[dakimber]

I'm restoring a 1986 model Jetmatic mill and using EMC2 to control it. I've installed and configured EMC2 using the Pluto-P as the interface. So far I've been able to accurately read the encoders to track position, but I'm having trouble with the H-bridge design. The servos on this machine are 100 volts and 9.5 amps, although the mill supply only outputs 90 volts. Still, the H-bridge needs to be pretty stout.

My first design used MTY55N20E MOSFETS (from Motorola originally, now from ON semi) that I had left from a previous project a long time ago. These are 55 amp 200 volt MOSFETs. I used a pair of LH1262 parts to drive the gates. The LH1262 uses an led to shine on an led stack to create the gate voltage. I liked this design because of the isolation between the control and the power circuits, but the LH1262 is pretty slow, leaving the MOSFET in the linear region too long, and the MOSFETs get very hot very fast (a few seconds is all I dared run them). I was able to get the X-axis going by reducing the voltage to 12 volts, but it is very slow!

So I redesigned the circuit using the IRS2304 from International Rectifier. I followed the example in the IR appnote, replacing the IR2110 with the IRS2304. This design proved disastrous! It blew the lid off the IRS2304 and destroyed my power supply. I haven't determined if the MOSFETS survived yet. Further investigation shows the two high-side MOSFETS are both being driven simultaneously - one with about 8V and the other with about 4V. This only happens if the servo supply voltage is on. If the servo supply is off, only one MOSFET gate is driven, and it is driven with a full 12V. Removing the servo motor from the circuit makes no difference.

Does anyone with experience building H-bridges have some ideas? I suspect I've overlooked something obvious, but I'm not sure what...

Here are links to the relevant data sheets:

MTY55N20E: http://www.onsemi.com/pub/Collateral/MTY55N20E-D.PDF
LH1262: http://www.vishay.com/doc?83802
IRS2304: http://www.irf.com/product-info/data...rs2304spbf.pdf
IR App note: http://www.irf.com/technical-info/appnotes/an-978.pdf

Thanks!

Doug

[samco]

I have played around a little bit. I have had good luck with the ir2111. I made a h-bridge without current limit to play with.. It seemed to work great.

http://www.electronicsam.com/images/...rt/ampmess.JPG
http://www.electronicsam.com/images/...art/almost.JPG
http://www.electronicsam.com/images/...vostart/schem/
you can read a bit about it here
http://www.cnczone.com/forums/showth...=large+brushed

I ran this at around 60v 10a for testing - mosfets never got warm to the touch attached to the heatsink. The mosfets are 44a 600v iirc but the Rds and thermal transfer of the silicon would only allow a max of 20a.

the next version is only on paper at the moment - I have done some limited logic testing but have not had time to mill a board. (it also has a drive enable)

http://www.electronicsam.com/images/...boardagain.png
http://www.electronicsam.com/images/...schemagain.png

same basic design but with current limit. so I am using ir2184 as they had an disable pin unlike the ir2111.

sam

[dakimber]

Sam-

Thanks for the reply! Your original thread:
http://www.cnczone.com/forums/showth...=large+brushed

is part of what inspired me to try this. Like you, I discovered the parport wasn't fast enough to catch all the encoder pulses, so I got a Pluto-P. (Too bad - the dual parallel port card worked well and had just enough I/O for what I wanted to do!).

I haven't had time to completely digest your new schematic, but I did notice you addressed the isolation issue with the TLP2530 photocouplers. I picked up some Sharp PC900v0nszxf optos, but they are 1 opto per 8-pin dip, so maybe I'll try to find some TLP2530's and save some space.

I see why you prefer the IR2184 over the IR2111 - the 2111 always had one half of the bridge being driven, while the 2184 lets you turn both off (via the /SD pin). I picked the IRS2304 for a similar reason - it has independent control of the upper and lower channels. But, it doesn't include a shutdown pin. My plan was to connect the HIN of one 2304 to the LIN of the other 2304, and vice versa. Then, driving one pair of pins is forward, the other reverse, and neither being driven leaves the motor off.

