|
Post by hexpod on Oct 13, 2016 20:31:09 GMT
Hi,
After playing a lot with Pid, I could eradicate the oscillations but not in a very efficient way.
Actually more you push the "P" parameter, more you increase delay in the system.
When I got some satisfactory compromise in one position, I've remarked that it is NOT working for all arm angles.
That's why I am suspecting that the issue could come from our chipo sensors.
That's why I would like to go for high quality sensors.
Without making any advertisement the only place where I found quality products for 5V sensors is here:
www.contelec.ch/en/home.html
Because their catalogue is huge I would like to have an advise about what could eventually be the replacement for our Mexican stuff (the black ones) If it worth the invested money it's another question.
For ex they have a "Vert-X" contactless series and Vert-X E "non contacting" series which comes basically in three types:
5V 10-90 (05-.4.5) 5V PWM 5V SPI
Vert-X13
Vert-X16
Vert-X22 MH-C
Vert-X22 MH-C2 (14bit)
Vert-X28 MH-C
Vert-X28 MH-C2 (14bit configurable)
Vert-X 37 MH -C
Vert-X 37 MH -C2 (14bit configurable)
www.contelec.ch/nc/en/products.html?action=list&typ=Vert-X
We also have the "vert-X serie E" (non contacting) in two pieces
www.contelec.ch/nc/en/products.html?action=list&typ=Vert-XE
Thanos, could you please tell if the Vert-X and Vert-X E, both in 5V 10-90 (05-.4.5), will work "out of the box"? Could you also please comment if the AMC will support the PWM and PSI sensing technology in the future?
Thanks a lot
cheers
|
|
|
Post by tronicgr on Oct 14, 2016 2:08:40 GMT
The analog versions 0-5v of the above sensors will work same as the BI-technologies ones. You may see the same behavior though, it won't increase much the accuracy even if you use lots of oversampling. You will still need shielded cable for the noise too.
The PWM ones are our of the question too, it takes a lot to accurately decode the timing of the PWM signal with the poor timer capture features of the Atmega. It doesn't even have enough to do six PWM sensors.
I wish I could get my hands on one or two of the SPI versions, I could try to interface it and get all the 12 or 14 bit resolution without a hint of noise from it. I could multiplex for all six sensors using the extra six analog ports on the amc1280usb (that are not populated).
How much these cost anyways?
Thanks Thanos
|
|
|
Post by paulg100 on Oct 14, 2016 10:48:54 GMT
"That's why I am suspecting that the issue could come from our chipo sensors." I am coming to the same conclusion. The Bi Technologies sensors seem very sensitive to noise which is preventing me from getting the start frequency of the VFD,s low enough for 1:1 response of the motion cues and eliminating lag. I looked into optical encoders some but the cost!! the AMC dosent support them at present anyway?
|
|
|
Post by hexpod on Oct 14, 2016 11:02:41 GMT
The analog versions 0-5v of the above sensors will work same as the BI-technologies ones. You may see the same behavior though, it won't increase much the accuracy even if you use lots of oversampling. You will still need shielded cable for the noise too. The PWM ones are our of the question too, it takes a lot to accurately decode the timing of the PWM signal with the poor timer capture features of the Atmega. It doesn't even have enough to do six PWM sensors. I wish I could get my hands on one or two of the SPI versions, I could try to interface it and get all the 12 or 14 bit resolution without a hint of noise from it. I could multiplex for all six sensors using the extra six analog ports on the amc1280usb (that are not populated). How much these cost anyways? Thanks Thanos You are suggesting that my issues (unable to hold position at some arm angles) are noise related and have less to do with the sensor quality or his resolution. Did I got you right?
Thanks
|
|
|
Post by tronicgr on Oct 14, 2016 14:25:19 GMT
The analog versions 0-5v of the above sensors will work same as the BI-technologies ones. You may see the same behavior though, it won't increase much the accuracy even if you use lots of oversampling. You will still need shielded cable for the noise too. The PWM ones are our of the question too, it takes a lot to accurately decode the timing of the PWM signal with the poor timer capture features of the Atmega. It doesn't even have enough to do six PWM sensors. I wish I could get my hands on one or two of the SPI versions, I could try to interface it and get all the 12 or 14 bit resolution without a hint of noise from it. I could multiplex for all six sensors using the extra six analog ports on the amc1280usb (that are not populated). How much these cost anyways? Thanks Thanos You are suggesting that my issues (unable to hold position at some arm angles) are noise related and have less to do with the sensor quality or his resolution. Did I got you right?
