DIY Servo controller

[kreutz]

Yes, I am referring to the "O" parameter. There is some speed threshold above which I start to get errors. The more the speed the more the errors I get for a given time of travel.

My encoder cable is taken from an industrial machine - many pins in it - around 12 - A A_ B B_ C C_ + -. The pairs are twisted I think, I have not cut so much to see but they tend to come in color pairs. It is shielded one end only at the side of the drive. The encoder is bulgarian license of fanuc, very old one still used here.

Today happened a strange thing while playing with the PID parameters. I moved my keyboard a little away from the computer to be able to see the behavior of the axis and it started to go crazy hitting some buttons without me touch it. I was just clicking left and right to jog. I don't know if a keyboard cable is shielded but it was not even close to the motor power. Never had this while bench testing and is some kind of strange.

I continue to suspect EMI/RFI (produced by cross-conduction) as the cause, the weird behavior of the keyboard also supports that theory...

[rokag3]

Can you tell me the motor specs ands well as your power supply voltage?

Could it be that the minimum/maximum PWM limits are keeping you from reaching higher speeds?

no my problem is with the error count the encoder can go to 300khz (cui)
my program can send 15 khz
with M parameter at 5
less than 75 khz no error and the motor after 200 revolution go exactly to his position 1/2048 of the circle
with M=6
90 khz error append and the motor cannot reach is exact position
with M=8
120khz for a short travel it is ok but often the error count is so great that the controller give up
if i increase the error limit i manage to do it but the motor loose his position dramatically
it is this value 75 khz who make me think and this is the reason why i would like to have some feed back with guy saying Eh Lucien no problem i reach 120 or 140 with no problems
just to be sure that I am the guilty
I have to go in my origine policestate country(FRANCE or what is left of it) for 1 week hope it's the last time of my life so i will be out of the list. Wish me good luck since i am a smoker anarchiste and i cannot shut up

[kreutz]

no my problem is with the error count the encoder can go to 300khz (cui)
my program can send 15 khz
with M parameter at 5
less than 75 khz no error and the motor after 200 revolution go exactly to his position 1/2048 of the circle
with M=6
90 khz error append and the motor cannot reach is exact position
with M=8
120khz for a short travel it is ok but often the error count is so great that the controller give up
if i increase the error limit i manage to do it but the motor loose his position dramatically
it is this value 75 khz who make me think and this is the reason why i would like to have some feed back with guy saying Eh Lucien no problem i reach 120 or 140 with no problems
just to be sure that I am the guilty
I have to go in my origine policestate country(FRANCE or what is left of it) for 1 week hope it's the last time of my life so i will be out of the list. Wish me good luck since i am a smoker anarchiste and i cannot shut up

Lucien;

Your problem seems to be different. If the error count increases, it means that it is receiving pulses from the encoder and processing them OK, but is unable to catch up with the commanded position or speed. It seems like a PID tuning problem. Or that the motor is underpowered for the application, and unable to drive the load at the acceleration or speeds set by the PC software. It also could be due to using a mismatched power supply for that motor (p.s. power and/or low capacitance / high E.S.R.).

Kreutz.

[rokag3]

power supply auto transo 240>110 1600W
transfo 115 >>134V 1600W
bridge 50A
capacitor 11 MegaF 350V
and 15 V 3A 65 MF
Make 185 to 186V dc
Motor 180V 8A 42 contacts 2 brushs
weight 9 kg +mass inertia 3.5 kg 140 mm
motor from luxuous tuntori threadmill bought for 30 us$ fleamarket

[LZ1TWB]

I continue to suspect EMI/RFI (produced by cross-conduction) as the cause, the weird behavior of the keyboard also supports that theory...

Is there something I could try to change in my setup that will solve this problem? Different mosfets, different values of the RC group.

Todor

[rokag3]

Your problem seems to be different. If the error count increases, it means that it is receiving pulses from the encoder and processing them OK, but is unable to catch up with the commanded position or speed. It seems like a PID tuning problem. Or that the motor is underpowered for the application, and unable to drive the load at the acceleration or speeds set by the PC software. It also could be due to using a mismatched power supply for that motor (p.s. power and/or low capacitance / high E.S.R.).

