Home Made Bearpaw Spine Tester

Stretch

Well-known member
Remember that how important an arrow set matches for spine is more important the better you are. If you are good enough to find the differences then stock Easton coding may not be tight enough for you. (Hence some of the top shooters bulk shooting loads of x10s at 70m and discarding the bare shafts that don’t group bang on.) Mortals and Demi-gods need worry less. But also bear in mind that not ALL the top shooters do this. I’ve shot a 331 70m (in practice :mad:) and 302 90m (FITA *:giggle:) and never had to discard an arrow because of spine issues. (Nowadays it is more of a pointy end goes that way kind of thing...)

Without touching a thing on my bow the difference between an x10 450 and a 410 is not as much as you’d think down at the target when fletched - even if the weights don’t match. If it is a 455 or a 445 and matched weight most of us wouldn’t notice.

Stretch
 

dvd8n

Supporter
Supporter
AIUK Saviour
Well, that was disappointing. For some arrows it worked fine and for others it was, not to put too fine a point on it, a shambles.

In theory, whatever position I took to be zero degrees, an arrow should give consistent results. But I had one arrow that was getting results in every direction possible. Average spine was ok, but the direction of the maximum spine was all over the place.

IMG_20210512_193644.jpg

To try and figure out what was going wrong I took spine readings every 30 degrees.

Graphing the results was interesting.

IMG_20210512_193705.jpg

I did have the feeling that this pattern was meaningful, but I couldn't see why.
 

dvd8n

Supporter
Supporter
AIUK Saviour
I decided to do the readings again, in grams rather than ATA spine, plus with decimal points. This would reduce rounding errors. It'd also make things easier to think about as higher numbers mean higher spine.

I'd also do it really carefully, and plot it on polar graph paper.

IMG_20210512_193731.jpg

That's interesting. Look at that figure-of-eight shape. It's also not evenly centered on the zero point.

Note that zero on the paper isn't zero grams - the scale is chosen to highlight the shape. The middle is at 1105g. Drawn to scale the graph would look like a circle with almost imperceptible flattenings at 0 degrees and 150 degrees.
 
Last edited:

dvd8n

Supporter
Supporter
AIUK Saviour
My first thought was that the arrow was bent or misshapen, but I checked it with a dial gauge, and there was only a couple of thou run-out, offset slightly in the direction of the bottom lobe, 240 degrees.

So not that then. The spines really are like that. With the big notch at 0 degrees.
 

dvd8n

Supporter
Supporter
AIUK Saviour
My theory is that the carbon wrap is starting at about 30 degrees, wrapping round anticlockwise a few times, finishing at about 300 degrees, resulting in one layer less between 300 degrees and 30 degrees. Or something else similar. This would result in a dramatically lower spine at 0 degrees, and a corresponding, but lesser spine reduction at 108 degrees, mirroring it.

But anyway, the weird shape was really compromising the algorithm to find the centre of the notional spine circle depending upon the position of the cock feather (ie where 0 degrees is):

IMG_20210512_193746.jpg

The centres of force/spine ended up being up in 4 different directions, all of them wrong.
 

dvd8n

Supporter
Supporter
AIUK Saviour
One idea that I had was to use six readings.

Then take the two maximum readings to identify the two lobes, then use the circumcentre algorithm to identify the centres of the lobes. A line between the centres gives the line of maximum spine really well; you can even see the misalignment of the line of maximum spine and the centre of the arrow (see the graph above).
 
