PePiPoo Helping the motorist get justice

Welcome Guest ( Log In | Register )

LTI 20-20 self calibration - how does it work?
potkettleblack
post Mon, 27 Nov 2006 - 00:22
Post #1


Member
Group Icon

Group: Life Member
Posts: 448
Joined: 6 Nov 2006
Member No.: 8,772



Techie time. I've been through some of LTI's patents and can report that what they call 'self calibrating' is actually just the basic operation of the device. What is does is turn a measuring device that would be completely useless, into one that just requires normal calibration, weekly, monthly, annualy depending upon how accurate you want it to be. I will try to explain below.

The basic design of the machine is to measure distance. Your speed is calculated from the change in distance to you over time, so a bunch of distance measurements are used. Let's consider for now just a single measurement of distance.

Speed x Time = Distance eg: 10mph for 10hours = 100 miles.
The LTI has a good reference for speed, that of light through air (approx 300,000km/s).
The LTI has a good reference for time, a quartz crystal.

So on the face of it, the LTI can produce an accurate distance measurement. But there is a problem: light is too quick. Say the target is 50m away: it takes light only 0.33us (millionths of a second) to reach it and return. The crystal reference (in the patent) was producing pulses every 0.125us. That's like trying to measure 103 seconds using a digital clock with just hours and minutes. No accuracy there. What to do?

Let's use an F1 pit stop analogy. Let's say a car fills up in 8s and then goes on to race for 40 minutes. Next pit stop, we don't see how long it takes to fill, but it races for 30 minutes. We can infer that it filled up for 6s: each 2s of fuel lets it race for 10 mins.

This is how the LTI 20-20 turns the fast flight time of the laser pulse into something slow that it can measure. It fills up a capacitor with electrons during the flight time, then empties it at a slow rate in order to measure a longer but proportional time.

So what about the much vaunted self-calibration? Well the LTI fills the capacitor for a known short time, then measures the corresponding longer time. It can do this because although it can't measure a short time accurately, it can generate one. In the analogy above, with a digital clock you could accurately set an accurate time so long as it's a whole number of minutes.

In detail, there is a zero reference discharge time from one fixed voltage to another. Then the capacitor is charged for the calibration period and the new discharge time measured. Finally the laser flight time sets the charge period and the flight discharge time is measured. This is from patent US6445444:
Page 9 Fig. 7A: two reference voltages, decay V1 to V2 and measure time. Zero time = T3-T0.
Fig 7B. Bump up the voltage (after a known delay) and decay to V2. Cal time = total - Tzero.
Fig 7C. Pulse bumped up by flight time T12 to T13. Laser time = end time T14 - start T10.
Go to the top of the page
 
+Quote Post
2 Pages V   1 2 >  
Start new topic
Replies (1 - 19)
Advertisement
post Mon, 27 Nov 2006 - 00:22
Post #


Advertise here!









Go to the top of the page
 
Quote Post
Dr Science
post Mon, 27 Nov 2006 - 02:20
Post #2


Member


Group: Members
Posts: 1,397
Joined: 16 Oct 2006
From: S.Yorks
Member No.: 8,288



QUOTE (potkettleblack @ Mon, 27 Nov 2006 - 00:22) *
Techie time.

So what about the much vaunted self-calibration? Well the LTI fills the capacitor for a known short time, then measures the corresponding longer time. It can do this because although it can't measure a short time accurately, it can generate one. In the analogy above, with a digital clock you could accurately set an accurate time so long as it's a whole number of minutes.


So, the LTI's self-calibration is an internal check on the capacitor (the capacitance of which will change with temperature and as the electronics get old).

A very good check to have, the device would be very very questionable without it, but it is an addition to, not a substitute for the start/end session checks and the annual calibration, all of which either check different things or check many more things.


--------------------

Dr.S
Telephone calls may be recorded for the purpose of detection and prevention of crime.
I am an engineer/physicist, not a lawyer. My answers are based on The Laws 'O Physics (which ya' can 'ne change, Cap'n).
The law of the land is a much more slippery and changeable thing.

