IO that does this. But to read more than 8 sensors, you’ll need additional circuitry. A chip called a multiplexor can load successive banks of sensors into the μP for processing

Using a multiplexor requires log2(N) additional pins to address all sensors.

As far as the timing itself, you can use the Arduino functions to capture milliseconds or microseconds. The latter overflows in 70 minutes and would require extra coding. I would hope that 1/1000ths of s second would suffice and recommend that 1/100ths would be sufficient as microseconds might be affected by code branches.

Let me know if you have any questions.
 

Thread Starter

Mike Brickman

Joined Mar 12, 2019
18
Thanks for your contribution.

Unfortunately a sufficiently sturdy mounting is not an option. We run on very uneven often muddy surfaces and equipment needs to be as light and portable as possible. Also the equipment is setup be volunteers with little or no training - getting the transmitter and receiver pointing roughly at each other is the best we can hope for. And everything will wobble around. Welcome to my world! Everything about this project has been extremely challenging.

To answer your questions, the gate is between 1500mm and 2000mm wide (sometimes narrower of not at a jump). We time runs to 3 decimal places. I agree sunlight needs to be addressed - this is an issue with the current generation of timers. We avoid pointing the receiver into the sun as this swamps the sensors (and reverse the uprights if the sun moves round).

I am proposing the crisscross beam approach as it has a number of benefits. We do not have to worry about making sure that sensors only receive from their matching partner. We can be much more relaxed about alignment as we take advantage of the over spill to other sensors. We can have greater precision with fewer sensors and emitters - excluding the edges, the maximum distance between beams is half that of a parallel system. Fewer sensors means less power and at a 10 ring show we have 40 uprights to power which is a lot of batteries to worry about if they need changing frequently.

I'm not sure whether parallel or crisscross is more complicated to implement as with parallel you need to ensure that a sensor only detects its partner and this has to be done electronically as optical solutions (lenses or hoods) will require precises alignment.

Does any of this make sense. As I have said I am an application programmer and am well out of my depth here. Arduino starter kit ordered from Amazon today though.

Hope that helps - really looking forward to your continued input.
 

djsfantasi

Joined Apr 11, 2010
9,156
I’m still thinking out loud at this stage.

I have a couple of questions. To create the cross cross effect, it looks like you’ll have three separate emitters at each height. Is this correct?

I am assuming that you’re flashing the emitters to conserve battery power. This does also add complexity as you will have to synchronize reading the sensor with flashing the emitter. This extra time may prove to be costly later in the design.

Funny, I see using optics as a way to simplify alignment. For a second, assume the emitter is an LED laser. Using a diffuser on the emitter and a planoconvex lens on the sensor will increase the area in which the poles can be aligned.

Non-electronic or software considerations aside, I’d design a tripod base with adjustable legs, into which is inserted a gimbal. The gimbal would support a pole with a heavy base. This assembly would make a plumb bob for the pole. The gimbal axles would be locked in place with a set screw, guaranteeing the pole is upright

Also, a unique function could be implemented in software that can assist during the alignment process.

Forty poles! That’s a lot. Are they all going to communicate with a master processor. To return their results. To start all forty jumps timing at the same time.
 

djsfantasi

Joined Apr 11, 2010
9,156
You might need an Arduino Mega for its processing power. Also, there are many shields which can be added. A real time clock. An LCD display. Wireless communications over a variety of protocols.
 

Thread Starter

Mike Brickman

Joined Mar 12, 2019
18
My design is actually much simpler than you think. I plan just 9 emitters spaced 100mm apart with the receiver having 9 sensors at the same spacing. The uprights would typically be 1500mm apart. These would be regular infrared devices. For the theoretical design I am envisaging Vishay VEMD2000X01 VSMB2000X01 parts (although the final choice may be different).

The emitter has an angle of half intensity of 12 deg (and drops off very quickly after that). Using basic trigonometry (glad I paid attention in school) that equates to a spread of 319mm over 1500mm. So when on, it will illuminate its partner plus the three above and the three below. The sensor has a 15 deg angle of half sensitivity, so matches nicely. From the charts supplied for the devices the intensity x sensitivity factors works out as follows: main sensor 1.0, first offset sensor 0.95, second offset sensor 0.73, third 0.35. All this assumes perfect alignment, so in the real world the factors will be different and variable - but we will still get a spread across several sensors.

Only one emitter will be on at a time. To achieve 1/1000 sec accuracy the emitters will fire in sequence for .1 ms every 1ms. This leaves an empty time slot when nothing fires which can be used to synchronise the two uprights.

Most of the time there will be no dog in the way, so the gate will 'learn' which sensors are triggered on each firing. We will ignore weak or fluctuating sensors to create a 'null profile'. This is just a list of sensors we expect to be triggered during each firing. Any variation in this will be investigated and the software will decide whether this is a dog, butterfly or a re-alignment.

