measuring distance between two cars using infrared circuits

[Anthony Fremont]

It also adds an unknown and significant delay into the path.

I disagree. In the grand scheme of a signal that's usually taking tens
of mS to make its journey, a few uS of turnaround overhead is not, IMO,
significant nor does it have to be an unknown.

 

[Anthony Fremont]

It's *is* done without any transponder, which would make the idea

useless.

It's not a useless idea, theoretically it could be put to fairly
practical use provided the obvious problems of using sound to
communicate on the freeway. Provided that you could transmit/receive a
signal approx 500' I could envision a system whereby:

1) the car in front could know fairly accurately how far the car behind
it was
2) it could also know how fast the car is going (i.e. how fast it is
encroaching)

If the front transponder was appropriately calibrated to a specific
frequency, the rear transponder of the leading car could use the
frequency of a received signal (and it's own speed) to calculate the
speed of the vehicle in behind it.

To calculate the distance, the onboard GPS's (all new vehicles will
eventually have one) 1 second pulsed output can be used to determine
when pings should be sent from the front transponder. By analyzing the
time shift of the start of the received signal, the rear transponder of
the leading vehicle can determine the distance to following vehicle up
to about 500'. A single bit of data could be embedded into the signal
to indicate that it was transmitted on odd vs. even numbered second to
extend range to about 1000'.

To distinguish transmitted return signals from reflections, a different
frequency is used. Standard compensation techniques for atmospheric
variations would be applied of course.

 

[Mac]

Yes, pick your poision.

It looks like it doesn't really matter, anyway. The Fourier transform is
just a sum of two sinc() functions, one shifted right and one shifted left
by the carrier frequency. The pulse duration controls the magnitude of the
FT.

I believe the total bandwidth is infinite, but any finite signal has
infinite bandwidth, so that doesn't really help us.

Unfortunately, I'm not sure I know how to answer the question myself.

I'll try to remember to ask some people who might know tomorrow and get
back to you. (It also might pay to ask in the radar/sonar newsgroup.)

But the more you constrain the bandwidth, the more difficult it will be to
identify exactly where the pulse starts or stops. So for precise ranging,
you need more BW, regardless of pulse duration.

[snip]

I admit that the transponder is not essential. It just makes it easier
to detect the signal, and increases the range over which the system
would work.

It also adds an unknown and significant delay into the path.

Depending on exactly how the system is set up, the delay could be
completely neutralized by using a PLL.

--Mac

 

[Ken Smith]

People have mentioned triangulation. Does a license plate retroreflect
laser pointer light?

Why not just image the plate with a camera and measure the image size with
software? When the cars get further apart the image of the plate gets
smaller. No LEDs needed.

 

[keith]

I disagree. In the grand scheme of a signal that's usually taking tens
of mS to make its journey, a few uS of turnaround overhead is not, IMO,
significant nor does it have to be an unknown.

We're talking abotu cars a few tens of feet apart. Thus it's nanoseconds,
and *any* turn-around in the transponder is *significant*.

 

[keith]

It looks like it doesn't really matter, anyway. The Fourier transform is
just a sum of two sinc() functions, one shifted right and one shifted left
by the carrier frequency. The pulse duration controls the magnitude of the
FT.

Sure. I'm looking at launching a ~2GHz (wherever the FCC allows) CW pulse
and measuring its time in flight. At a ns/ft that's 6"/ns round-trip.
Some tricks should be able to get this down significantly less than this.
A ns is a long time these days.

I believe the total bandwidth is infinite, but any finite signal
has infinite bandwidth, so that doesn't really help us.

Sure. I don't see a few kHz on either side of 2GHz to be a big deal
though. It might be a challenge to gate an uwave tranmsitter on in a
millisecond, but...

Unfortunately, I'm not sure I know how to answer the question myself.

I'll try to remember to ask some people who might know tomorrow and get
back to you. (It also might pay to ask in the radar/sonar newsgroup.)