Thanks
With current resolution of the ADC, you can achieve 0.175 degrees steps of angle if you are using 180 degrees sensor. If your motor cannot hold on certain angles it might be also mechanical reasons or motor losing torque from misaligned coils at that angle. Remember that once the motor reaches its target position, the speed is set to 0, which in turn might let the motor coast to a lower point where the PID increases the speed setting to get the motor position back to correct target. So the higher the setting is for the P, it will be less likely to drift off its target too much. The noise from the sensor might disrupt this and add vibration while trying to hold position with higher P values. In my opinion using SPI digital absolute sensors would be the best bet right now. I'll see if I can get my hands on them. Quadrature encoders are hard to use and are not absolute, they need homing every time you want to use them and if they loose count for any reason, there is no way to detect it and correct it unless you hit the limit switches (which has to be also actively read from the microcontroller). Thanks Thanos
|
|
|
Post by hexpod on Oct 14, 2016 16:03:22 GMT
|
|
|
Post by hexpod on Oct 14, 2016 16:13:18 GMT
"In my opinion using SPI digital absolute sensors would be the best bet right now. I'll see if I can get my hands on them."
For USA people, it seams to be the place to ask about contelec products:
www.novotechnik.com/Rotary-Overview/rotary.html
cheers
|
|
|
Post by paulg100 on Oct 14, 2016 16:50:39 GMT
I'm not sure if it is the same issue. All i know that after many hours of experimenting with PID , in order to improve response and achieve better repeatability of position, it really helps to get starting frequency of VFD as low as possible to achieve 1:1 with motion cue, but because of signal noise I cant get lower than 0.6hz at present (0.5hz is default with wj200).
Did you ever check signal of pot with oscope? especially with all 6 running? would be interesting to see how clean the pot signal is.
"When I got some satisfactory compromise in one position, I've remarked that it is NOT working for all arm angles."
Did you try the new firm ware with individual PID for each motor? That did help improve my situation, mostly a slightly different I for some motors.
|
|
|
Post by hexpod on Oct 14, 2016 16:58:56 GMT
I'm not sure if it is the same issue. All i know that after many hours of experimenting with PID , in order to improve response and achieve better repeatability of position, it really helps to get starting frequency of VFD as low as possible to achieve 1:1 with motion cue, but because of signal noise I cant get lower than 0.6hz at present (0.5hz is default with wj200). Did you ever check signal of pot with oscope? especially with all 6 running? would be interesting to see how clean the pot signal is. "When I got some satisfactory compromise in one position, I've remarked that it is NOT working for all arm angles." Did you try the new firm ware with individual PID for each motor? That did help improve my situation, mostly a slightly different I for some motors. Thanks for the advise. I am not using the inverters. I am on DC side with brushless servos. I will check the individual Pid. I did not tried yet.
cheers
|
|
|
Post by tronicgr on Oct 14, 2016 17:28:03 GMT
I found locally this Absolute encoder with SSI interface (for multiplexing) www.digikey.com/product-detail/en/broadcom-limited/AEAT-6012-A06/516-2341-ND/2210458Its 12bit (4096 positions) with resolution 0.0879 degrees within 360 rotation range. I could limit it in software to read only 180 degrees with less resolution (2048) or gear it down mechanically. I'll try this asap,it's simple to interface with no changes on the amc1280usb hardware. I'll be looking for other solutions too, the sensors in the links you gave, does not have any information on the interface ( is it actually SPI or SSI?) Thanks Thanod
|
|
|
Post by hexpod on Oct 14, 2016 17:36:32 GMT
I'll be looking for other solutions too, the sensors in the links you gave, does not have any information on the interface ( is it actually SPI or SSI?) check this link. you will find everything you need. you have to ask for prices
www.novotechnik.com/Rotary-Overview/rotary.html
cheers
|
|
|
Post by tronicgr on Oct 14, 2016 18:29:03 GMT
I'll be looking for other solutions too, the sensors in the links you gave, does not have any information on the interface ( is it actually SPI or SSI?) check this link. you will find everything you need. you have to ask for prices
www.novotechnik.com/Rotary-Overview/rotary.html
cheers
I asked them for samples and prices for the vert-X 1300 with SSI interface that is more simple than full SPI. I'll have to wait for their answer. It uses the same SSI interface as the one I found here locally on Digikey. Thanks Thanos
|
|
|
Post by hexpod on Oct 14, 2016 18:37:21 GMT
"With current resolution of the ADC, you can achieve 0.175 degrees steps of angle if you are using 180 degrees sensor."
are you talking about AMC analog-digital converter? I thought it was 16 bit capable.
cheers
|
|
|
Post by hexpod on Oct 14, 2016 18:56:45 GMT
"I asked them for samples and prices for the vert-X 1300 with SSI interface that is more simple than full SPI. I'll have to wait for their answer.
It uses the same SSI interface as the one I found here locally on Digikey."
Thanks for digging into it!
Indeed 14bit digital 360deg. sensing would finally close the topic. :-)
|
|
|
Post by tronicgr on Oct 14, 2016 19:24:51 GMT
"With current resolution of the ADC, you can achieve 0.175 degrees steps of angle if you are using 180 degrees sensor."
are you talking about AMC analog-digital converter? I thought it was 16 bit capable.
cheers
Its selectable resolution from 10bit to 16bit, but the more you increase the resolution, the slower the controller gets as it takes multiple readings to achieve the higher resolution.
|
|