Thanks for the thoughtful replies. A/C & D/C grounds in place... Sral, I think we've stated our positions and aren't likely to change each others minds on the points we've discussed, so I'll just leave it here. Happy to contribute further if needed.
Can reef-pi be PH controller?
- Thread starter genericfishname
- Start date
Well that very much depends on how many values one uses for the average and the same would be true for a delay of 60s. That‘s where a median of e.g. 5 values is superior for example.In theory but what happens if you have lots of bad readings or just simply messing around, the average will be way off and will take time to smoothen out. Your right about the charts being averaged, sometimes I get people wondering why the chart value is different then the raw value going to my point that averaging can lead to inaccurate values.
If the water temp rises .1 or more in 60 seconds the user might have bigger problems.60 seconds is a little on the high side for a heater, just put it as an example. I use this on something I'm working on and it works great, couldn't be without it.
Here's a nice troubleshooting guide someone sent me the other day, the last page it talks about ground loop issues.
But maybe we are also misunderstanding each other. I was talking about completely disregarding erroneous readings, so they never appear in the graph. The delayed temperature control sounds more like a special condition for the control loop behavior, which is something I can see as useful.
- Joined
- Feb 14, 2020
- Messages
- 946
- Reaction score
- 1,259
- Review score
- +0 /0 /-0
- Location
- Edmonton, Canada
Yeah I know you mentioned disregarding values all together and that's a good thing too when they are far out of range as an example for pH anything under 4.0 and 10.0 can be tossed unless a person is calibrating. Doing that would help with the averages. One could even tighten that window. The delay would strictly be for controls only to remove the hair trigger and would be optional.
That’s would be along my earlier lines, separate error mode for sensor failures and a filtering function. In this case limited to the control loop, which is probably superior to also manipulating the actual data with an average or similar.
Looking at a delay function in the way @robsworld78 described I would absolutely be in favor of it. Easiest way might be to use a setting of „number of readings“ X, meaning X consecutive readings have to continuously satisfy the condition for the control loop to trigger.
In the case of pH this would likely mean a value of only 2, right ?
Looking at a delay function in the way @robsworld78 described I would absolutely be in favor of it. Easiest way might be to use a setting of „number of readings“ X, meaning X consecutive readings have to continuously satisfy the condition for the control loop to trigger.
In the case of pH this would likely mean a value of only 2, right ?
Yes, I believe Rob's approach would work fine... Frankly, I hadn't really thought through the finer details, it was initially just a broad suggestion ...
Not to make things more complicated, but this is interesting as well .. why not machine learning? ;-)
Anomaly Detection in Machine Learning
In this article, you will read about different types of outliers and machine learning techniques that help to find anomalies. Learn how you can apply ML anomaly detection techniques to fraud prevention, medical diagnosis, and more.
serokell.io
jb
Similar threads