RADAR was my primary interest here. Measuring ns delays is rather trivial
these days. ...and that gets us to 6" distance resolution. Put enough of
these together with a (very) little computation and we get velocity. I
don't see how the mechanics of a couple of cars will exceed the physics or
computational needs.

But the more you constrain the bandwidth, the more difficult it will be
to identify exactly where the pulse starts or stops. So for precise
ranging, you need more BW, regardless of pulse duration.

Ok. We can measure more points of the envelope. The question is where is
the bandwidth limitation. I suspect it will be in the transmitter,
though I don't know. Again, a few kHz isn't a lot of bandwidth.

The same rules would apply to a modulated IR signal. There is no way
the OP is going to get any kind of high resolution ranging using IR
alone because there is just not enough bandwidth. (Some laser diodes
have more than enough bandwidth to do this, but I don't think they
put out enough power)

[snip]

Ultrasound might work well. You could have a transponder on the back
of the car in front and a range-finder on the front of the car in
back.

It's *is* done without any transponder, which would make the idea
useless.

I admit that the transponder is not essential. It just makes it easier
to detect the signal, and increases the range over which the system
would work.

It also adds an unknown and significant delay into the path.

Depending on exactly how the system is set up, the delay could be
completely neutralized by using a PLL.

How? The PLL has to capture the signal and then re-launch the "answer".
That's time. If we're measuring the round-trip delay of two cars ten
meters apart on the Autobahn, the capture/retransmit time is an error I'd
rather not make.

 

[Dave VanHorn]

Doppler or direct time of flight, is doing it the hard way.

Chirp the transmitter at some number of MHz/uS
The reflection, even at a few nS delay, will be your carrier frequency of
that number of nS ago (can't be anything else!) so, the mix frequency
product at the receiver, will be proportional the the distance.

Smooth the output a bit to eliminate jitter, and you're there.

 

[Anthony Fremont]

We're talking abotu cars a few tens of feet apart. Thus it's nanoseconds,
and *any* turn-around in the transponder is *significant*.

This portion of the thread is about using ultrasound transponders, not
radar or IR. Sound doesn't get very far in a nS.

 

[Robert Monsen]

Why not just image the plate with a camera and measure the image size with
software? When the cars get further apart the image of the plate gets
smaller. No LEDs needed.

That is a pretty good idea. I'm not sure it's workable as is, though.

A license plate is a foot across. Thus, the function of angle given
distance is

f(x) = 2*arcsin(.5/x)

At 50', that translates to about 0.02 radians. Not too bad. However,

f'(x) = about 1/x^2

For a change of 1 foot at 50', that would be an angular change of
1/2500. It's better at 10', where a foot change would cause 1/100 change
in angle. If the maximum width were 500 pixels, then you could detect a
foot change (ie, 1/500 = 1 pixel) at 22 feet.

However, another possibility would be to use the camera to measure the
entire car width. You could use the license plate as a standard measure,
and measure the width in license plates (which are 1 foot in width in
the US.) That way, you could use the wider angles, but also have an
absolute measure. Also, that way you could autozoom the camera without
losing registration, and get better angular resolution for longer distances.

One problem is that license plates change colors, so identification
might be an issue. Thankfully, they are almost always in the same place
on the vehicle.

--
Regards,
Robert Monsen

"Your Highness, I have no need of this hypothesis."
- Pierre Laplace (1749-1827), to Napoleon,
on why his works on celestial mechanics make no mention of God.

 

[Dave VanHorn]

For a change of 1 foot at 50', that would be an angular change of
1/2500. It's better at 10', where a foot change would cause 1/100 change
in angle. If the maximum width were 500 pixels, then you could detect a
foot change (ie, 1/500 = 1 pixel) at 22 feet.

Well, the good news is that as it becomes more important, it becomes easier
to see the difference.

