LabRadar: Great Data - Terrible Implementation. (partial solution inside)

[MLN1963]

I hope the folks that make LabRadar are made aware of this and are open to making improvements. Anything that makes their product more accurate should be welcomed. It makes their one of a kind product even more desirable.

 

[Birchbark]

Very interesting and will download the spreadsheet. Not saying bad about the company, but I do hope they look closely at what you've brought to light. Thank you for your time and effort.

 

[shinbone]

Good work entoptics!

After working with my Labradar for a few sessions, I realized that the muzzle velocity number was calculated rather than measured. I asked Labradar about the accuracy of that calculated number, was told that there is no error in that number. (note - I am just relating what I was told.)

Hopefully, the Labradar folks will see that the flaws in their data fitting algorithm are becoming know by their customers, and will come up with a fix. Keeping obviously bad data points in the curve fitting routine is tragically dumb and easily resolved.

 

[jebel]

I asked Labradar about the accuracy of that calculated number, was told that there is no error in that number.

"Calculated" and "no error" don't work together. I appreciate the research into accuracy here. I've downloaded the file and when I have a chance to look closely, I intend to. I'm a fan of the LabRadar and an even bigger fan of refining it further. Thanks.

 

[entoptics]

After working with my Labradar for a few sessions, I realized that the muzzle velocity number was calculated rather than measured. I asked Labradar about the accuracy of that calculated number, was told that there is no error in that number. (note - I am just relating what I was told.)

"Calculated" and "no error" don't work together. I appreciate the research into accuracy here. I've downloaded the file and when I have a chance to look closely, I intend to. I'm a fan of the LabRadar and an even bigger fan of refining it further. Thanks.

"No error" doesn't work with just about anything, calculated, speculated, measured, prayed, remembered, or wished.

Anyone who downloads my spreadsheet, please post your experience. I'd like to know how it works out for others, and I'll help if I can.

I'm traveling for a couple weeks, so not sure how much I'll be watching this thread.

 

[Tulsa Reiner]

Hope this is useful to someone. Based on the lukewarm reception so far, I may be wasting my time, but oh well.[/QUOTE]


Entoptics: I'm fascinated by what you've discovered, and curious about how it came about? (you don't have to answer that question.)
The only answer I can think of re why LabRadar engineers didn't do what you did is that they are doppler wonks, but don't give a **** about fairly simple statistics. I am not a "statistician," having taken only an introductory course in college, but I think I can understand what your algorithm is doing and why. This whole project of yours is just an extension of what we are all doing with velocity and BC when working up the optimal hand loads for our toys.
THANK YOU!

 

[Rover]

Great work Entoptics. Hope you continue this and please contact Labradar and get the Labradar team working on upgrading their software. I will definitely use your spreadsheet and hope that lessons learned will be implemented by Labradar.

 

[tbrice23]

I doubt that they will have any motivation to improve until they have market competition.

 

[shinbone]

I wonder if those first-collected data points would be better numbers if the Labradar was triggered by an external accelerometer trigger (i.e., not a microphone trigger), such as by Piet or JKL Precision.

The Labradar would probably be triggered a little sooner by the accelerometer, which may give the circuitry a little more time to stabilize before seeing the bullet and collecting the first data points.

 

[LongBomber]

What is the biggest variance that you have seen between the labradar data and the excel calculated velocity?

I have run mine through something much simpler running through an pivot chart, not kicking out the outlier readings, I have only done this on about twenty shot strings, with around 120 shots total, somewhere around 70% of the shots are within 1 fps, almost all within 2, and I don't remember any over 5fps. Now if I had pulled out those outliers maybe the change would be larger?

 

[Tulsa Reiner]

What is the biggest variance that you have seen between the labradar data and the excel calculated velocity?

I have run mine through something much simpler running through an pivot chart, not kicking out the outlier readings, I have only done this on about twenty shot strings, with around 120 shots total, somewhere around 70% of the shots are within 1 fps, almost all within 2, and I don't remember any over 5fps. Now if I had pulled out those outliers maybe the change would be larger?

