Machsupport Forum
Mach Discussion => General Mach Discussion => Topic started by: Sargon on August 13, 2011, 09:14:12 AM
-
I want to use an ultrasonic sensor for height control as my sheets are not always perfectly flat. Blowing compressed air will clear the chip so I get accurate readings from the sensor. The ultrasonic sensor is ideal for this application as it can be used on transparent material and doesn't contact/damage the material.
The sensor output is 0-10V. I can use an op-amp and adjust the gain to condition the signal for THC or similar device. Or I could use it to compare to a set-point to generate THC UP or THC DOWN, but that doesn't solve the problem of no PID loop. I need PID to smooth out the signal so it's not constantly moving up and down. Any recommendations that don't involve electronic design, or at least very simple with minimal development time?
Has anyone developed a PID that works well for this application? If not, can anyone recommend a THC for this application?
Any thoughts?
Someone must have done something like this already!
Thanks,
Chris
-
Any of the affordable THC's out there (hobby level) do not have built in PID loops. They simply monitor tip voltage and send THC UP/DOWN commands to Mach. The lack of a PID loop is exactly why you get excessive oscillation if the THC feed rate is set too high. If you want to spend some money, you can get a stand alone THC that operates outside of Mach3. These THC's will have built in PID loops and control the Z axis independently of Mach3.
Which way do you want to go? If you're looking for hobby level THC that can just command Mach3 to make the THC UP/DOWN moves, you can build something easily. An stand alone industrial THC with PID capability can cost a few grand. Thermal Dynamics gets just over $10k for theirs.
One thing to consider is that once you move the Z over to an external THC, you will no longer be able to control it using Mach3 directly. When plasma cutting, other than setting initial torch height and retracting prior to rapids, the THC controls the Z. There are no Z feedrate moves in the program.
Are you looking to just measure the material height and apply a Z offset to compensate for variations in material height?
-
Any of the affordable THC's out there (hobby level) do not have built in PID loops. They simply monitor tip voltage and send THC UP/DOWN commands to Mach. The lack of a PID loop is exactly why you get excessive oscillation if the THC feed rate is set too high.
That sucks!
An stand alone industrial THC with PID capability can cost a few grand. Thermal Dynamics gets just over $10k for theirs.
That is absolutely ridiculous! PID is bordering on ancient technology! We're still in 2011, right? Good. When you said that I thought maybe I had gone back 50 years.... PID has been around almost as long as electronics... Wow... Thanks for the insight. If I need it I'll be happy to develop it myself at those prices. Then sell it for a few hundred and steal the market.... hmmm..... lol
Are you looking to just measure the material height and apply a Z offset to compensate for variations in material height?
That is exactly what I need, but if it does see a chip I don't want it pulling Z up.... I suppose using a slow feedrate for compensation may work alright... Any suggestions?
-
The low cost THCs DO have the pid loop built in and cost around $500 or less complete . It is a basic function of the system. You can also use a simple PLC and IF you can do the programming build in a PID loop to do it as welll.
THE advantage of using an existing DTHC to do this IS????????? They react VERY VERY VERY fast to changing conditions and that is REquired IF you are going to try and keep up with cutting and Z height changes.
Now that brings up another related question HOW FAST is the sonic sensor and HOW accurate is it. A plasma torch is a very course user of the Zheight compared to routing . A plasma torch can have minor changes in Z height and you would probably not ever notice it in the cut quality. Do that too much with routing and you have an ugly mess.
SO are there any specs out there on the Sonic height sensor?? I personelly would not make a sonic sensor my first choice of sensors as it will be effected by swarf conditions and speed of Cut.
ANOTHER thought would be to use a linear scale to monitor the Zheight and slave it to the Z heightaxis so that the Quad ouput of the sensor DRIVES the Z axis to react like a CNC control BUT follow the surface directly.
Just some thoughts, (;-) TP
(;-) TP
-
(;-) This all sounds cool but I bet it would be better in th elong run to just invest in a good vacuum hold down system and get back to cutting(;-)
(;-) TP
-
I said the material is not flat, not that I have a problem holding the material. I do indeed use a vacuum system strong enough to pull down 3/4" HDPE. That definitely isn't the problem. The material itself can have variations, and some materials are either too thick or too ridgid to hold with vacuum. That's why I want height control.
