The pricelist is going to look like this
TDC $20.00
APD $140.00
APD circuity $50.00 needs a reverse bias of 100+volts
75Watt Diode $45.00
Diode driver $100.00
Microcontroller $10.00 assuming propeller
plus lensing and some form of mounting that will allow for precision alignment
Just phoned a few suppliers and in small quantities got the following prices:
TDC = $20.00 (the timing chip)
APD = $37.25 (high quality type)
APD circuitry = $10 (the controller is on the chip)
Laser = 75W - $24.12; 25W - $17.30
Laser driver circuit = $15.00 (the controller is on the chip)
Propeller = $8.00
Total (25W, excluding PCB's and optics) = $107.55
What I find exciting here is that logicmax knows what he's talking about and his prices are very much the industry standard ($365.00). But take a look at the impact on the system cost of having the right kind of controller chip:
Saving on APD circuitry = $40.00
Saving on laser circuitry = $85.00
Total saving = $125.00
I would be very interested to know how much logicmax is paying for his timing circuitry. There may be further savings there as well.
As for handling the issues of paralax and focus, if the optical system is reasonably well designed in terms of lens diameters and focal lengths then the optics become "self adjusting" in that a graph showing signal strength against distance will be fairly flat. This means that it is not necessary to add any hardware to compensate for the parallax and focus imperfections.
. got the following prices:
Total (25W, excluding PCB's and optics) = $107.55
Total saving = $125.00
if the optical system is reasonably well designed in terms of lens diameters and focal lengths then the optics become "self adjusting" in that a graph showing signal strength against distance will be fairly flat.
I've worked for outfits that did not do this kind of investigation until the third generation of the product. Doing lots of investigation and making adjustments during the design phase seems to be paying off.
I looked at the Parallax proto boards. What about mating the LRF to the Gadget Gangster - so it plugs straight on top making use of the available power supplies and port pin connections.
The laser and receiver modules could connect to the LRF board via two separate ribbon cables. I think that we may just be able to get away without using any RF connectors if we impedence match the ribbon.
Easiest and simplest for you is best. Does it really need to be externally located and connected by ribbon cables? If not, I would think it would be easiest to mount the laser and receiver directly to the LRF board and make the entire rig a single module; wouldn't it be easier to impedance match and manage in general if there were no cable at all? Rule of thumb is unless the design specifically requires it, leave it off.
If your intent is to allow greater flexibility in placing the sub assemblies into custom robot configurations at the expense of more parts and a touchier design, I would advise saving that for later versions. Because you won't really know what configuration options folks will want till they start using it and asking for specific capabilities.
This is an interesting engineering decision - compact or modular? There are pro and cons to both methods.
Compact pros:-
1. The matching of the high frequency signals is easier and there are less parts - especially connectors and cables which can be unreliable.
2. The cost will almost certainly be lower, but perhaps not by much.
Compact cons:-
1. Imagine how much light a 75 Watt laser gives off. This will be in very close proximity to the detector. I've worked on several projects where measurement errors in the return signal were traced to light leakage BETWEEN the layers of the circuit board. This happens because the laser is rated by the amount of light that comes out of the front face of the semiconductor. However, the laser is in fact symmetrical with an equal amount of light coming out of the back. Even metal packages leak an enormous amount of light through the insulating material around the connecting pins. Rule of thumb - encapsulate the laser in a black potting compound with just the front face visible. This is best done in a separate module.
2. The range of the LRF is determined mostly by the physical size of the optics. In order to reduce the effects of parallax you want the laser optics and the receiver optics as close together as possible. A number of manufacturers actually cut flats on the sides of the lenses so that they fit closer together. If the laser and receiver are on the same board then the spacing between them will restrict the kind of optics that they can use to one size only. If the lenses are any smaller then the laser and receiver will be too far apart. If the lenses are any bigger then the laser and receiver will be too close together.
A compromise solution:-
Most designs have either separate laser and receiver modules, or the timing circuit as part of the receiver with a separate laser module.
More research is needed I think - greater clarity on typical applications. I've read other threads where people wanted to go a long way - that means bigger optics. But for a small robot you want things to be as compact as possible. Compact or flexible/modular?
encapsulate the laser in a black potting compound with just the front face visible
Can the layout be design such that the use "paints" the back surfaces of the laser as the user-finishing step? If this results in another careful step for the user, but a significant simplification in physical design and decreased costs, this might be somthing to explore. Otherwise, if its too touchy and just has to be done right, make is a separate module and an have them potted.
physical size of the optics.
I would guess an equal number would want extremely close as extremely far; and equal number to both of these will want somewhere in the middle. A design that allows a choice of optics would be most flexible even if there are only two choices. A design goal could be "simplest flexibility for optics".
Most designs have either separate laser and receiver modules, or the timing circuit as part of the receiver with a separate laser module.
...greater clarity on typical applications. I've read other threads where people wanted to go a long way - that means bigger optics. But for a small robot you want things to be as compact as possible. Compact or flexible/modular?
So it looks like at least one subsystem must be on a separate board or module. I can't really get the trade-off points, can you post or PM or link some examples? Short of that, I would say to just do what easiest for the prototype. The novel bit is your TDC chip, once that is in a rig we can see what the implications of the particular implementation are.
I don't know it this makes any sence, but here goes: "Laser and reciever optics must be as close as possible". Why not use the same optics for both laser an reciever? You can get any closer than that. List the reasons why its impossible, maybe we can see where each item contradicts another, and design around it. This works sometimes, I think its called the "Triz" method.
Why not use the same optics for both laser an receiver?
There are almost as many optical configurations as there are manufacturers of LRF's. Some manufacturers even use different optics on different products like these two from Optech out of Canada.
