Open Source LIDAR

[Mark Womack]

I saw this project on hackaday, and thought others might be interested.

https://hackaday.io/project/163501-open-source-lidar-unruly

-Mark

[Mark Johnston]

First of all, a nice little gizmo.  Thanks for posting link.

Second (Crotchety old dude mode):   I so wish that Lidar was NOT used for a single point ToF sensor.  The term LiDar has sadly come to mean two things.

1) The real 'Radar' type of thing that scans many points  or at least more than one.  He discusses putting it on a servo but that is yet another issue and NOT fast like a 'real lidar'.

2) A laser diode ToF sort of approach to a single distance measurement 'thingy' such as you are showing.

We need a some way to state 'multi-point lidar' or something.  The Neato is 360 points at workable inside distances.   Maybe it has to be called 'multi-point near realtime Lidar' or something.   ... whatever ...

Anyway, back on topic:

The huge range is QUITE impressive on his spec.  That is really nice.  Up to 100 meters on non reflective surfaces outside????   I kind of drop into 'prove it' mode on that statement.

I timed out on looking for one of the most key specs on these things besides distance and that is what is the width of the field of detection in degrees?   Maybe I missed it.  that is EXTREMELY important for many uses like putting it on a rotating platform or even use of many like is stated is supported. If it is there, sorry, I just lost interest due to 'single point' and also rather expensive but good to know for extreme distances (if it really does it).

anyway, yet another interesting thing.  Expensive BUT if it really can do even 20 meters and has a narrow light cone like sub 5-10 degrees at least, that is very useful gizmo.

Thanks,

Mark

[Chris Albertson]

We don't make is distinction with Radar.   The radar sensors on car autopilots don't scan

It is really just a matter of terminology.  You can add scanning to any Lidar by either moving the entire unit or aiming the laser at a spinning mirror,

This Lidar was very impressive because they claim 100 meter range outdoors in normal sunlight and cost under $100.    

I really want to see how they measure TOF at few pico-seconds resolution with such an inexpensive device.

--
You received this message because you are subscribed to the Google Groups "HomeBrew Robotics Club" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hbrobotics+...@googlegroups.com.
To post to this group, send email to hbrob...@googlegroups.com.
Visit this group at https://groups.google.com/group/hbrobotics.
For more options, visit https://groups.google.com/d/optout.

--

Chris Albertson
Redondo Beach, California

[Maxim Kulkin]

What do you mean radars on cars don’t scan?! They surely do: most popular lidars - by Velodyne - are of spinning type.

I think terminology here is whether it is 1D, 2D or 3D lidar. E.g. Neato vacuums use 2D (planar) lidar, self-driving cars use 3D lidars and the one described here is 1D.

Lidars are more of a product, not a technology. E.g. there are time-of-flight lidars, frequency or amplitude modulated waves.

Here is a nice summary on different types of lidars:

https://arstechnica.com/cars/2019/02/the-ars-technica-guide-to-the-lidar-industry/

[David Wyland]

Lidar starts out as point Lidar.

You mount it on a rotating platform or use rotating mirrors to make a circular scanning, 2D Lidar.

Velodyne uses multiple (16-64) point Lidars mounted on a spinning platform to get a coarse grained 3D Lidar.

Flash Lidars like the Micdrosoft Hololens use a powerful flash infrared light source plus an area sensor for true 3D Lidar.

There are new Lidars waiting in the wings. They use 1550 nm light instead of 850 nm light.

850 nm light must be limited to 1 mW peak power *independent of duty cycle* to prevent eye damage.

Note: this is for a collimated light beam wiith a diameter pf 5 mm or less, they typical Lidar case.

1550 nm light does not have this problem. Intensity could be increased to Watts (!) safely.

There are 1550 nm Lidars out there for ships that have a range of 4 km by using higher power lasers.

Luminar among others is developing this type of Lidar.

The catch is that 1550 nm lasers/LEDs are expensive: $35.00 instead of $0.35.

They require exotic, expensive semiconductors such as Indium Phosphide not in high volume, low cost production.