Last edited:

Nictrix

Member
When you place the arrow on your tester is it at the same points every time?
Do you have some sort of stopper at one end that the arrow can butt up against so that when it is rotated it is the same distance and same contact points on your tester.
Not sure really how to describe what I mean but hope you can understand.
 

dvd8n

Supporter
Supporter
AIUK Saviour
So, what are the issues with all this?
  1. The basic circumcentre algorithm works fine on 'nice' arrows. But it's problematic devising a strategy to detect more unusual arrows and switch to a more complicated algorithm.
  2. The reasoning behind the circumcentre algorithm was to use only three readings, but to determine if three readings are enough, you need to take six...
  3. The more advanced algorithm needs six readings too...
  4. The centres of the lobes strategy gives a good line, but given that you need to identify the positions of the lobes, the line position is kind of self-fulfilling. The offset of the line is interesting to see, but what use is that information? Really, the only use that you could make of it is to throw the arrow away if it's too bad - but if it's that bad then I think that it'd be an obvious problem arrow anyway.
  5. And the big elephant in the room is that none of the complicated calculations give answers that are qualitatively much better than taking six readings, averaging them and giving the direction of the highest one. This is likely more reliable too.
So, the plan is to rip out all the complicated code and replace it with simple averaging.
 
Last edited:

dvd8n

Supporter
Supporter
AIUK Saviour
When you place the arrow on your tester is it at the same points every time?
Do you have some sort of stopper at one end that the arrow can butt up against so that when it is rotated it is the same distance and same contact points on your tester.
Not sure really how to describe what I mean but hope you can understand.
I was making an effort to keep the position consistent when testing, but yes an end-stop is probably a good idea worth implementing.
 

malbro

Instinctive Archer
Supporter
AIUK Saviour
#8 - The Hypothetical Mk II

If I were to build a Mk II, it would have the following changes:
  • The end columns would be less sturdy and less elaborate.
  • The beam would be stiffer – extruded aluminium would provide lightness and stiffness.
  • I’d use two strain gauges, either 1 kg or 2 kg rather than 5kg – 1 kg would probably be enough if the end supports were less bulky (as all weight in the supports takes away from the capacity of the strain gauges). 2Kg would give a better margin for stiffer arrows but it’d have half the precision.
  • I’d thread the central pillar so that it could be adjusted in height by twisting it, which would be a much better solution for adjustment.
All in all it’d look a lot more like the Bearpaw.

Funny that.
Well that appears to be a Bearpaw Analyzer, mine packed up so I opened it up to see what was inside, Atmega128, two HX711 with what appear to be 750gm load cells of the type you can get on Amazon already on a metal bar fix one end, weight the other and a 0.96" OLED I2C display. I though t it was the load cells but it actually appears to be the I2C connection to the HX711 as my logic analyser cannot make head nor tails of the signals to the HX711s.
Best of luck with the V2, but you may need another lockdown and that may not be for a long time.
 

dvd8n

Supporter
Supporter
AIUK Saviour
Well that appears to be a Bearpaw Analyzer, mine packed up so I opened it up to see what was inside, Atmega128, two HX711 with what appear to be 750gm load cells of the type you can get on Amazon already on a metal bar fix one end, weight the other and a 0.96" OLED I2C display. I though t it was the load cells but it actually appears to be the I2C connection to the HX711 as my logic analyser cannot make head nor tails of the signals to the HX711s.
Best of luck with the V2, but you may need another lockdown and that may not be for a long time.
Well, thanks for having a look inside your analyser.

I was initially surprised how similar it sounds to mine, but then, when you think about it, there aren't really that many ways that you can implement the functionality as described by Bearpaw.

I'm a little surprised that Bearpaw didn't implement a custom board with all the circuitry on it, but on the other hand with HX711s being so incredibly cheap, it maybe didn't make sense for their production volumes.

Bearpaw will win on two counts. They can probably get access to strain gauges of guaranteed quality at sensible prices as opposed to the no-name unknown quality hobbyist things that I can get from Amazon and Ebay. And they will have time and access to test materials (loads of quality arrows) to fine tune the software to give good results. Their software is almost certainly of higher quality than mine if only because they have the resources to spend time on it.

You didn't take any photos of the insides, did you?
 
