Repository navigation
Conversation
Add optional bypass_min_move for the next guide-image update. Carry the override with one ordinary guide move without changing configured MinMo values. Keep the setter immediate and the original behavior when the flag is omitted. Repeated requests at the same coordinates rearm the override without restarting lock-change handling.
|
I don't see any sense in this, is this just a theoretical problem? Let's say you do this and you end up with a zero displacement from the lock point. In the very next guide camera frame, the target can have moved off target due to seeing conditions and various mechanical errors, At that point, no correction will be issued less than the MinMove and you will no longer have a zero displacement. What image scale are you using? Is you need this kind of precision, you need to be guiding at a fine image scale, probably less than 1 arc-sec/px. MinMove values are expected to be at or below the typical size of the seeing fluctuations and can be as low as 0.1 px. I don't think we should pursue this. |
Thanks for the feedback, but I believe there is a fundamental difference between steady-state seeing rejection and external deterministic positioning (e.g., fiber/slit injection). 1. DC Systematic Offset vs. AC Seeing JitterSeeing fluctuations are zero-mean stochastic noise (
2. Why not just lower
|
|
I'm not willing to spend more time trying to educate your Chatbot. You can just pursue this using a private, unsupported version of PHD2 that you build yourself. |
Crying chatbot is just an easy excuse to run away when you can not defend your bad logic. You can not even tell the difference between random seeing shake and a real tracking error when an instrument needs a precise locking position, and you clearly do not understand slit physics or photon loss. Do not talk about educating anyone. Go back to your backyard and take pretty pictures for blurry desktop wallpapers. Keep your toy software for amateur eye candy. Real instruments will run on a real fork. Forked and closed. |
|
@Atlas8m wrote:
@Atlas8m I see the problem as failure to personally and adequately explain the theory and reasoning behind your submission in the description of the PR. The only thing that is explained is the code change itself while the reasoning for it is given only a very weak gesture in the Motivation section. Only after being prompted to did you supply an obviously agent-generated word-salad explanation of the whys and hows. "Zero-mean stochastic noise?" "Systemic DC bias?" Seriously? Nobody real talks like that outside of dry papers, and tossing back an agent-generated word salad is rather disrespectful. The PR would be far more believable and the motivations for it far more understandable if you could advocate for the change and explain its theory in your own words. After all, if you can't explain the change you're offering, why should anyone else take time out of their day to divine that explanation themselves? This leads anyone who is a competent developer to conclude:
On top of that, you seem to be unfamiliar with basic development norms and etiquette. That can catch up to you. |
@daleghent Let me make this as simple and plain as possible so nobody cries about "AI words" again.
I truly respect the original PHD2 authors for building and maintaining this open-source tool for so many years. That respect does not extend to this kind of gatekeeping lecture. Honestly, I am glad this happened. It broke my inertia of trying to patch an amateur tool, and now my own headless guiding system is running cleanly on Windows without needing anyone's permission. Perhaps I will open source my whole software bundle once they are mature enough after completing multiple rounds of Hardware-in-Loop testing with real gears and real observing condition. Done here.
|
You neglected to provide any supporting materials to support your proposal. This would be a non-starter even before the age of AI. There were no logs or anything of substantive nature to demonstrate the change has benefits in a context that extends to more than just yourself and your specific operational circumstances. Merely showing that the code compiles does and the boolean gets fiddled via API call does not constitute adequate testing in any setting. You're introducing an option that affects guiding methodologies and performance and may have unintended ramifications or effect on systems other than your own, so it should have been obvious that you will need to supply evidence to back it up. How you came across is that it seems you presumed your change would be unquestionably accepted with the paltry amount of information submitted with it, and are annoyed that you were asked to further explain it. Your response was to throw an agent-generated word salad back in the face of the requestor, as if you were annoyed at being asked to do so. It's not a genuine act, to start, and it's seen as disrespectful. Bruce asked for your explanation, not an LLM's. Bruce could have just prompted any AI himself if he wanted an AI-based breakdown of the change. But he asked for yours. That's not gatekeeping. It's his project after all and he has to answer for everything that goes into it. Then there's your rather unprofessional personal mocking at the end. It's good that you're here under an alias rather than presenting in any professional capacity. Speaking of which, since you claim to work in the research space, it would be good to supply your account with your ORCiD code. As for, Meatproxy, it's now part of the engineering lexicon and is wholly descriptive of a process. Recruiters even use it now, usually to explain what they're not looking for. |
@daleghent Let me address your points one by one and wrap this up:
It is pretty funny watching someone preach about opensource "etiquette" while demanding personal real world info on a closed GitHub PR just to satisfy their curiosity. Whatever happened to respecting privacy? Especially when you still have not addressed a single technical point about slit photon loss, decoupled headless architecture, or why a one-shot optional RPC flag needs a video recording when I already have full SIL simulation results. I have said everything I need to say. Finally, I don't want to keep arguing with you about how to be polite or your HR style words. Now I just feel that amateur astronomy (like those using NINA) and serious astronomy are two different things, especially when talking to someone who does not understand spectroscopy requirements. Terms like slit, IFU, random error and systematic error are basic concepts in astronomical instrumentation. Anyway, I am open to technical discussions, not your political correctness. |
The saying goes, "On the Internet, nobody knows you're a dog." Which is to say that the veil of anonymity can be a double-edged sword. Given the lack of any other clues and given the context of the current times, something that presents itself like an AI agent and talks like an AI agent can be easily assumed to be an AI agent. And if someone in the flesh is being asked to engage in an unsolicited conversation, it's still pretty safe to assume that they are expecting to have it with something that plausibly passes as another real person and not some automata. Whether you just copying+pasted your responses from an agent or actually wrote it out with "AI polish" done afterwards, is an aspect that is neither here nor there. Google Translate works just as well now as it did before Claude. Documentation helps explain a feature in more plain terms and submitting it contemporaneously ensures it's not forgotten about and leaving it up to someone else to figure out and (hopefully) correctly document when it's realized to be missing. As far as PHD2 contributions go, I've done that myself in furnishing both new API functions and the documentation for them. It shows you're aware of what's required to help ensure the success of your contribution for users and that you're more invested things and not just trying to slam your own code in.