I suspect you are right, Long Bomber

 

[entoptics]

Entoptics: I'm fascinated by what you've discovered, and curious about how it came about? (you don't have to answer that question.)...

Excel and large data sets are a big part of my living. The LabRadar's ability to generate data was one of the prime reasons I purchased one. I'm not sure what triggered me to really investigate the calculations, but it was probably one of those readings we all see, like example 1 in my first post. A ludicrous V0, with subsequent distances being more believable. With the tracking files, it was easy enough to investigate what was going on.

I wonder if those first-collected data points would be better numbers if the Labradar was triggered by an external accelerometer trigger (i.e., not a microphone trigger), such as by Piet or JKL Precision.

The Labradar would probably be triggered a little sooner by the accelerometer, which may give the circuitry a little more time to stabilize before seeing the bullet and collecting the first data points.

It's plausible, but I still think the problem is that LabRadar calculates the velocity based on a few points nearest the distance you've asked for (with V0 always being "asked for"). If it started tracking sooner, I think that will just move the artifacts in closer. It's possible it would help, but I consider that sort of solution to be treating the symptom and not the disease.

What is the biggest variance that you have seen between the labradar data and the excel calculated velocity?

See the first example in my first post. That one was off by about 4800 fps. Generally though, 9 outa 10 shots are within 5 fps of the Excel regression, and 8 outa 10 are within 2 fps.

If you take a look at my OP, you'll see a table with the last few load tests I did. I took the raw output from the LabRadar, then ran the same shots through my spreadsheet. You can see that in 4 outa 5 cases, the ES/SD of the LabRadar was more than the regressed velocities, and in some cases substantially so.

In my opinion, it's a problem with enough of the data, that for important work, I run every shot through the spreadsheet. If I'm just trying to get an idea how fast Tula 556 is going in my AR, I shoot 15 rounds and consider the average just fine. When developing a load for potential extended range work, with only 3-4 rounds per recipe, getting one bad reading by 5 fps can muck up the statistics pretty significantly.

 

[LongBomber]

See the first example in my first post. That one was off by about 4800 fps. Generally though, 9 outa 10 shots are within 5 fps of the Excel regression, and 8 outa 10 are within 2 fps.

If you take a look at my OP, you'll see a table with the last few load tests I did. I took the raw output from the LabRadar, then ran the same shots through my spreadsheet. You can see that in 4 outa 5 cases, the ES/SD of the LabRadar was more than the regressed velocities, and in some cases substantially so.

In my opinion, it's a problem with enough of the data, that for important work, I run every shot through the spreadsheet. If I'm just trying to get an idea how fast Tula 556 is going in my AR, I shoot 15 rounds and consider the average just fine. When developing a load for potential extended range work, with only 3-4 rounds per recipe, getting one bad reading by 5 fps can muck up the statistics pretty significantly.

Thats why I asked the question, I have something over a thousnad round shot past the labradar and have yet to see one with a large variance like that. Other than trying a pistol without swapping to pistol mode, that showed a 3900fps from 9mm ball ammo. Other than that I have yet to see a result that makes you go hmm.

I track pretty much all my long distance shooting past the labradar, and combine several shooting sessions into one excel file later. Your file is much more than I do, I just run it in a simple average line. Thanks for putting the file together.

Why labradar hasn't done some software upgrades and brought their app into 2019 instead of apple IIe age I just don't get. A combined ballistic app (i like the hornady 4dof) that could also grab data from the labradar and a blue tooth rangefinder and weather station would make a useful app.

 

[OutLawJoec]

I'm interested in your work here sir. I have had readings as high as 10,000 fps so yes I realized something was up and wondered until now. I will try your spread sheet and probably questions about it. I'm just starting out in the ELR phase of my shooting career and find your solution way more useful than what it being provide in the Labradar provided wonder window....