Flypt Filter Research (or why doesnt my rig move right?)

Flypt Filter Research
91517 91518

Sir, can you tell me how to set it up? Flypt Filter Research

Hmmm,... not really. I have no experience with these racing rigs. I only use 6DOF Stewart platforms. What I would recommend is to experiment and play around. Initially, use only a single DOF (like 'pitch' for example) and apply a EMALP(EMAHP(VALUE)). 'Value' in this case should be 'Acceleration Longitudinal with Gravity', but play around and find what you like best.

I know, this is not a satisfying answer, but think of it like this:
It's like asking "I want to motivate people. What should I say?" The only real answer would be "Depends on what you want to motivate them for." So, the question to ask is not: "What should I say?" but rather "How do I become a compelling speaker?".

And in the context of motion sims that question would be: "How do I understand filters?" to which I would recommend: Build them yourself in Excel and look at the curves. Get a feeling for what those curves mean.

Sorry, for not being of much help here :-/

Maybe there are others who use such a rig who can share their profile. That would give you a good starting point.

Finally made a few more :-D This time in full-on 4K :-D

Here's the first one:

Hello!

Does someone know where I could find all these filters for excel?

What do I want to do /why do I want them?

I want to record the telemetry of a few laps, then export it to excel (there is where I need the filters formula) and then I should have the output...

As I'm using multiple direct modules, I want to see "live" how changing the parameters of the filters affect the graph of the motion (that I'm going to make with excel)

Is true, we have a graph viewer, but I can't analyze it and see the real change, I cant pause it and compare one graph to another...

Just need the formulas in excel and import the data...

BTW... another question...
The Multiple Direct, to calculate the output, just "adds" all the outputs of each parameter/variable? Regular addition of each value?

These are the values that I want to "play" with:



So... moving each parameter I want to understand how the "total" curve moves, and by looking at them (a static graph on time) I can make tune it to the way it should be...

Thanks!!!!

I think that's a cool idea!

Essentially all you have to do is extract the data, write it into an Excel spreadsheet, then model the filters you want to use (one collumn each) and you can plot the output graph that shows you which values your DOF would've had if you had used those filters. And then you can play around with the filters and see the result directly.

That would actually be a cool software on its own already. Even better if you integrated it into a motion cueing software.

Cheers,...

One more This time about high pass filters:

Maybe you could record and export the telemetry with different filters, than view the end result in excel.
Not better ?

If I understoood correctly, I think the idea was to export the inputs to the filter in order to be able to then go and try out different filter settings and observe in an instant how the output would've changed.

Instead of having to actually drive/fly the sim, you could try out different setting and you would see the effect directly. By "seeing" I mean in the graphs.

Cheers

Dirty, if you feel this is off-topic, please let me know to erase it...

What happened? I have an active harness (sim racing). A few months ago, I pick up a friend's track car, install a device that captures the tension of the belt, shoulders, and waist, and drive some laps. After that, for making it simple I pick up a big turn, a big breaking point and with those values, I did a ratio (for example 7:4.. this is the shoulders receive 3/4 more strength than the waist). So I build a harness that keeps that ratio when it pulls the harness (is not that specific one.. I don't remember right now).

I have some idea of the forces that are involved in the tension of the harness, longitudinal accel, lateral accel, and vertical accel (when you brake, turn, and go through a bump for example...); and of course when you have it tight, some vibrations of the car you can also feel them...

So an exclusive Mega board for the belt, a servo motor, FlyPt and... A multiple direct where I pick up all those values and apply some filters and violà...

The result is "mmmmmm"... I know what I should feel in some scenarios, I know how much force or strength I should feel and how should I feel it.... If I make a graph strength/time in a known curve, for example, I have some idea of how that graph should be..

Now, the real thing is that I really don't know how will affect changing a parameter, the graph function of Flypt does not let me for example analyze what happens to the movement of the actuator if I change a parameter because when you hit pause, all goes to 0.

So... here comes my idea...

