Reply with QuoteI would think that at a minimum you want one sense resistor just to make sure you aren't blowing things up. Three might be overdoing it for this style of controller.
SMD discrete MOSFETs
IRAMS10UP60A power module
Discrete through-hole MOSFETs with heatsink
SMD discrete MOSFETS + possibility to connect own custom external power stage
Controller logic only without integrated power stage. An external power stage must be used.
Some other solution (please specify)
How much research have you done with the current/torque loop. I have no idea how advanced you want the drive to be. High end drives have a way to detect phase voltage/current. While this may not be necessary, designing this into the hardware in the first place will make adding different control algorithms much easier in the future. Below is a very good intro on some advanced control techniques. Some other drives also auto-tune and have feedfoward and pole-placement control.
http://www.worldservo.com/html/arc.htm <-Vector Drive (some different control ideas)
http://www.yetmotion.com/YetIsrael/P...tails.asp?ID=7 <-adaptive feed forward (copy of Yaskawa maybe??)(check out the pdf)
But don't let feature creep get you, but designing the hardware to sense the voltage and current shouldn't be a problem since the dsPIC is so awesome
I would think that at a minimum you want one sense resistor just to make sure you aren't blowing things up. Three might be overdoing it for this style of controller.
Very interesting links! It takes some time to study that dsPIC bootloader stuff, there is code but no good documentation. I have to read the code and simultaneously look datasheets to get idea how it actually works. I refuse to use code that I don't understand thoroughly.
CNCAddict, there are companies in both sides. Others claim that vector control is better than current regulated and the others claim the opposite. It seems that I have to try both to find out the truth.
My current work is using current regulated drive. It has current sensors on two phases. Third phase current can be solved by calculating sum of two first currents and therefore third sensors isn't necessary. One sensor is not enough, it would make only trapezoidal drive possible.
WOW, it sounds like you have everything under control (wow, what a corny joke). I think all the super high end drives have Vector control, and according to World Servo, it allows peak torque at much higher rpms since motor timing is automatically advanced to account for motor inductance, etc.
Changing the subject, do you plan on using a lookup table for the Sinewave commutation? I don't think the dsPIC is fast enough to actually calculate the duty cycle on the fly for each encoder count. But that would be a bit extreme anyway. I think a fixed number of commutation points (64, 128, etc) would be more than smooth enough. I know some people aren't very happy with the trapezoidal control on the Rutex and it sounds like your drive will be far more advanced and higher performance. I'm very excited to see progress going so fast. Keep it up
David B.
CNCAddict, are their post stating people arnt happy with the Rutex controls?
I recall Tom posting that the drives aren't truely trapazoidal in that they don't have the typical cogging at low speeds.
have you seen anything from someone who has one of these drives or is it just hearsay?
thanks
Michael T.
"If you don't stand for something, chances are, you'll fall for anything!"
I haven't seen a single post from anyone actually using the Rutex, other than Rutex US (Tom). There was some speculation that they wouldn't work well, but I think everyone was witing for the new drives to actually try them. Tom stated they have a money back guarantee if you're not happy with them.Originally Posted by CNCAddict
Gerry
UCCNC 2017 Screenset
[URL]http://www.thecncwoodworker.com/2017.html[/URL]
Mach3 2010 Screenset
[URL]http://www.thecncwoodworker.com/2010.html[/URL]
JointCAM - CNC Dovetails & Box Joints
[URL]http://www.g-forcecnc.com/jointcam.html[/URL]
(Note: The opinions expressed in this post are my own and are not necessarily those of CNCzone and its management)
Hmm, now that you mention it I can't say for certain where the post came from. But I believe it was in the Sanyo Denki thread. Sinewave commutation is always better than trapezoidal for motion control but on real applications the difference may be nill. I'm so excited about this DIY AC servo drive that I may have mis-spoken about the rutex
Yes, I'm using lookup tables. 64 or 128 points are well enough for smooth commutation. Using calculated sine is not impossible either, I have written my own "fixed point" sine function that should be smaller and faster than the one provided by math library.
I'm getting to point where I need RS232 communication with dsPIC. Today I scetched a schematics for optoisolated RS232 interface. It draws power from serial port for optos like serial mice do:
http://www.tkk.fi/Misc/Electronics/c...s/rspower.html
Optos are also acting as level converters so no MAX232 like chips are needed. TX and RX lines can be directly connected to MCU.