Thanks in advance for any help you can offer!
Chris
-
OK before we go on lets clear somthing up. You seems to want a proven design to do this.
"I would much rather just use an existing, proven design then reinvent the wheel"
BE aware there is no such known function for MACH3 that currently does this. You are treading on unproven ground. Both hardware and software wise.
THe only KNOWN solutions for this ARE machine the surface flat before you work with it OR use a floating Zhead to set Z cut height.THesae are industry standard methods and are proven to work. THe rest are SWAG in nature and may not give the results you require.
(;-)TP
-
Sorry I didn't have time earlier to respond to everything.
I wanted to use something existing just to reduce the time required for the project. I'm not quite sure what you mean by a floating Z... Does it have a depth guide that contacts the material?
The ultrasonic sensor has a response time of 2.5ms.... should be plenty fast for a table with max speed of 130ipm. That translates to a response roughly every 0.0055" at max speed. The swarf should be cleared with compressed air, but may randomly be picked up by the sensor if they happen to move past it at the wrong time, but should be rare. That is one reason for wanting PID.
What do you mean by a linear scale? Please elaborate, I'm not sure what you're referring to.
Thanks for all the help/ideas/insight. It's much appreciated.
I thought surely someone would have done this by now... but maybe not...
Chris
-
"The low cost THCs DO have the pid loop built in and cost around $500 or less complete ."
Which ones? Anything that interacts with Mach3 through THC UP/DOWN commands doesn't have a PID loop, or if it does, it doesn't do any good. The idea behind the PID loop is to adjust the feed rate based on the distance of travel and to control acceleration to reduce/eliminate over-travel. Most of these low cost THC's simply trigger the Mach3 THC commands, which in turn move the Z at the set THC feed rate with no acceleration applied until the input is turned off. when it's turned off, there is always some over-travel, which is dealt with by reducing THC feed rate to where the "bounce" is acceptable, which as you pointed out, is not as big a deal in plasma as it would be in routing.
-
"That is exactly what I need, but if it does see a chip I don't want it pulling Z up.... I suppose using a slow feedrate for compensation may work alright... Any suggestions?"
Then you don't need a THC. You want to read the material height, do some math and apply a height offset to the Z to compensate for changes in material thickness. A THC controls the Z axis. Sounds like that's not what you're after.
The only way this could work is if you wanted to take say....a consistent 1/4" off the top of the material. Then you could adjust the THC to run the Z at a constant - 0.25...Which would be only as accurate as you ultrasonic sensor.
PID loops have been around forever, but there's more to an industrial controller that the PID loop. It's all relative. When you drop $200k on a brand new Koike MasterGraph, $30K for a controller and $10K for a THC doesn't sound so bad.
-
Linear scale is a measuring device linear in nature that outputs vaues OR signals that are used in measuring a comon signal is a quadrature output where there are 4 signals generated for each unit of movement. Basically same as a servo encoder but only it is linear not rotary.
It may be possible to have those ouputs Drive the Z axis motor movement up and down instead of directly from MACH3 It is called electronic gearing or slaving.
Yes the floating Z uses a mechanical platform mounted to Z in order for the cutter to follow the surface, not be controlled by the COntroller WHILE it is cutting.
Another thing about sonic sensors is the doppler effect of moving objects tends to distort distance measurements. Flying kerf and moving table may cause problems.
(;-)TP
-
MIght want to check again on the pid on low cost DTHCs THE thcup/down is only the fairly high speed digital INPUT to control the Z thru MACH.
IF those signals are erratic then the Z will be erratic. The PID filtering takes place in the DTHC.
Being that Mach3 is NOT a real time controller your on the fly height correction calculations are not going to work well. Yes you can use #vars for Zheight control but ONCE the moves are loading into the buffer then that is it. IF you try to calculate the variable at the point the #var is loaded into the buffer you will find you cannot comp something that has not happened yet. the next problem is the only way you have to update those vars are thru either the brains or the macropump. At a MAX of 10 updates per sec you will fall FAR short of the speed to do it well.