E.S.R. !!! sorry i do not know what is it
i will lower my encoder resolution to check it but my feeling is that the motor take is load very easy and do not go high with Amp.
also i do not see a high lost of volts
bye lucien

[kreutz]

power supply auto transo 240>110 1600W
transfo 115 >>134V 1600W
bridge 50A
capacitor 11 MegaF 350V
and 15 V 3A 65 MF
Make 185 to 186V dc
Motor 180V 8A 42 contacts 2 brushs
weight 9 kg +mass inertia 3.5 kg 140 mm
motor from luxuous tuntori threadmill bought for 30 us$ fleamarket

capacitor 11 MegaF 350V 11,000,000 F !!! How do you charge it?

Motor 180V 8A Here there is a limitation... It will never get to see 180V with a 185V power supply. On the original UHU manual it specifies 15% higher power supply voltage than the maximum specified for the motor (to compensate against the output stage voltage drop and the limited PWM range of 13 to 87%).

HP UHU will need about 5% more voltage due to the 1 uSec dead-time added before and after the PWM pulses. So, in order to drive your motor to the maximum speed it will need a 216V power supply (HP UHU has been tested to 185V only).

Note: in order to increase the P.S. voltage over 185V it will be necessary to replace the on board Capacitor by a 330 uF 300 V Snap-on type capacitor as the first step, then 216V is too near the maximum specified Vds for the Mosfets (250V) and it might not work reliably anyways).

With the configuration you have now you should get about 80% of maximum speed without load. If you don't get to that speed it will be necessary to investigate a little more...

Kreutz.

E.S.R. means "Equivalent Series Resistance"

[kreutz]

Is there something I could try to change in my setup that will solve this problem? Different mosfets, different values of the RC group.

Todor

If the interfering noise is due to the cross-conduction you will need the HP UHU board.

[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...

[jcdammeyer]

[QUOTE=kreutz;457029]I have been thinking...If the new version does not include the UHU chip, then it won't belong on this thread....

dsPic30F4012 $4.47**

QUOTE]


Don't know why it couldn't stay on this list. DIY Servo Controller does't specify which processor. One reason to start a new list might be just to reduce the number of postings. But if a new list is started I think it should also be open source so perhaps start a new list GPL DIY Servo Controller.

As to processor, I prefer the dsPIC only because a while back I bought the compiler and have a demo board. I've also got demo boards for the TI32F2812 so that's another contender for me. The up side of the PIC is that the student compiler is available for free.

But at the moment my plate is full so I'm staying with the UHU for a while yet. Which reminds me... must order two more processors...

John

[kreutz]

[quote=jcdammeyer;457050]

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

dsPic30F4012 $4.47**

QUOTE]


Don't know why it couldn't stay on this list. DIY Servo Controller does't specify which processor. One reason to start a new list might be just to reduce the number of postings. But if a new list is started I think it should also be open source so perhaps start a new list GPL DIY Servo Controller.

As to processor, I prefer the dsPIC only because a while back I bought the compiler and have a demo board. I've also got demo boards for the TI32F2812 so that's another contender for me. The up side of the PIC is that the student compiler is available for free.

But at the moment my plate is full so I'm staying with the UHU for a while yet. Which reminds me... must order two more processors...

John

If I had the full compiler I would like to try it, the only problem with the free version is that after 60 days it gets limited to non-optimized code + some size limit... , that time would be enough if I did not have to work, on the other hand, I already own the full compiler for the Arm7.

Kreutz

[jcdammeyer]

[quote=kreutz;457053]

If I had the full compiler I would like to try it, the only problem with the free version is that after 60 days it gets limited to non-optimized code + some size limit... , that time would be enough if I did not have to work, on the other hand, I already own the full compiler for the Arm7.

Kreutz

Yes. I understand. I ran into the same problem with the E-Leadscrew. Since I knew I was going to release that under GPL the votes on the list were for either an ATMEL or PIC variety because of the tools.

Isn't there a free GPL ARM7 cross compiler available for Windows along with a ICE debugging setup? That was the other reason for the PIC. Inexpensive incircuit debugging even if it was only for two or three breakpoints. Handy for interrupt driven code.

The last time I looked both the ARM7/9 and F2812 require $500+ ICE JTAG pods for debugging and programming. OTOH, I built a 13.5V PIC programmer circuit right onto the ELS board.

