And we have a winner and in record time, nice job. Those big caps were the last two parts I had installed, I must have been excited to power the board up.
-Ben
Thanks for sending the data Michael. It is clear that the Scribbler's IR sensors are blind to the black banner material. I suspect that you could swap out the LED with a visible one as you had suggested and also use a visible light sensitive photo diode. The circuit would work with a small amount of tuning. You would probably just need to adjust the series LED resistor. Have you tried coloring the black banner lines with a black ink marker to give it more contrast?
-Ben
Hello, Ben!
Using of optical sensors of a visible range is, certainly, possible, but the problem is in incorrect location of the optocoupler elements. In the robot the light diode and the photodiode are situated at an angle of 90 degrees to each other symmetrically vertical: led90.png
In such location the light stream from the light diode reflecting from the surface practically fully comes on the photodiode. Though the degree of the reflection depends on the characteristics of the surface very much.
If the surface is white and satin (e.g. usual paper), then all the light of the emitter practically fully comes into the photodetector. If a piece of black colour is printed on the paper in ink of a jet printer or painted in gouache paint then all the light stream is swallowed and the photodiode fixes the presence of the dark piece.
The surface of the banner cloth is not satin, it's nearly reflective. Solvent ink penetrates under the surface of the cloth and doesn't influence the reflected light stream. The light being polarized is nearly fully reflected from the surface even on the black places. As a result the difference between the strain on the dark and light places almost disappears. So the programs of moving along the line don't work on the banner cloth.
The same character of reflecting can be seen in printing a black line on the paper with a laser printer. In this case the light reflects the toner. But the surface made by the toner is very smooth and the light is well reflected. So, in spite of the black colour of the line the robot can see it as a white one. You may sure in it looking at such line at some angle: a glimpse of the light source will be seen.
To correct this situation you may put the sensors in the following way: led45.png
In this case the character of the received reflected light stream is as following.
For white satin paper it doesn't differ from the previous one for the white pieces and for the dark ones.
While using the banner cloth a remarkable part of the light stream comes to the white piece with an amplitude enough for the identification of the piece as a white one.
On the black piece only a small part of the light stream comes to the photodetector (the character of light is unpolarised). As a result the sensor quite perfectly identifies this piece as a black one, so the ability of the moving programs reconstructs itself.
For installing the photodiodes in a vertical position the following changes in the robot have been done:
1. Two holes d= 3.2 mm have been made between the holes of the optocouplers in the hull of the robot. (1.jpg)
2. The sleeves (the holders of the photodiods) and the holders of the sleeves have been made. (2.jpg & 3.jpg)
3. From the inner side of the hull over the new holes some parts of the plastic divider with the width of 8 mm have been deleted for loose coming of the photodiods sleeves.
4. Semicircle holes have been cut in the detail fixing the light diodes to let the sleeves to move loosely in the hull. (4.jpg)
The optocoupler is installed in the hull of the robot in the following way: 5.jpg
Unfortunately during the reconstruction one of the outputs of the light diode was broken. Diego from the technical support service hasn't told me the trademark of the used light diodes and I have installed a pair I could get which is close in metrics: L-7113FC (L-53F3C) Kingbright.
The experimental checking showed that after the reconstruction the robot moved correctly along the line printed on the banner cloth. If it is necessary you can match the reaction of the line sensor in a programming way.
I'll send you the tables of strain on the outputs of the operational amplifiers and video recordings. On one of the video recordings comparative movements on the paper and the banner cloth are shown before the reconstruction (the paper ground is on the left side, the banner one is on the right side); on the other one they are shown on the banner cloth after the reconstruction.
Thanks for sending the data Michael. It is clear that the Scribbler's IR sensors are blind to the black banner material. I suspect that you could swap out the LED with a visible one as you had suggested and also use a visible light sensitive photo diode. The circuit would work with a small amount of tuning. You would probably just need to adjust the series LED resistor. Have you tried coloring the black banner lines with a black ink marker to give it more contrast?
As Ken says, you have a bunch of us out hear waiting with itchy fingers on the "Buy" button.
Good, good, great actually! There's lots of greedy hand-wringers at Parallax scheming to separate you from your money! Oh, I'm not one of them. . .I mean. .nevermind.
3x bi-color
1x power
2x phototransistors up front
2x infrared LEDs underneat
1x infrared LED on the tail wheel
2x infrared LEDs for the motor encoder
2x SMT LEDs for the XBee stuff
=======
13 is my answer
Is there some trick to properly answering this question?