The beauty of this design is that (other than very close to the sides) we get a vertical resolution of 50mm or better. If we used single beams the vertical resolution would be a fixed at 100mm. The maths works out so that for any vertical separation of the devices the worst case vertical resolution is half that distance. Put another way criss-cross beams have at least twice the accuracy of parallel beams (except at the edges).

To eliminate background infrared, the signals will be modulated. I have no idea how this is achieved or how we get the sensors to check for a modulated signal, but know this bread and butter to circuit designers. The remaining problem is direct sunlight hitting the receiver and swamping the sensor. I guess this comes down to choosing the correct (quality) parts that are designed for outdoor use and avoiding pointing into direct sun where possible.

Regarding tripods, bases, mounting etc. Competitions are run in a very imperfect world. Ideally we would use permanent venues with laser gates planted in concrete. But this is not where we are. We rely on unpaid and untrained helpers to setup rings at competitions (often in appalling weather conditions – this is England). They have enough to do getting everything ready first thing in the morning so we have to keep things as simple as possible. The option we will go for is to use velcro straps to attach the uprights to the wings of the jump. This is quick and easy and makes for a nice portable solution (which is extremely important). The gimbal idea is very clever and would work well for other sports but not in my world.

Thanks for the tip on the Arduino Mega. My uno arrives today, but once I have got the hang of that I will invest in a Mega.

As for communications. Each ring has a control box containing a raspberry pi. This stores that data and provides the WiFi access point for the tablets used on the ring. The control boxes exchange data amongst themselves using a mesh network. The timing gates would just need to talk to their ring’s control box.
 

Ya’akov

Joined Jan 27, 2019
9,070
I would encourage you to consider mounting the emitters and detectors on the same uprights and for the other one, using retroreflectors.

This would make alignment much easier as well as simplifying wiring. You can use your same calibration method.
 

BR-549

Joined Sep 22, 2013
4,928
Does only one dog run at a time?

I would look for a much simpler detection method. That's way too much overhead. What material are the poles? And can it be changed to wood or fiberglass?
 

Thread Starter

Mike Brickman

Joined Mar 12, 2019
18
I would encourage you to consider mounting the emitters and detectors on the same uprights and for the other one, using retro reflectors.

This would make alignment much easier as well as simplifying wiring. You can use your same calibration method.
Having emitters and detectors on the same upright would clearly make life much easier - not just from a design point of view, but also operationally. Better portability, fewer devices to go wrong, easier to set up , fewer batteries etc.

The main problem I see is accuracy. With a direct light source the detector is viewing a pin point object (eg 1.5mm dia) so we can tell when a dog crosses the line to mm accuracy. With reflected light our target is potentially bigger so we may lose accuracy.

That said, I think this is worth experimenting with as the trade off may well be worth it. It boils down to the physics of the reflector material used - how good is it at infrared wavelengths, how efficient it is (% of light returned), how much is the returned light spread?

I see that tape is available designed specially for infrared sensors so will buy some to try out.

Thank you very much for the tip.
 

Thread Starter

Mike Brickman

Joined Mar 12, 2019
18
Does only one dog run at a time?

I would look for a much simpler detection method. That's way too much overhead. What material are the poles? And can it be changed to wood or fiberglass?
When you are dealing with sport, the competitors want reliability and accuracy - even at amateur levels. So this trumps complexity. A single beam system would be much simpler, but times would depend on where or how or if the dog crossed the beam (dogs have very different jumping styles).

The jump poles are basically made from uPVC plumbing pipe. They need to be lights to avoid injury if the dog knocks them. Before you get too carried away though, the course does not have to start or end on a jump. It might start with a tunnel - rare, but permitted under the rules.
 

djsfantasi

Joined Apr 11, 2010
9,156
My design is actually much simpler than you think. I plan just 9 emitters spaced 100mm apart with the receiver having 9 sensors at the same spacing. The uprights would typically be 1500mm apart. These would be regular infrared devices. For the theoretical design I am envisaging Vishay VEMD2000X01 VSMB2000X01 parts (although the final choice may be different).

The emitter has an angle of half intensity of 12 deg (and drops off very quickly after that). Using basic trigonometry (glad I paid attention in school) that equates to a spread of 319mm over 1500mm. So when on, it will illuminate its partner plus the three above and the three below. The sensor has a 15 deg angle of half sensitivity, so matches nicely. From the charts supplied for the devices the intensity x sensitivity factors works out as follows: main sensor 1.0, first offset sensor 0.95, second offset sensor 0.73, third 0.35. All this assumes perfect alignment, so in the real world the factors will be different and variable - but we will still get a spread across several sensors.

