Current Controller for RC ESCs

[nieles]

i have a very weird problem. the voltage regulator is not working properly, it is puting out ~2v, but when i put my finger on R1 and R3 it does work.

it it not the pressure on the Resistors that makes it working, so it probably has something to do with the capacitance from my finger

i tried an extra resistor in parralel with R1 but nothing changed

 

[rhitee05]

My guess is that the PNP transistor Q1 is probably not being turned on, or maybe not turned on completely. What input voltage are you feeding the board? A couple of suggestions:

Probe between R1 and R3 at the base of Q1 to look at the voltage. I'd also look at the base of T2 (pin 1 of the op-amp) to see if the op-amp is doing what it's supposed to, and at the emitter of T2 (between T2 and R3) to make sure that it's being pulled down. If it's working correctly, Q1 should be switching on and off and holding the input of the regulator around 10 V.

The first possibility that comes to mind is that maybe R3 is too large. Does it work if you just put your finger on R3? You might be giving a little bit of capacitive coupling that lets more current flow from the base of Q1. I might drop R3 by a factor of two and see if that works.

I'm more inclined to the first possibility, since touching R1 and R3 makes it work, but you can adjust the value of R2 if it doesn't seem like the regulator is starting up correctly. R2 allows some current to flow past the regulator on start-up. There needs to be enough voltage to power the op-amp and let it turn on Q1 to power up the main regulator. If there's too much load, you might need to reduce R2.

The other thing is you could adjust the frequency of switching. It's not a fixed-frequency design, but you can adjust the value of C4 to make it faster or slower. Mostly, that will affect how much ripple voltage is allowed at the input to the main regulator. Larger C4 will slow the switching down and allow more ripple.

 

[nieles]

i lowered R1 to 50k, and the voltage is now 2.3v.

if i put the ohm meter across B and E of T2 it works,

it also works if i put 500K between the base of Q1 and GND (found out by touching gnd and R3 at the same time)

 

[nieles]

i think i know what the problem is.

i ordered a LMV931 for the opamp, as i couldnt find a ad8541 on mouser.

i thought i checked the pin compatibility of the opamps, but i guess not because the pinout is totally different.

 

[rhitee05]

That would probably do it!

If you want to keep testing, you could place a resistor to pull the base of Q1 down to ground (1k should be fine), and then it will just stay on. You should be fine as long as you keep the input voltage modest so the regulator doesn't overheat. I think 40 V is the maximum allowed (but don't trust me without checking the datasheet).

 

[nieles]

allright i have the powersupply working.. with 15v it is not getting too hot, so good enough for testing.

next problem. when the current controller is not connected to an ESC it is putting out an 15ms puls signal(1ms high, 14ms low) . i added a 100k to r14 and it is now 22ms (again 1ms high, but 21ms low).
the amplitude is 4.2v.

now when i connect it to an ESC the signal is only putting out 2-3v amplidude, and the pulswith stays the same. as a result the ESC goes to fault mode, i am pretty sure it is because of the

 

[rhitee05]

You might need to decrease the value of R18. Dropping it to 1k instead of 10k might do the trick.

 

[deVries]

allright i have the powersupply working.. with 15v it is not getting too hot, so good enough for testing.

next problem. when the current controller is not connected to an ESC it is putting out an 15ms puls signal(1ms high, 14ms low) . i added a 100k to r14 and it is now 22ms (again 1ms high, but 21ms low).
the amplitude is 4.2v.

now when i connect it to an ESC the signal is only putting out 2-3v amplidude, and the pulswith stays the same. as a result the ESC goes to fault mode, i am pretty sure it is because of the

Thanks Nieles (and Eric) for taking on this project to complete testing if possible. This is one of the top of the list problems to solve for ESC & RC motor use, and it seems the main problem is always finding someone with enough technical experience AND time to complete these prototypes to be usable as kits or DIY duplication of the pioneering work.

Thanks to you & Eric we may get there... 8)

 

[nieles]

i got some new opamps with the correct pinout.
i also changed R1 of the interface board to 50k, because with 100k the voltage regulator will not start.

the regulator is buzzing a bit.. can this be eliminated?

tomorrow i am off from work, so the plan is to mount the sensors and make a flux ring and do some real bench testing.

i will post when i have some more progress

Niels

 

[Thud]

Nieles,
you have not recived the flux rings i mailed out yet?