I would recommend leaving the concentric optics to those who want to play around with polarizers and mirrors. There are some fun configurations that the telescope makers might try, like putting the laser in front of the receiver optics. This is brilliant for long range work but has problems close up. So how do we decide on the type of optics?
the "Triz" method
Now that is my kind of development philosophy! One repeating theme of our LRF design seems to be that the range requirement is poorly defined. That means that the optical system is also poorly defined and consequently, the spacing of the laser and receiver is in question. We have mentioned the possibility of light weight systems for aircrft use, short range systems for small bots and long range systems perhaps just for the fun of it. Ideally, the final solution should work for all of them.
At this point my intuition is ringing alarm bells. Flexibility implies modularity. This thing needs to be more like a Lego set than a Meccano set. Also, the chip needs to directly support the aspirations of the hobbyist. If I was playing with this thing what would I want to do? I would like to find the limits. That means minimum range, maximum range, best resolution, highest speed, smallest size, highest power and so on. The more limited the system is, the less exciting the learning experience will be. If it could do lots of cool stuff then I might also be willing to pay a few dollars more.
So what is the cost of this flexibility?
As a start I've extended the range of the chip to 370m (1100') so that hobbyists who have an old telescope lying around can configure a long range instrument with the laser mounted where the sighting scope goes and the receiver using the primary. This extra range is for free.
Next, I've looked at the cost of using RF cables to link the laser module and receiver module to the timer board. These are only a few dollars each and many companies now custom make any length that you want for very little charge. We could provide a standard set of cables in the kit but someone who wants to make a tiny system can get really short cables made, whilst someone who wants to push the range limit with large optics can get much longer cables made.
One trick is to make sure that the laser and receiver cables are the same length. That way altering the cable length will not affect the calibration of the system. This is because the "zero" signal and "return" signal will have to travel the same distance down the cables and so cancelling out the electronic delays.
I feel that a modular system offers the highest "fun factor" and so I propose that the laser, receiver and timer are all separate modules.
If we agree to a modular solution then the laser module becomes the next target for design "optimisation". We must be very careful about shipping lasers across national boundaries. Lasers are regulated by the FDA and the process of getting them to approve a product for import or export is tedeous and expensive. However, if the laser is already in the country then there are no restrictions.
It may be worthwhile designing the laser module in such a way that the end user can easily solder on their own laser. That way they can choose a low power or high power device to suit their application. Of course, we'll have to design the module in such a way that the "light leakage" mentioned earlier is not a problem. Perhaps prof_braino's "paint" solution could be used.
I feel that a modular system offers the highest "fun factor" and so I propose that the laser, receiver and timer are all separate modules.
agreed
the end user can easily solder on their own laser.
agreed
This sub forum, the sensors forum, does not get as much attention as for example the main propeller forum. There are several threads in other forums that discuss laser projects. The participants in the other threads among them seem to know everything about everything, except lack your particular experience with lasers.
Would this be a good time to invite these other folks to join this discussion? I would guess that these are the people that would be the earliest users of your results.
Now that we have discussed much for the highest level parameters, these folk might provide insight to actual "use-cases".
Having the initial signal pulse detected by the receiver is sometimes very helpful. This can give a start signal for the timer also know as T-zero. So then the receiver will detect an initial pulse and then after the TOF it will detect a second pulse. The time between the 2 pulses being twice the time it takes for light to travel from the receiver to the target.
This is a great approach and definitely gives the best temperature stability. I've tried it both ways - with T-zero going straight to the timing chip or T-zero going to the receiver and I'm undecided as to which works better and which is simpler. Do you have any experiences one way or the other?
I would like to hear from logicmax on this one. In any event it doesn't alter the design of the chip because the user can select if they want to read the first signal (T-zero) or the last signal (return).
I've attached the first pages of the draft datasheet. There are many timing diagrams that need finishing and my editor wants to change everything about the format (as usual) but the basic idea is there.
I would like to thank prof_braino for some of the ideas and, believe it or not, alex123 who suggested that some of the latest CMOS interfaces may be up to the task. He was correct. There is a type of CMOS called LVDS that is almost identical to LVPECL but with a slightly longer gate delay (200ps). Using this should be fine for the signal interface and will reduce the cost of the chip by around $4 (final prices still subject to confirmation).
OK, I sent a PM to a couple folks to see if they are interested. These are folks that either have worked on other laser projects, know about timing, do board design, and/or are just nice. This is just a small set of the folks that fit this description, only five names are allowed per message. Hopefully if the thread has merit it will be high enough in the list to attract attention.
One thing about the previous laser projects, they all start out considering time of flight, but encounter issues and and switch to another method. So there should be a bunch of folks that are familiar and can comment on what you have put together.
StefanL38 - the version that we will be playing with is going to be in an FPGA. It's not worth going to custom silicon or ASIC for small volume production. Initial production costs will be zero.
All the internal components are designed and tested but I want to make some changes to suit the hobby market. For example, today I have been working on bringing out some of the more interesting internal signals onto test pins. This will allow someone with a relatively low frequency 'scope to see what's going on inside so that they can fiddle with the external components and try to improve their LRF design. I don't normally do this in an industrial design because the external components (laser, receiver etc) are fixed, and once in production don't change from unit to unit. But for a hobbyist, I think that playing with the design is half the fun.
Another thing that prof_braino has been emphasizing is the correct use of terminology - avoiding jargon etc. I've been rewording the datasheet and other support documents to get some consistent and acceptible words that are meaningful to someone with no experience in LRF design.
This is the next revision of the Datasheet for the LRF chip. There's still plenty to do. I've reworded a few sections to give a little more clarity to the operation.
The images are fuzzy because they haven't been converted to vectors yet. That will only be done in the final draft because I keep changing them.
Below is a photo of two LRF's. The little one works up to 10m and was designed as a level measuring instrument for small silos. The larger one has a range of 120m and was used underground in the gold mines of South Africa.
The optics of the small unit are concentric whilst the larger unit uses a side-by-side configuration. I think that the new LRF design needs to be able to work in both types of device so flexibility and modularity are very important.