I'm starting to wonder now if my problem was a bad mosfet right from the start. I haven't been able to find a logic or wiring mistake, and don't see the reason both upper mosfets were being turned on. Since this happened with the motor disconnected, it wasn't related to the large inductance. Looking at other half-bridge drivers (e.g., from National), they all seem roughly similar. My plan is to rebuild with different parts and try again (without applying the 90V if I don't get good results!).

What do you use for schematic capture and board layout? I've been using the ExpressPCB stuff since it was free. I can order boards from them over the internet (I don't have a PCB router - that will be another project....) and the quality seems decent, but it still costs more than if I was able to make my own.

Doug

[samco]

I use eagle. I am getting used to it finally. It seems a bit counter intuitive - but once you get used to it....

Yes - the ir2111 always has either the upper or lower fet on. This really isn't a problem or atleast it wasn't a problem for me. They call this 'slow decay' iirc. This is a picture of a so-so tuned pid loop. aprox .0005 max following error. (I am pretty new to tuning)

http://www.electronicsam.com/images/...art/ferror.png

I also have the pluto - I will use it for testing. I actually have a mesa 5i20 for the final solution.

I am hoping for around 150v 20a.

sam

[dakimber]

Maybe I'll have to spring for Eagle - it seems fairly popular. I used HiWire and OrCad eons ago, but last I checked OrCad was way beyond hobby prices.

Now I finally get it on the ir2111's. When the motor is not being driven both lower mosfets are on, providing braking. And that's why it's ok to always have one of the two mosfets on in each half-bridge.

When you tuned your system to the 0.0005 following error did you use the auto tune feature of at_pid? Or did you do it manually?

I'm thinking now my H-bridge problem was a bad bootstrap cap. One of the caps was 67 ohms (yes ohms) when I checked it. Anyway, I rebuilt with all good parts and it seems to work! I ran the X-axis back and forth a number of times using the full 90V and the Mosfets never got warm. I used a 555 timer circuit with adjustable pulse width and frequency, since I don't have isolation yet to keep my computer safe. That's the next step, after which comes the current limiting.

Doug

[samco]

That is awesome! I think the at_pid works the best with velocity mode amps. It does get you close - but from my experience a bit squirley(sp). Pid tuning seems to be something that you just need to do. The more you play with it - the better you get. Don't forget ff1 and ff2. <- seems to do the best for current mode amps.

sam

[dakimber]

Sam-

I'm just getting back to this - had boards made so I didn't have all that spaghetti wire...

In looking at your schematic I see you used a pair of caps (0.1uF and 10uF) in parallel for the bootstrap capacitance. How did you determine the values? When I followed the IR app notes I got values that were much less (granted, I'm using different MOSFETs and drivers). I used a 1.5 uF cap, but that was way more than I calculated was necessary. I don't see a downside to going larger (too small could pose problems), but I was curious.

Another schematic question - you used a cap across the series resistor on the opto (TLP2530) inputs. It seems this will slow the rise and fall times of the opto inputs, but why is that needed? Is it just to filter out noise? I used the TLP2531 in my design (Digi-Key didn't have the TLP2530's).

I'm starting to try and tune, and you mentioned using the feed-forward parameters (FF0, FF1, FF2). How do you select values for these? I didn't see anything other than the descriptions on the manual pages. The examples mostly have them set to 0, with an occasional FF1=1. The Wiki doesn't add much about them under tuning, and the manual doesn't say anything at all that I saw.

Thanks!

Doug

[samco]

Some one I trust recomended the 10uf cap. The only issue brought up is that it might take more than 1 pwm cycle to charge totally. (I am not worried.)

