The Logic of Autolearn

[Monoxcide]

I've been annoying Escort through emails trying to learn the exact logic of autolearn. They won't disclose alot of details but I thought I would share what I have discovered.

1) "The auto learn needs to see the same signal three times in the same place before it locks it out. If the signal changes because the source has gone out of calibration, or if there is another signal present, it won't lock it out, to prevent the chance of locking out a genuine police signal. It does not block out all bands in a location."

Looking at that quote it looks perfect. It blocks a signal, if it sees a second signal on the same band it alerts, regardless of frequency. If the cop was using the same frequency as the door his readings would be inaccurate. Yet they would not answer directly if the detector would alert to the added signal. They said it would not be stored in memory but I'm assuming that also means the warning will happen.

2) I asked the question why I can drive by an autolearned location with the signal source off and when it comes back the next day it instantly blocks it. This was the response.

"It will delete an autolearned location if there is none there; but if it comes back, it will remember it."

This is what scares me. Say you pass the same cop, same spot and same frequency 3 times in a row. Autolearn would naturally lock that signal out. The next time he is not there. Autolearn deletes that location but still keeps it on 2 of 3 strikes. If it sees the signal again, it will instantly without a "Stored" warning, filter the signal. I know this situation is extremely unlikely but it can exist.

I'm not a engineer but I think the logic of autolearn would be fairly simple to implement. The detector has the ability to determine the exact frequency, so if it sees that exact frequency 3 times it blocks it in the logged location. The only problem I see dependent on how accurate the detector is. If it estimates the signal, or only view it in the 30mhz blocks this could be why the false lockouts are happening.

I just thought I would share this information to help more people decide if autolearn is right for their situation. For now I'm keeping it on but leaving the detector in expert mode for the multiple signals alert. I do not use the manual lockout function.

Feel free to correct me if anything is inaccurate.

[SlapMyFro]

"It will delete an autolearned location if there is none there; but if it comes back, it will remember it."

Are you sure about that? if thats the case, I am assuming is pretty smart, just in case you lock out a real threat.

[Monoxcide]

Yeah its in the 9500ix manual and that is a direct quote from an escort support email.

[AirMoore]

The 9500ix does not have the ability to recognize EXACT frequencies... Belscort models are widely known to fluctuate in this regard, so they do implement and essentially 'lock out' blocks of the frequency.


I believe Jim did tests by putting his RX65 in the fridge (freezer)? in the name of science. He found that essentially as the unit 'warms up' or goes through 'cycles' in one respect: It sees different frequencies... most (all?) a bit off from the actual frequency.


This is a widely know critique of the system.



Most of what you have said is commonly held material (though I do thank you for putting it together)... The autolearn/lock feature makes me a bit nervous as well. Its something that is still a 'budding' technology,and it seems as if Belscort is responsive to some of the critiques of the units that we talk about on here.


There aren't very many features on any detector that I know of that don't normally involve some margin of risk/error/cost-to-benefit issues.

[SlapMyFro]

Yeah auto learn is a bit risky in my opinion.
After all it doesn't take long to lock out manually all the locations that you are aware of.
Unless you live midtown manhattan which in that case you cant have a rd on period

[Monoxcide]

I know it is common knowledge but the biggest thing I wanted to point out is when autolearn stores the frequency block its going to reject it will never unlock that area unless you manually do it. The manual reads as if it sees something 3 times, locks it out and then doesn't see it, it is 100% deleted. Meaning you would have to see it 3 more times for another lockout. That is not true, if it deletes it but sees that block again it is instantly rejected, without the "Stored" text and beep when something is first locked out.

I also assume it can have the same ghosting issues if it blocks out a very strong signal, which it will. If anything, its a Truelock that you have zero control over. In my opinion it may be smarter in locking out some situations but it is no better if not worse than manual Truelock after reading the sticky in the forum.

With more testing today I'm using autolearn in heavy false areas and watching my speed. On the highway or aggressive driving, GPS filtering gets turned off. Thank god the silent 2 beeps still functions under 20mph with the filtering disabled.

[mswlogo]

I agree it seems a bit odd/dangerous (and new information) that if it does not see it it only removes 1 hit. Meaning a single hit by a real threat would get locked out. Even if they removed at least 2 hits I think it would be much safer.

I always read it that it would have to see it 3 times again.

I'm not hugely concerned. I'm sure they will tweak it over time.

Perhaps they meant it removes a hit for EACH miss. That wouldn't be so bad, in fact I think it would be great. It's just a hit counter (up down on hit vs miss) and when it's 3 or above it locks out and below 3 it does not. And it maxes out at 3.

[Cips]

...

1) "The auto learn needs to see the same signal three times in the same place before it locks it out. If the signal changes because the source has gone out of calibration, or if there is another signal present, it won't lock it out, to prevent the chance of locking out a genuine police signal. It does not block out all bands in a location."

Looking at that quote it looks perfect. It blocks a signal, if it sees a second signal on the same band it alerts, regardless of frequency. If the cop was using the same frequency as the door his readings would be inaccurate. Yet they would not answer directly if the detector would alert to the added signal. They said it would not be stored in memory but I'm assuming that also means the warning will happen.

...

I drive by an area everyday that has anywhere from 4-6 falses. Some on K, at least three on X. How would AutoLearn treat that atrea? Will it not lock anything out because there's more than one signal present?