The ONLY access you have at that point IS the THC functions up/down to comp the surface heights based on what the DTHC sees. AND it is NOT value based it is like playing ping pong slapping the ball back and forth. THe Z at that point will not even use the motor vel/accel parameters it just basically slaps the motor wide open in the direction it is to move and lets the DTHC deal with it being correct in height. Woops overshot slap it back in the other direction. That is where the PID has to step in. Same basic princible as a servo motor drive.
Just a thought, (;-)TP
-
Thank you for the lesson rrc1962 and br549, it has been extremely enlightening, and thanks for the quick responses... I now see why this has never been done before.
It should be possible to alter the buffer in Mach3 via plugin dll, right? Timing would be a huge issue though, and it's probably impossible as the buffer will be emptying and being filled very very quickly. Hmmmm..... just thinking out loud...
It appears near impossible at this point due to the buffer delay. If you designed a board or used PLC to intercept the pulses to modify them "on-the-fly" you would have to delay all motor pulses by the same delay as is introduced by adjusting Z to keep all motions in sync.
Interesting.... Now I see why BR549 was suggesting PLC....
Chris
-
THE thcup/down is only the fairly high speed digital INPUT to control the Z thru MACH.
THC UP/DOWN is the only input to control Z through Mach...Unless you're talking about a THC that controls the Z independent of Mach.
IF those signals are erratic then the Z will be erratic. The PID filtering takes place in the DTHC.
You're talking about signal conditioning, not PID control. If the tip voltage is all over the place, then the height will be erratic.
Being that Mach3 is NOT a real time controller your on the fly height correction calculations are not going to work well. Yes you can use #vars for Zheight control but ONCE the moves are loading into the buffer then that is it. IF you try to calculate the variable at the point the #var is loaded into the buffer you will find you cannot comp something that has not happened yet. the next problem is the only way you have to update those vars are thru either the brains or the macropump. At a MAX of 10 updates per sec you will fall FAR short of the speed to do it well.
I'm not talking about doing the math inside Mach. I agree that this is way to slow. The purpose of a PID loop is to attempt to look ahead. If the there is a large difference between the set voltage and the tip voltage, then the loop moves the Z at a faster feed rate than if there were a smaller difference in an attempt to "catch up". As the two voltages get closer the PID loop slows the feed rate. It works very well, but it only work if the THC is controlling the Z axis motor drive directly.
The ONLY access you have at that point IS the THC functions up/down to comp the surface heights based on what the DTHC sees. AND it is NOT value based it is like playing ping pong slapping the ball back and forth. THe Z at that point will not even use the motor vel/accel parameters it just basically slaps the motor wide open in the direction it is to move and lets the DTHC deal with it being correct in height. Woops overshot slap it back in the other direction.
This is exactly how the low end THC's do it. Just turn your THC feed rate up to about 25 and watch the Z axis bounce. If these $500 THC's have a PID loop, then it's not working. They simply filter the tip voltage to remove any noise and compare that to a set voltage. If there is a difference, it triggers the THC UP or DOWN commands in Mach. If there were some sort of PID control, then we would be able to run the THC feed rate at 100% as the PID would adjust feed rate and prevent over shooting. As it stands, to reduce over shoot, we have to run the THC feed rate at 5-10%. even that doesn't eliminate it. It just reduces it to an acceptable level.
That is where the PID has to step in. Same basic princible as a servo motor drive.
How exactly does the PID step in? We're talking about a system with ON/OFF UP/DOWN inputs. You can not vary the feed rate based on conditions, which is the purpose of a PID loop. The only possible way this could happen is if the THC were to pulse the THC inputs sort of like a DC speed control, but then I doubt Mach would be able to handle that.
Which THC are you running? I have a Sound Logic and a CandCNC on the bench and I can tell you that neither of them have build in PID control. I know this because (1) They operate through Mach's THC commands and (2) Turning the THC feed rate past 15 turns the Z into a pogo stick.
-
You cannot look ahead any further than you can see data for, other than that you and the control are just guessing. If it gets faster than the the stiffness of the system you get the pogo so you slow the max velocity down to compensate.
The pid is used as PART of the sensoring of the arc voltages. IT tries to average out the input stream and make a smooth control output signal not just filter out high /low.