Opinions?
Last edited by Xerxes; 12-07-2005 at 12:52 PM.
I stared at that for a while before I realized what you were doing. Too simple I guess. That looks good. I can't imagine an optocoupler would draw more current than a 7805.
The PCRX circuit looks funny to me. It looks like you are deriving negative voltage from the PCTX line, shouldn't you hook it to an RS232 control signal?
PCRX is a recieve line of PC so it has high impedance and power can not be drawn there. Positive voltage comes from PCDTR and PCRTS lines which normally have +12V all the time. Negative voltage comes from PCTX line which has -12V when not transmitting data. Negative voltage capacitor must hold enough charge to keep circuit operational during data transmission.
I modified the circuit little bit, now it's hopefully easier to read. I also swithed PCRX line to be pulled up to +12V with resistor because it spends most of its time at +12V.
Last edited by Xerxes; 12-07-2005 at 01:59 PM.
I realize now that it saves you a wire. The website you listed used a control pin, they probably didn't care about the extra wire. Very nice. Looks like a nice conservative design.
I like the change in the schematic.
Hi folks!
I was working on the schemas & layout when I noticed that the PCB is getting way too crowded even when it's only half populated. I'm limited to this eagle 100x80 mm board. So it seems that I have to switch on SMD's.
SMD's seems to be very effecient on saving some footprint so I can keep some parts thru hole. I'm planning to keep power stage thru hole design so it's easier to avoid fatal short circuits. Only logic side would have SMD's.
I already switched most passives to 0805 size chips. What do you guys think is 0805 too small / small enough?-)
0805's aren't too hard to deal with. I've only done a limited amount of soldering with them, but even with little experience it wasn't hard at all. I think going smaller might cause problems though. Some comments from the pro's would be welcome
Xerxes, the guys over at the yahoo OSMC group are very experienced with power/brain board designing. If you haven't allready, it would be smart to sign up and tell them what you're doing. They would be more than happy to help.
http://groupyahoo.com/group/osmc/
OK, CNCzone doesn't like that link, it should read "groups.yahoo"
Thanks CNCAddict! That link might be helpful if I encounter problems. I will posts schematics and layout for open inspection when they are complete enough.
Here is an another circuit I drew. HV+ input is the high voltage DC bus voltage and SOURCE goes directly to power stage MOSFETS. The circuit has two functions:
1) L1, L3 and D2 forms one kind of shoot-thru current reduction circuit. It is impossible to avoid all power stage cross conduction no matter how long dead time is being used so this circuit might improve efficiency and reduce EMI.
I'm not sure if any kind of snubber circuit would be necessary here to damp oscillations because inductors and diode capacitace tends to form an LC circuit. L1 and L2 actually are 7 Amp rated ferrite bead filters which typically intentionally have high HF losses to avoid oscillation.
2) Rest of the circuit forms short circuit / over current detection logic. When voltage over RS2 grows over 0.5 volts then transistor Q4 pulls OC pin up that tells controller to disable all outputs. D5 clamps the base-emitter voltage of Q4 to avoid destroying that small signal transistor. BTW, do anyone know a good low power transistor for 200 volts? This one is only for 150V.
Comments?
MPSA92 makes a good 300V small-signal PNP transistor.
I would change the sense circuit though. The attached Isense.jpg works very well. The left MPSA92 has a 470 Ohm resistor in its emitter circuit and a 4.7K resistor in the collector circuit for a circuit gain of 10.
The right MPSA92 has only its base-emitter junction used. Its Vbe closely matches the left MPSA92 Vbe and more importantly temperature-compensates the left MPSA92's Vbe. The 100K resistor provides the base bias current.
When no current flows (left to right) thru the 0.025 Ohm sense resistor, the left MPSA92 is just on the verge of conducting. If 1A flows, then the collector current will be 0.025V / 470 Ohms or 53.2uA. This will produce 0.25V across the 4.7K collector load.
20A thru the 0.025 Ohm resistor will produce 5V across the 4.7K for a 1.06 mA collector current. This circuit will produce a linear voltage proportional to current thru the sense resistor; being temp-compensated, the voltage will be independent of component temperature over a very wide range (-55C to 125C). Keep both MPSA92s in close physical proximity (same temp).