I've been fighting with port pin assignments - trying to make the pcb layout as easy as possible but handling all the different logic protocols. As a reference, it looks like the power supply requirements for the different logic types are:
core = 1.5V
SPI and general I/O = 3.3V
LVCMOS = 2.5V or 3.3V
LVDS = 2.5V
LVPECL = 3.3V
There are three "differential outputs" on the chip - Laser PWM, Laser fire and APD PWM. At a minimum, the Laser fire pulse needs to use a high speed protocol - either LVPECL or LVDS. By comparison, the PWMs are relatively slow speed so they could (theoretically) be driven by two complimentary LVCMOS pins instead of genuine, differential signals. Not perfect perhaps, but I can stagger the switching points to avoid SSOs (simultaneous switching output glitches on the power rails).
This means that the chip needs three power rails:
core = 1.5V
SPI and general I/O = 3.3V
LVDS = 2.5V
This bugs my sense of engineering esthetics - I'd really like to throw out one of these power rails. The core voltage has to stay - no way around that. But having both 2.5V and 3.3V for the I/O ports seems messy. Perhaps these could be combined into one. But which one?
After trying all kinds of different combinations it seems that the determining factor is that the Propeller needs a 3.3V SPI interface to work properly. It will not function correctly with a 2.5V SPI interface. Making all the I/O ports 3.3V means changing from the 2.5V LVDS in favor of the 3.3V LVPECL and viola! Only two power rails.
core = 1.5V
SPI, general I/O and LVPECL = 3.3V
The compromise is that LVPECL draws more power than the LVDS (6mA @ 3.3V instead of 3mA @ 2.5V) but I think that is an acceptible price to pay for getting rid of an entire power supply. From a functional point of view, either protocol would have worked fine, but the final selection criterion has turned out to be the power rails.
Here's the first pinout diagram of the chip.
Vcc is 1.5V and is the power to the core.
VCCB's are 3.3V and are the power to each side of the chip for the I/O's.
Even though there are four VCCB's, they will all be attached to the same 3.3V power supply. They will be independently decoupled to reduce cross-talk between the port pins.
Just got the first optical parts for a "long range" LRF. A mechanical engineering friend of mine did the design - fantastic job!
The lenses are acrylic (plastic) and in the picture you can see the first part of the holder that will carry the lenses, laser module and receiver module. This holder is molded in a hard plastic with glass fiber embedded in it. This will ensure that the optics and laser/receiver modules will be mechanically stable.
As discussed earlier with prof_braino, we've gone for a modular design that will allow for either "long range" or "short range" optics. For simplicity, we may be able to use the same laser lens (the smaller one in the pic) but a large or small receiver lens for the two different options.
The receiver lens shown here is 2" in diameter. It is really thick in order to give it a short focal length. This short focal length will help with the close distances (it has a massive depth of field), whilst the relatively large diameter will help with the long range performance (a larger diameter can collect more photons of returned laser light).
If we're lucky, a 14W laser with an APD detector should reach 50m whilst a 75W laser may reach over 200m with this optical configuration.
There will be a few more mechanical and optical parts needed to make a complete LRF but these first ones were the most complicated to design.
In the meantime, I'm learning how to use the Eagle PCB package so that I can make a demo board for the chip. Also, I'll try to get my hands on a P8X32A QuickStart board to test the SPI interface.
Will the emitter/detector be one module, and the DS00VQ100 chip mounted on its board be the a separate module?
I got the impression that the emitter and detector have very specific constraints, and would best be manufactured as a single unit.
Or, will the laser and lens be a sub module on the emitter/detector module?
At first I had the idea that a user would supply their own laser, but now it seems the laser, optics, and detector have to be specifically matched.
Is the power supply a specified module, or is this open to user implementation, such as "4.8v battery 4@AA or equivalent"?
For maximum flexibility I thought of using four electronic modules:
1. The Propeller board
2. A timer board carrying the chip
3. A laser module
4. A receiver module
The interconnections between these modules need to be as simple and reliable as possible so I've been reviewing the literature on ribbon cables. It may be possible to run some of the high speed signals through a ribbon using alternating ground connections. The impedance is about 100 ohms and can be matched at the termination. As long as the bandwidth of the signals doesn't get too high then maybe short ribbon cables will work.
Or, will the laser and lens be a sub module on the emitter/detector module?
The key to getting the optical components to work is a combination of the design of the lenses themselves and a mechanical structure to hold them stable. This requirement for excellent stability means that the laser lens and receiver lens need to be held together as a single entity - an optical module. The electrical laser and receiver modules will plug into this optical module. There will have to be provision for focus and alignment of the electrical modules so I thought of building each of them onto their own carrier that mates up with the optical module.
At first I had the idea that a user would supply their own laser...
Ideally, the laser module should be made in such a way that the user can solder in their own laser. Remember that these are high power, pulsed lasers, not the kind that are used in pointers. I'm working on the laser module design right now (at least I will be when the power comes back on) and there are some tricks to getting it to work properly. The physical layout seems to be critical with some components carrying very high peak currents. For the prototypes it may be best to build the laser into the laser module just to keep the number of ways that the smoke can come out down to a minimum.
... but now it seems the laser, optics, and detector have to be specifically matched.
There will have to be some matching in as much as particular types of laser and detector components will be used together. However, there is no particular constraint on who manufactures the parts and changes to the modules to fit a different size or shape of component should be relatively easy.
Is the power supply a specified module, or is this open to user implementation, such as "4.8v battery 4@AA or equivalent"?
I'm trying to stick to the design constraint of making each electronic module work on 4.8V. This is becoming a problem for the laser module where some of the circuitry operates above 5V. I might have to add a step-up supply and may end up putting this on a separate board - still busy with this.
The challenge for the last few days has been to design a small and simple laser module. This has been a kind of "unlearning" process because there are many complications in making a stable, high performance laser and what I wanted to do was strip out all the mumbo-jumbo and only include the absolute essential parts. To make things more interesting, I've tried only to use components that are available from RS Components. Other suppliers may have better and cheaper parts so this is just a starting point.