 

[rhitee05]

the regulator is buzzing a bit.. can this be eliminated?

You're talking about audible buzzing? I ran a SPICE simulation, and the operating frequency of the pre-regulator ends up being around 1 kHz, although this will vary with the load. My simulation shows that with the default resistor values the ripple should be around 1 V. Probably the best way to adjust that is to change the value of R5, which controls the hysteresis. Increasing R5 will tend to make the ripple smaller and the frequency higher, decreasing it will make the ripple larger and frequency lower. I suspect that increasing it is the way to go - a higher frequency would be more annoying, but lower ripple should make it less audible. Another option is to adjust the value of C9-C12, or change how many of them are populated. More capacitance will decrease the frequency, less will increase it - but the amplitude of the ripple should remain the same.

There are other changes that could be made if necessary, but those are the easy things to try first.

Also, I'm curious how clean the output from the 5V regulator ends up being.

 

[nieles]

you have not recived the flux rings i mailed out yet?

no haven't received them yet.

You're talking about audible buzzing? I ran a SPICE simulation, and the operating frequency of the pre-regulator ends up being around 1 kHz, although this will vary with the load. My simulation shows that with the default resistor values the ripple should be around 1 V. Probably the best way to adjust that is to change the value of R5, which controls the hysteresis. Increasing R5 will tend to make the ripple smaller and the frequency higher, decreasing it will make the ripple larger and frequency lower. I suspect that increasing it is the way to go - a higher frequency would be more annoying, but lower ripple should make it less audible. Another option is to adjust the value of C9-C12, or change how many of them are populated. More capacitance will decrease the frequency, less will increase it - but the amplitude of the ripple should remain the same.

There are other changes that could be made if necessary, but those are the easy things to try first.

Also, I'm curious how clean the output from the 5V regulator ends up being.

yes the buzzing is audible, today i tested with an battery instead of a powersupply and it is a lot better.
right now i have all c9-c12 capacitors populated.

here is a shot of the scope connected to the 5v line. settings: 5mv/div and 1ms/div

sorry could not get the picture any clearer

i also need to to some "high voltage" tests, as Q1 is 70deg C with a 5s (~20v) (all PCB's connected, throttle pluged in but no esc)
IC2 is around 50 deg C

 

[rhitee05]

I'd suggest trying to increase R5. Something in the 20-30k range should decrease the ripple amplitude. It looks like the ripple on the 5V line is around 10-15 mV, which isn't bad but it would be nice to get it a little smaller and that might help with the noise.

i also need to to some "high voltage" tests, as Q1 is 70deg C with a 5s (~20v) (all PCB's connected, throttle pluged in but no esc)
IC2 is around 50 deg C

That will be a good thing to check. Can you measure the input current? I'd expect the temp of IC2 to stay constant with a higher input voltage - if the circuit is working correctly, it's input voltage should stay around 10 V. You should also check to make sure that Q1 is being operated in saturation - look at Vce across the transistor and see how low it gets when the transistor is switched on. As long as it's being saturated, it shouldn't get too much hotter for higher input voltages. If it gets too hot though, you can swap out the zener diode for one with a higher Vz, which will shift more of the power dissipation to the linear regulator. Ideally they should split the dissipation.

 

[nieles]

the input current draw is 70mA.


this is what the signal on the Collector of the Q1 looks like
the bottom horizontal line is the 0v line for both channels


this is the same signal on the Collector, and on the 2nd channel the Vbatt.

scope settings: ch1 5V/div ch2: 5v/div and 0.2ms/div

with a voltmeter across C E of Q1 i measure 9.8V
when the Transistor is in saturation the signal should be a square wave beteween 20v and 0v right?

i also did a quick test with double the input voltage and Q1 is getting to 100deg C in like 10-20 seconds

 

[rhitee05]

the input current draw is 70mA.

Really? That seems much higher than I would've expected. That means that the 5V regulator is supplying a little less than 300 mA? Much higher than what I would have expected. How much of that is the boards and how much is the throttle or other external devices?

