PDA

View Full Version : LV-MaxSonar-EZO sensor



cross1933
04-08-2009, 06:20 AM
Anyone have experience with the LV-MaxSonar-EZO sensor,http://www.maxbotix.com/uploads/LV-MaxSonar-EZ0-Datasheet.pdf. (http://www.maxbotix.com/uploads/LV-MaxSonar-EZ0-Datasheet.pdf) ,


Using output "B" would be my choice for a beam pattern.What I like is the two types of output signal, low current and internal trigger.

the sensor I am comparing this to is the Devantech SRF05.

robologist
04-08-2009, 06:29 AM
Since this is a question about another sensor other than the GP2D12, it probably deserves it's own thread rather than hijacking this one. [mod note: moved into new thread - Alex]

I really enjoy the Max sonars, actually prefer them to the Devantechs. I think Adrenalynn prefers the Devantechs though, so each to their own.

Checking over the datasheet linked, the B pattern is how a 1" diameter rod would be detected, and the other patterns are for different size objects and how they are detected. So all 4 patterns exist with a given sonar, just depends on the size of the reflecting object.

lnxfergy
04-08-2009, 07:55 AM
Anyone have experience with the LV-MaxSonar-EZO sensor,http://www.maxbotix.com/uploads/LV-MaxSonar-EZ0-Datasheet.pdf. (http://www.maxbotix.com/uploads/LV-MaxSonar-EZ0-Datasheet.pdf) ,


Using output "B" would be my choice for a beam pattern.What I like is the two types of output signal, low current and internal trigger.

Darn, you moved this just as I posted -- I'll put this here just for later reference.

What output mode do you plan to use? I've found that the analog one is pretty jumpy, you'll have to add filtering - and even then your environment might be too noisy for it to work. I know Connor has had pretty good results with the Serial mode.

Also note, you'll likely have to put a filter on the power input if your robot's power supply isn't perfect (which it's almost certainly not going to be).

Personally I prefer the Devantech sensors, I've just never gotten the Maxbotix ones to work like I want.

-Fergs

cross1933
04-08-2009, 08:18 AM
Darn, you moved this just as I posted -- I'll put this here just for later reference.

What output mode do you plan to use? I've found that the analog one is pretty jumpy, you'll have to add filtering - and even then your environment might be too noisy for it to work. I know Connor has had pretty good results with the Serial mode.

Also note, you'll likely have to put a filter on the power input if your robot's power supply isn't perfect (which it's almost certainly not going to be).

Personally I prefer the Devantech sensors, I've just never gotten the Maxbotix ones to work like I want.

-Fergs

I am not fixed on which output, but the option to have either is appealing. I will probably opt for the digital output.

I have not measured the noise from the regulator circuits, a good idea though.

lnxfergy
04-08-2009, 08:54 AM
There are actually 3 outputs though: analog, digital pulse, and serial data. In reality, if your sensor is more than a few inches away from your microcontroller, the analog mode probably won't be very useful.

I've found that the Maxbotix sensors are extremely finicky as to power supply, where the Devantech ones are quite robust under really cruddy power supplies.

I think something to point out is that I demand a lot from my sonar sensors -- if you are just using them as a non-contact bumper (is something closer than 1ft?) than you'll probably be fine with any of the sensors out there. On the other hand, I'm trying to use these for map building... which means that random outputs are not acceptable. Sensors I've tested over the past few years for various projects (and the quantity of each I've used for testing): EZ-1 (4), EZ-3 (2), SRF-04/SRF-05 (4), SRF-08 (2), SRF-02 (6). In the end, I still like the SRF-05 best - then the I2C SRF-02/SRF-08 next (I2C itself has some issues with bus wire length becoming problematic)

-Fergs

cross1933
04-08-2009, 09:45 AM
I am using them for contact avoidance. I need to make sure the vehicle will not come in contact with object in the environment of use. During operation the possibility of object contact is reduced(operator error would be the cause), what concerns me is if the wireless signal is lost. The vehicle will continue on its current direction if the wireless signal is lost. Objects to avoid consist mostly of metal with some concrete walls.

robologist
04-08-2009, 10:45 AM
Hey Fergs, do you have any links to your sonar tests? It would be interesting to see the data. And probably useful to quantify errors.

lnxfergy
04-08-2009, 11:23 AM
Hey Fergs, do you have any links to your sonar tests? It would be interesting to see the data. And probably useful to quantify errors.