It's probably all a moot point anyway. These types of projects end up being done by one individual rather than a group unless you have one writing user interface, another CAN Stack, another USB interface code etc. Even then...

John

[kreutz]

[quote=jcdammeyer;457058]

Yes. I understand. I ran into the same problem with the E-Leadscrew. Since I knew I was going to release that under GPL the votes on the list were for either an ATMEL or PIC variety because of the tools.

Isn't there a free GPL ARM7 cross compiler available for Windows along with a ICE debugging setup? That was the other reason for the PIC. Inexpensive incircuit debugging even if it was only for two or three breakpoints. Handy for interrupt driven code.

The last time I looked both the ARM7/9 and F2812 require $500+ ICE JTAG pods for debugging and programming. OTOH, I built a 13.5V PIC programmer circuit right onto the ELS board.

It's probably all a moot point anyway. These types of projects end up being done by one individual rather than a group unless you have one writing user interface, another CAN Stack, another USB interface code etc. Even then...

John

Yes, the JTAG debugger and programmer is a little expensive, I am planning to use a boot-loader to do firmware updates using the RS232, USB or CAN connection. The most expensive investment will be the P&P machine for SMd assembly...The compiler I have is a commercial version...there is a GNU Compiler (ARM GCC) but I have never used it.

Open source soft/hardware projects, once published, benefit mainly a few people, those who want to learn. Those projects are generally more expensive to build than DIY kits because of the cost of parts and PCBs in low quantities. On the other hand, collaborative projects require everyone involved to work on schedule, otherwise some of the participants will lose interest or get frustrated...

[jcdammeyer]

[quote=kreutz;457071]

Open source soft/hardware projects, once published, benefit mainly a few people, those who want to learn.

Those projects are generally more expensive to build than DIY kits because of the cost of parts and PCBs in low quantities. On the other hand, collaborative projects require everyone involved to work on schedule, otherwise some of the participants will lose interest or get frustrated...

Not sure I agree totally. Open source projects provide a second set of eyes and can add tremendous improvements benefitting all. Linux is an example. Windows is an example of what happens when it's not open source.

That said, I've kept the ELS closed source until I release the hardware. Once I start shipping I'm releasing the code and I hope to see a few people port it to other processors. Will it be stolen and perhaps released as a far east clone? Perhaps. But it's a specialty product unlike the DC servo so I doubt it. Can't say the same for a DC servo project which might be stolen and cloned into a cheap copy for sale.

As for DIY costs. The ELS is going to be sold with all SMD parts installed. The cost to kit 210 parts is higher than assembly costs for the SMD and the boards end up nicer. Instead I have to kit only 86 parts of which 70 are the buttons and button caps.

I'm building 200 to start. Hopefully they will all sell. At least then I'll break even. I've had some pretty tremendous support for it. Ohmikron and AVR_STMD stepper drivers were donated by their respective companies. Paul has graciously donated two HP UHU boards for testing with the ELS.

I did buy and build up an Embeddedtronics.com UHU board and although I could get the DC servo to move with the ELS the board was just too flakey with the noise coupling and the processor resets all the time. I have yet to find the time to try Manjeet's UHU version.

It's been interesting building these various kits. A lot different from the Heathkit days where everything was step by step. Hopefully, when people start assembling the ELS the feedback will be positive.

I think what I'm saying is any sort of DIY kit is expensive compared to a 'product'. Embeddedtronics.com only sells the bare boards. If it doesn't work they provide some email support but obviously can't afford to accept the assembled board back and repair it (if even possible).

Paul normally sells assembled boards I believe. (I just whined and snivelled a lot).

Manjeet's UHU board had some wrong components and backordered the FET drivers or gave me the chance to just accept a refund. I chose the refund. (Must check into why it's not back in my paypal account). But otherwise it looks like the original UHU board including the wrong hole size for the DB connector. It's a mid level difficulty assembly project.

I wonder if Maris at Gecko had made the G100 open source whether others would have jumped in and made it work. As I understand, that project has fallen by the wayside. The people who bought an early prototype are SOL.

I'm obviously babbling now so I'll leave it here. I'd like to see an open source and open hardware project including the complete reasoning behind component selection. But it's obviously not the best way to go if one intends on making a product for profit.