I pity the engineer tasked with writing the test procedure for the S3. These LEDs are just the beginning.
You guys are getting close but you have forgotten a few. They are hiding all over the place in that 20 page schematic!
1x bi-color for the USB Rx/Tx
1x bi-color which is red when the battery is charging, blue when the battery completes charging and flashes red/blue if there is a battery fault.
3x bi-color (user LED's)
1x Prop controlled power (blue)
2x obstacle IR LED's up front
2x infrared LEDs underneath for line following
1x infrared LED on the tail wheel
2x infrared LEDs for the motor encoder - good job Ken, I thought you guys would miss that one.
1x Xbee commissioning
For a total of 19! I couldn't believe it when I added them up.
From what I've read, Mikhail is presenting an alternate geometry to the S2's IR transmitter and receiver. I gather these changes are being made so the robots works with courses they had printed (or made).
I'm guessing Mikhail was having the same trouble many of us do when using shinny materials such as black electrical tape.
Has Parallax explored alternate geometries for the sensor?
I wonder if Mikhail's 45 degree setup would work with a wider range of materials than the conventional 90 degree arrangement. As I think about this a bit, I'm inclined to think Mikhail's geometry would offer an advantage over the present configuration of sensors.
It's pretty easy to tweak the IR sensitivity in the object to make the S3 work with electrical tape. Ken played with this and had success by changing the line_thld value to 80 in the object. It remember working on a floor cleaning robot 15 years ago that couldn't see chrome obstacles in grocery stores despite avoiding everything else perfectly.
Electrical tape is shiny or a specular reflector put more precisely which is always the worst type of reflector for sensors to detect, you want a diffuse reflector.
There is also a paper type tape which is a diffuse reflector that works great.
From PhiPi: The best tape I've found for line following is a flat black "crepe" paper masking tape from McMaster-Carr. Once there, do a search on 76265A41 (1/2"), or 76265A42 (3/4"). You will have to have cookies enabled to do the search.
From what I've read, Mikhail is presenting an alternate geometry to the S2's IR transmitter and receiver. I gather these changes are being made so the robots works with courses they had printed (or made).
I'm guessing Mikhail was having the same trouble many of us do when using shinny materials such as black electrical tape.
Has Parallax explored alternate geometries for the sensor?
I wonder if Mikhail's 45 degree setup would work with a wider range of materials than the conventional 90 degree arrangement. As I think about this a bit, I'm inclined to think Mikhail's geometry would offer an advantage over the present configuration of sensors.
Yes, I had forgotten that Parallax started selling the tape as well. It was over 10 years ago when I first developed the line sensor but I do remember playing with different angles for the sensors quite a bit. I don't remember the angle being critical but the what is critical is to align the beam angle of the emitter and detector so that they are focused on the same point. The two beam angles create an elliptical overlap on the paper which is the sweet spot for the detector. Of course you want the long side of the ellipse aligned with the long side of the line which is why the sensor pairs are orientated as they are. One of the challenges with the Scribbler is the large ground clearance and corresponding large distance between the sensor and the paper. If the sensor is close to paper, the alignment of the LED and sensor isn't very critical at all. The Scribbler required narrow beam angle sensors and precise alignment. The other constraint bounding the sensor angles was the mold pulls. I hope that helps. -Ben
@Ben and Ken - The hacker port and pins are exactly what I was hoping for! I think everybody will love it too.
Might it have a cover? - Could be black rubber or maybe something more fun... Here is a very quick sketch...
Go with kind of a industrial robotic look.
If Ken says it is too expensive, I at least want Gareth to "print" me one...
I wonder if Mikhail's 45 degree setup would work with a wider range of materials than the conventional 90 degree arrangement. As I think about this a bit, I'm inclined to think Mikhail's geometry would offer an advantage over the present configuration of sensors.
The info in your posts is really interesting Mikhail! I have Scribblers (Blue) and S2s - may have to give this a try! Great videos too! Thanks for all the great information and documentation of the changes.
I have Scribblers (Blue) and S2s - may have to give this a try!
We need a video of all 3 Scribblers playing together nicely. I still have a healthy respect for "old Blue", and I have several of them in various hacked configurations. I've nabbed several off Ebay for $25 and given to kids to start them off on the right path.
In the GUI Scribbler robot one error is noticed (I don't know, whether someone noticed it earlier).
We will assume, creating the program we created the yellow subroutine. After – orange where used the yellow (Images 1 and 2):
Attention: the yellow subroutine at first appears, and then is used.