Only one emitter will be on at a time. To achieve 1/1000 sec accuracy the emitters will fire in sequence for .1 ms every 1ms. This leaves an empty time slot when nothing fires which can be used to synchronise the two uprights.

Most of the time there will be no dog in the way, so the gate will 'learn' which sensors are triggered on each firing. We will ignore weak or fluctuating sensors to create a 'null profile'. This is just a list of sensors we expect to be triggered during each firing. Any variation in this will be investigated and the software will decide whether this is a dog, butterfly or a re-alignment.

The beauty of this design is that (other than very close to the sides) we get a vertical resolution of 50mm or better. If we used single beams the vertical resolution would be a fixed at 100mm. The maths works out so that for any vertical separation of the devices the worst case vertical resolution is half that distance. Put another way criss-cross beams have at least twice the accuracy of parallel beams (except at the edges).

To eliminate background infrared, the signals will be modulated. I have no idea how this is achieved or how we get the sensors to check for a modulated signal, but know this bread and butter to circuit designers. The remaining problem is direct sunlight hitting the receiver and swamping the sensor. I guess this comes down to choosing the correct (quality) parts that are designed for outdoor use and avoiding pointing into direct sun where possible.

Regarding tripods, bases, mounting etc. Competitions are run in a very imperfect world. Ideally we would use permanent venues with laser gates planted in concrete. But this is not where we are. We rely on unpaid and untrained helpers to setup rings at competitions (often in appalling weather conditions – this is England). They have enough to do getting everything ready first thing in the morning so we have to keep things as simple as possible. The option we will go for is to use velcro straps to attach the uprights to the wings of the jump. This is quick and easy and makes for a nice portable solution (which is extremely important). The gimbal idea is very clever and would work well for other sports but not in my world.

Thanks for the tip on the Arduino Mega. My uno arrives today, but once I have got the hang of that I will invest in a Mega.

As for communications. Each ring has a control box containing a raspberry pi. This stores that data and provides the WiFi access point for the tablets used on the ring. The control boxes exchange data amongst themselves using a mesh network. The timing gates would just need to talk to their ring’s control box.
Very complete description. Brilliant actually.

Hooding the emitters and sensors might help, but not in all cases.

Have you analyzed the necessary alignment range for successful operation? E.g., the worst case is when the jump is slightly twisted and the poles are crossed slightly wrt each other. How many degrees of twist is allowed before the functionality of the detector is compromised? A similar situation is when the poles lean back from each other.

The Uno has 13 GPIO pins; the Mega has 50+. FYI
 

Ya’akov

Joined Jan 27, 2019
9,070
Having emitters and detectors on the same upright would clearly make life much easier - not just from a design point of view, but also operationally. Better portability, fewer devices to go wrong, easier to set up , fewer batteries etc.

The main problem I see is accuracy. With a direct light source the detector is viewing a pin point object (eg 1.5mm dia) so we can tell when a dog crosses the line to mm accuracy. With reflected light our target is potentially bigger so we may lose accuracy.

That said, I think this is worth experimenting with as the trade off may well be worth it. It boils down to the physics of the reflector material used - how good is it at infrared wavelengths, how efficient it is (% of light returned), how much is the returned light spread?

I see that tape is available designed specially for infrared sensors so will buy some to try out.

Thank you very much for the tip.
Retroreflectors are amazingly efficient, do a little research on that. Also, good filtering on the dectector to block all bury wavelengths of interest will be critical if this is used in daylight. Baffles might also be necessary. While I loathe premature optimization, and this isn’t something I have done or seen, I think you might also be able to use some sort of differential detection using a second sensor not sheilded from the daylight. Just a thought.
 

Thread Starter

Mike Brickman

Joined Mar 12, 2019
18
Very complete description. Brilliant actually.

Hooding the emitters and sensors might help, but not in all cases.

Have you analyzed the necessary alignment range for successful operation? E.g., the worst case is when the jump is slightly twisted and the poles are crossed slightly wrt each other. How many degrees of twist is allowed before the functionality of the detector is compromised? A similar situation is when the poles lean back from each other.

The Uno has 13 GPIO pins; the Mega has 50+. FYI
I am expecting a lot of trial and error testing with a prototype to nail down the right compromise between sensitivity and flexibility. One benefit of attaching the uprights to the jump wings is that these have to be aligned to a certain degree else the pole falls down.
 

Thread Starter

Mike Brickman

Joined Mar 12, 2019
18
You might find this https://ita.ovh/files/rcwl-0516.pdf or a similar product, of which there are several, interesting. An alternative approach, might need testing, but possibly simpler.
My inclination is to stick with the light curtain technology as I know this will deliver what I need. Long term there are other technologies that look interesting - especially ToF 3D cameras. These are progressing rapidly because of drones, self drive cars and smart phone face recognition.

I will definitely be trying your reflector idea though.
 
Top