DIY Servo controller

[LZ1TWB]

Kreutz,

You say my encoder signals could be out of quadrature. Since there are pots for adjusting the pulse width how critical is the timing of the signal? I know it should be 50%/50% but measuring it with a scope is not so accurate. I've tried to make them as equal as possible.

As for the new processor you said something about analog tacho-generator input. I was thinking that the position and speed are taken from 2 different places and then a glass scale encoder could be used as in many industrial machines. This eliminates any backlash in the system.

I wonder if a parameter for the backlash could be entered in the drive's CPU thus not relying on software from the PC side to correct it as this problem is not so good solved in Mach3.

Todor

[kreutz]

Kreutz,

You say my encoder signals could be out of quadrature. Since there are pots for adjusting the pulse width how critical is the timing of the signal? I know it should be 50%/50% but measuring it with a scope is not so accurate. I've tried to make them as equal as possible.

As for the new processor you said something about analog tacho-generator input. I was thinking that the position and speed are taken from 2 different places and then a glass scale encoder could be used as in many industrial machines. This eliminates any backlash in the system.

I wonder if a parameter for the backlash could be entered in the drive's CPU thus not relying on software from the PC side to correct it as this problem is not so good solved in Mach3.

Todor

Todor;


Not only equal pulses but also both at 90 degrees out of phase, are important.

The analog tacho-generator input will be useful for speed control with or without quadrature encoder. Probably good news for those who want spindle speed control and already own a servo with integrated tacho-generator.. The same applies to those who want to replace the +/-10V amplifiers without having to replace the control system. Both are options that I would like to add in the near future.

The better way to control backlash is on the mechanism itself, software control is not that useful... Having a glass scale encoder would be very nice if it is fast enough to serve as servo-encoder input, but if some amount of backlash is present, the servo motor could oscillate or be unstable and proper tuning will be difficult.

Normally the speed information could be derived from the position encoder, having an independent speed sensor (tacho-generator) could improve the servo behavior, that is something I would like to test.

Do you have access to another motor/encoder to test your UHU drive?, that way you could find out if the encoder is the source of the problem.

Kreutz.

[kreutz]

I have been thinking...If the new version does not include the UHU chip, then it won't belong on this thread....

Since it would be nice to have a torque regulation loop inside the position/speed PID loop, it will be necessary to include a "heavier weight" (16/32 bits) CPU or a combination CPU + CPLD. I have been considering the following CPUs (prices are in qty=100, not including shipping):

LM3S811 $5.45*
LPC2103 $3.08*
LM3S601 $3.58*
LPC2141 $4.86*
dsPic30F4012 $4.47**
LM3S818 $4.27
56F8322 $6.70**

* means that it needs additional hardware.
** means that I need additional software.

So far the main contender is the LM3S818.

I also decided to buy a "pick and place" machine to help during the assembly...

Found another contender:

dsPic33FJ32MC202 $2.96 (Qty=100)
Same problem with the full MPLAB_C30 compiler, it is a little expensive... I am tempted to try the GNU Compiler.


Kreutz.

[sigma relief]

Well I may as well stop lurking, I have been following this thread almost daily since December, but have not had a reason to post. I am currently living in a tiny east coast appartment and have been forbiden from building a machine shop by SWMBO, so instead I have been compiling a list of common issues for the UHU so I can use the design when if ever I get shop space. My notes list is very rough and probably not useful to anyone else, but I did spend a lot of time focusing on parts, sources, and substitute parts.

I ended up updating the "official" parts list by FromDusk with tolerances, manufacturer part numbers, and other vendor information. I have not spoken with the origional author (I don't know German), but have kept the file intact and merely added additional information. Like the origional author, I make no claims to accuracy (I am a Mechanical Engineer, I just have a passion for electrical hobbies) and have never even built a UHU controller. The tolerances on this list are based either on the origional parts, or a basic knowlegde of the circuit theory. PLEASE post corrections, additions, or alternate parts, I will try to keep the sheet up to date for a month or so if people are interested, then a more stable version can be added to the wiki if others find this useful.