The SPL_LL85 hybrid laser from Osram includes some critical, built-in components that make the design much easier - a discharge capacitor and a FET. Firing a high power, pulsed laser takes more than 10Amps. Getting this current with a fast enough rise time (a few nano-seconds) can only be done using an avalanche driver transistor. These come in two types, ordinary BJTs (bipolar junction transistor) and FETs (field effect transistor). The BJTs are easy to use but tend to work at high voltages. The FETs can work at low voltage (9V in this case) but need more than 1A of gate current to get them to turn on.
Now before everyone starts shouting about how FETs don't need gate current, remember that we are working in nano-second time frames here and capacitance plays a much greater role than resistance. The gate resistance of a FET is very high but so is the gate capacitance. That means it takes a huge current to swing the gate voltage fast. In the proposed design there is a dedicated FET driver chip that can deliver 5A into the gate (ZXGD3003).
To eliminate switching ground currents between the timer board and the laser module the incoming fire signal is differential. There is a fast differential receiver chip to convert the fire signal into the single ended signal needed for the FET driver (AD8561). There are much cheaper parts but I couldn't find one at RS.
The laser fires when the capacitor, C, discharges through the FET. With this design there is very little control on the "shape" of the optical pulse coming out of the laser. Once the avalanche starts it runs until the capacitor is empty. The capacitor recharges through a resistor, R, which, whilst horribly inefficient, is very simple.
The timer chip (on another board) needs a signal from the laser module to indicate when light starts emitting. This defines the "zero" time of measurement. Most LRFs measure the light itself by feeding a small part of the outgoing pulse into the detector. This takes some fancy optics and electronics which I'd really like to avoid. In this design there is a small capacitor hooked onto the discharge line of the capacitor. As the FET begins to avalanche there is a sudden, steep fall in the voltage on the capaitor that is a prety good proxy for the outgoing light. This is the ZeroN signal.
The only remaining trick is to lay the PCB out in such a way that track lengths are really short. Otherwise the combination of high speed and high current can create all kinds of unwanted spikes and ringing.
10 Amps from a BC107?
The idea that transistors can generate enormous currents in avalanche breakdown mode has always fascinated me. It's not the kind of thing that gets taught in electronics courses or books but it is really easy to test out. Most transistors will avalanche without destroying themselves provided that the right conditions are met.
An easy experiment to perform is to take a transistor with a fairly low operating voltage (NPN). Connect it up with a 1ohm resistor between the emitter and ground (Re) and a 1k resistor (Rc) between the collector and an adjustable power supply (V). Put a capacitor of 0.01uF (C) between the collector and ground. It is important to keep track lengths really short. This circuit works well when using SMD components soldered directly onto each other with no PCB. Ordinary, leaded parts also work.
Drive a 1kHz square wave signal into the base of the transistor.
We would expect to see the collector voltage drop quickly when the transistor turns on and then climb back slowly as the capacitor recharges through Rc when the transistor turns off. There would be very little signal across the 1ohm resitor, Re, because the current flowing through the transistor is small.
Now gradually increase the voltage on the circuit. Depending upon your transistor you might have to go quite high and you will probably have to be above the rated breakdown voltage.
When you reach the avalanche point, the discharge time of the capacitor suddenly drops down to the nano-second time frame and a short spike shows on Re. You might even hear the circuit screaming...
Because of the speed you might need a fast scope (>100MHz) to see the signals in any detail. You can play around with the value of C and see what effect it has on the avalanche breakdown voltage - a bigger value reduces the voltage but increases the likelihood of the smoke getting out.
I've tried many different types of transistor in avalanche mode and some clearly work better than others. Zetex makes transistors specifically for this purpose. Other than for lasers and specialized scientific devices I can't think of any ordinary products that make use of avalanche mode transistors. Perhaps that's why few people know about them.
For maximum flexibility I thought of using four electronic modules:
1. The Propeller board
2. A timer board carrying the chip
3. A laser module
4. A receiver module
Good choice
reviewing the literature on ribbon cables. It may be possible to run some of the high speed signals through a ribbon using alternating ground connections.
Many of use have lots of 80 pin IDE hard disk drive cables laying around. Would these be possible? These should be easy to get in surplus. What is the recommended length? How many conductor would be needed, could we take one IDE cable and use it for several connection in the rig? I would not be against having the user cut the cable and solder all the conductors individually to the boards, if this would work.
the laser lens and receiver lens need to be held together as a single entity - an optical module. The electrical laser and receiver modules will plug into this optical module. There will have to be provision for focus and alignment of the electrical modules so I thought of building each of them onto their own carrier that mates up with the optical module.
This is the part where most of us have little of no experience, "best in your opinion" is the preferred criterion over cost, etc.
For the prototypes it may be best to build the laser into the laser module just to keep the number of ways that the smoke can come out down to a minimum.
Thanks, you have just saved many of us from several strings of failures.
a step-up supply and may end up putting this on a separate board - still busy with this.
Is this an area where you next extra help? You had mentioned that you usually have the support of a team, what team member would you like. There are a couple folks interested, but you seem to have everything well under control. If you can identify an area you want to hand-off or ask for input, I will start looking.
Laser Module
only to use components that are available from RS Components. Other suppliers may have better and cheaper parts so this is just a starting point.... FET driver (AD8561). There are much cheaper parts but I couldn't find one at RS.
....Other than for lasers and specialized scientific devices I can't think of any ordinary products that make use of avalanche mode transistors. Perhaps that's why few people know about them.