Such situation leads to two undesirable effects:
1. The yellow subroutine isn't cleared by the “New worksheet” button from the first (Image 3).
2. When loading the program containing the yellow subroutine (without cleaning of a work sheet), contents of the yellow subroutine vanish. For example, when loading the same program, the work sheet will look so (Image 4):
Comments
Last-in, first-out? LIFO Caps!
Hello, Ben!
Using of optical sensors of a visible range is, certainly, possible, but the problem is in incorrect location of the optocoupler elements. In the robot the light diode and the photodiode are situated at an angle of 90 degrees to each other symmetrically vertical: led90.png
In such location the light stream from the light diode reflecting from the surface practically fully comes on the photodiode. Though the degree of the reflection depends on the characteristics of the surface very much.
If the surface is white and satin (e.g. usual paper), then all the light of the emitter practically fully comes into the photodetector. If a piece of black colour is printed on the paper in ink of a jet printer or painted in gouache paint then all the light stream is swallowed and the photodiode fixes the presence of the dark piece.
The surface of the banner cloth is not satin, it's nearly reflective. Solvent ink penetrates under the surface of the cloth and doesn't influence the reflected light stream. The light being polarized is nearly fully reflected from the surface even on the black places. As a result the difference between the strain on the dark and light places almost disappears. So the programs of moving along the line don't work on the banner cloth.
The same character of reflecting can be seen in printing a black line on the paper with a laser printer. In this case the light reflects the toner. But the surface made by the toner is very smooth and the light is well reflected. So, in spite of the black colour of the line the robot can see it as a white one. You may sure in it looking at such line at some angle: a glimpse of the light source will be seen.
To correct this situation you may put the sensors in the following way: led45.png
In this case the character of the received reflected light stream is as following.
For white satin paper it doesn't differ from the previous one for the white pieces and for the dark ones.
While using the banner cloth a remarkable part of the light stream comes to the white piece with an amplitude enough for the identification of the piece as a white one.
On the black piece only a small part of the light stream comes to the photodetector (the character of light is unpolarised). As a result the sensor quite perfectly identifies this piece as a black one, so the ability of the moving programs reconstructs itself.
For installing the photodiodes in a vertical position the following changes in the robot have been done:
1. Two holes d= 3.2 mm have been made between the holes of the optocouplers in the hull of the robot. (1.jpg)
2. The sleeves (the holders of the photodiods) and the holders of the sleeves have been made. (2.jpg & 3.jpg)
3. From the inner side of the hull over the new holes some parts of the plastic divider with the width of 8 mm have been deleted for loose coming of the photodiods sleeves.
4. Semicircle holes have been cut in the detail fixing the light diodes to let the sleeves to move loosely in the hull. (4.jpg)
The optocoupler is installed in the hull of the robot in the following way: 5.jpg
Unfortunately during the reconstruction one of the outputs of the light diode was broken. Diego from the technical support service hasn't told me the trademark of the used light diodes and I have installed a pair I could get which is close in metrics: L-7113FC (L-53F3C) Kingbright.
The experimental checking showed that after the reconstruction the robot moved correctly along the line printed on the banner cloth. If it is necessary you can match the reaction of the line sensor in a programming way.
I'll send you the tables of strain on the outputs of the operational amplifiers and video recordings. On one of the video recordings comparative movements on the paper and the banner cloth are shown before the reconstruction (the paper ground is on the left side, the banner one is on the right side); on the other one they are shown on the banner cloth after the reconstruction.
Yours faithfully, Michael
Video S2 before alteration of the gauge of a line: http://vk.com/video44245607_171134126
Video S2 after alteration of the gauge of a line: http://vk.com/video44245607_171134127
Great work, and thanks for the updates. As Ken says, you have a bunch of us out hear waiting with itchy fingers on the "Buy" button.
Amen Brother! Ready to Scribble in another color!
How many LED's does the new S3 have?
-Ben
Good, good, great actually! There's lots of greedy hand-wringers at Parallax scheming to separate you from your money! Oh, I'm not one of them. . .I mean. .nevermind.
Ken Gracey
3x bi-color
1x power
2x phototransistors up front
2x infrared LEDs underneat
1x infrared LED on the tail wheel
2x infrared LEDs for the motor encoder
2x SMT LEDs for the XBee stuff
=======
13 is my answer
Is there some trick to properly answering this question?
I pity the engineer tasked with writing the test procedure for the S3. These LEDs are just the beginning.