My main reason for all of this work was to archive the project for future personal use just in case places like CNCzone dissapear. I also wanted to convey a little bit of the design intent, ie. how important the parameters of certain parts are. The problem with my archiving information is I am assuming Uli is still selling the chips when I need them. I have an awful lot of respect for him and his project and can't blame him for keeping the project closed source. My one request to Uli would be if he ever stops selling the chips, to open the project up so others can continue to use the design. Even if only binary files are posetd, it would be of great help.

With that said, my vote for the new controller would be to make compiled binary code for the processor available so people can program the chips themselves if they so choose, even if they can not easily modify the code. A fully open code is nice in theory, but I agree it stinks to have your work ripped off.

On a side note, I'm sure some people would be able to benefit from actual part numbers for the HPUHU design. I have no intention of doing this for that design, but in future parts lists, a little information about each part, specifically the most important specs and suggested brand/model numbers could be useful.

John K.

[kreutz]

Well I may as well stop lurking, I have been following this thread almost daily since December, but have not had a reason to post. I am currently living in a tiny east coast appartment and have been forbiden from building a machine shop by SWMBO, so instead I have been compiling a list of common issues for the UHU so I can use the design when if ever I get shop space. My notes list is very rough and probably not useful to anyone else, but I did spend a lot of time focusing on parts, sources, and substitute parts.

I ended up updating the "official" parts list by FromDusk with tolerances, manufacturer part numbers, and other vendor information. I have not spoken with the origional author (I don't know German), but have kept the file intact and merely added additional information. Like the origional author, I make no claims to accuracy (I am a Mechanical Engineer, I just have a passion for electrical hobbies) and have never even built a UHU controller. The tolerances on this list are based either on the origional parts, or a basic knowlegde of the circuit theory. PLEASE post corrections, additions, or alternate parts, I will try to keep the sheet up to date for a month or so if people are interested, then a more stable version can be added to the wiki if others find this useful.

My main reason for all of this work was to archive the project for future personal use just in case places like CNCzone dissapear. I also wanted to convey a little bit of the design intent, ie. how important the parameters of certain parts are. The problem with my archiving information is I am assuming Uli is still selling the chips when I need them. I have an awful lot of respect for him and his project and can't blame him for keeping the project closed source. My one request to Uli would be if he ever stops selling the chips, to open the project up so others can continue to use the design. Even if only binary files are posetd, it would be of great help.

With that said, my vote for the new controller would be to make compiled binary code for the processor available so people can program the chips themselves if they so choose, even if they can not easily modify the code. A fully open code is nice in theory, but I agree it stinks to have your work ripped off.

On a side note, I'm sure some people would be able to benefit from actual part numbers for the HPUHU design. I have no intention of doing this for that design, but in future parts lists, a little information about each part, specifically the most important specs and suggested brand/model numbers could be useful.

John K.

Good Job, John.

Thanks!!

Kreutz.

[kreutz]

.......With that said, my vote for the new controller would be to make compiled binary code for the processor available so people can program the chips themselves if they so choose, even if they can not easily modify the code. A fully open code is nice in theory, but I agree it stinks to have your work ripped off.
......
John K.

I am planning to do that. The only problem with that approach is that in order to program the chips, a JTAG/ISP programmer is needed. Even if I use a boot-loader (in order to do upgrades over the serial/USB connection), the boot-loader code should be programmed that way. So for the average guy that wants to program only his 3 or 4 chips, is it not cost effective.

I have not decided yet which chip will be the next protagonist. The main reason is not only chip cost and availability, but development system cost and availability.

Kreutz.

[tenmetalman]

Hello,
I was just re-reading the posts on this page & zeroed in on the possible auto tune feature. that really, really made my ears perk up. for people like me with a limited electronics background & three left thumbs that would be a feature that would really enhance saleability. I know we're talking about "in the future" but if this feature is doable without a huge amount of time, It's a MUST HAVE as far as I'm concerned .........
Paul

[automationtechinc]

I like to oreder to test my DC servo motor
http://www.kelinginc.net/ServoMotor.html

[kreutz]

I like to oreder to test my DC servo motor
http://www.kelinginc.net/ServoMotor.html

Hello;

Please, contact Tenmetalman or Contactirfu (Paul or Irfan).

Thanks,

Kreutz.

[Xerxes]

Found another contender:

dsPic33FJ32MC202 $2.96 (Qty=100)
Same problem with the full MPLAB_C30 compiler, it is a little expensive... I am tempted to try the GNU Compiler.