I've pondered this for a few days, and came to the conclusion that I probably goofed up a bit on the regulator design. I thought I could get away without the inductor for a really low-power converter, but it's the inductor which allows the transistor to saturate. Otherwise, the transistor always sees the difference between the input voltage and the pre-regulated voltage (in this case, 20-10 = 10 V). Even though it's pulsed, it's still the same average power dissipation as a plain old linear regulator. It looks like this will probably work okay for low input voltages, say 5s or 6s systems, but almost certainly not for 10s or 12s. I don't think there's any reasonable way to hack the board to put in an inductor (feel free to correct me if you see a way), so that probably means a re-spin for the interface board. On the plus side, we can probably get away with just a little inductance, so a simple little homemade air-core inductor might be enough. Since the layout would have to be re-worked to add an inductor anyway, I might consider swapping a P-FET in for the PNP transistor as well.

Let me know what you think. I'm excited to see how the rest of the circuitry works out.

 

[nieles]

I don't think there's any reasonable way to hack the board to put in an inductor

i will see what i can do with some inductors. i have some 22uh and 100uh i think, though those will be overkill because they are rated for around 3A
the inductor needs to be between the C of Q1 and pin 1 of IC1 right?
what if i lift pin 1 and connect the inductor between the via and the pin1. and also cut the trace between the the pin and via of course.
this will only be for my proto though i think.

since i installed the 3 hall sensors the board won't start up without disconnecting the interface board. when disconnected the powersupply will start and i can plug the board back in.
how low can i go with r1, to help with the startup?

How much of that is the boards and how much is the throttle or other external devices?

i will disconnect everything and measure again.

i have tried to decrease r3 to 500ohm, i thought that could be the problem. but i will change that back to 1k again.


i will continue testing the rest of the circuits on 5s for now.

 

[rhitee05]

i will see what i can do with some inductors. i have some 22uh and 100uh i think, though those will be overkill because they are rated for around 3A
the inductor needs to be between the C of Q1 and pin 1 of IC1 right?
what if i lift pin 1 and connect the inductor between the via and the pin1. and also cut the trace between the the pin and via of course.
this will only be for my proto though i think.

No, that won't work. The inductor needs to be between Q1 and the capacitors, so lifting the input pin of IC2 won't help. The problem there is that the collector of Q1 is also connected to the package tab and it relies on the copper fill for cooling... I've attached a picture of the only way I can think of to do it. The copper fill has to be cut at the two yellow lines, then the inductor and a wire jumper placed as shown in orange.

how low can i go with r1, to help with the startup?

R3 is what you want to adjust for startup. This resistor allows some current to bypass the regulator until it's up and running. For a board re-spin, it might be a good idea to add a control FET on the 5V output. The FET would only turn on and supply current to the 5V bus after the regulator is running, which should make the startup much easier. Although, if reducing R3 takes care of it, that's probably good enough.

 

[nieles]

R3 is what you want to adjust for startup. This resistor allows some current to bypass the regulator until it's up and running. For a board re-spin, it might be a good idea to add a control FET on the 5V output. The FET would only turn on and supply current to the 5V bus after the regulator is running, which should make the startup much easier. Although, if reducing R3 takes care of it, that's probably good enough.

now i am confused. R1 is 100k originally and R3 is 1k. so i want to lower R1 right?

so i added the two 100uh inductors in series that i had laying around. this is working a bit, but i think i need more inductance.
i could not get the scope to trigger correctly so i could not take a picture. is was in a bit of a hurry though so i will try again tonight.

a also measured the current draw again. with all boards connected (no external devices) the current draw is 63mA. the interface board alone draws 14.7mA when not connected to anything. (all at ~20v)

the temperature of Q1 is also lowered quite a bit. it is now at 54 deg C after running half an hour. this could also have to to with the extra thermal mass of the coils. the coils are way overkill (rated for 3A)
i will post some pictures tonight.

the startup problem is also still there. so the FET might be a good solution to ensure it will start 100% of the time.

 

[rhitee05]

Oops, sorry, I mis-spoke. What I should have said is that you want to lower R2 for better startup. R2 allows some current to flow directly from the input to the 5V rail, so the op-amp can become active and turn the regulator on. R1 serves to pull the base of Q1 up to turn it off, but it can't be too low or T2 will need to sink more current to turn Q1 on. R3 only serves to limit the current in T2 and needs to be much smaller than R1. Putting in a little FET start-up circuit is probably a worthwhile thing for a revision.

a also measured the current draw again. with all boards connected (no external devices) the current draw is 63mA. the interface board alone draws 14.7mA when not connected to anything. (all at ~20v)

It occurs to me that, if you populated the LED, that's probably where almost all of that current is going. You could probably increase R6 to cut down on the idle current draw. Or just take out the LED altogether!

so i added the two 100uh inductors in series that i had laying around. this is working a bit, but i think i need more inductance.
i could not get the scope to trigger correctly so i could not take a picture. is was in a bit of a hurry though so i will try again tonight.

