I will take a look at the paper. I have, all along, assumed independence of the axes which indeed allows for simplification.

@Alex Baum, ok great. BTW it's possible your 2axis subsystem already takes care of that. Ping me ("@Pete W") if you want any further discussion, I find this stuff interesting...

@PeteW As far as I know, the ship itself doesn't have any response associated with it. That is, its motion is characterizable purely through the sea state specified. In this case, sea state 5 is expected to have a maximum excursion of around 19 degrees (amplitude) over a period of 10 seconds. This translates to a maximum velocity of ~12 deg/sec, acceleration of ~7.5 deg/sec/sec.

The control object is to track a moving target. It's not just stabilization, but stabilization with the added feature of having a user dynamically adjust the pointing location.

I would be interested in how to precisely characterize the zeros and poles of this system. While I'm not new to control systems, I am new to self-stabilization.

The system's maximum velocity/acceleration is around 20 deg/sec, 20 deg/sec/sec respectively.

@AlexBaum - That's a start. You or a collaborator would need access to the actuator ("the plant" in the control loop) on land, or on the ship when it is really still. You would also need a way to measure its output (EL/AZ angles) with a meaningfully faster bandwidth than the actuator's own response.

* If possible, collect data for sinusoidal input, with small amplitude.

* One input at a time, measuring both outputs.

* From this, extract gains (amplitude ratio of each out/in) and phases.

* It might be useful to check for nonlinearity, like if the output shape is not a pure sine wave (can just eyeball it)

* if visibly nonlinear, reduce amplitude.

* When collecting data, if you have limited time to do it, I'd vary the frequencies logarithmically by factors of sqrt(10). Then graph it like a bode plot, and go back and measure more frequencies near the breakpoints.

....Then... If the sensor you will use in you loop doesn't have significantly faster bandwidth than the actuatior, it would also be fairly important to do all the same stuff for the sensor. You might need to get creative to generate a reference measurement with significantly faster bandwidth than the sensor, for characterizing it

What it sounds like you're suggesting is characterizing the system's bandwidth.

yes, constructing a bode plot

not just a single figure

Does this sound familiar at all?

...Anyway, to finish this stream of thought... there is the hull movement. Hopefully you already have data for this.

Do an FFT , to get a spectrum of each E/A angle's amplitude vs frequency. Hopefully it's possible to ignore their coupling here. It might be interesting to see if this spectrum is different for the same sea state, but different speed of boat. (I suspect it might be)

Of course. This is what I'm working towards right now. I'm going to create a test signal with constant amplitude and a frequency which increases over time. Pass that signal into the system and record the output. I can get phase data by comparing the resultant phase of the system with each successive sine in the test pattern, and check amplitude of sine waves from the system versus the constant amplitude.

I have no boat data, only sea state data.

I have a strict specification to make it within the target sea state, but no practical data. I am hoping that by testing it under worst-case sea state it meets specification all around.

Ok. Lets back up a sec to the test pattern.

You're looking at a frequency sweep. That should be excellent, as long as it is fairly slow.

What kind of "sea state data" do you have?

Yes, a frequency sweep from low to high. This will allow me to map almost directly to a bode plot

👍

This is part of the "either meet the spec or say why I cannot based on the hardware"

In terms of sea state data all I have is amplitude and period.

These are 19 degrees and 10 seconds, respectively.

So expect 38 degrees pp over a 10 second period.

Ok. To me, that sounds like a sort-of worst case bounding thing, like a generic specification that was maybe meant for other purposes

It's more like: "Here's the hardware, stabilize this as best you can, and here's what the stabilized output must look like in order for it to be in spec, or have a shot at working"

Like what happens if the boat goes thru a giant wave? Is there no reaction at anything <0.1Hz at all?

There absolutely is a reaction under this circumstance. I have access to Euler angles of the IMU.

Slow waves are easy. The error increases in direct proportion to the amplitude of the disturbance.

So in my mind, there would be some spectrum that represents the worst-case scenario that would challenge your controller. I am guessing on this, and it would depend on the mass of the boat and the waves, but if it had some amplitude at, say, 1Hz, even if that amplitude was a lot smaller than the 0.1Hz amplitude, it might be just as important when designing the control loop

This may be something that could be researched, too.

Anyway, here's what you would do with that spectrum: it would tell you how much attenuation the control loop response must have vs frequency

or rather noise-rejection, to be more specific

Yes, that was a second idea that I had... but perhaps it would help if I had some way of giving you the data so that you could see what this looks like.

Sounds good. If I may ask, what kind of project is this?

It's a EL over AZ pointing system.

I mean, is it commercial? academic? military?

It is commercial

Ok, was just curious. Wasn't sure what level of involvement you were looking for, if any. I can give you my email, but will delete it shortly after so it doesn't stay on here

That sounds good

please let me know when you get this so I can delete it.

Copied

Ok apparently I can't delete it after some time limit.

Anyway, feel free to email me.

I have