Looks like Microchip has new interesting dsPIC models.

You might not want to use compiler optimizations anyway since they may not produce correct code. I have found severe optimization bugs from all MCU compilers that I have used (AVR gcc, MPLAB C30, Freescale...). Sometimes compiler optimizes main loop away, does simple arithmetics wrong or other stupid things like these. The VSD firmware doesn't work at all if I compile it with any optimizations.

[kreutz]

Looks like Microchip has new interesting dsPIC models.

You might not want to use compiler optimizations anyway since they may not produce correct code. I have found severe optimization bugs from all MCU compilers that I have used (AVR gcc, MPLAB C30, Freescale...). Sometimes compiler optimizes main loop away, does simple arithmetics wrong or other stupid things like these. The VSD firmware doesn't work at all if I compile it with any optimizations.

Xerxes;

Thank You!!, good to know. I already found a few bugs in the MPLAB_C30 3.10B (student edition), and I am not even trying to do serious programming yet, just trying to initialize and test some dspic30 and dspic33 peripherals.

Best Regards,

Kreutz.

[kreutz]

Hello,
I was just re-reading the posts on this page & zeroed in on the possible auto tune feature. that really, really made my ears perk up. for people like me with a limited electronics background & three left thumbs that would be a feature that would really enhance saleability. I know we're talking about "in the future" but if this feature is doable without a huge amount of time, It's a MUST HAVE as far as I'm concerned .........
Paul

It will take some time, but it is doable. I would like to test first an "off-line" approach in order to make the solution cheaper. On that approach the PC will run the auto-tuning program, the motor controller board will send the results of the run, and receive new set of parameters from the PC.

The second approach: "adaptive auto-tuning", done on the board, will take more resources, but is also doable.

[vrsculptor]

Hello,
How hard would it be to produce an analog (lobotomized) HP UHU that would respond to +/- 10 volt signals? EMC2 supports some very inexpensive DSP based devices that include D/A's to provide the analog voltages and read the encoders. Existing high powered (160 volt, 25 amp) analog servo amps are VERY expensive.

I really like Mach3 and have used it since Master5, but I'm beginning to think that EMC2 may have it right in that it is not using step/direction and is closing the motion loop through a software PID loop. Big commercial CNC machines are almost all analog based.

PS. I finally received my 2KV isolation transformer and am in the process of final assembly and testing of my HP UHU board. I hope to have it ready to demo at the CNC Workshop June 16-23 in Illinois.

[kreutz]

Hello,
How hard would it be to produce an analog (lobotomized) HP UHU that would respond to +/- 10 volt signals? EMC2 supports some very inexpensive DSP based devices that include D/A's to provide the analog voltages and read the encoders. Existing high powered (160 volt, 25 amp) analog servo amps are VERY expensive.

I really like Mach3 and have used it since Master5, but I'm beginning to think that EMC2 may have it right in that it is not using step/direction and is closing the motion loop through a software PID loop. Big commercial CNC machines are almost all analog based.

PS. I finally received my 2KV isolation transformer and am in the process of final assembly and testing of my HP UHU board. I hope to have it ready to demo at the CNC Workshop June 16-23 in Illinois.

Hello;

That is one of the options I would like to have on the next design because it will require a new microcontroller. Nothing is really hard to do if you have the resources, knowledge and time... (I think I have the first, don't know yet about the other two...)

I already ordered a few parts in order to prototype a development board for that purpose. The chosen chip is the dsPic33FJ32MC202. Do not assume the new drive is going to be out soon, this one is going to be a lot of fun to develop, and I want to enjoy it by testing a few new features not seen on hobby level drives.

Unless a charitable person/company is willing to sponsor this design by letting me use (as a loaner) a high power servo with high resolution encoder, the first prototype will drive a low power servo (24v 2.9A 3000 rpm with 500 ppr encoder) I already have.

Don't forget to send us videos of your tests.

Thanks,

Kreutz.

[DennisCNC]

Kreutz,

How big of a servo do you need for testing latter? What count encoder? I assume it needs to be brush.

[kreutz]

Kreutz,

How big of a servo do you need for testing latter? What count encoder? I assume it needs to be brush.

