Altimeter Inaccuracies?

I am having an issue that I have been trying to solve for the last 100 jumps or so. Every time I pull my altimeter typically says it takes 800 – 1200’ to deploy my main. This is how I obtain that data:

- Lock onto altimeter at 4000’ with hand on PC handle
- Pull handle at 3500’
- Look at altimeter as soon as slider is down, subtract the difference in altitudes

I’ve done that almost 100 times now and the readings are consistently 800 – 1200’. More recently I started filming my openings. I have filmed my openings about 10 times and from the time that the PC is out of the mesh pocket until the time the slider is down is 5 seconds on average. The canopy only snivels for 4 seconds on average with 1 second in freefall. . I normally jump with a Viso 2+, I have borrowed other Visos to try and I have even tried an old school analog Galaxy. They all say 800 – 1200’ despite the fact that my videos clearly show 5 seconds total, 4 seconds snivel time which means a more realistic opening distance of maybe 500’

This is an issue I cant figure out. I ran it by L&B and the only answer I got was that I can mail my altimeter in for a refund if I want… Very useful.

What canopy (size and model)?

Short answer, use a GPS to double check your readings, if it agrees with the altimeters, your math is wrong (likely how much the snivel is actually slowing you down in the beginning), or your vertical speed is much faster than you think (I'm a big guy and my relaxed arch is about 140mph).

There could be differences between freefall and canopy pressure due to airspeed moving around the altimeter or burbling the altimeter. For instance, AADs are set to fire slightly offset to make up for the burble they are in (like ~+300ft If I remember right). Also as an example, I've used a AltiForce GoPro altimeter (embeds altitude data in the video) when you watch the video, in freefall there is variation in the altitude, it even goes up sometimes, but under canopy it mellows out and agrees with other altimeters on the ground and in the plane. They just aren't perfect systems for ascending or descending rapidly; but more than "good enough" for our purposes.

Which brings me to my next point, it doesn't matter beyond a "curiosity."

From personal non-empirical data experience, I'm under the impression that a "brisk" deployment is <=500ft, "normal" is 500-700, and "long" is 800-1200. I've even had a couple 1,500+footers.

If the GPS doesn't agree with the altimeters, my guess would be the altimeters are being burbled a little in freefall (or there is a slight delay in their sample rate), and are telling you that you are a couple hundred feet higher than you actually are at PC throw.

To clarify the AAD thing, set to fire ~+300 ft because if you're on your back vs. belly, there will be a difference in altitude reading (higher on belly, lower on back at the same AGL) so they want them to fire in any position by (say 800ft Pro Setting) so if it fires at exactly 800ft on your back, fine, but if it waits too long when you're on your belly (because it thinks it is a couple hundred feet higher) then there is more potential for a splat (discounting the difference in quality of deployment on your back of course) but as we all know, AADs aren't perfect either.

This is a neat little graph I put together playing with the Altiforce a long time ago (it also has a G sensor) anyway, it isn't perfect, but lets you see that deceleration during deployment isn't linear.

Assuming the device is perfect (it probably isn't), the collection method is not. Helmet mounted (so head movement) and I aggregated the G force in all 3 axises (since precise head position is unknown). The results though were very consistent in the shape of the deceleration curve from jump to jump.

Parts of the curve were matched with the video. First tiny bump is PC inflation/bag extraction, second bump is line stretch, big bump is inflation, artifacts after are canopy surge/turn.

EDIT: Second chart is aggregate G force experienced, or "total deceleration" expressed as a % (it took several more seconds to return to 1 G after canopy inflation which it why it doesn't reach 100% in 5 seconds). Again, not perfect by any means, but the thing to consider is that in the 1st whole second about 15%, 2rd whole second 35%, and 3rd whole second, 65%. So deceleration doubles almost every second, you get almost as much in the 3rd whole second as you did in the 1st 2nd second combined. i.e. you're retaining speed longer that a linear deceleration and most of the actual slowing down occurs during inflation, not the snivel.

Also, yes I threw this together really fast in excel, it isn't perfect, and the data collection isn't perfect. Just ball parking due to nerd curiosity.

I think that digital altimeters likely do some "smoothing" of the chaotic/noisy data that is their input. This smoothing means that they don't respond as quickly compared to no smoothing, and their accuracy could be expected to suffer especially during changes in speed.

Without smoothing, we would see a lot more back and forth of the digits as we climb, but I rarely see, for instance, my alti reflect the noise that is of course present in the pressure input.

Your analysis of time based on video - how many frames/frame rate has got to be quite accurate.

Agreed about the alti's and time calculation, and without a back facing camera would be very tricky.

Also, if anyone cares enough to play with it or wants to make a decent graph I can send you raw CSV files from some jumps (you'll see the wonky altitude readings in freefall, the AltiForce is not buffered or smoothed).

which gadget has the accelerometer/G sensor? I don't recognize Altiforce, and it isn't on the L&B website

Here is the link

It is it's own thing, plugs into a GoPro, creates a CSV file that auto plays with the video (as a "subtitle") or you can export the raw data into Excel or whatever you like.