WITH the DTHC i am using now the first sign of unstable responce from the Z axis IS missed steps not motor oscillation. The system can run faster than the motor can phsically keep up. SO that tells me the pid is doing its job(;-).
HIYA CHris yes it is a complex subject. The DTHC is basicaly an advanced PLC system specifically engineered for the specific job of THC. IF it was as easy as it looks everyone would be building one and they would be $10 a copy(;-).
COntour on the fly machining can be done BUT it takes alot of FAST control and responce to make it work well. Probably well beyond MACH3 to do well
With a buffered system you CAN run it 1:1 and fill as you go BUT the machining MAY get choppy or slow as the drives may end up waiting on Data
Your milage may vary, (;-) TP
-
RRC , it is very easy to get the MACHINE to operate outside of the preset PID range. IF you use a feed screw that is too fast it is possible to outrun the pid because is not adjustable in most THCs and most pids are range specific to tuning SO if the PID is not use adjustable you MUST engineer the feedsytem to OPERATE in the boundaries of the set PID.
SAME goes true for the other end as well too slow or fine.
Get everything balanced and you have a very good working system.
Just a thought, (;-)
-
The pid is used as PART of the sensoring of the arc voltages. IT tries to average out the input stream and make a smooth control output signal not just filter out high /low.
That's not PID, That's signal conditioning. Our THC does not have a PID loop built in, but it does have a signal conditioner...and it's as stable as the day is long. A PID loop is used to dynamically change motion parameters based on conditions, such as temperature, flow, voltage, etc. They are used when a simple on/off won't work. That's why the higher end THC's have PID control loops. They control the Z axis directly and based on conditions adjust the feed rate accordingly. That's why a THC with a PID control loop can respond at a much faster feed rate that one of these $500 THC's that use the Mach3 THC commands.
-
RRC , it is very easy to get the MACHINE to operate outside of the preset PID range. IF you use a feed screw that is too fast it is possible to outrun the pid because is not adjustable in most THCs and most pids are range specific to tuning SO if the PID is not use adjustable you MUST engineer the feedsytem to OPERATE in the boundaries of the set PID.
If the PID is programmed properly, it adjusts the feed rate to meet conditions. You're confusing signal conditioning with PID loops. For instance, the Thermal Dynamics THC on their controller running a Koike machine with an Ultra-cut 400 plasma will run the Z at 100IPM. The PID controller has control of that. If the gap suddenly widens, the Z can go as high 100IPM to close the gap. The actual feed rate depends on the amount of correction needed.
-
(;-)No confusion here.
(;-)TP
-
To clarify for others reading this, signal conditioning and PID are quite different things. Signal conditioning removes noise and/or modifies the signal to be accepted by the next stage. PID stands for Proportional, Integral, Derivative and is used to improve the accuracy of the error correction by speeding up or slowing down the corrections to make it match the actual error as closely as possible with minimal overshoot. Without it the error correction cannot be accurate, as overshoot will always occur to some degree. The PID controls both the velocity and acceleration of the correction. According to what's been posted here this is impossible with the standard THC controls in Mach3 as there are no provision for controlling the feedrate, and thus no control of velocity or acceleration, aside from using a fast feedrate and sending the signals as pulses. This is likely to cause choppy movement on Z.
-
I guess the engineer that designed and built the unit did not know what he was talking about then. Shame to waste all that time on education. I'll let him know that he should not waste all his time working on such silly things that could never be possible.
But then again MAYBE that is why it works so good.
You guys are funny,
(;-) TP
-
I have no experience with THC at all, just trying to piece the information from the two of you together. The possibility does exist that the unit has PID and Mach3 can't utilize it. I'm not saying that is or isn't so, I have no idea. I don't run torch or plasma and am trying to learn if it would be applicable in my application.
Could you recommend such a unit for me to research?
Thanks, Chris
-
I guess the engineer that designed and built the unit did not know what he was talking about then. Shame to waste all that time on education. I'll let him know that he should not waste all his time working on such silly things that could never be possible.
But then again MAYBE that is why it works so good.