"The only way to deal with bureaucrats is with stealth and sudden violence" - Boutros Boutros-Ghali, Secretary General of the United Nations
NOTICE The content of this post and of any replies to it may assist in or relate to the formulation of strategy tactics etcetera in a legal action. This post and any replies to it should therefore be assumed to be legally privileged and therefore must not be disclosed, copied, quoted, discussed, used or referred to outside of the PePiPoo forum on which it was originally posted additionally it must not be disclosed, copied, quoted, discussed, used or referred to by any person or organisation other than a member of PePiPoo appropriately paid up and in full compliance with the PePiPoo terms of use for the forum on which it was originally posted. The PePiPoo terms of use can be found at http://forums.pepipoo.com/index.php?act=boardrules. For the avoidance of doubt, if you are reading this material in any form other than an on-line HTML resource directly and legitimately accessed via a URL commencing "http://forums.pepipoo.com" then it has been obtained by improper means and you are probably reading it in breach of legal privilege. If the material you are reading does not include this notice then it has been obtained improperly and you are probably reading it in breach of legal privilege. Your attention is drawn to the Written Standards for the Conduct of Professional Work issued by the Bar Standards Board particularly under heading 7, "Documents".
Go to the top of the page
 
+Quote Post
Whizz
post Mon, 27 Nov 2006 - 10:14
Post #3


Member


Group: Validating
Posts: 306
Joined: 11 May 2006
Member No.: 5,675



Thank you for that information. Please can you tell me what the "start/end session checks" should consist of?
Go to the top of the page
 
+Quote Post
Prof
post Mon, 27 Nov 2006 - 12:04
Post #4


Member


Group: Members
Posts: 253
Joined: 25 Mar 2006
Member No.: 5,095



mmmmm, so if it has been calibrated at base, switched off (went cold)

then switched back on at location, its going to have a different setting???

Its also going to be different on a cold January morning to a July Afternoon.

But the temperature delta has been validated as part of the type approval.

BR Prof

This post has been edited by Prof: Mon, 27 Nov 2006 - 12:04
Go to the top of the page
 
+Quote Post
madmoe
post Mon, 27 Nov 2006 - 12:25
Post #5


Member


Group: Members
Posts: 219
Joined: 18 Nov 2006
From: wales
Member No.: 9,066



how to know if it is working 100% corect is there is a way to know or the lti20-20 have self manag system that if the lti20-20 is not 100% corect it wont work like motherboard on pc if any thing go wrong it wont boot in .

and dos the lti20-20 have any chips that need or relay on software or any BIOS if it is what the chance that this software or bios some time get bug that or can cose any miss calibration at times .

any one know where I can look to find out

thank you


--------------------
I am not mad I am only insane
making mistakes is not a crime keep making same mistakes is
Go to the top of the page
 
+Quote Post
potkettleblack
post Tue, 28 Nov 2006 - 22:20
Post #6


Member
Group Icon

Group: Life Member
Posts: 448
Joined: 6 Nov 2006
Member No.: 8,772



madmoe - the machine has self check; it can't test everything but proving it wasn't working as intended will be nearly impossible without the machine. Same with the software: it doesn't 'get bugs' like a PC because it doesn't change. There's a question over version number that you can read about elsewhere, but the biggest flaws in the machine are errors with the basic operation like targetting and slip. This post is delving into the details.

whizz - you want the ACPO code of practice / guidlines





prof n Dr S. I can't imagine the LTI wouldn't self calibrate every second or less. No reason to do otherwise. I expect the real drift with temp will be a) the linearity of the charge/discharge circuit and b) the crystal frequency, both of which are pretty small errors.

I can think of 4 points that the forum might consider though.

1) Self calibration: is it a good name?
2) Calibration certificate results: why all the same?
3) Distance versus speed accuracy - discrepancy in claims
4) Speed reading is a probability not a result


1) Self calibration: is it a good name?
Is the operation really self-calibration? Well the unit effectively marks it's own scale on the ruler. However it only needed the ruler in the first place because the reference it has isn't accurate over the timescales it wants to measure. So unless it makes itself a ruler, it would be hoplelessly inaccurate. The trouble with the term 'self-calibrating' is it has connotations of a machine that can fix itself, which is not true.