Last edited:

dvd8n

Supporter
Supporter
AIUK Saviour
I though t it was the load cells but it actually appears to be the I2C connection to the HX711 as my logic analyser cannot make head nor tails of the signals to the HX711s.
By the way, I'm not sure that the interface to HX711s is IC2. It's definitely serial, but I think it's bespoke to the HX711 chips. That may be why you found the logic analyser results confusing.
 
Last edited:

dvd8n

Supporter
Supporter
AIUK Saviour
Best of luck with the V2, but you may need another lockdown and that may not be for a long time.
I've already implemented some of the V2 features, by the way. Threading the centre post ended up being vital to save my sanity when trying to calibrate it. And I had to convert it to two load cells to get repeatable results. One cell at one end should have been fine in theory but in the real world it just wasn't good enough. I used 2kg cells; I'm surprised that Bearpaw could get away with 750g ones.
 

malbro

Instinctive Archer
Supporter
AIUK Saviour
You didn't take any photos of the insides, did you?
The main tube is a 30 x 20mm rectangle, wall thickness around 2mm, 2 L shaped brackets one at each end held by two screws hidden under the foam 'feet', the fixing screws hold the bracket in place and fix the weighing bars at one end, the other end of the bar has another L bracket fixed to it, this is the arrow rest.

The weigh bars are marked 750 hence my reason for suggesting 750gms, thats 1.5kg with two, surely more than enough for arrows, even wooden ones, mine only come in aroud 450grains or about 30gms.

The main PCB is not actually fixed in place the tabs on the side of the PCB fit with slots in aluminium tube and two screws hold the plastic cover in place.

In image1 the HX711 is the black blob, its fixed in place with heat shrink, image2 is the weigh bar and image3 is the post bracket and weigh bar disassembled.
 

Attachments

Last edited:

malbro

Instinctive Archer
Supporter
AIUK Saviour
By the way, I'm not sure that the interface to HX711s is IC2. It's definitely serial, but I think it's bespoke to the HX711 chips. That may be why you found the logic analyser results confusing.
I had a look at the Arduino HX711 libraries and it is a customised version of I2C with a data pin and a clock pin, but it does not use the I2C libraries instead custom code to bit bang the lines.

Sorry if this is getting off subject for an archery forum.
 
Last edited:

malbro

Instinctive Archer
Supporter
AIUK Saviour
I've already implemented some of the V2 features, by the way. Threading the centre post ended up being vital to save my sanity when trying to calibrate it. And I had to convert it to two load cells to get repeatable results. One cell at one end should have been fine in theory but in the real world it just wasn't good enough. I used 2kg cells; I'm surprised that Bearpaw could get away with 750g ones.
From a measurement perspective a single cell is going to be prone to inaccuracies due to position of the point of force especially when measuring at the lower end of a load cells range.

The standard test is defined around a 2lb weight, which is around 900gms, that is then split between two cells so with a centre point of force it is around 450gms each, 750gm load cells sound like a reasonable margin of safety. I have purchased a couple of 1kg cells as the nearest I could get to see if I can resurrect my device, might decide on a more powerful CPU as well so I have control and write my own code.
 
Last edited:

dvd8n

Supporter
Supporter
AIUK Saviour
I had a look at the Arduino HX711 libraries and it is a customised version of I2C with a data pin and a clock pin, but it does not use the I2C libraries instead custom code to bit bang the lines.

Sorry if this is getting off subject for an archery forum.
Ah, ok. I hadn't looked into what the library was doing. Makes sense though.
 

dvd8n

Supporter
Supporter
AIUK Saviour
The standard test is defined around a 2lb weight, which is around 900gms, that is then split between two cells so with a centre point of force it is around 450gms each, 750gm load cells sound like a reasonable margin of safety.
At 28" and 1/2" depression you will, as you say, get 440g load on each cell from a 500 arrow. But remember that the 1/2” depression is used for all arrows so a 250 arrow will impart 880g force on each end, which blows the 750g limit of the cells by a fair margin. 250 arrows aren't common but they do exist. Is the depression used by the Bearpaw less than a half inch by any chance?
 
Last edited:
Top