They are being used for fiber optics in data centers, where $200 for a 10 GHz laser is not a problem.

[sdey...@hotmail.com]

Maxim, I think he meant the look ahead radars for collision avoidance.( I believe they work around 70Ghz.)

[Chris Albertson]

"

On Fri, Feb 8, 2019 at 12:27 PM Maxim Kulkin <maxim....@gmail.com> wrote:

What do you mean radars on cars don’t scan?! They surely do: most popular lidars - by Velodyne - are of spinning type.

Unless they make a product I don't know about  Velodyne makes LIDAR, with an "L"  not "radar or an "R".   Yes the Velodyne Lidars do scan.  but I wrote that auto RADAR does not scan.

The basic Lidar sensor is really the same if it scans or not.  Sanners are simple things and easy to add.    What is important here is that this guy has a sub $100 Lidar that works OUTDOORS and in SUNLIGHT.    If you want it to scan then buy a $12 motor.  But before this all the low-cost Lidars were low power indoor only units

[Mark Johnston]

Right on the head there Maxim.   I feel this is 'spot on' and what I was looking for in terms of basic catagories.  Thanks

Refering To:    terminology here is whether it is 1D, 2D or 3D lidar.

As stated in my initial reply, 100 meters is really amazing.   Again, I am in 'show me' mode and can you do 100 meters outside and stay within light emission standards for lasars or whatever? (If in public 100m detection taken in at 1 foot may be a 'bad' thing.  But even 20 to 50 meters on 'any' surface at that price seems a good thing to keep track of.  Still don't know it's 'beam width' but one of the many replys may have that already.  No reply needed.

Again the gizmos is a new price point for '1D' lidar.    Thanks again Maxim for the terminology, I just had not seen it much in a lot of posts and my poking on the issue was to encourge others posting in future to perhaps specify this 1D, 2D or 3D simply to make the reader more focused in his/her interest by subject line info.

[Mark Johnston]

Technology put aside and just focusing (pun) on very basic types in terms of what output you get I think I would call a  camera 3d type x-y pixel 'gizmo' would then be a 3D Lidar.

In that 3D lidar creates some sort of point-cloud over some field of view (be that field in a camera frame or a rotating multi-y 3D lidar which happens to be a large cylindar segment of some height.

[Chris Albertson]

There is one more important parameter.   The number of potential returns per point.    This Open Source Lidar can return up to 5 (or 6?) distance measurements per pulse.    I guess when you get out to 100M distance the illuminated spot is large and might hit several objects each a different distance away.

The other parameter is the "class"

The site says the laser power is adjustable via software from class 1 to class 3.   I assume to get the maximum range of 100M in the daytime you might need to bump the power up to class 3.  But could you run a Lidar on a public street at class 3 power?    I don't think that would be a good idea.  

How are companies like Waymo able to get 100M range on public streets?   I assume they'd be using class 1 only.   Does making the optics large diameter help?

On Sat, Feb 9, 2019 at 5:46 PM Mark Johnston <mjst...@gmail.com> wrote:

Technology put aside and just focusing (pun) on very basic types in terms of what output you get I think I would call a  camera 3d type x-y pixel 'gizmo' would then be a 3D Lidar.

In that 3D lidar creates some sort of point-cloud over some field of view (be that field in a camera frame or a rotating multi-y 3D lidar which happens to be a large cylindar segment of some height.

--
You received this message because you are subscribed to the Google Groups "HomeBrew Robotics Club" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hbrobotics+...@googlegroups.com.
To post to this group, send email to hbrob...@googlegroups.com.
Visit this group at https://groups.google.com/group/hbrobotics.
For more options, visit https://groups.google.com/d/optout.

[David Wyland]

Velodyne gets 100M by a combination of a large (2") lens and photon counting.

The larger lens gathers more light. The return intensity is roughly:

     Laser intensity*target reflectivity*(lens diameter/return hemisphere diameter)^2

The return can be modeled as a hemisphere expanding from where the laser dot hit.

The return intensity at 100M with a class 1, eye safe laser gets down to individual photons.