Laser module design is completely outside my area, but I have a thought on parts. Just as establishing requirements is criitical, finding parts to perform the specified function is critical. My question is about the impact of using one part in the development of the design, and using a differnt part later on. What impact will this have one the deisgn? I suspect that the specific part selected can have a large impact, changing a critical part may trigger a complete redesign. If this is the plan, then ok; otherwise, I would consider relaxing the "RS Components Only" guideline, or asking RS Components to get the part you really want. (For free, since your prototype is going to generate traffic for them?) The rule of thumb is to avoid half measures in the earlier part of the project, as these can compound to cuase trip-ups near the end. Just my 2 cents.
Many of use have lots of 80 pin IDE hard disk drive cables laying around
I started with a similar idea - use a standard type of ribbon cable rather than the exotic RF stuff. But then I discovered that RS stocks FFC (flat flexible cable). This is cool stuff with small connectors and it is really easy to use. It's also reasonably priced with 6" of cable at $1.50 and the connectors at $1 each. More importantly for this project, it has good high frequency characteristics. But, in truth, it's more of an emotional decision that an engineering one. I just like cool looking designs.
Is this an area where you next extra help?
Thanks for the offer (again). When I started this project I was very nervous about putting in the energy needed to get it finished. I'm sure that many of you out there have faced the demon of an unfinished project. I mentioned to prof_braino that whenever I finish a project I look in the mirror and can see how much older I've become. There is often a heavy emotional toll to pay when trying to push the boundaries of your own experience and knowledge and, as simple as this design may seem, it will still take hundreds of hours of work.
For now I'm enjoying learning how to do the electronic basics - finding parts and laying out PCBs. When I start to get tired I'll call for help.
Comments
Noted. Waiting for parts list
Noted
TOF laser chip $20
Propeller chip $8
75W laser $?
APD $?
TDC $20.00
APD $140.00
APD circuity $50.00 needs a reverse bias of 100+volts
75Watt Diode $45.00
Diode driver $100.00
Microcontroller $10.00 assuming propeller
plus lensing and some form of mounting that will allow for precision alignment
TDC = $20.00 (the timing chip)
APD = $37.25 (high quality type)
APD circuitry = $10 (the controller is on the chip)
Laser = 75W - $24.12; 25W - $17.30
Laser driver circuit = $15.00 (the controller is on the chip)
Propeller = $8.00
Total (25W, excluding PCB's and optics) = $107.55
What I find exciting here is that logicmax knows what he's talking about and his prices are very much the industry standard ($365.00). But take a look at the impact on the system cost of having the right kind of controller chip:
Saving on APD circuitry = $40.00
Saving on laser circuitry = $85.00
Total saving = $125.00
I would be very interested to know how much logicmax is paying for his timing circuitry. There may be further savings there as well.
As for handling the issues of paralax and focus, if the optical system is reasonably well designed in terms of lens diameters and focal lengths then the optics become "self adjusting" in that a graph showing signal strength against distance will be fairly flat. This means that it is not necessary to add any hardware to compensate for the parallax and focus imperfections.
I've worked for outfits that did not do this kind of investigation until the third generation of the product. Doing lots of investigation and making adjustments during the design phase seems to be paying off.
I looked at the Parallax proto boards. What about mating the LRF to the Gadget Gangster - so it plugs straight on top making use of the available power supplies and port pin connections.
The laser and receiver modules could connect to the LRF board via two separate ribbon cables. I think that we may just be able to get away without using any RF connectors if we impedence match the ribbon.
If your intent is to allow greater flexibility in placing the sub assemblies into custom robot configurations at the expense of more parts and a touchier design, I would advise saving that for later versions. Because you won't really know what configuration options folks will want till they start using it and asking for specific capabilities.
Compact pros:-
1. The matching of the high frequency signals is easier and there are less parts - especially connectors and cables which can be unreliable.
2. The cost will almost certainly be lower, but perhaps not by much.
Compact cons:-
1. Imagine how much light a 75 Watt laser gives off. This will be in very close proximity to the detector. I've worked on several projects where measurement errors in the return signal were traced to light leakage BETWEEN the layers of the circuit board. This happens because the laser is rated by the amount of light that comes out of the front face of the semiconductor. However, the laser is in fact symmetrical with an equal amount of light coming out of the back. Even metal packages leak an enormous amount of light through the insulating material around the connecting pins. Rule of thumb - encapsulate the laser in a black potting compound with just the front face visible. This is best done in a separate module.
2. The range of the LRF is determined mostly by the physical size of the optics. In order to reduce the effects of parallax you want the laser optics and the receiver optics as close together as possible. A number of manufacturers actually cut flats on the sides of the lenses so that they fit closer together. If the laser and receiver are on the same board then the spacing between them will restrict the kind of optics that they can use to one size only. If the lenses are any smaller then the laser and receiver will be too far apart. If the lenses are any bigger then the laser and receiver will be too close together.
A compromise solution:-
Most designs have either separate laser and receiver modules, or the timing circuit as part of the receiver with a separate laser module.
More research is needed I think - greater clarity on typical applications. I've read other threads where people wanted to go a long way - that means bigger optics. But for a small robot you want things to be as compact as possible. Compact or flexible/modular?
This is the fun part
Can the layout be design such that the use "paints" the back surfaces of the laser as the user-finishing step? If this results in another careful step for the user, but a significant simplification in physical design and decreased costs, this might be somthing to explore. Otherwise, if its too touchy and just has to be done right, make is a separate module and an have them potted.
I would guess an equal number would want extremely close as extremely far; and equal number to both of these will want somewhere in the middle. A design that allows a choice of optics would be most flexible even if there are only two choices. A design goal could be "simplest flexibility for optics".
So it looks like at least one subsystem must be on a separate board or module. I can't really get the trade-off points, can you post or PM or link some examples? Short of that, I would say to just do what easiest for the prototype. The novel bit is your TDC chip, once that is in a rig we can see what the implications of the particular implementation are.
I don't know it this makes any sence, but here goes: "Laser and reciever optics must be as close as possible". Why not use the same optics for both laser an reciever? You can get any closer than that. List the reasons why its impossible, maybe we can see where each item contradicts another, and design around it. This works sometimes, I think its called the "Triz" method.