I haven't done any calculations yet on what a good value might be. If L is small it will operate in discontinuous mode, but I think that's okay. Also, I totally forgot that it will also need a free-wheeling diode now, too (anode = ground, cathode = collector of Q1). You could even lift one of the caps and replace it with the diode. Sorry, it's the end of the semester here so things are crazy.

 

[nieles]

some more progress.

You could even lift one of the caps and replace it with the diode. Sorry, it's the end of the semester here so things are crazy.

no problem, school is more important than this project
the diode cant be in stead of a capacitor right? looking at the picture below, so i added beteween C of Q1 and GND

i killed the first interface board yesterday so i had to make a new one today. i left on spot open for capacitors C9-C12.
not sure what happend i was probing around and all of a sudden the regulator only put out ~2v.. no sparks or smoke, it just stoped working.


i increased r5 to 22k, added the coil of 100uh. and also added a 1n4148 diode for freewheeling.
to lower the power consumption i increased r6 to 330r. the interface board alone draws 10.5mA now.

Q1 is staying a lot cooler with all these mods. around 30C. i used a different coil this time.
some shots of whats going on at Q1.


settings: 5v/div and .2ms/div


settings 5v/div and 50us/div

 

[nieles]

addition to my last post:

the inductor i used for the last interface board has an inductance of 100uh and a resistance of 50R. this inductor was running pretty hot after 5 min of testing, around 50deg C

i did the same test with a different inductor. inductance 100uh and resistance of ~0.2R

Settings 5v/div and 0.1ms/div (previous pic was 0.2ms/div)

 

[rhitee05]

the diode cant be in stead of a capacitor right? looking at the picture below, so i added beteween C of Q1 and GND

Yes, entirely correct and a stupid mistake on my part. Glad at least one of us is paying attention. It sounds like we're making progress.

Am I correct that I see a bit of oscillation in those scope plots? It looks a little more prominent in the first set, with the 50R inductor and less so with the 0R2 inductor. Are these plots of collector voltage that you're showing us? I'm seeing some similar effects showing up in SPICE simulations. It doesn't really seem to affect the circuit much, but I'm looking to try and isolate exactly the source and make some slight changes to eliminate it if possible. I'm trying to figure out exactly what rev 2.0 of this circuit should look like. The changes I'm looking at are probably things that we don't need to worry about hacking this one for, just little improvements for the next version.

Also, how you attached a flux ring to the sensor board, yet? I put that huge U-cutout in the interface board to make sure it doesn't interfere with the flux ring. But if we don't need the cutout, I can gain back a lot of board real estate that would make the layout easier, and also use more copper area to heatsink the regulator. Just something to think about as we're making lists of what to change.

 

[nieles]

the diode cant be in stead of a capacitor right? looking at the picture below, so i added beteween C of Q1 and GND

Yes, entirely correct and a stupid mistake on my part. Glad at least one of us is paying attention. It sounds like we're making progress.

Am I correct that I see a bit of oscillation in those scope plots? It looks a little more prominent in the first set, with the 50R inductor and less so with the 0R2 inductor. Are these plots of collector voltage that you're showing us? I'm seeing some similar effects showing up in SPICE simulations. It doesn't really seem to affect the circuit much, but I'm looking to try and isolate exactly the source and make some slight changes to eliminate it if possible. I'm trying to figure out exactly what rev 2.0 of this circuit should look like. The changes I'm looking at are probably things that we don't need to worry about hacking this one for, just little improvements for the next version.