You guys are funny,
(;-) TP
Just because it works well doesn't mean it has an internal PID controller. Sargon nailed the definition of PID to a T. PID control of the Z axis through the Mach THC commands is not possible. PID is not necessary for a THC to work well. It's simply a matter conditioning the incoming signal and tuning the THC feed rate to reduce over-shoot to an acceptable level.
Get your hands on high end THC and controller like a TD, Hypertherm or Burny and you'll see the difference. The TD THC can track corrugated material at 150IPM. Maybe faster, but the material we were cutting called for 150IPM.
-
Could you recommend such a unit for me to research?
If I understood correctly, what you're trying to do is read the TOM and make Z offset adjustments on the fly to maintain accurate Z cutting depths even though TOM height may change. If you use a THC, you will lose control of the Z axis while the program is running. Maybe a better approach would be to read the TOM and use a brain or macropump to adjust the tool length offset on the fly. Never done that, so I can't tell you how to do it.
You could use Peter Homanns MODIO to get the analog signal from the sensor into Mach.
-
We tried that with a Modio and the overall system responce (CB) was too slow at 10hzs.The reason I suggested the THC functions is that the LPT inputs sample at kernal speeds.
The second probem was getting the correction data to the Gcode buffer fast enough. You can run 1:1 on the buffer filling BUT now the machine was limited in overal performance and speed because of waiting on the buffer.
To run a full buffer with a lot of look ahead (best for smoothness) you lost the ability to do the comp because the buffer loaded BEFORE the machine got to the point you needed to measue and apply the comp. ( catch 22).
The last option we tested was to take a linear encoder and use it to measure the height and use the quad output to drive the stepper drive directly that way it was geared to the surface measurements. With a modern digital servo drive running the Z this worked fairly well . BUT you had to swapaxis back and forth to have full control of Z and this made it VERY clunky to code and use. ALSO you woul dloose control of the positional state of Z each time and needed to refhome each time.
I still think the THC method could be made to work. WHen TOMc and Art developed the function Along with MUXing the ports it could be classed as a work of genius's in the DIY field of CNC.
BUT it would have to be fully investigated and tested. The nice part withte hTHC metod is you never loose postional control of the Z no matter how much the DTHC slapps it around(;-) MACH always knows where it SHOULD be. SO Gcoding is never effected and it runs on the fly.
You mileage may vary based on local conditions, (;-) TP
-
If Z is at say...0 and if you move up say .25" using the THC UP input, does Mach offset the axis up .25 or just make a rapid move to .25? If it does an offset, it should work for what he's trying to do. If it just makes a move, then the next Z move in the G-code would just move it back where it was.
-
Thanks for the responses, it has been incredibly helpful. It seems the only sure-fire way of making a robust dynamic height control would be to send the pulses into a PLC and apply the THC there, delaying all signals equally. A PID loop could also be implemented in the PLC, and wouldn't be too difficult. The result would be a kick-a$$ dynamic height controller.
To me, it looks like there are too many barriers in Mach3 for this application (i.e. no velocity/acceleration control for THC, macros way, way too slow, applying offset directly to motor causes lost position, 1:1 buffer results in buffer under-runs, etc...). While I think it's likely possible, it doesn't look simple at all, while the PLC approach is straight forward and extremely robust.
Thanks,
Chris
-
Hi Chris
Though I agree with a lot (but not all) of what's been said here I think your conclusions are perhaps unneccessarily pessimistic about THC with Mach.
I've built several DTHCs which work just fine with (and without) Mach and I've thus far not seen the need to add PID control. I guess I'll have to try it to see if there's any appreciable difference in performance, but frankly I'm not holding my breath.
I think that sometimes we can get a tad carried away with theory and forget the real world. I can accept (maybe) that PID could be usefull in plasma THC if there were DRAMATIC and SUDDEN changes in the level of the metal but I'd argue that in 99% of cases this just isn't the case. THCs do NOT need to "rocket about" at great speed. Watch one doing it's thing and you'll see that more often than not they just wander up and down gently over a few mm or so.
That said, I think the success you'll have with your particular requirement, however you implement a height control, will depend ultimately on whether your sensor device has the response speed, resolution and accuracy commensurate with the cut quality you require.