Let's hope so.
There are almost as many optical configurations as there are manufacturers of LRF's. Some manufacturers even use different optics on different products like these two from Optech out of Canada.
I would recommend leaving the concentric optics to those who want to play around with polarizers and mirrors. There are some fun configurations that the telescope makers might try, like putting the laser in front of the receiver optics. This is brilliant for long range work but has problems close up. So how do we decide on the type of optics?
Now that is my kind of development philosophy! One repeating theme of our LRF design seems to be that the range requirement is poorly defined. That means that the optical system is also poorly defined and consequently, the spacing of the laser and receiver is in question. We have mentioned the possibility of light weight systems for aircrft use, short range systems for small bots and long range systems perhaps just for the fun of it. Ideally, the final solution should work for all of them.
At this point my intuition is ringing alarm bells. Flexibility implies modularity. This thing needs to be more like a Lego set than a Meccano set. Also, the chip needs to directly support the aspirations of the hobbyist. If I was playing with this thing what would I want to do? I would like to find the limits. That means minimum range, maximum range, best resolution, highest speed, smallest size, highest power and so on. The more limited the system is, the less exciting the learning experience will be. If it could do lots of cool stuff then I might also be willing to pay a few dollars more.
So what is the cost of this flexibility?
As a start I've extended the range of the chip to 370m (1100') so that hobbyists who have an old telescope lying around can configure a long range instrument with the laser mounted where the sighting scope goes and the receiver using the primary. This extra range is for free.
Next, I've looked at the cost of using RF cables to link the laser module and receiver module to the timer board. These are only a few dollars each and many companies now custom make any length that you want for very little charge. We could provide a standard set of cables in the kit but someone who wants to make a tiny system can get really short cables made, whilst someone who wants to push the range limit with large optics can get much longer cables made.
One trick is to make sure that the laser and receiver cables are the same length. That way altering the cable length will not affect the calibration of the system. This is because the "zero" signal and "return" signal will have to travel the same distance down the cables and so cancelling out the electronic delays.
I feel that a modular system offers the highest "fun factor" and so I propose that the laser, receiver and timer are all separate modules.
If we agree to a modular solution then the laser module becomes the next target for design "optimisation". We must be very careful about shipping lasers across national boundaries. Lasers are regulated by the FDA and the process of getting them to approve a product for import or export is tedeous and expensive. However, if the laser is already in the country then there are no restrictions.
It may be worthwhile designing the laser module in such a way that the end user can easily solder on their own laser. That way they can choose a low power or high power device to suit their application. Of course, we'll have to design the module in such a way that the "light leakage" mentioned earlier is not a problem. Perhaps prof_braino's "paint" solution could be used.
This sub forum, the sensors forum, does not get as much attention as for example the main propeller forum. There are several threads in other forums that discuss laser projects. The participants in the other threads among them seem to know everything about everything, except lack your particular experience with lasers.
Would this be a good time to invite these other folks to join this discussion? I would guess that these are the people that would be the earliest users of your results.
Now that we have discussed much for the highest level parameters, these folk might provide insight to actual "use-cases".
If these are not mutually exclusive, could both options be provided? That is, if it adds flexibility and does not increase cost.
I've attached the first pages of the draft datasheet. There are many timing diagrams that need finishing and my editor wants to change everything about the format (as usual) but the basic idea is there.
I would like to thank prof_braino for some of the ideas and, believe it or not, alex123 who suggested that some of the latest CMOS interfaces may be up to the task. He was correct. There is a type of CMOS called LVDS that is almost identical to LVPECL but with a slightly longer gate delay (200ps). Using this should be fine for the signal interface and will reduce the cost of the chip by around $4 (final prices still subject to confirmation).
DS00VQ100 Datasheet Rev0-00.pdf
Sent PM with review comments.
One thing about the previous laser projects, they all start out considering time of flight, but encounter issues and and switch to another method. So there should be a bunch of folks that are familiar and can comment on what you have put together.
How much money is nescessary to produce this chip?
If I understood right the chip is designed.
But who will gonna produce the chip?
Or are you using a CPLD or FPGA?
keep the questions coming
best regards
Stefan
All the internal components are designed and tested but I want to make some changes to suit the hobby market. For example, today I have been working on bringing out some of the more interesting internal signals onto test pins. This will allow someone with a relatively low frequency 'scope to see what's going on inside so that they can fiddle with the external components and try to improve their LRF design. I don't normally do this in an industrial design because the external components (laser, receiver etc) are fixed, and once in production don't change from unit to unit. But for a hobbyist, I think that playing with the design is half the fun.
Another thing that prof_braino has been emphasizing is the correct use of terminology - avoiding jargon etc. I've been rewording the datasheet and other support documents to get some consistent and acceptible words that are meaningful to someone with no experience in LRF design.
This is the next revision of the Datasheet for the LRF chip. There's still plenty to do. I've reworded a few sections to give a little more clarity to the operation.
The images are fuzzy because they haven't been converted to vectors yet. That will only be done in the final draft because I keep changing them.
Below is a photo of two LRF's. The little one works up to 10m and was designed as a level measuring instrument for small silos. The larger one has a range of 120m and was used underground in the gold mines of South Africa.
The optics of the small unit are concentric whilst the larger unit uses a side-by-side configuration. I think that the new LRF design needs to be able to work in both types of device so flexibility and modularity are very important.
core = 1.5V
SPI and general I/O = 3.3V
LVCMOS = 2.5V or 3.3V
LVDS = 2.5V
LVPECL = 3.3V
There are three "differential outputs" on the chip - Laser PWM, Laser fire and APD PWM. At a minimum, the Laser fire pulse needs to use a high speed protocol - either LVPECL or LVDS. By comparison, the PWMs are relatively slow speed so they could (theoretically) be driven by two complimentary LVCMOS pins instead of genuine, differential signals. Not perfect perhaps, but I can stagger the switching points to avoid SSOs (simultaneous switching output glitches on the power rails).