I haven't really posted anything. Here goes a quick description, although not exactly empirical data. The OP asked for our opinion, so here goes.... YMMV:

Test setup for each was: 12" lead (hitec servo pigtails for maxbotix/srf-05, 22ga stranded for I2C based devices) from AVR controller board, pointed at wall (90 degrees, 45 degrees), distances from 12" to 8'. I was taking about 200 readings in each location (just let it run and read) and seeing what the error rate was. The devantech SRF-02/SRF-05 were just plugged right into the the 5V bus and would consistently have only 5-10 "bad" readings out of 200 (plus or minus 3 inches was deemed acceptable for what i was doing). The maxbotix gave terrible data without any filters. By adding a filter to the analog output I was able to get it to about 50-75% reliable, adding a filter to the power input didn't seem to improve it any further (filters were added as per the spec sheet, I know Connor has also had trouble with the power supply filtering, but I think has made it work adequetly). I was pretty much just fed up at that point and switched to my SRF-02s. I also had a maxbotix ez-3 on GMR originally, and found that it wildly oscillated when he was driving around... switched to a SRF-02, problem fixed...

I had used the maxbotix sensors since it seems lots of people were raving about them... but I just couldn't make them work to my liking. It seems most people are using them for non-contact bumpers, I'm not finding many people using them for mapping/etc (whereas I have seen the SRF sensors used).

I'm also not entirely sold on these single transducer models. I personally feel that the 2-transducer models are just more reliable, especially if you don't want to run over people. While single transducer models allow you to "back the range right up to the sensor" in my testing I feel that they tend to miss clutter and people. An interesting point to note is that at the Trinity Fire Fighting contest last weekend, nearly everyone was using devantech sensors (and almost all the 2-transducer models).

-Fergs

robologist
04-09-2009, 04:48 PM
Had a bit of a delay in reply, but interesting data gathering, Fergs. My own experience is with some Polaroid camera hack sonars(2), Sonaswitch Mini S (2), Devantech SRF04 (2), the original MaxSonar EZ1 (2), Devantech SRF05, Devantech SRF08 (1), and the LV-Maxsonar-EZ0 (2). Most of these were on 24 guage wire, typically 7". For the MaxSonars, I'd vary with either analog and pulse width measurements. The Pulse width seemed the most stable, but the readings that were analog still were consistently better than the other sensors used. The Devantechs would show an occasional glitch reading in the midst of some steady values. The Sonaswitch were all over the place, but there may have been some power supply issues there. The camera modules needed more stability as well, but these last 2 would be good for longer range sensing. The I2C of the SRF08 had to be bit-banged since the micro being used didn't natively support it, and it was borrowed from a friend. I still use the code on a Devantech compass, but really don't like I2C to sensors, much prefer direct reading by the processor.

lnxfergy
04-09-2009, 05:24 PM
I completely concur on the I2C comments.. I'd much prefer direct connection (and thats pretty much the route I plan to go in the future).

What sort of surfaces where you testing against?

-Fergs

robologist
04-10-2009, 12:47 AM
Most testing was against paint surfaced sheetrock, moving the "sensor-pod" various distances from touch to 6 feet or so. Initially I'd wire something up, fire it at the tile ceiling (fiber like material typically of drop ceilings) that you'd expect were a little absorbtive of the sound, but seemed to reflect fine to get some initial data, then aim at a wall for more tests . Another setup was using a 1x6, to duplicate someone elses tests observed. He had mounted an SRF04 and an early Maxsonar on a 3 foot 2x4, and had another 3 foot 2x4 as a target which was set at various distances. He had gotten good results with the Maxsonar in comparison, and I had wondered about it, so tried my own duplication of his tests, but running a different micro and adding the Sonaswitch. Other tests were sitting on various robots, where sonar height and aim were discovered to matter a bit when the Maxsonars started picking up cracks in the floor. So they and other sonars were either mounted a little higher off the ground, or aimed slightly up to lessen these false obstacles. I used a DSP56F05 (Freescale/Motorola) in both analog (12 bit) and pulse width measurements.

maxbotix
04-14-2009, 03:14 PM
What I would like to state here is that we at MaxBotix Inc., understand the issues when using ultrasonic sensors. With that we are in a constant state of improving our products and expanding our product line. Our company is in constant growth, with sales doubling every year, and even with the economy we are in now (Spring 2009) we have had to move to a larger facility just to handle our growth (and we did the same thing last July!). (Our new address will make it onto the website soon.) This not just because of our product’s quality & low-price, but it also because we hear and respond to our customer’s needs. In addition, we help anyone with an issue, from the school kids, to the design engineers. (To date we have handled every email request for help that we get with the same care. In addition, we try to scope the web and try to handle even the things that are not brought to our direct attention.)