It also implies unquestionable accuracy. One thing the LTI *cannot* do is improve upon the accuracy of the main reference. Of course a crystal is an accurate reference in the first place. A typical crystal might have a maximum error of +/-0.005%, which is about 4s per day were it in a watch. However the LTI *must* be less accurate: in translating time to distance, there are other errors such as the linearity of the charge/discharge circuit, and jitter in the oscillator causing error in the calibration pulse, which are likely to be more significant than the crystal error. Usually these types of errors would be no better than 0.1%. On the lasertech website the ultralyte has a stated accuracy of +/-15cm over a range of 610m, which is an optimistic 0.024%.

My vote goes for a name goes to 'self-scaling'.



2) Calibration certificate results
I wonder about the simulated speed measurement during the annual calibration. By chosing the environment (for example measuring close to the 'self scaling' distance, using optics for an impossibly square return light pulse) it would be possible to eliminate some of the real world errors. So it may be possible to develop a test setup that resulted in a reading of 20mph for 20, 30 for 30 etc. as we see on the annual calibration certificates. However I thought I read somewhere that the machine rounds down (29.9mph reads as 29mph)? If this is true, then calibration results that always show 30 for the 30mph test implies 1 of 2 things:
1) All LTIs overread somewhere between 0-1 mph
2) The calibration test equipment is inaccurate (on purpose), putting out say 30.5mph for the 30mph test.

Otherwise you would expect an equal number to read 29mph as read 30mph.



3) Distance versus speed accuracy
None of the error figures quoted in the initial post is large enough to cause doubt on a measured speed. I think there may be cause to question one aspect though, and that is the accuracy claimed for distance versus that claimed for speed. Here's some example numbers:

50ft = exactly 15.24m
1mph = exactly 0.44704m/s

A vehicle is traveling at 30mph
The LTI measures the speed over 6 pulses then tracks it for another 40 at 125 pulses per second. That's 0.368s.
In this time the vehicle travels 4.935m.
The LTI accuracy is +/-15cm. Say the first measurement is short by 15cm, the last one long by 15cm. The LTI would read a distance of 5.235m and thus calculate a speed of 31.8mph.
Isn't that odd? The LTI claims to measure speed to a greater accuracy than the distance measurements on which it bases those speeds. Before you say 'ah, but it averages', well it could/should average the distance measurements too to get to the 15cm figure.


4) Probability of a speed
Slightly off topic, I also read that the speed is based on method of least squares. It's a straightforward averaging scheme (for a techie), but differs from the usual arithmetic average in that it is a statistical method. This means there *must* be some uncertainty in the result. Without knowing the input data we can't work out how big the level of uncertainty is, but I would question any machine that has to take 40 readings in order to provide an answer. Surely that means by definition it is working on 'balance of probability' rather than providing an answer 'beyond reasonable doubt'? Well done to any legal expert who made it this far and can answer that one.
Go to the top of the page
 
+Quote Post
wilbye999
post Wed, 29 Nov 2006 - 09:25
Post #7


Member


Group: Members
Posts: 57
Joined: 21 Apr 2005
From: Bristol
Member No.: 2,795



My guess from reading the patents and general available information about this device is that it is unlikely that attacking its electronics will be effective. As a distance measuring device it is quite accurate. The problems appear to come from poor operating procedures, including the distance checks, misalignment, two reflective surfaces very close to one another and slip. The lasers have 3 beams which spread out so close proximity of reflecting surfaces is a problem and despite their claims, the algorithms they use to trap errors will not be perfect. You can judge this from the reports on this site where it takes a long time to acquire the targets even in what appears to be correct operation with an uncluttered target. The software/firmware is probably in some form or programmable memory rather than battery backed up ram so is unlikely to become corrupted.
Go to the top of the page
 
+Quote Post
Rockbanger
post Wed, 29 Nov 2006 - 16:07
Post #8


Member


Group: Members
Posts: 13
Joined: 29 Sep 2006
Member No.: 7,959



I hate to nit pick Potkettleblack, but the LTI doesn't measure distance. The only parameter it actually MEASURES is time. Everything else is calculated based on the time delay in the return of the laser pulse reflected from the target and the assumed speed of light (which is probably substantially less than 300,000 km/s for the speed of light in a vacuum).