This means that the chip needs three power rails:
core = 1.5V
SPI and general I/O = 3.3V
LVDS = 2.5V
This bugs my sense of engineering esthetics - I'd really like to throw out one of these power rails. The core voltage has to stay - no way around that. But having both 2.5V and 3.3V for the I/O ports seems messy. Perhaps these could be combined into one. But which one?
After trying all kinds of different combinations it seems that the determining factor is that the Propeller needs a 3.3V SPI interface to work properly. It will not function correctly with a 2.5V SPI interface. Making all the I/O ports 3.3V means changing from the 2.5V LVDS in favor of the 3.3V LVPECL and viola! Only two power rails.
core = 1.5V
SPI, general I/O and LVPECL = 3.3V
The compromise is that LVPECL draws more power than the LVDS (6mA @ 3.3V instead of 3mA @ 2.5V) but I think that is an acceptible price to pay for getting rid of an entire power supply. From a functional point of view, either protocol would have worked fine, but the final selection criterion has turned out to be the power rails.
Back to the drawing board...
Vcc is 1.5V and is the power to the core.
VCCB's are 3.3V and are the power to each side of the chip for the I/O's.
Even though there are four VCCB's, they will all be attached to the same 3.3V power supply. They will be independently decoupled to reduce cross-talk between the port pins.
The lenses are acrylic (plastic) and in the picture you can see the first part of the holder that will carry the lenses, laser module and receiver module. This holder is molded in a hard plastic with glass fiber embedded in it. This will ensure that the optics and laser/receiver modules will be mechanically stable.
As discussed earlier with prof_braino, we've gone for a modular design that will allow for either "long range" or "short range" optics. For simplicity, we may be able to use the same laser lens (the smaller one in the pic) but a large or small receiver lens for the two different options.
The receiver lens shown here is 2" in diameter. It is really thick in order to give it a short focal length. This short focal length will help with the close distances (it has a massive depth of field), whilst the relatively large diameter will help with the long range performance (a larger diameter can collect more photons of returned laser light).
If we're lucky, a 14W laser with an APD detector should reach 50m whilst a 75W laser may reach over 200m with this optical configuration.
There will be a few more mechanical and optical parts needed to make a complete LRF but these first ones were the most complicated to design.
In the meantime, I'm learning how to use the Eagle PCB package so that I can make a demo board for the chip. Also, I'll try to get my hands on a P8X32A QuickStart board to test the SPI interface.
Will the emitter/detector be one module, and the DS00VQ100 chip mounted on its board be the a separate module?
I got the impression that the emitter and detector have very specific constraints, and would best be manufactured as a single unit.
Or, will the laser and lens be a sub module on the emitter/detector module?
At first I had the idea that a user would supply their own laser, but now it seems the laser, optics, and detector have to be specifically matched.
Is the power supply a specified module, or is this open to user implementation, such as "4.8v battery 4@AA or equivalent"?
Quickstart board is an excellent choice.
The protoboard http://www.parallax.com/Store/Microcontrollers/PropellerDevelopmentBoards/tabid/514/CategoryID/73/List/0/SortField/0/Level/a/ProductID/423/Default.aspx
also is a good choice, it has the typical power and eeprom configurations commonly used with the propeller chio.
The schmart board http://www.parallax.com/Store/Microcontrollers/PropellerDevelopmentBoards/tabid/514/CategoryID/73/List/0/SortField/0/Level/a/ProductID/568/Default.aspx
and Gadget Gangster Propeller Platform
http://www.parallax.com/Store/Microcontrollers/PropellerDevelopmentBoards/tabid/514/CategoryID/73/List/0/SortField/0/Level/a/ProductID/711/Default.aspx
are also popular implementations of the common configuration.
1. The Propeller board
2. A timer board carrying the chip
3. A laser module
4. A receiver module
The interconnections between these modules need to be as simple and reliable as possible so I've been reviewing the literature on ribbon cables. It may be possible to run some of the high speed signals through a ribbon using alternating ground connections. The impedance is about 100 ohms and can be matched at the termination. As long as the bandwidth of the signals doesn't get too high then maybe short ribbon cables will work.
The key to getting the optical components to work is a combination of the design of the lenses themselves and a mechanical structure to hold them stable. This requirement for excellent stability means that the laser lens and receiver lens need to be held together as a single entity - an optical module. The electrical laser and receiver modules will plug into this optical module. There will have to be provision for focus and alignment of the electrical modules so I thought of building each of them onto their own carrier that mates up with the optical module.
Ideally, the laser module should be made in such a way that the user can solder in their own laser. Remember that these are high power, pulsed lasers, not the kind that are used in pointers. I'm working on the laser module design right now (at least I will be when the power comes back on) and there are some tricks to getting it to work properly. The physical layout seems to be critical with some components carrying very high peak currents. For the prototypes it may be best to build the laser into the laser module just to keep the number of ways that the smoke can come out down to a minimum.
There will have to be some matching in as much as particular types of laser and detector components will be used together. However, there is no particular constraint on who manufactures the parts and changes to the modules to fit a different size or shape of component should be relatively easy.
I'm trying to stick to the design constraint of making each electronic module work on 4.8V. This is becoming a problem for the laser module where some of the circuitry operates above 5V. I might have to add a step-up supply and may end up putting this on a separate board - still busy with this.
The challenge for the last few days has been to design a small and simple laser module. This has been a kind of "unlearning" process because there are many complications in making a stable, high performance laser and what I wanted to do was strip out all the mumbo-jumbo and only include the absolute essential parts. To make things more interesting, I've tried only to use components that are available from RS Components. Other suppliers may have better and cheaper parts so this is just a starting point.