Ken Gracey
If Ben sticks some RGB's in there, that would be 19,
1x bi-color for the USB Rx/Tx
1x bi-color which is red when the battery is charging, blue when the battery completes charging and flashes red/blue if there is a battery fault.
3x bi-color (user LED's)
1x Prop controlled power (blue)
2x obstacle IR LED's up front
2x infrared LEDs underneath for line following
1x infrared LED on the tail wheel
2x infrared LEDs for the motor encoder - good job Ken, I thought you guys would miss that one.
1x Xbee commissioning
For a total of 19! I couldn't believe it when I added them up.
-Ben
From what I've read, Mikhail is presenting an alternate geometry to the S2's IR transmitter and receiver. I gather these changes are being made so the robots works with courses they had printed (or made).
I'm guessing Mikhail was having the same trouble many of us do when using shinny materials such as black electrical tape.
Has Parallax explored alternate geometries for the sensor?
I wonder if Mikhail's 45 degree setup would work with a wider range of materials than the conventional 90 degree arrangement. As I think about this a bit, I'm inclined to think Mikhail's geometry would offer an advantage over the present configuration of sensors.
A few S3 renderings. I'll send this off to the SLA Printer next week.
I also added in a 3D PDF for anyone who wants to really get a good look.
-Ben
It's pretty easy to tweak the IR sensitivity in the object to make the S3 work with electrical tape. Ken played with this and had success by changing the line_thld value to 80 in the object. It remember working on a floor cleaning robot 15 years ago that couldn't see chrome obstacles in grocery stores despite avoiding everything else perfectly.
Electrical tape is shiny or a specular reflector put more precisely which is always the worst type of reflector for sensors to detect, you want a diffuse reflector.
There is also a paper type tape which is a diffuse reflector that works great.
From PhiPi: The best tape I've found for line following is a flat black "crepe" paper masking tape from McMaster-Carr. Once there, do a search on 76265A41 (1/2"), or 76265A42 (3/4"). You will have to have cookies enabled to do the search.
http://forums.parallax.com/showthread.php/117686-11-year-old-s-question-about-Scribbler-line-following-abilities?highlight=scribbler+electrical+tape
-Ben
Parallax sells this:
https://www.parallax.com/product/900-00020
I have several rolls myself. I haven't used it much but I believe it works well.
I wasn't really after the best material to make lines. I still wonder if the 90 degree sensor geometry is the optimal arrangement for the sensors.
Do you know if other geometries have been tried?
Yes, very much. Thanks for the explanation.
Now my mouth is really watering! So excited about this new version! Thanks for keeping us whipped into a frenzy! What is Ken paying you?
Edit - Just saw the ipad GUI video too! HOW DID I MISS THAT? Now I really want one!
Can I call first in line here?
Might it have a cover? - Could be black rubber or maybe something more fun... Here is a very quick sketch...
Go with kind of a industrial robotic look.
If Ken says it is too expensive, I at least want Gareth to "print" me one...
Confirming to your words look video
S2 Before alteration of the gauge of a line: http://youtu.be/5c0Rpz_aQ6Y
S2 After alteration of the gauge of a line: http://youtu.be/Cj3ee-rvGqg
Tons of money. Ben will never have to work another day in his life after the S3 is finished.
You, Whit, shall help us recover our return on investment
But on a truthful note, do you have an iPad?
Ken Gracey
Yes! And the GatorBots have access to the school's - we are covered! We will eat this up!
Edit - My original post was a bit confusing - I meant I really want an S3 after seeing the ipad programming demo video!
Seems fair to me!
We need a video of all 3 Scribblers playing together nicely. I still have a healthy respect for "old Blue", and I have several of them in various hacked configurations. I've nabbed several off Ebay for $25 and given to kids to start them off on the right path.
A great way to begin!
In the GUI Scribbler robot one error is noticed (I don't know, whether someone noticed it earlier).
We will assume, creating the program we created the yellow subroutine. After – orange where used the yellow (Images 1 and 2):
Attention: the yellow subroutine at first appears, and then is used.
Such situation leads to two undesirable effects:
1. The yellow subroutine isn't cleared by the “New worksheet” button from the first (Image 3).
2. When loading the program containing the yellow subroutine (without cleaning of a work sheet), contents of the yellow subroutine vanish. For example, when loading the same program, the work sheet will look so (Image 4):
Yours faithfully, Mikhail.
Trade two? C'mon, you haven't even STARTED your swarm collection yet.
And that's all for the S2 - Ben, this is your calling! We're fully committed to the S3 now.
KG