I'm currently fighting a case where the distance quoted on the camera is approximately 25% greater than the actual distance on the ground, as MEASURED with a 50m tape measure and checked against readings from a hand-held GPS. Why this particular instrument cannot calculate the distance correctly, I don't know. However, if it cannot calculate the distance accurately, it most certainly cannot calculate the speed.

Rockbanger
Go to the top of the page
 
+Quote Post
wilbye999
post Wed, 29 Nov 2006 - 18:16
Post #9


Member


Group: Members
Posts: 57
Joined: 21 Apr 2005
From: Bristol
Member No.: 2,795



When I said distance yes of course it is measuring the time delay of the optical pulses and that allows it to display distances based on the speed of light.

There have been cases where the beam has bounced off the target hit another reflective target and then made its way back to the sensor on the device, so in this case the computed distance would be greater than the actual distance to the target. Are there any such reflective surfaces around where you were allegedly speeding?
Go to the top of the page
 
+Quote Post
madmoe
post Wed, 29 Nov 2006 - 18:36
Post #10


Member


Group: Members
Posts: 219
Joined: 18 Nov 2006
From: wales
Member No.: 9,066



I did find this .
http://www.dailymail.co.uk/pages/live/arti...ource=&ct=5

I am not sure if this a new news or old one .

and I also read this .

QUOTE
What happens? When LRC 100 acquires a laser gun, it immediately returns a 5-second jamming signal , gives the driver a visual warning, and emits a 100dB audible alert. For 5-seconds, the laser gun’s display will not produce a speed reading, more than enough time for the driver to check, and if necessary, adjust vehicle speed safely. The diode then shuts down, and the jammer acts as a passive detector for the next 60 seconds while you clear the laser speed trap. After recycling, the LRC 100 is ready to do its job again


is this legal or not as if this true then the laser gun is not 100% corect


--------------------
I am not mad I am only insane
making mistakes is not a crime keep making same mistakes is
Go to the top of the page
 
+Quote Post
potkettleblack
post Wed, 29 Nov 2006 - 22:05
Post #11


Member
Group Icon

Group: Life Member
Posts: 448
Joined: 6 Nov 2006
Member No.: 8,772



I agree wilbye, this is not the machine's weakness. But until you check, well, who's to know? You'll note my points to question are not about self cal, only #4's on operation at all. So what's the advantage of 3 beams? It's harder to do, spreads more and results in greater liklihood of multipath. They must have done it for some reason.

I agree too Rockbanger, but perhaps you can forgive the slip since I'd just written two pages on how it converts one to the other! Unfortunately I think it makes quite a good optical tape measure; the crystal ref. is good and sadly light is a pretty constant speed in air, albeit 0.9997 of the vacuum speed. If I had to measure distance to the whitewashed side of a house I'd trust it, and I say that as an engineer on your side. So long as the house stays still, it's fine.

In your case there are two possibilities:
- the LTI was not working. Difficult to prove, especially if the video looks reasonable, and you have witness and certificates to the contrary.
- the LTI was working. Then it wasn't you (bad aim or alignment) or reflections were involved, in which case the speed is unlikely to be correct. I think this is more likely and easier to show.

madmoe - thanks, you'll find the full article scanned on here somewhere but I can't read the date
Go to the top of the page
 
+Quote Post
blackdouglas
post Wed, 29 Nov 2006 - 23:15
Post #12


Member


Group: Members
Posts: 685
Joined: 8 Mar 2005
Member No.: 2,524



QUOTE (potkettleblack @ Tue, 28 Nov 2006 - 22:20) *
3) Distance versus speed accuracy
None of the error figures quoted in the initial post is large enough to cause doubt on a measured speed. I think there may be cause to question one aspect though, and that is the accuracy claimed for distance versus that claimed for speed. Here's some example numbers:

50ft = exactly 15.24m
1mph = exactly 0.44704m/s

A vehicle is traveling at 30mph
The LTI measures the speed over 6 pulses then tracks it for another 40 at 125 pulses per second. That's 0.368s.
In this time the vehicle travels 4.935m.
The LTI accuracy is +/-15cm. Say the first measurement is short by 15cm, the last one long by 15cm. The LTI would read a distance of 5.235m and thus calculate a speed of 31.8mph.
Isn't that odd? The LTI claims to measure speed to a greater accuracy than the distance measurements on which it bases those speeds. Before you say 'ah, but it averages', well it could/should average the distance measurements too to get to the 15cm figure.