Also, how you attached a flux ring to the sensor board, yet? I put that huge U-cutout in the interface board to make sure it doesn't interfere with the flux ring. But if we don't need the cutout, I can gain back a lot of board real estate that would make the layout easier, and also use more copper area to heatsink the regulator. Just something to think about as we're making lists of what to change.

yes 100% correct. below some more pics of of the oscillation.
yes this are the measurements between the collecotor of Q1 and gnd

this is the 50R inductor.
settings: 5v/div and 5us/div


with the same inductor but now 1us/div


and this is with the 0R2 inductor
settings: 5v/div and 5us/div

the horizontal line is 0V.

i have made a flux ring, but it didnt turn out very well. the only material i had acces to was some 3mm steel sheet. so i had to cut the circle my self also making the cutout beteween the 1/4inch holes wasnt easy(with the limited tools i have acces to anyway). when all the testing is done i will see if i can get some of these flux rings made at work. i will ask if i can add these parts to one of the orders they send out every week. could you get me a dxf or any other file format with some measurements? i have some problems with the scaling beteween imperial and metric and form dxf to inventor to be able to order them. if you have the important measurements for me i can redraw it if necessary.
the flux ring is not attached to the board yet, but it is just clamped beteween the hall sensors. some m3x5mm screws could be used to attach it permanently.
no, the big U-cutout is not needed at the moment. it could be just the 1/4inch holes just like on the current controller board. but if you ask me i would not put any parts there, atleast not on the bottom side. my reason for this: when testing is done and we can switch to a 100% smd version we can mount the boards closer together and make it a little more compact. also i dont think the pot meter on the hall board is needed for future designs.if we mount the boards closer together we can not use the arduino style headers. but once the design is tested it is not needed i think, and instead of the headers we use long pins and solder the boards directly together (maybe with some sort of spacer).

next i will test the powersupply @ 40v. let me know when you calculated a better value for the inductor. i borrowed a inductance meter from a friend so i can wind my own inductor with some thin magnet wire. and then test it.
what would you suggest for a min value for R2 to help with the startup?

Niels

 

[rhitee05]

I've attached what I think is the correct DXF file. I can export it from FEMM, but I don't actually have anything installed on my computer to read DXF... I know, lazy. You should be able to check against the BRD file for the dimensions, and if you need you can pull the hole locations and re-draw it yourself. The critical dimensions are the flats where the 3 Hall sensors lie, as they should be flush against the steel. The center of the holes is obviously fixed, but the exact diameters could vary a little if needed.

The oscillation showed looks fairly similar to what I'm seeing in SPICE, so that's good that the simulation is fairly accurate. I'll fiddle when I get a little time and work up a v2 design for regulator circuit that will hopefully fix that, include the inductor, and maybe a couple of little things. For the time being, 100 uH seems to be working so I'd say you can probably keep that if it works for you. I'll use the SPICE model to see if there's a better value for the v2 circuit. I think I'm going to design it so the oscillation frequency is quite a bit higher, so the inductor value will probably be lower.

no, the big U-cutout is not needed at the moment. it could be just the 1/4inch holes just like on the current controller board. but if you ask me i would not put any parts there, atleast not on the bottom side. my reason for this: when testing is done and we can switch to a 100% smd version we can mount the boards closer together and make it a little more compact. also i dont think the pot meter on the hall board is needed for future designs.if we mount the boards closer together we can not use the arduino style headers. but once the design is tested it is not needed i think, and instead of the headers we use long pins and solder the boards directly together (maybe with some sort of spacer).

The boards could definitely be mounted closer. The arduino-style headers are nice because it's easy to separate them, but if the design is stable it might be okay to have the boards soldered together using pins. I will still avoid placing any components on the bottom side within the flux ring, but having the top side available and being able to use the bottom copper will make things much easier.

what would you suggest for a min value for R2 to help with the startup?

The absolute minimum value for R2 would be about (20-5)/10e-3 = 1.5 k, assuming 20 V input and 10 mA idle current. With that value for R2 basically all the current is bypassing the regulator and flowing through R2. R2 might get warm, but it should be okay for a 1/4 W resistor. You'd have to increase the value for higher input voltages, otherwise the 5V bus will lose regulation and be pulled higher. The main fix here is just to add that turn-on FET so the regulator can power-up without that large current draw. I'll definitely put that on v2.

 

[nieles]

well for some reason my ESC (turnigy monster-2000) isn't working. i it was working 2 weeks ago when i last used it, oh well it was a crap controller anyway.
so i am in the maket for a new controller. what should i get? the ICE2 HV 160 would be a good option, but expensive and out of stock everywhere.

are there any other ESC out there that have good startup? like the HV 160?


bit off-topic;

eric, how hard would it be to also have the option to use a "rc 20ms signal" for the throttle input? this could be handy for mountainboards and rc cars.
maybe add this option on its own pcb and add it to the stack if you want this option.