We would also like the opportunity for everyone to bring to our attention, any issue, so we can eliminate the issue either in our current product line, or make a new product that overcomes the issue. For all of our current products, all of our outputs are 100% tested at the factory, to real targets at long range, and all must match and all must be stable or the product does not ship. In addition, the beam patterns are calibrated into each part at the factory, again, just to make the part to part beam width consistent. This is why we are climbing, upward in sales.

This said, I would like to address the things brought up here with the same care that we handle our emails.

Long Cables to our MaxSonar Product
Our sensors can be wired to places a long distance away. All of the outputs are very good and have been tested through 1000 feet of cable and all are very good even with this cable length. There was no degradation to the analog output (or any of the outputs). (This 1000-foot test was also performed and worked just as well for our WR1 sensor.) While it is not yet on our website, we sell seven conductor shielded cable to those that would like to run the long distances. (We in fact could not see why the sensors would not work well, out many 1000s of feet.)

Dirty Power Supply
MaxBotix Inc., highly recommends clean power to our sensor, but we ran this test just because of this posting. For our noise source, we used the Sharp GP2D05 sensor and connected it directly to our sensor. For power we used an EXTECH power supply and powered both the Sharp GP2D05 and the LV-MaxSonar-EZ1. We looked at the +5V power with an oscilloscope measured a maximum of 3.24V peak to peak noise on the +5V line, with regular continuous excursions of 3V peak to peak. (To us at MaxBotix Inc., 3V of noise on a 5V line seems like a lot of noise, but this is the normal noise output that the Sharp IR range sensor has.)

Seeing is believing, so we took an oscilloscope plot of this sensor powered with our regular EZ1 sensor and tried to understand just how bad the sensor might be in the presence of this kind of noise.
http://app4.websitetonight.com/projects/3/4/4/0/34404/uploads/Scope1.jpg

CH3 - Input power of 5V, together with the noise produced by the Sharp distance sensor.
CH4 – Our Pulse Width output. (Note that the input noise is plainly visible on the top portion of the pulse width output even with the 1.5V/div setting on the oscilloscope.)
CH2 – Our internal processed analog signal showing just how little effect the outside noise has on the actual signal we use to determine the range.

This is a real world test and the real world results show that although the outside noise does affect the sensor, and the noise might be considered “jumpy”, is the LV-MaxSonar-EZ1 really expected to operate with this noise on the input power? Yet the LV-MaxSonar-EZ1 actually does remarkably well!

Our normal EZ1 sensor will have some abnormal readings with this very high noise being present. (The WR1 is OK with this type of noise, mainly due to the internal 100uF capacitor.) For the LV-MaxSonar-EZ1 we found that the sensor produced the correct output reading (within one inch) 91.3% of the time, even when powered with 5V power this dirty. The serial and pulse width outputs were equally affected, and the analog output was worse than the others were.

Analog Output
Our analog output matches, and is as good as the other outputs in a noise free environment. When noise is on the +5V line, the noise is conducted directly into the microcontroller. Since the microcontroller sets the analog voltage it is more affected by this outside noise. If you really must operate within this noise, we have recommended the 100-ohm resistor and the 100uF capacitor (Digikey part number P803-ND) will work very well and no false readings will occur with this filter. Please Click this link the view the complete write-up on our website http://www.maxbotix.com/MaxSonar-EZ1__FAQ.html#New_Flash_UAV_and_Mobile_Robotic_Use rs (http://www.maxbotix.com/MaxSonar-EZ1__FAQ.html#New_Flash_UAV_and_Mobile_Robotic_Use rs) . The analog output is not any worse or “jumpy” as some might say. This is readily seen if using a four-channel oscilloscope and if you put all outputs on the scope at the same time.

Even so, the majority of our users do in fact use the analog output, and do in fact use no filtering on the analog output, (but they assure they have OK power where Vcc can range from 2.5V to 5V, but it should be stable).

High end sensors
The MaxSonar product line has consistently been evaluated to the others and we have consistently. High-end companies have stated to us “your sensors work perfectly”, and they are buying high quantities for collusion avoidance, room mapping and such. The fact that they can get consistent part to part beam widths, from lot to lot means a great deal to these companies.

So while there are a number of reasons that a sensor might or might not perform a certain way, MaxBotix Inc., does produce a sensor that works well and we calibrate it and verify it at our factory before it ships. Please, if you have any trouble with the sensor, please send us an email and we will support you. (We do get a number of emails each and every day, so sometimes we may take a day or two to response, but if it takes longer than two days please resend it, as it was lost in the delivery.)

Regards,

Bob Gross
CEO of MaxBotix Inc.
Web: www.maxbotix.com (http://www.maxbotix.com/)
Email: [email protected] ([email protected])