No, you don't get to hide behind this. The open mockery and outright personal attack you exhibited in your response to Bruce had all the echoes of a disappointed adolescent who is unaware of how they present themselves, not of a working professional. It was pretty vicious. I normally would have passed over this kind of PR discourse but your comment alone made me stop to register my disappointment. You don't seem to think it warrants any apology or contrition, so I made sure to quote you on it so that it sticks around. It sure doesn't help your case and deflecting won't work.
I never mentioned ROR ID, which is completely different than an ORCID. You ought to know that RORs identify institutions and ORCiDs identify individuals. An ORCiD is required for each author that participates in research that is sponsored by any major national and EU research funding arm, as well as thousands of journals in order to submit. If you don't have one now, you will certainly find yourself needing one in the future - especially if you reference your own code in a paper. GitHub enables one to be associated with one's account.
You can engage in all these things, but if you're doing it through a proxy, people who you're trying to get the attention of will tend to want to know who or what they're dealing with. If it talks like AI and doesn't look like a real human, it likely is an AI. This is something you should understand in an age were open source developers in particular often find themselves dealing with PRs and issues being opened by Agents with no known provenance, affiliation, or (in some cases) clear motive.
What Bruce (and everyone else) wants to understand is why you think your method of attacking the problem is universally viable, with demonstrative proof beyond closed-environment simulations. Saying that the code compiles and the simulators seem ok with it just says to the observer that the code has never left your desktop and has not been used in real-world situations, let alone across an acceptably diverse population of mounts. A change like this has to work for more people than just "Atlas8m on some unknown telescope".
I, personally, understand all these things. I'm part of a team that does speckle interferometry on the 100" and 60" telescopes at Mt. Wilson - storied instruments, as you might know. Getting a fibre-fed spectroscope added to our existing instrument package is one of our upcoming goals so that we can also measure radial velocity in addition to the theta and rho of the subordinate star. This is so we can start filling in data on double stars that even Gaia was not able to resolve, which can include placing the secondary on the H-R diagram. A tough problem to solve for us is that the control systems for these telescopes are electric with encoders but manually directed, so we have to devise a secondary mechanical system to position target double star beyond what the telescope's own fine movement gears are able to perform. So, yes, I actually do understand spectroscopy requirements. Consider putting a check on the hot-shot attitude you're displaying because you might not be fully aware of the company you're in, or who else might be silently watching here. I understand you recognize me as being affiliated with NINA and you're wasting no opportunity to get an amateur vs. professional dig in on that as well. We're also running our speckle sessions using NINA, just so you're aware. If you want to exude an aura of professionalism, you're doing everything but that. |
As an engineer, you don't even understand basic mutual respect. Look back at the thread and ask yourself who started demeaning the other side and provoked this whole argument in the first place, dismissing a valid requirement on sight and calling me a You wrote four long paragraphs about ORCIDs, Mt. Wilson, and veiled threats about Doing millisecond speckle interferometry on a 100 year old mount with NINA is completely different from continuous closed loop guiding. One takes short lucky imaging exposures, the other holds a target inside an IFU for 15 minutes or longer, requiring no deadband drift, and, has to do it fully automatically across a large scale high resolution spectroscopic survey. The PR is closed, keep your legacy code, and I'm done wasting time here. If you want to keep arguing instead of having a constructive discussion about instrumentation, maybe pointing a bot at this thread really would be a better use of my time. Anyway, I have my own custom RC mirror set to fix and mount, an echelle spectrograph to build, and direct-drive mount control boards and firmware to finish, wasting time on this just ruins my mood. You owe me 20000 bucks for all the engineering time you've wasted :) |





Summary
Add an optional boolean parameter
bypass_min_move(defaultfalse)to the
set_lock_positionRPC method.Motivation
Automated target acquisition, such as placing a star on a spectrograph
slit or fiber feed, can require a final adjustment smaller than the MinMo
used for normal guiding. The deadband can prevent that adjustment.
This option allows the controller to request it without changing the
saved MinMo settings.
Changes
bypass_min_moveparameter toset_lock_position.If that frame cannot produce a correction, another request is needed.
bypass_min_move: truerequiresexact: trueand active,unpaused guiding.
Other guiding rules and pulse-duration limits still apply.
Example
{ "method": "set_lock_position", "params": { "x": 120.5, "y": 240.5, "exact": true, "bypass_min_move": true }, "id": 1 }Testing
enabled, followed by a normal call that obeyed MinMo again.
a 1-pixel lock change produced one sub-MinMo correction, followed by
normal MinMo behavior. Repeating the request allowed another correction.
Saved algorithm settings were unchanged.
parameters, and rejection of invalid bypass arguments.