The SPL_LL85 hybrid laser from Osram includes some critical, built-in components that make the design much easier - a discharge capacitor and a FET. Firing a high power, pulsed laser takes more than 10Amps. Getting this current with a fast enough rise time (a few nano-seconds) can only be done using an avalanche driver transistor. These come in two types, ordinary BJTs (bipolar junction transistor) and FETs (field effect transistor). The BJTs are easy to use but tend to work at high voltages. The FETs can work at low voltage (9V in this case) but need more than 1A of gate current to get them to turn on.
Now before everyone starts shouting about how FETs don't need gate current, remember that we are working in nano-second time frames here and capacitance plays a much greater role than resistance. The gate resistance of a FET is very high but so is the gate capacitance. That means it takes a huge current to swing the gate voltage fast. In the proposed design there is a dedicated FET driver chip that can deliver 5A into the gate (ZXGD3003).
To eliminate switching ground currents between the timer board and the laser module the incoming fire signal is differential. There is a fast differential receiver chip to convert the fire signal into the single ended signal needed for the FET driver (AD8561). There are much cheaper parts but I couldn't find one at RS.
The laser fires when the capacitor, C, discharges through the FET. With this design there is very little control on the "shape" of the optical pulse coming out of the laser. Once the avalanche starts it runs until the capacitor is empty. The capacitor recharges through a resistor, R, which, whilst horribly inefficient, is very simple.
The timer chip (on another board) needs a signal from the laser module to indicate when light starts emitting. This defines the "zero" time of measurement. Most LRFs measure the light itself by feeding a small part of the outgoing pulse into the detector. This takes some fancy optics and electronics which I'd really like to avoid. In this design there is a small capacitor hooked onto the discharge line of the capacitor. As the FET begins to avalanche there is a sudden, steep fall in the voltage on the capaitor that is a prety good proxy for the outgoing light. This is the ZeroN signal.
The only remaining trick is to lay the PCB out in such a way that track lengths are really short. Otherwise the combination of high speed and high current can create all kinds of unwanted spikes and ringing.
10 Amps from a BC107?
The idea that transistors can generate enormous currents in avalanche breakdown mode has always fascinated me. It's not the kind of thing that gets taught in electronics courses or books but it is really easy to test out. Most transistors will avalanche without destroying themselves provided that the right conditions are met.
An easy experiment to perform is to take a transistor with a fairly low operating voltage (NPN). Connect it up with a 1ohm resistor between the emitter and ground (Re) and a 1k resistor (Rc) between the collector and an adjustable power supply (V). Put a capacitor of 0.01uF (C) between the collector and ground. It is important to keep track lengths really short. This circuit works well when using SMD components soldered directly onto each other with no PCB. Ordinary, leaded parts also work.
Drive a 1kHz square wave signal into the base of the transistor.
We would expect to see the collector voltage drop quickly when the transistor turns on and then climb back slowly as the capacitor recharges through Rc when the transistor turns off. There would be very little signal across the 1ohm resitor, Re, because the current flowing through the transistor is small.
Now gradually increase the voltage on the circuit. Depending upon your transistor you might have to go quite high and you will probably have to be above the rated breakdown voltage.
When you reach the avalanche point, the discharge time of the capacitor suddenly drops down to the nano-second time frame and a short spike shows on Re. You might even hear the circuit screaming...
Because of the speed you might need a fast scope (>100MHz) to see the signals in any detail. You can play around with the value of C and see what effect it has on the avalanche breakdown voltage - a bigger value reduces the voltage but increases the likelihood of the smoke getting out.
I've tried many different types of transistor in avalanche mode and some clearly work better than others. Zetex makes transistors specifically for this purpose. Other than for lasers and specialized scientific devices I can't think of any ordinary products that make use of avalanche mode transistors. Perhaps that's why few people know about them.
Many of use have lots of 80 pin IDE hard disk drive cables laying around. Would these be possible? These should be easy to get in surplus. What is the recommended length? How many conductor would be needed, could we take one IDE cable and use it for several connection in the rig? I would not be against having the user cut the cable and solder all the conductors individually to the boards, if this would work.
This is the part where most of us have little of no experience, "best in your opinion" is the preferred criterion over cost, etc.
Thanks, you have just saved many of us from several strings of failures.
Is this an area where you next extra help? You had mentioned that you usually have the support of a team, what team member would you like. There are a couple folks interested, but you seem to have everything well under control. If you can identify an area you want to hand-off or ask for input, I will start looking.
Laser module design is completely outside my area, but I have a thought on parts. Just as establishing requirements is criitical, finding parts to perform the specified function is critical. My question is about the impact of using one part in the development of the design, and using a differnt part later on. What impact will this have one the deisgn? I suspect that the specific part selected can have a large impact, changing a critical part may trigger a complete redesign. If this is the plan, then ok; otherwise, I would consider relaxing the "RS Components Only" guideline, or asking RS Components to get the part you really want. (For free, since your prototype is going to generate traffic for them?) The rule of thumb is to avoid half measures in the earlier part of the project, as these can compound to cuase trip-ups near the end. Just my 2 cents.
I started with a similar idea - use a standard type of ribbon cable rather than the exotic RF stuff. But then I discovered that RS stocks FFC (flat flexible cable). This is cool stuff with small connectors and it is really easy to use. It's also reasonably priced with 6" of cable at $1.50 and the connectors at $1 each. More importantly for this project, it has good high frequency characteristics. But, in truth, it's more of an emotional decision that an engineering one. I just like cool looking designs.
Thanks for the offer (again). When I started this project I was very nervous about putting in the energy needed to get it finished. I'm sure that many of you out there have faced the demon of an unfinished project. I mentioned to prof_braino that whenever I finish a project I look in the mirror and can see how much older I've become. There is often a heavy emotional toll to pay when trying to push the boundaries of your own experience and knowledge and, as simple as this design may seem, it will still take hundreds of hours of work.
For now I'm enjoying learning how to do the electronic basics - finding parts and laying out PCBs. When I start to get tired I'll call for help.