If on one hand, I have all the formulas of the filters in excel, if I know the way multiple variables (long accel, vert accel, lat accel for example) are "joint" together to have the output of the actuator and, by the other hand I have the way to get all the telemetry (raw data of long accel, vert accel, lat accel), I could load those values to the excel sheet, apply the formulas with the parameters that I have and I would have the output to the actuator (it should be same as the actuator behaves because it is "calculated" the same way. That way I will have the behavior of the actuator in the period of time that the values of the telemetry have (I know that each value is calculated every 2ms).

Now.. that's my baseline. Then I copy/paste those same values and start changing the parameters of the filters and I could compare how the actuator would behave under the same situation. As I know how it should feel (or the actuator does under that circumstance) I could twitch the parameters of the filter or change the filter to get the behavior that I'm looking for under that scenario.... Once I'm ok with the result, apply it and for sure I will have the feedback I'm looking for.

I can apply ceteris paribus, one input at a time, for example under breaking...with abs active is different than without the abs acting and the ideal max braking force is the top of the strength I should feel, then down...

Then I can go to a turn, and then to Eau Rouge and pick up the telemetry... and at the end... I pick up a curve and break in the middle...and so on.. that way I could find a nice setup for, in this case, my belt... but this could be applied to ANYTHING...

So what do I need:
-The math of each filter, is ideal if there is a complement for excel (faster)
-how to export telemetry (recording) to excel, I have tried but it's alien to me

This is my dream, but if I don't know the math of the Flypt filters...all that we could take from the excel would be nothing because the parameters in excel would not make the same in Flypt (output)

I think is a good idea and it will help me a lot... what do you think... maybe it's a very stupid idea...or maybe impossible... but the creator of Flypt is in this forum also.. he could give us the info...

If someone gets here reading, MASSIVE THANKS for your time... please leave your opinion... and sorry for my English, not my mother language, hope I manage to explain myself in the correct way

Have a great week!!

All the book I wrote before...
Here you can read it in one paragraph...

Here's an introduction to the idea of convolution:


Essentially all it says is:

1. Any descrete-time signal can be considered as a series of impulses. The response of the filter to the signal is then just going to be the sum of the impulse responses to each individual impulse.
2. Similarly, any descrete-time signal can be considered as a series of steps. The response of the filter to the signal is then just going to be the sum of the step responses to each individual step.

I will stick to using the step response as an example, because I find it easier to imagine a situation in motion cueing where the signal makes a step, rather than imagining a series of impulses.

The video is more of the nice-to-know than the need-to-know sort

...and trying to put everthing into context. Here's the" bigger picture" of what motion cueing is and what the filters' job is.

Finally done with conversion training on the A340, so I had a few hours to spare. There's a lot in this video, so it might be a bit convoluted. I described how the "classical" filter based algorithm turns acceleration signals into translation commands. I used longitudinal acceleration --> surge in this example, but it is just as valid for lateral and vertical acceleration signals. Enjoy

I am using FlyPt Mover and trying to achieve the behavior as shown in the red curve (screenshot taken from the video in the above posting):

However, the best I can get is something like this:

This is more like the yellow curve than the red one.
Here's a screenshot of my FlyPt Mover setup:

The filter in use is DEMALP(EMAHP(VALUE;100.000);5000.000).
Am I doing something totally wrong here?

Yes, you have to use “filter ordering”. Just put a two or 3 EMALP in cascade to eradicate the entry corner

Hello. Thanks for all the great information. I was wondering if someone could help me. I built a 6DOF platform. Then I broke it due to huge shock loading after hitting the wall exiting Radillon. True story. So now I’ve rebuilt it replacing 3D printed CF nylon parts with billet machined parts. But I still have a problem with crashes sending huge shocks and tripping Odrives with over current. My dilemma is if I make the system pass high frequency so it replicates texture like kerbs and grass, how do I filter spikes from when I hit the wall? Any help would be appreciated. Thank you.

So I think I found it. Do I need to use
SPIKE, EDYNLP and LDYNLP?

With EMALP(EMALP(EMAHP(VALUE;80.000);1000.000);1000.000) I get this:

There is still a sharp corner at the beginning.