With OK (50%) reflectivity, you are talking of ~1,000 photons for a 1 msec integration time.

To do this, you need avalanche photo diodes to count the individual photons.

If you are designing a long distance, Class 1, 850 nm Lidar, you need to do a photon budget.

This takes into account all the pieces along the way from the laser diode to the photo detector.

-----Original Message-----
From: Chris Albertson
Sent: Feb 9, 2019 9:06 PM
To: hbrob...@googlegroups.com
Subject: Re: [HBRobotics] Re: Open Source LIDAR

[Chris Albertson]

Thanks, that is what I figured, larger diameter optics would allow for using a lower power beam.  The APD detectors are very good, kind of like the old PMTs

Next to figure out a cheap way to scan the beam.   I think Velocidyne uses a spinning mirror and the mirror has a complex 3D shape.

Just one other thing I need to figure out from the Open Source Lidar is how he measures the time of flight using only low cost electronics.  There are a couple ways I know of, the cheapest method is basically analog.    You charge a capacitor with a constant current.  Voltage is proportional to time.

[Maxim Kulkin]

> I think Velocidyne uses a spinning mirror and the mirror has a complex 3D shape.

I think Velodyne and others spin the whole assembly, not just a mirror.

Here is 10 year old Velodyne teardown video:

http://www.hizook.com/blog/2009/01/04/velodyne-hdl-64e-laser-rangefinder-lidar-pseudo-disassembled

Here is teardown of Neato's 2D Lidar:

https://www.sparkfun.com/images/newsimages/XV11Teardown/TheGoodStuffUnmasked.jpg

[David Wyland]

Spinning a mirror is nice but complicated. You have to deflect two beams: outgoing laser and incoming detector.

If you are talking 100 meters, the intensity of the two beams can differ by 10^6: milliwatt out to nanowattt receive.

You have to be careful that *none* of the outgoing beam leaks into the incoming beam.

Detecting time of flight can be simple but not easy.

A simple way is to use phase shift of the return beam to the outgoing beam.

Assume a laser that puts out a 1 mW pulse with a width off 100 ns and a 200 ns period.

This 50% duty cycle pulse will get delayed by the trip out and back at 6.7 ns/meter.

Note: Remember for a class 1 laser, you get 1 mW PEAK, independent of duty cycle.

The only way you get more energy in the beam is to increase the duty cycle.

50% is usually the best trade off.

AND the received signal with the inverse of the transmit signal, i.e. the last half of the cycle.

If the received signal has very little delay, very little of it will be in the last half of the cycle.

The result is a narrow pulse.

If the received signal was delayed by a full ns (~14 meters), you get almost a full 100 ns pulse.

The result scales linearly with distance, 6.7 ns = 1 meter.

Now for the fine print.

Problem 1.

There will be delay in the detector between receiving the photons and generating an electrical pulse.

This delay can be calibrated out, but will probably drift with temperature.

Note that a 7 ns total delay = 1 meter of error (!).

If you want an accuracy of 2 cm, you have some work to do.

Problem 2:

Return intensity falls off as the square of the distance.

Simple Lidars have min (near) and max (far) range values.

If max/min = 10, then the return at max will be 1%of the return at min.

The net result is a tradeoff between not saturating the detector at min range

   and having enough signal at max range to have acceptable signal to noise ratio.

This is true for simple as well as avalanche photodiode detectors.

That is why you mat modulate the laser amplitude for very close targets.

Unfortunately, you have to receive the too-bright signal first,

   *then* reduce the laser power at that range point for following cycles.

A subtlety of this problem is that signal intensity decreases as the square of the distance,

   but we are using phase time delay to measure the distance.

If we look at the received pulse, its amplitude decreases with distance but not time.

The 5 MHz square wave is received at a specific distance from the Lidar,

   and its amplitude is determined by this distance.

But the amplitude of the signal remains constant throughout the pulse train.

You have to keep this in mind if you are going to try sophisticated compensation schemes,

Good luck,

Dave

-----Original Message-----
From: Maxim Kulkin