Looks like they have gotten more sophisticated since the first generation (what I have) but the basic model is the same price ($60).

Can you tell us what kind of performance and results you expect ?

Basically his argument is that 5 seconds from pitch to inflation shouldn't = a 1,000ft loss in altitude, as that would be an average of 136mph vertical. I saw a video, good camera angle, deployment is at most 5 seconds (have to take his word on the altitude readings though). He thinks (reasonably so) it should be more like 500ft rather than 1,000ft.

It is an interesting point and either the snivel is near terminal the whole time (he doesn't think so), or the altimeters (analog & digital) are giving false readings, likely being burbled or something. Or I guess some other weird atmospheric phenomenon at his DZ, but that is getting pretty speculative and almost impossible to verify w/o a plane and some fancy equipment.

Only way to really verify if it is speed of snivel vs. altimeters being wonky would be to use a GPS like an FlySight that is indifferent to barometric pressure.

I wear a Viso 2+ on my lower forearm during tandems with handcam. I'll routinely stare at it during the end of the freefall portion just for the hell of it. Some jumps it seems to smoothly scroll altitude readings, while other jumps it will pause at certain altitudes, maybe even click up 100ft, then rapidly descend to catch up with actual altitude.

As someone who works daily with very expensive airplane test equipment, I can tell you it takes a lot of money and high quality parts to get the accuracy you're expecting. These things are measuring extremely small changes in pressure which are changing at high rates and very susceptible to position and atmospheric properties, it's just the nature of it.

I'm sure someone could make an altimeter for a few thousand dollars that would meet the accuracy you're expecting, but it wouldn't sell well. For a couple hundred dollars, it's going to be "good enough" for altitude awareness, but it's not going to give you good raw data for analysis.

OldGregg

This is a neat little graph I put together playing with the Altiforce a long time ago (it also has a G sensor) anyway, it isn't perfect, but lets you see that deceleration during deployment isn't linear.

Assuming the device is perfect (it probably isn't), the collection method is not. Helmet mounted (so head movement) and I aggregated the G force in all 3 axises (since precise head position is unknown). The results though were very consistent in the shape of the deceleration curve from jump to jump.

Parts of the curve were matched with the video. First tiny bump is PC inflation/bag extraction, second bump is line stretch, big bump is inflation, artifacts after are canopy surge/turn.

EDIT: Second chart is aggregate G force experienced, or "total deceleration" expressed as a % (it took several more seconds to return to 1 G after canopy inflation which it why it doesn't reach 100% in 5 seconds). Again, not perfect by any means, but the thing to consider is that in the 1st whole second about 15%, 2rd whole second 35%, and 3rd whole second, 65%. So deceleration doubles almost every second, you get almost as much in the 3rd whole second as you did in the 1st 2nd second combined. i.e. you're retaining speed longer that a linear deceleration and most of the actual slowing down occurs during inflation, not the snivel.

Also, yes I threw this together really fast in excel, it isn't perfect, and the data collection isn't perfect. Just ball parking due to nerd curiosity.

This is the kind of thing that makes nerd girls like me happy.

If you have a Vigil and have access to the Vigil interface, you can download on a computer the last 16 minutes of your last jumps, therefore at least 10 jumps. The download gives you the actual graphs of those 10 jumps. Those graphs show the curves, vertical distance versus time. If you print those graphs, with a rule you can figure out the end of the straight part of the free fall just before the deceleration part (a curve). When the deceleration is over, note with a rule the begining of another straight line with a slope less than the free fall one. It is the part of your canopy descent.

Report on the left side of the graph on the vertical axis (vertical distance) both the end of the free fall part and the begining of the second straight line (descent).

The distance between the 2 points marked on the vertical axis (graduated in feet), indicates the actual opening vertical distance.

If you have a Protrack and its interface, you can do the same exercice by downloading the data of your last 10 jumps including the graphs.

I did it and can tell you that my former canopy (a Sabre 2-170) had an average (out of 10 jumps) opening vertical distance of 384 feet while my Katana 170 has an average (out of 10 jumps) of opening vertical distance of 600 feet (way smoother).

IMO this method using graphs of basic physics the most accurate.

erdnarob

If you have a Vigil and have access to the Vigil interface, you can download on a computer the last 16 minutes of your last jumps, therefore at least 10 jumps. The download gives you the actual graphs of those 10 jumps. Those graphs show the curves, vertical distance versus time. If you print those graphs, with a rule you can figure out the end of the straight part of the free fall just before the deceleration part (a curve). When the deceleration is over, note with a rule the begining of another straight line with a slope less than the free fall one. It is the part of your canopy descent.

Report on the left side of the graph on the vertical axis (vertical distance) both the end of the free fall part and the begining of the second straight line (descent).