4) Probability of a speed
Slightly off topic, I also read that the speed is based on method of least squares. It's a straightforward averaging scheme (for a techie), but differs from the usual arithmetic average in that it is a statistical method. This means there *must* be some uncertainty in the result. Without knowing the input data we can't work out how big the level of uncertainty is, but I would question any machine that has to take 40 readings in order to provide an answer. Surely that means by definition it is working on 'balance of probability' rather than providing an answer 'beyond reasonable doubt'? Well done to any legal expert who made it this far and can answer that one.


Er. Much of this information is incorrect, or misleading.

The acquisition phase is 4 samples, not 6. The measument phase requires a set of 40 samples, of which 35 must be valid.

To be saved, the 40 samples must be +/- 0.6m of the expected position.

Once 35 <= valid samples <= 40 are acquired, the device uses a least squares algorithm to calculate an average speed over the samples.

An example formula is given in the patents - a correction to this forumla applies - it works, I've got an EXCEL spreadsheet which implements the formula.

If the first and last samples are out by 0.15m, the speed is not neccessarily out by 31.8mph as you claim, nor the distance. The actual speed calculated depends equally on ALL of the samples and cannot be slewed that much by just the first and the last.

The displayed range is simply calculated by averaging the distance obtained from each of the samples.

Once a speed is calculated - which is an average speed over the sample period - a further check is performed. The expected position of each of the samples is calculated, using a reference (say the start distance), and the calculated speed. A delta of calculated distance for each sample, versus expecpted distance calculated from speed for each sample can be calculated.

If ANY of these samples is out by a given amount, the speed reading can be rejected.

It is claimed, that in this final validation phase, distance errors of 0.02m can be detected.

I do not know where you got your information from.

The flaws in the LTI 20.20 are much more fundamental than the issues presented here.
Go to the top of the page
 
+Quote Post
potkettleblack
post Thu, 30 Nov 2006 - 01:38
Post #13


Member
Group Icon

Group: Life Member
Posts: 448
Joined: 6 Nov 2006
Member No.: 8,772



Thanks for the more accurate information, I did not know the details you provided.
My sources are:
memory of posts (6 instead of 4 initial samples, then 40)
Lasertech.com specification sheet
LTI patent US6445444 (directly, checked a bunch others)


I'm not trying to rattle your cage but it doesn't actually make a difference to the error. See if you agree:

Take the error conditions I mentioned - first and last sample - and put them at the limits. First distance short by 15cm, last long by 15cm. This is the distance registered in software for that sample, not that on the display. Imagine a line between the two points. Now put another 38 samples on that line. The vehicle's only going 0.10m between samples so they easily meet the 0.6m criteria. All samples are linear so all are accepted. Linear regression is used to calculate a speed: it returns exactly the drawn line. The final validation checks for distance error. 30cm of error over 40 samples - all are within the 2cm threshold so no distance error is detected. And the speed is - 31.8mph.


Yes, it's very contrived but it does show the calimed speed accuracy of +/-1mph is a statistical probability, not a certainty. And yes, I beleive the technology has more fundamental errors, I've said that in every post.


I did track down the quote for rounding down the speed:

www.safetycameraswiltshire.co.uk/files/Information_relating_to_LTI_Speedscope_Camera.pdf

"The digital display on the equipment is in whole numbers only and the equipment will automatically ‘round down’ the measured reading to the nearest whole figure. Thus a speed measured at 34.9mph will display as 34mph to the advantage of the driver."