Brushed, 8-20Amps 90-150V about 3000 rpm with 1000 ppr differential interface encoder (or more). It will be used only during the tests and sent back when finished.

I still have a lot of things to do until it will be needed.

Thanks for asking.

Kreutz.

[jcdammeyer]

Hi all,

I need a transformer to supply power to my two DC Servos which are rated at 90V 7.8A Peak Current 40A. When I look at the Digikey or Mouser Catalog I really don't see any transformers that will give me 75VAC RMS * 1.414 = 106V Peak. That value is based on no more than 85% duty cycle from the UHU.

I'd like to have no more than 5V ripple so with 16A draw * 8.3mS = 132mCoulombs. Taking 132 mCoulombs / 5V ripple = 26.5mF or 26,500 uFarads.

Does anyone have a source for a 1.2KVA transformer 120VAC or 240VAC 60Hz in and 75VAC RMS 16 Amps out?

Same with 120 to 150 V capacitors 27,000 uF.

Thanks

John

[kreutz]

Hi all,

I need a transformer to supply power to my two DC Servos which are rated at 90V 7.8A Peak Current 40A. When I look at the Digikey or Mouser Catalog I really don't see any transformers that will give me 75VAC RMS * 1.414 = 106V Peak. That value is based on no more than 85% duty cycle from the UHU.

I'd like to have no more than 5V ripple so with 16A draw * 8.3mS = 132mCoulombs. Taking 132 mCoulombs / 5V ripple = 26.5mF or 26,500 uFarads.

Does anyone have a source for a 1.2KVA transformer 120VAC or 240VAC 60Hz in and 75VAC RMS 16 Amps out?

Same with 120 to 150 V capacitors 27,000 uF.

Thanks

John

If you are going to use the HP UHU drive, I suggest using a 80Vac (secondary) transformer, same rating. HP UHU uses dead-time that is added to the maximum and minimum PWM periods. For calculation purposes use a 25% higher voltage than the motor rated voltage.

Other general power supply considerations:

Use capacitors rated about 50 % higher dc voltage than the peak dc voltage of the power supply in order to account for normal line voltage variations and regenerative voltage. Use several capacitors in parallel instead of big higher capacitance single capacitors in order to improve ESR and ripple current handling capacity, at least half of them should be rated for high ripple current at 100Khz. Use 1 uF/400 V film capacitors in parallel with electrolytic capacitors.

Use regenerative power dumping in order to protect your drive and power supply.

Use two copper bars to wire the capacitor's bank instead of copper wire.

[contactirfu]

Kreutz, I have 5000x5 (200v) caps for a DC 110v 20 amps supply. all connected in parallel using aluminum bars, the motors are 110v 7 amps contineous each. hope the above is ok. I am planning on a regenerative power dump circuit soon.

Its ok if I get a little less performance bcos I have chosen a same PS as that of motor rating bcos the ps is so expensive.

Hope the above means I am ok except for the 1 uF/400 V film capacitor in parallel. Do i need to use one on each electrolytic or just one for all is ok?

RGDS
IRfan

[jcdammeyer]

If you are going to use the HP UHU drive, I suggest using a 80Vac (secondary) transformer, same rating. HP UHU uses dead-time that is added to the maximum and minimum PWM periods. For calculation purposes use a 25% higher voltage than the motor rated voltage.

Good point. My Motor is 90VDC so I need 112VDC for the 25% over factor. 112 * .707 = 80VAC RMS.

Other general power supply considerations:

Use several capacitors in parallel rated about 50 % higher dc voltage than the peak dc voltage to improve ESR. Use 1 uF/400 V film capacitors in parallel with electrolytic capacitors.

Use regenerative power dumping in order to protect your drive and power supply.

Use two copper bars to wire the capacitor's bank instead of copper wire.

Also good suggestions. I've redesigned the regenerative circuit and replaced the comparator with a PIC processor. That allows me to set the hysterisis so that an over voltage can hold the clamp on longer than the over voltage event if needed. But that wasn't the main reason. The circuit also monitors the power supply AC input and if it vanishes, the PIC turns on the clamp and discharges the cap bank very quickly. That way I don't have to have a larger bleeder resistor always drawing power. Plus I have a few 8 pin PICs in my bin.

Back to original question though. Anyone have a good source for the transformers?

Thanks
John