Efficiency

[micha.d]

ok, I will try to sum up some things i heard:

---

There are different criteria for optimization for different locations. So if we want a station to be optimized for certain criteria, then we need a separate value for that. This location specific value should be high when the station is optimized for it's position in the network.
That means for stations at the border of the network:

1. They should send many signals, even if only a few more strikes are detected.This helps to get the valuable new strikes. So (detected strikes) / (detected signals) should be low. Sounds strange, but that might be helpful.
   
2. In this area we do not know how many strikes are there actually and thus calculating the ratio of (detected strikes) / (actual strikes) makes no sense. It will be always close to 100%.
   
3. Stations at the border might be those, where let's say, there are not more than the magic 11 stations within 3000km.
   

... and for stations in the center of the network:

1. They should send NOT too many signals. So (detected strikes) / (detected signals) should be closer to 1. Here the graph proposed a lot of times is looking very promising. If a station is too far away regarding this, then we can assume it will not help the network at all.
   
2. As we know nearly every actual strike, here (detected strikes) / (actual strikes) makes a lot of sense. (actual strikes) should of course only be the strikes within the expected area, not all worldwide.
   
3. Stations like this have let's say more than 11 other stations within 500km.
   

This value, as we can see above, has completely contradictory optimization strategies. It is therefore usable to measure the contribution to the network, but not for the local optimizations. And the owner of each station should be aware of what to optimize for.

---

For the local optimization the criteria are different. We all cannot change the position of our detector, but what we want to achive is a good signal to noise ratio. So we need some help to find the right amplification, threshold, filter frequency or quality of the power supply. For that we should get help e.g. like this:

1. When there are repeated signals normally we have noise, even if the numbers are below the interference. That means that (sporadic signals) / (all signals) would be a good ratio. So we need to measure the sporadic signals. The first attempt that is already built into a receiver is the interference detection, but I think we can get better data there. We can measure the time difference delta_t between the signals and see if delta_t is constant several times. Also (2 x delta_t) should be detected as repeated. All those with a completely different delta_t should be sporadic. I know this requires a bit more calculation effort, but if we find a proper algorithm, then this would be a very valuable input for the auto mode.
   
2. Wheneven a signal has a lot of amplitude BEFORE the trigger point, it is very likely to be noise as well. Looking at the "bad" stations (I'm allowed to say bad because mine is one of them sometimes), I see a lot of signals like that SOMETIMES. So (bad signals with noise before trigger) / (all signals) should be low. It would be a good input for the auto mode as well.
   
3. We could measure the V_in several times in the background. If it changes with high slopes, then something could be wrong with the power supply. That was the case for me.
   
4. ONLY if a station has a lot of noise all day, then a manual change makes sense. So we should not punish someone if we don't know it's a permanent problem.
   
5. And we should mark the TEST SYSTEMs in all the graphs. When looking at some of the "real bad ones", I also found a couple of test systems, which makes sense but is not obvious in the graphs. And stations that are not maintained actively anymore (let's say the user did not change anything or login within the last 1 year) could also be excluded (not marked because we don't want to blame someone), at least if they are not performing at all.
   

All in all, most of the points are too much changing over time and depending on CURRENT noise, so it makes more sense to improve the auto mode (or manual mode with partly auto) to solve them. It also makes no sense to put everything into ONE number. Giving the numbers I proposed above could help to get an overview but is not the solution. And I guess that users, that are not ABLE to optimize their station right now, will only slightly improve. When I had the problem with the power supply, I nearly gave up. It does not help to give them bad numbers, only contructive advices help. So go for a better auto mode to solve the problems with the settings and go for real advices, e.g. also by mail, for proposed hardware changes (shielding, relocation, power supply, ...).