The circuit freq response is over 10MHz so it's plenty fast. The 1mA collector current allows for easy clamping during gross overcurrent situations and keeps the transistor dissipation to under 300mW even at a 300V supply to ground differential. The 0.25V per Amp transconductance linearity is much better than 1% over a 1A to 20A range. I designed this circuit for the G202 short-circuit sense protect front end so it's well tested.
Mariss
I didn't mention why this circuit is better. The other circuit has a few problems:
1) If the current thru the 0.025 Ohm resistor exceeds 28A (0.7V), it will destroy D5 thru overcurrent.
2) Q4 is not temp-compensated. Figuring -2.2mV / degree C Vbe temp coefficient and Vbe = 0.6V @ 25C, Q4 will turn on at 31A at -55C and at only 15A at +125C. This is not a precise, predictable response.
3) R22 gets the brunt of the dissipation; it is 9W at 300V, still 1W at 100V.
4) There is no scaling for the desired current limit trip point; it's either 'on' or 'off' somplace between 15A and 31A depending on circuit temperature! That broad range can only be changed by changing an expensive low-Ohm sense resistor instead of a cheap 1/4 "penny" resistor.
Mariss
you could also use a linear tech
lt1787 current monitor ( 8 pin sop)
readily available, fast, scaleable and very linear
shoot them directly into an adc pin
i have used them and they seem almost bulletproof
maybe i am lucky ( so far)
Mariss, a very clever circuit. Thousands thanks! How small sense resistor value can be used without making it too inaccurate?
Maybe this kind of circuit could be used also as phase current sensing when modified to add some output DC-bias on zero amp current. Any ideas for modification? The goal is to make it sense also negative currents.
I thought it would be easiest to put zeners on emitters so the biasing transistor base would have about 0.25V lower potential causing some DC current thru the 470 ohm resistor when voltage over the sense resistor is zero. For example 2.4 and 2.7V zeners. Of course temperature stability is affected by those zeners but hopefully not too much.
I think those aren't very big problems:
1) BYV28 diode can survive 90A current for 10ms. And some resistor could be added before diode to protect it.
2) Fortunately the temperature instability seems to happen in the favourable direction. MOSFET current rating derates while temperature rises so lowering current protection threshold would actually compensate this. :-)
3) Resistor shouldn't fry because short circuit duration is so short. Only microjoules would be dissipated.
But the bad thing is that MPSA92 is not very suitable for this type of circuit because it has Vbe of 0.9 volts making sense resistor larger.
Odd that Linear doesn't mention anything about bandwidth in the datasheet. That kind of chip would probably make a very good phase current sensor if it had higher voltage rating (LT1787HV is only for 60V). Currently I'm planning to use ASC704 current sensors.
1) I guess my biggest objection to the original circuit is the parallel diode doesn't really do anything useful. Your sense resistor can be thought of as a voltage source with an impedence of 0.025 Ohm, your diode can be looked at the same way, a .8 to 1V voltage source with very low impedece as well. Two low-impedence voltage sources do not play together nicely in parallel.
2) The PNP Vbe doesn't really matter so long as it's balanced (cancelled). You can look at the circuit as a fixed gain differential amplifier with a ground referenced output.
Let's say 1A flows thru your 0.025 Ohm resistor left to right. The left end of the resistor will be +0.025V relative to the right end. Let's say Vbe happens to be 0.5V for both MPSA92s. The emitters will be at the same potential. If Vbe increases to 1V for both MPSA92s, the emitters will still be at the same potential.
The diode-connected (right) MPSA92 has its emitter at the same potential as the right end of the sense resistor. This makes the left emitter that same voltage. The 470 Ohm resistor has 0.025V differential across it so it carries 53uA of collector current.
In effect the emitters become differential inputs. It doesn't matter if the left end of the sense resistor is at 24VDC or 240VDC; the same 53uA will still flow if 1A is passing thru the sense resistor. This makes the circuit meet the other criteria of a differential amplifier which is common-mode voltage rejection. The output voltage is only a function of sense resistor current. A changing power supply voltage has no effect.
Mariss
Posting Permissions