So if this were true, it seriously questions the calibration results. The LTI claims to be centred (+/-1mph), so we would expect a normal distribution around zero error. When calibrated, half the machines would read high and half low, even though the distribution is tight around zero, say +/-0.02%. With rounding down, on a simulated 40mph test half would show 40mph and the other half 39mph. On the dozen I've looked at, all speeds read exactly on target.
Go to the top of the page
 
+Quote Post
Dr Science
post Thu, 30 Nov 2006 - 05:14
Post #14


Member


Group: Members
Posts: 1,397
Joined: 16 Oct 2006
From: S.Yorks
Member No.: 8,288



QUOTE (Prof @ Mon, 27 Nov 2006 - 12:04) *
mmmmm, so if it has been calibrated at base, switched off (went cold)
then switched back on at location, its going to have a different setting???

No, because as soon as its switched on again it will re-set itself

QUOTE
Its also going to be different on a cold January morning to a July Afternoon.

Again no, the whole point of this self-test is that it sets itself to whatever is right for today.

The problem comes when its switched on inside a nice warm car and sets itself up for those conditions, then it gets taken outside, next to a cold windy road, where the "correct" setting would be a bit different. Or if it gets taken out of the nice warm car and immediately switched on (while its still warm from the car). After a few minutes outside it will have cooled down (or on a hot day, warmed up) and the setting it gave itself when it was first switched on isn't appropriate anymore.

But I agree with potkettleblack, its real weakness is alignment/steadiness and the operator's ability to give it a stable platform to work from.

This post has been edited by Dr Science: Thu, 30 Nov 2006 - 05:28


--------------------

Dr.S
Telephone calls may be recorded for the purpose of detection and prevention of crime.
I am an engineer/physicist, not a lawyer. My answers are based on The Laws 'O Physics (which ya' can 'ne change, Cap'n).
The law of the land is a much more slippery and changeable thing.

"The only way to deal with bureaucrats is with stealth and sudden violence" - Boutros Boutros-Ghali, Secretary General of the United Nations
NOTICE The content of this post and of any replies to it may assist in or relate to the formulation of strategy tactics etcetera in a legal action. This post and any replies to it should therefore be assumed to be legally privileged and therefore must not be disclosed, copied, quoted, discussed, used or referred to outside of the PePiPoo forum on which it was originally posted additionally it must not be disclosed, copied, quoted, discussed, used or referred to by any person or organisation other than a member of PePiPoo appropriately paid up and in full compliance with the PePiPoo terms of use for the forum on which it was originally posted. The PePiPoo terms of use can be found at http://forums.pepipoo.com/index.php?act=boardrules. For the avoidance of doubt, if you are reading this material in any form other than an on-line HTML resource directly and legitimately accessed via a URL commencing "http://forums.pepipoo.com" then it has been obtained by improper means and you are probably reading it in breach of legal privilege. If the material you are reading does not include this notice then it has been obtained improperly and you are probably reading it in breach of legal privilege. Your attention is drawn to the Written Standards for the Conduct of Professional Work issued by the Bar Standards Board particularly under heading 7, "Documents".
Go to the top of the page
 
+Quote Post
blackdouglas
post Fri, 1 Dec 2006 - 00:33
Post #15


Member


Group: Members
Posts: 685
Joined: 8 Mar 2005
Member No.: 2,524



QUOTE (potkettleblack)
First distance short by 15cm, last long by 15cm.

QUOTE (potkettleblack)
all are within the 2cm threshold so no distance error is detected.


That doesn't exactly compute. Have another go.

This post has been edited by blackdouglas: Fri, 1 Dec 2006 - 00:35
Go to the top of the page
 
+Quote Post
Max Damage
post Fri, 1 Dec 2006 - 21:08
Post #16


Member


Group: Members
Posts: 39
Joined: 1 Apr 2003
Member No.: 8



The simple truth is that the accuracy of any scientific instrument should never be gauged solely on its internal self-check mechanisms. It should always be calibrated against independent external standards and a calibration curve should be established for the desired measurement range, ideally before each usage. In a Corporate lawsuit, if you were attempting to defend product quality determined solely by "self-calibrated" instruments, a sharp prosecution lawyer would easily tear your case apart.

The case against the so-called "self-calibration" of laser speedmeters was upheld in Canada and this link provides the details.