 

[Rich Grise]

Well, the good news is that as it becomes more important, it becomes easier
to see the difference.

I wonder what the original intent was. Does the OP want to have a machine
to substitute for paying attention to the road? To decide on a safe
following distance for him? Is this the next stage in "Cruise Control?"

If you're looking for the distance to a specific car, or relative speed,
then take a spring-loaded reel of wire, or string, and attach the end to
the car in front, then sense the position/speed of the tether wire.

What's the actual goal here?

If he wants to drive a car without driving a car, then he should either
take cabs or just drive off a bridge.

Thanks,
Rich

 

[peterken]

hello,
i want to construct a circuit which will be able to measure distance
between two moving cars.The two cars are needed to maintain a fixed
distance between each other.
if the distance between them increases or decreases the circuit
should be able to detect this change and notify the amount of change
in the distance to both cars.
i want to construct this circuit using infrared LEDS ,so if any one
can help me out with this(circuit idea) please mail it to me .

anyone ever thought of a "visual measurement" and using a moire-effect ?

Say the leading cars has a pattern of concentric circles or parallel lines,
and the trailing cars' visual system looks to first through also the same
kind of pattern, then a moire pattern will be seen
(same principle as looking through curtains, or on a computer monitor if
pixelsize gets close to dot-pitch)

interpretation of the moire is related to the effective distance

if defining the system well, the moire may even be a "growing shape" related
to the actual distance

 

[Dave VanHorn]

I wonder what the original intent was. Does the OP want to have a machine
to substitute for paying attention to the road? To decide on a safe
following distance for him? Is this the next stage in "Cruise Control?"

It is in fact the next thing in cruise control.
The current ones are blind, and will not compensate for changes in the speed
of the car you're following. There are a few systems out now, that handle
that, plus braking if they sense rapid downward change in separation. The
theory is that they can react faster than you do.

 

[Robert Monsen]

hello,
i want to construct a circuit which will be able to measure distance
between two moving cars.The two cars are needed to maintain a fixed
distance between each other.
if the distance between them increases or decreases the circuit
should be able to detect this change and notify the amount of change
in the distance to both cars.
i want to construct this circuit using infrared LEDS ,so if any one
can help me out with this(circuit idea) please mail it to me .

http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&category=73916&item=7129299161&rd=1&tc=photo

Mount it on your car so it fires directly forward. Hack the output
section so it gives a signal which can be fed into your cruise control...



--
Regards,
Robert Monsen

"Your Highness, I have no need of this hypothesis."
- Pierre Laplace (1749-1827), to Napoleon,
on why his works on celestial mechanics make no mention of God.

 

[Mac]

Sure. I'm looking at launching a ~2GHz (wherever the FCC allows) CW pulse
and measuring its time in flight. At a ns/ft that's 6"/ns round-trip.
Some tricks should be able to get this down significantly less than this.
A ns is a long time these days.


Sure. I don't see a few kHz on either side of 2GHz to be a big deal
though. It might be a challenge to gate an uwave tranmsitter on in a
millisecond, but...

It is not difficult to turn on a low power transmitter in a millisecond.
But I don't think a few kHz of bandwidth is anywhere near enough. I was
too busy today to talk this over with people who would know.

RADAR was my primary interest here. Measuring ns delays is rather trivial
these days. ...and that gets us to 6" distance resolution. Put enough of
these together with a (very) little computation and we get velocity. I
don't see how the mechanics of a couple of cars will exceed the physics or
computational needs.

Well, measuring a ns delay can be somewhat challenging in a digital
circuit. It is easy for a good oscilloscope, of course. But even if you
use an ADC, followed by a DSP, the ns resolution implies a sample rate of
1 GHz, in some sense.

I have seen programmable delay circuits which were adjustable in
small steps (picoseconds) but they incorporated clever analog stuff
along with digital clocks.

In the application you are talking about, you would need to have some kind
of analog detection (time to voltage circuit, perhaps) which would then be
sampled.