John

[tenmetalman]

John D,
Just to clear up the facts before the orders start flowing in........ Irfanulla & I sell kits that include most everything except Uli's UHU chip & crystal, also not included, the heatsink, & small misc things for $150/00 ea or three kits for $425.00 or the bare PCB'S at one for $25 or three for $65. We do NOT sell completed drives. although after getting my feet wet in this hobby area ......................... It only takes a couple of missing orders to put you back in the red. If we were paying ourselves by the hour we'd get one hell of a raise if we went on welfare ;^)
But look at al the fun we'd miss.
Paul

[jcdammeyer]

John D,
Just to clear up the facts before the orders start flowing in........ Irfanulla & I sell kits that include most everything except Uli's UHU chip & crystal, also not included, the heatsink, & small misc things for $150/00 ea or three kits for $425.00 or the bare PCB'S at one for $25 or three for $65. We do NOT sell completed drives. although after getting my feet wet in this hobby area ......................... It only takes a couple of missing orders to put you back in the red. If we were paying ourselves by the hour we'd get one hell of a raise if we went on welfare ;^)
But look at al the fun we'd miss.
Paul

Ain't it the truth!!!! I figure with all the reels of parts I've bought and 200 PC boards (to get a decent cost) I'll have to sell all 200 in order to finally break even. I know I'll never directly get paid back for the software R&D especially with releasing it under GPL. I'm also doing the boards in gold plate and all parts are RoHS so the European market won't have an issue with partially assembled stuff.

The up side though is I have an ELS on my home built Gingery Lathe and soon on my South Bend. And believe me, given the choice between using the dinky Gingery with all its' inherent vibration and chatter and the South Bend 10L with 1.375" bore and 5C collets, I'll use the Gingery+ELS for threading every time. I may prep the job on the South Bend but I'll put it between centres and whip out the threads on the Gingery.

Life is fun except when I forget to order some parts and have to do it again. Or worse, do a double order.... sigh...

John

[Xerxes]

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):

Torque regulation for DC motor isn't that heavy. It's just reading ADC and doing PI-regulation and writing result to PWM. It easily gets 10-20 times heavier on AC motor since commutation and vector math gets involved.

[kreutz]

Torque regulation for DC motor isn't that heavy. It's just reading ADC and doing PI-regulation and writing result to PWM. It easily gets 10-20 times heavier on AC motor since commutation and vector math gets involved.

Right. The extra processing power will be used on a few extra features, like power supply ripple compensation, P.S. voltage clamping control, electro-mechanical brake control, intermittent torque area use. Optional: +/-10V speed input, analog tacho-generator input, and maybe... some kind of adaptive auto tuning...

[tenmetalman]

Good Morning Kreutz & All,
As usual I'm reading the posts between you all, enjoying the heck out of them & understanding very little. I figure as I read I'll learn.
I just wanted to say that in my opinion, a new drive without Uli's chip could go just about anywhere you wanted to place it. it could go on this chat list as a natural progression to the development of DIY drives. Although there would be some cross talk. It could be it's own list under DIY servo controller. under it's own heading the start probably would be slow at first, then I'm betting the following and posts would increase at a impressive rate. every time i write on this topic I vacillate, today I'm for a new list.
Paul

[kreutz]

Good Morning Kreutz & All,
As usual I'm reading the posts between you all, enjoying the heck out of them & understanding very little. I figure as I read I'll learn.
I just wanted to say that in my opinion, a new drive without Uli's chip could go just about anywhere you wanted to place it. it could go on this chat list as a natural progression to the development of DIY drives. Although there would be some cross talk. It could be it's own list under DIY servo controller. under it's own heading the start probably would be slow at first, then I'm betting the following and posts would increase at a impressive rate. every time i write on this topic I vacillate, today I'm for a new list.
Paul

It will take some time to start that design, have a few projects that I need to finish before starting a new one. The Pick and Place machine will help a lot with the SMD bipolar micro-stepper boards, and will also be helpful for the new DC servo boards when the time comes... I am making some design decisions meanwhile because the P&P machine will ship in a few weeks and I want to make good use of the time until it arrives.

Best Regards,

Kreutz.