Just my 2 whatnots.
Cheers
Ian
-
Thanks for the responses, it has been incredibly helpful. It seems the only sure-fire way of making a robust dynamic height control would be to send the pulses into a PLC and apply the THC there, delaying all signals equally. A PID loop could also be implemented in the PLC, and wouldn't be too difficult. The result would be a kick-a$$ dynamic height controller.
That is the best way to control torch height, but keep in mind in that application, Mach3 has no control of the Z axis during feed rate moves. Mach only makes Z clearance moves. All of the high end THC's use an internal PID controller. It's the only way to get a high feed rate with out bouncing the Z like a pogo stick, but as Stirling said, that sort of speed in a THC is not needed most of the time.
While I think it's likely possible, it doesn't look simple at all, while the PLC approach is straight forward and extremely robust.
In a THC application it is fairly straight forward. In your application, you'll be controlling the Z axis drive with the PLC, which means you'll need to send the Z step and direction pulses from Mach to the PLC, modify them and send them out to the drive. Keeping the PLC and Mach in sync might prove to be a challenge.
You keep talking about a THC, but that's not really what you're doing.
-
Just a noted point(;-) if You are in a commercial cutting business to use theTHC and plasma it is imperitive that it control the z down to a few thou for control of the cuting process. The plasma arc is very sensitve to changes in the arc length. Loose control of that process and you will not be competitive in cut quality. Today everyone thinks plasma should be as precise as a milled profile. SO yes there is a Need for fast thc responce. Cutting coragated sheets it is a must or you end up dragging the head or a very sloppy cut. DOing it on thin sheets requires you to CUT fast and close. Try it with diamond plate. It probably is the worst at crashing torch heads. Remember cut quality come with very close arc length so Fast cutting at .050" height can be very tricky.
Now to balance out that in some cutting you could turn the thc off and never know it wasn't working.
If you are talking hobby plasma or plasma art? After you have cut with a high speed thc you will not want to go back.The quality of cut is very noticable and the level of cleanup is greatly reduced. It lowers the cost of the end product.
With a slow thc you can actually WATCH the torch lagging behind the process or minor almost not noticable pogo effect. That will leave a minor scalloping to the cut.
SO I guess it all depends on what you require out of the process. (;-)
(;-) TP
-
OK something to think about (;-) IF it was as easy to do "WELL" as buying a simple PLC do you think the only LOW COST solutions on the market today would cost $500 ???
I don't use market share as a main planing tool BUT ALWAYS keep it in the mix for a market decision such as a product feasibility. In this case WHY would I invest in a new product and manufactoring the electronics for a Lowcost DTHC at a market price of $500 IF someone could easily use a $25 PLC and do the same thing and sell it for $100????
Just a thought, (;-) TP
-
In a THC application it is fairly straight forward. In your application, you'll be controlling the Z axis drive with the PLC, which means you'll need to send the Z step and direction pulses from Mach to the PLC, modify them and send them out to the drive. Keeping the PLC and Mach in sync might prove to be a challenge.
You keep talking about a THC, but that's not really what you're doing.
Sending just the Z step/dir pulses through the PLC is definitely not enough. All signals must be kept in sync, therefore all signals must be routed through the PLC to impose equal delays.
I refer to THC only because it's the closest thing I've seen to my application.
OK something to think about (;-) IF it was as easy to do "WELL" as buying a simple PLC do you think the only LOW COST solutions on the market today would cost $500 ???
I didn't say it would be "easy", just not too difficult. It's all programming. The cost of the controller isn't the problem. Isn't it true that this job can be handled quite handily with PLC? I don't see anything here that would be beyond the capabilities of a PLC. Perhaps they have other motivations for custom electronics, such as protecting their product from being easily copied? After all, with PLC all you're really selling is a controller that can be obtained from anywhere, and some software. It would likely be harder to sell for maximum price than an integrated solution.
Maybe I'm going a little overboard with wanting full control of the height control process, but I don't see any reason not to use PLC, and I love the idea of having full control of the machine at both levels... It should allow me to use Mach3 the same as always, with full 3D capability as well as height control. Maybe someone can point out something that I'm missing.
Thanks again,
Chris