The Police used to check speed their speed measuring devices like the LTI against a vehicle with a calibrated speedometer, before and after each shift. However this requirement was subsequently downgraded by the Home Office - how convenient, eh?

Best regards,

Max
Go to the top of the page
 
+Quote Post
trick911t
post Fri, 1 Dec 2006 - 22:25
Post #17


Member
Group Icon

Group: Life Member
Posts: 59
Joined: 6 Sep 2006
Member No.: 7,496



QUOTE (potkettleblack @ Tue, 28 Nov 2006 - 22:20) *
madmoe - the machine has self check; it can't test everything but proving it wasn't working as intended will be nearly impossible without the machine. Same with the software: it doesn't 'get bugs' like a PC because it doesn't change. There's a question over version number that you can read about elsewhere, but the biggest flaws in the machine are errors with the basic operation like targetting and slip. This post is delving into the details.



_All_ software regardless of where its stored can have bugs.

This is why bios chips are updated, and its why there are such things as firmware upgrades for other devices what ever they might be.
Go to the top of the page
 
+Quote Post
potkettleblack
post Sat, 2 Dec 2006 - 00:29
Post #18


Member
Group Icon

Group: Life Member
Posts: 448
Joined: 6 Nov 2006
Member No.: 8,772



It's a shame science that can be independently verified doesn't hold more sway. I'm stunned that, for example, the wrong direction for a speed isn't seen as an unequivocal error.

All software can have bugs, heck yes. Can't *get* bugs though if it's an EPROM in a box. I speculate that the difference in legal terms is that if the software used is that approved by the Home Office (regardless of how buggy) it's considered flawless. If the version is different it's againt ACPO CoP, less buggy and hence is still flawless. I did not find a conclusive thread on here about type approval software version.


Had another go at the maths, this time in XL. Got the same result so I'll post for review. PM me for XL version (can't upload .xls). The speed error went up slightly, to 32.1mph, cos I cut the samples down to 40 from 46. This appears to poke a hole in the data sheet for the LTI. If distance accuracy is +/-0.15m, speed accuracy cannot be guaranteed to +/-1mph. That's not dependent upon operation, it's just distance and time. I chose a nearly worstcase error to get to this, but that's the difference between a probable error limit and a guaranteed one. Not a spec that would stand up in my domain of engineering.

Effectively all I did was put a slip error of -15cm to +15cm on a vehicle traveling 30mph. The LTI spec says this can happen without it declaring a fault. We shouldn't be surprised that 30cm of slip causes 2mph of error.
Attached File(s)
Attached File  LTI_algorithm_linear_error.pdf ( 17.91K ) Number of downloads: 62
 
Go to the top of the page
 
+Quote Post
Dr Science
post Sun, 3 Dec 2006 - 00:13
Post #19


Member


Group: Members
Posts: 1,397
Joined: 16 Oct 2006
From: S.Yorks
Member No.: 8,288



QUOTE (Max Damage @ Fri, 1 Dec 2006 - 21:08) *
The simple truth is that the accuracy of any scientific instrument should never be gauged solely on its internal self-check mechanisms. It should always be calibrated against independent external standards and a calibration curve should be established for the desired measurement range, ideally before each usage. In a Corporate lawsuit, if you were attempting to defend product quality determined solely by "self-calibrated" instruments, a sharp prosecution lawyer would easily tear your case apart.

Absolutely! If a self-check is going to be meaningful, then the internal reference used for the self-check must be independently calibrated, to an acceptable standard.

Inadequate calibration would not only take down a civil/corporate suit, in a criminal prosecution for Weights & Measures offences, inadequate calibration will bring the whole thing crashing down. This is not just down to sharp lawyers, its down to good guidance (which comes from the Dept. of Trade & Industry - which includes The National Measurement Systems Directorate, which includes The National Physical Laboratory and The National Weights & Measures Laboratory). These people know a thing or two about measurement, traceability, uncertainty and calibration, so they give very good guidance.

In speeding cases, the guidance (such as it is) comes from The Home Office. The HO's scientists know a lot about fingerprints, DNA, firearms and explosives residue, but they don't know shit about measuring time, distance and speed, or about calibration or quality assurance.