As far as pid tuning. Good luck . It is really something you have to do to get used to. Here is a little help.
http://wiki.linuxcnc.org/cgi-bin/emc.../HAL_PID_Loops
http://www.linuxcnc.org/handbook/part2/pidtune.html
http://linuxcnc.org/docs/html/motion_pid_theory.html

I have been testing
http://www.electronicsam.com/images/...ostart/amp.JPG
Maybe in a few days I will know for sure.. Time will tell.

sam

[samco]

http://www.electronicsam.com/images/...pmessagain.jpg
[ame="http://www.youtube.com/watch?v=eUmFKOVepYY"]YouTube - Servo h-bridge running with emc2[/ame]

sam

[acondit]

Sam,

I have been trying to embed a YouTube video on a post in my CNCLathe conversion log and I can't get it to work. How do you do it?

Alan

[acondit]

Sam,

Thanks for the help.

I watched your test video and it was really looking good. Does this version of your h-bridge have the current limiting on it?

Alan

[samco]

Yes - it is basically this circuit without R10.

http://www.electronicsam.com/images/...schemagain.png

[epineh]

Nice work Sam !

I couldn't quite make it out but are you using HAL Scope in the video by any chance ?

Russell.

[samco]

Yes - that is EMC2's halscope. Awesome little app - as is the whole program.

sam

[epineh]

Sweet, I am looking forward to using more of EMC

Russell.

[samco]

[ame="http://au.youtube.com/watch?v=fGF07oVkl6M"]YouTube - EMC2 closed loop test with simple h-bridge and pluto-p[/ame]

[dakimber]

I finally finished building the H-bridges and the Pluto break-out-board (see photo).

The H-bridges stay very cool, but I seem to have two new problems...

First, the servos don't move until I get to about 50% on the PWM. This sounds a lot like the problem discussed here:

http://www.cnczone.com/forums/showth...90997#poststop

There it is suggested that PWM be used - which I am doing. My H-bridges do not short out the motor (i.e., all 4 MOSFETS are off with no PWM pulses), which was also suggested, but that is a bigger change than I wanted to try. The original post was using PDM, so I thought by using PWM I would be OK. (I still don't completely understand why shorting out the motor keeps the average current up).

Does anyone have a simple way to compensate in HAL - I didn't quite follow how that was done in the post above.

The second problem is that once the servos do start moving they will suddenly jump, and then I lose everything (or, maybe they jump because I lose everything). I can no longer read the encoders until I kill and re-start axis. At that point the encoders read fine again. If I leave Axis running, it will not complain, but it also will not drive the servos anymore.

I'm not sure where to start tracking the second problem. I disabled the Pluto watchdog, just in case, but that didn't make any difference.

Any ideas?

Thanks!

Doug

[samco]

You might want to look at 'inverse deadband'. I don't think it is in emc - but you can build it.
http://www.anderswallin.net/2008/04/...nent-for-emc2/

What does your circuit look like? 50% seems really huge. My h-bridge does always short out the lower 2 mosfet when there is no pwm. It is very responsive down to 0 duty cycle.

sam

[dakimber]

Sam-

Thanks for the pointer to the inverse deadband component. I'll look this evening, and if it isn't in EMC I'll add it and try it. My estimate of 50% came from watching the PID bars in Axis and noticing when the axis started to move (via the Axis position counters). My motors are 90V, so I set the PID bars to range from -90 to +90, and the axis starts to move when the PID bar hits about 45.

Here's the schematic for my H-Bridge. I didn't populate the diodes at the MOSFET gates when I built them - didn't think I needed them.

My bigger concern is the loss of control once the axis does start to move. I looked at the 5V and 12V supplies with a scope and they seemed clean. Still, the whole thing goes out to lunch after moving just a little bit, and it won't come back until I restart Axis.

Doug

[kreutz]

.......
My bigger concern is the loss of control once the axis does start to move. I looked at the 5V and 12V supplies with a scope and they seemed clean. Still, the whole thing goes out to lunch after moving just a little bit, and it won't come back until I restart Axis.

Doug

How many seconds since the start of movement?

Kreutz.