The distance between the 2 points marked on the vertical axis (graduated in feet), indicates the actual opening vertical distance.

If you have a Protrack and its interface, you can do the same exercice by downloading the data of your last 10 jumps including the graphs.

I did it and can tell you that my former canopy (a Sabre 2-170) had an average (out of 10 jumps) opening vertical distance of 384 feet while my Katana 170 has an average (out of 10 jumps) of opening vertical distance of 600 feet (way smoother).

IMO this method using graphs of basic physics the most accurate.

I have a Mars M2. Also, I dont want to repack the reserve just to open it to get to the interface. :(



Anyway, here is a video of some of the openings. On all the jumps I threw the PC at 4000' (had my hand on it and pulled it right as the Viso said 4.00). Both the Viso and AAD reported deployment altitudes around 3200' and I was around 2800 - 3000' by the time the slider came down. However, that's impossible as if you count the frames in the video, the entire deployment process takes 5 seconds or less on every deployment. So on one hand I have the Viso and AAD which very dependably report very similar deployment altitudes on ever jump and on the other hand we have basic math which says a 5 second opening cannot take 1000' to open--it's impossible unless I was doing 190 MPH+ when I opened (which I wasent).

https://youtu.be/gVSTEILsIMM

Full Disclosure: I am working on a new AAD..

This discussion made me take a look at the AAD altimeter readings from the point where the deployment is detected and when the vertical descent rate is <10 meter per second on one of the test jumps.

I am currently jumping a Vector 3 with a Stiletto 120 ..

After looking at the data, the time from deployment to <10mps is 3.5 seconds, with an altitude loss of 229Ft.

From Deployment detection to Canopy Detection was 6.9 seconds.

Very interesting topic..

df8m1

Full Disclosure: I am working on a new AAD..

This discussion made me take a look at the AAD altimeter readings from the point where the deployment is detected and when the vertical descent rate is <10 meter per second on one of the test jumps.

I am currently jumping a Vector 3 with a Stiletto 120 ..

After looking at the data, the time from deployment to <10mps is 3.5 seconds, with an altitude loss of 229Ft.

From Deployment detection to Canopy Detection was 6.9 seconds.

Very interesting topic..

Please clarify the distinction between deployment and deployment detection, and how canopy detection differs from <10mps.

sundevil777

***Full Disclosure: I am working on a new AAD..

This discussion made me take a look at the AAD altimeter readings from the point where the deployment is detected and when the vertical descent rate is <10 meter per second on one of the test jumps.

I am currently jumping a Vector 3 with a Stiletto 120 ..

After looking at the data, the time from deployment to <10mps is 3.5 seconds, with an altitude loss of 229Ft.

From Deployment detection to Canopy Detection was 6.9 seconds.

Very interesting topic..

Please clarify the distinction between deployment and deployment detection, and how canopy detection differs from <10mps.

I am determining the "Start of Deployment" by analyzing the jump data. The "Deployment Detection" is done by the Deployment Detection algorithm, which identifies when a deployment has started.

The average time between when the data first indicates a deployment and when the Deployment Detection algorithm identifies the deployment is around 1.2 seconds.

The OP was using the AAD and altimeter identified deployment altitude and what the altitude was when his canopy was open, so I did as well.

How does Canopy Detection differ from <10mps?

They are two separate things. The OP was questioning the vertical distance from when the AAD / Altimeter said he deployed, to when he had slowed down under an open canopy. I chose <10mps for that point because my canopy can take a little bit to settle in so to speak lol..

The Canopy Detection algorithm decision is based on more than descent rate, and because the initial canopy flight characteristics are inconsistent and can be aggressive, the Canopy Detection algorithm requires more time to make the call. Usually between 7 to 10 seconds after the Deployment Detection algorithm has identified a deployment has started. Long after the canopy has settled down and is flying stable.

The purpose of the Canopy Detection is to prevent firing while under canopy, so it is important that the canopy be truly flying and not malfunctioning.

Does that answer your question?

Westerly

Anyway, here is a video of some of the openings. On all the jumps I threw the PC at 4000' (had my hand on it and pulled it right as the Viso said 4.00). Both the Viso and AAD reported deployment altitudes around 3200' and I was around 2800 - 3000' by the time the slider came down. However, that's impossible as if you count the frames in the video, the entire deployment process takes 5 seconds or less on every deployment. So on one hand I have the Viso and AAD which very dependably report very similar deployment altitudes on ever jump and on the other hand we have basic math which says a 5 second opening cannot take 1000' to open--it's impossible unless I was doing 190 MPH+ when I opened (which I wasent).

Do not underestimate the time between "seeing" 4000 ft at the altimeter," tossing" the pilot chute and having an open container.
You see an open canopy, the altimeter detects an airspeed below a certain threshold. These two actions do not happen at the same moment.
Just after the opening, the canopy is flying and descending faster than normal and is still slowing down.
You are using different start and endpoints to calculate or to measure the opening.