There is one standard for quality and reliability of evidence needed for criminal weights & measures prosecutions and another (or rather several others) for criminal speeding prosecutions.

This is not equal treatment before the law.


--------------------

Dr.S
Telephone calls may be recorded for the purpose of detection and prevention of crime.
I am an engineer/physicist, not a lawyer. My answers are based on The Laws 'O Physics (which ya' can 'ne change, Cap'n).
The law of the land is a much more slippery and changeable thing.

"The only way to deal with bureaucrats is with stealth and sudden violence" - Boutros Boutros-Ghali, Secretary General of the United Nations
NOTICE The content of this post and of any replies to it may assist in or relate to the formulation of strategy tactics etcetera in a legal action. This post and any replies to it should therefore be assumed to be legally privileged and therefore must not be disclosed, copied, quoted, discussed, used or referred to outside of the PePiPoo forum on which it was originally posted additionally it must not be disclosed, copied, quoted, discussed, used or referred to by any person or organisation other than a member of PePiPoo appropriately paid up and in full compliance with the PePiPoo terms of use for the forum on which it was originally posted. The PePiPoo terms of use can be found at http://forums.pepipoo.com/index.php?act=boardrules. For the avoidance of doubt, if you are reading this material in any form other than an on-line HTML resource directly and legitimately accessed via a URL commencing "http://forums.pepipoo.com" then it has been obtained by improper means and you are probably reading it in breach of legal privilege. If the material you are reading does not include this notice then it has been obtained improperly and you are probably reading it in breach of legal privilege. Your attention is drawn to the Written Standards for the Conduct of Professional Work issued by the Bar Standards Board particularly under heading 7, "Documents".
Go to the top of the page
 
+Quote Post
blackdouglas
post Sun, 3 Dec 2006 - 22:12
Post #20


Member


Group: Members
Posts: 685
Joined: 8 Mar 2005
Member No.: 2,524



QUOTE (potkettleblack)
Had another go at the maths, this time in XL. Got the same result so I'll post for review. PM me for XL version (can't upload .xls). The speed error went up slightly, to 32.1mph, cos I cut the samples down to 40 from 46. This appears to poke a hole in the data sheet for the LTI. If distance accuracy is +/-0.15m, speed accuracy cannot be guaranteed to +/-1mph. That's not dependent upon operation, it's just distance and time. I chose a nearly worstcase error to get to this, but that's the difference between a probable error limit and a guaranteed one. Not a spec that would stand up in my domain of engineering.

Effectively all I did was put a slip error of -15cm to +15cm on a vehicle traveling 30mph. The LTI spec says this can happen without it declaring a fault. We shouldn't be surprised that 30cm of slip causes 2mph of error.


Now you've changed what you said originally.

Originally you said just to change the first and last samples (you didn't mention changing any of the intermediate ones).

QUOTE (potkettleblack)
The LTI accuracy is +/-15cm. Say the first measurement is short by 15cm, the last one long by 15cm. The LTI would read a distance of 5.235m and thus calculate a speed of 31.8mph.


What you've now got is classic slip, and it's bloody obvious, was proved YEARS ago - and has even been on TV.

An even more obvious (bordering on the ridiculous) "thought experiment" is the following.

A car is driving along at 30mph.

200m in front of the car, also travelling at 30mph is a speed camera van with an LTI 20.20 being operated in the back.

The driver aims the laser at the car.

What speed does he get?

The answer is of course 0mph.

This is because the LTI 20.20 "method" can only measure relative speed between the reflection point of the signal and the device. And if you pan it, you are effectively altering that relative speed.
Go to the top of the page
 
+Quote Post

2 Pages V   1 2 >
Reply to this topicStart new topic
1 User(s) are reading this topic (1 Guests and 0 Anonymous Users)
0 Members:

 



Advertisement

Advertise here!

RSS Lo-Fi Version Time is now: Tuesday, 26th May 2020 - 00:48
Pepipoo uses cookies. You can find details of the cookies we use here along with links to information on how to manage them.
Please click the button to accept our cookies and hide this message. We’ll also assume that you’re happy to accept them if you continue to use the site.