Ok. We can measure more points of the envelope. The question is where is
the bandwidth limitation. I suspect it will be in the transmitter,
though I don't know. Again, a few kHz isn't a lot of bandwidth.

Well, the transmitter doesn't have to be high bandwidth. You just need a
high bandwidth (fast) switch between it and the antenna. The antenna does
need to have high bandwidth.

The receive chain, including the antenna, I think, does need to be
wideband. That opens you up to all kinds of noise, which is
problematic. I'm not saying its impossible, I'm just saying that it isn't
trivial or easy.

Another problem with 2GHz is that it is difficult to get a narrow beam
antenna that can fit unobtrusively into the car's styling.

[snip]

How? The PLL has to capture the signal and then re-launch the "answer".
That's time. If we're measuring the round-trip delay of two cars ten
meters apart on the Autobahn, the capture/retransmit time is an error I'd
rather not make.

You could use a dual frequency scheme where you send out a pulse
at f1, during which the PLL can lock, then abruptly change the
frequency to f2. As soon as the transponder PLL detects the step
in frequency, it can turn on its transmit gate.

The transmitter would use the frequency step as the synchronizing time. I
don't know that this would work it is just a thought.

--Mac

 

[Mark Jones]

It is in fact the next thing in cruise control.
The current ones are blind, and will not compensate for changes in the speed
of the car you're following. There are a few systems out now, that handle
that, plus braking if they sense rapid downward change in separation. The
theory is that they can react faster than you do.

React faster, probably. React better? No way. Unless this idea is
*very* well engineered, it will likely backfire with catastrophic
results. There are an infantessimal number of anomalous things that
can happen while a vehicle is in motion. Anything from equipment
failures to weather conditions, obstacles in the road, visibility
issues, even radiological accidents, etc... a human really is needed
to process all of that information during driving (and even so,
sometimes we get it wrong.) A device to further refine "cruise
control" likely needs to be more A.I. than hardware. For instance, if
you were cruising down the highway and the distance-to-car-in-front
sensor suddenly malfunctioned and indicated you were 0 feet from the
car in front of you and travelling at 70MPH, what would happen? Would
that in itself cause an accident? No per-vehicle cruise control system
could accomidate that kind of an issue. But if the cars were connected
to an A.I. network...


-- "It is said that a caterpillar must struggle to open it's own
cocoon. It is this struggle which makes it strong enough to survive
being a butterfly. If a person were to help the caterpillar open it's
cocoon, the butterfly will die." MCJ 20050119

 

[Dave VanHorn]

React faster, probably. React better? No way. Unless this idea is
*very* well engineered, it will likely backfire with catastrophic
results.

True, this is why it's only now becoming available, with work being done in
this direction since about the mid 80-s IIRC.

 

[Rich Grise]

True, this is why it's only now becoming available, with work being done
in this direction since about the mid 80-s IIRC.

How do those supermarket door openers work?
(not the footpad - the ones with the little box over the door)

Thanks,
Rich

 

[John Woodgate]

I read in sci.electronics.design that Rich Grise <[email protected]>
wrote (in <[email protected]>) about 'Car A.I.
[was: measuring distance between two cars using infrared circuits]', on
Tue, 25 Jan 2005:

How do those supermarket door openers work? (not the footpad - the ones
with the little box over the door)

I've always supposed they work the same as a security light, using a
pyroelectric ceramic.

 

[Jim Thompson]

I read in sci.electronics.design that Rich Grise <[email protected]>
wrote (in <[email protected]>) about 'Car A.I.
[was: measuring distance between two cars using infrared circuits]', on
Tue, 25 Jan 2005:

How do those supermarket door openers work? (not the footpad - the ones
with the little box over the door)

I've always supposed they work the same as a security light, using a
pyroelectric ceramic.

Doppler at ~10.25GHz... they will set off speed radar detectors when
set on "highway".

...Jim Thompson