Inductance What is it what does it do? Collosus has 8uH!

[Toorbough ULL-Zeveigh]

It's a little more complicated than the volt-second analysis.

then how about z-parameter analysis.

 

[John in CR]

Oh well. I had my fingers crossed that pulsing DC was sufficiently different. It would seem that to an inductor pulsing DC at 40khz is enough different that we'd get the rounded edges of the pulse that we need. I guess I'm going to have to come up with some kind of iron cores for my bigger motor.

Thanks again for the patient explanations.

John

 

[rhitee05]

Oh well. I had my fingers crossed that pulsing DC was sufficiently different. It would seem that to an inductor pulsing DC at 40khz is enough different that we'd get the rounded edges of the pulse that we need. I guess I'm going to have to come up with some kind of iron cores for my bigger motor.

It's the average current that gets you, not the pulses.

then how about z-parameter analysis.

I guess I'll assume you mean impedance? I don't see how that would be useful.

Saturation is a non-linear phenomenon. Most of the tools we use to simplify problems only work for linear systems, so here we have to do things the hard way. Although, the "hard way" doesn't really seem that hard to me. The equations are pretty straightforward... No need to complicate things.

 

[John in CR]

It's the average current that gets you, not the pulses.

So I take it that the saturation lingers until an AC switch of polarity gives the inductor a fresh start. If it cleared out with each DC switch off, then we wouldn't care so much about the average would we, since it's the delay in the beginning of the pulse that we're looking for from the inductor?

Does a sine wave controller make this issue go away?

 

[rhitee05]

I think you're making this more complicated than it needs to be.

Excluding commutation changes, the phase current never goes to zero during the PWM cycle. It increases and decreases as the PWM switches on and off, but ideally the ripple is fairly small so the current flow is relatively constant. That's the whole point of the inductors, after all. That constant (average) current is mostly what we care about when determining if the inductor's core will saturate. If you want to get really precise about it, we do care about the ripple because the maximum current is worst-case for saturation... but that level of detail isn't really needed.

Think of it this way. In a dramatically over-simplified analogy, an inductor behaves like an inductor below a certain threshold and like a plain wire above that threshold. If you stay below the threshold, you can forget about all this and go back to life as usual. In real life, it's not a sharp edge, but the inductance will gradually decrease in a non-linear fashion from the nominal value at zero current to zero inductance at infinite current. We like to assume that inductance isn't a function of current because that makes the math really messy, and that's a reasonable assumption for low currents. "Low", of course, is relative.

 

[John in CR]

Thanks Eric. I didn't realize the current still flows during the OFF portion of PWM. In that case, doesn't average current have to equal battery current? If so, then my cores should be fine and it's just a matter of getting to the right inductance value.

I stopped by the electronics shop and they have no tool to measure inductance, so now I'm playing the waiting game on shipping and Customs.

John

 

[rhitee05]

In that case, doesn't average current have to equal battery current?

No, this is the infamous, often mis-understood, and not well-named "current multiplication" effect. Ignoring losses, power in must equal power out. Since the average voltage out is lower than the battery voltage due to PWM, the current must be higher.

 

[Arlo1]

I just found out about the Tritium Wavesculptor motor controller, Min for their controller is 50uH
http://www.tritium.com.au/products/TRI50/TRI50.016_Datasheet_v6.pdf

The WaveSculptor requires a minimum amount of inductance in each motor
phase to properly regulate current. Not providing this inductance may result
in an out-of-regulation condition of the motor current control loop, possibly
resulting in an undesired self-protection shutdown, or failure of the
controller. Please ensure that both the motor inductance, and any external
inductors (if used), are still providing at least the minimum required
inductance, even at full rated current, and at elevated temperatures.
7. As long as the minimum inductance per phase requirement is met, the
WaveSculptor will regulate current and operate successfully into a shorted
connection.
8. The WaveSculptor can report inductance and resistance present on it’s
output when running the configuration / setup program. This will provide a
figure for the complete output circuit, including motor, external inductors (if
any), wiring, and connectors. This can be used to verify these values meet
the datasheet requirements, but only for low current operation.

 

[John in CR]

Nice, that almost identically matches the 110uH minimum that you came up with as what we need, 55uH per phase.

 

[Lebowski]

I think you guys got it totally wrong with the inductance... These controller's you're using, are they
hall sensor based or some sort of simple sensorless (windmilling / back emf sampling) ?

I have the suspicion that by adding an external L you're actually adding more R than L
(so you're increasing the overall R/L ratio of the windings). Increasing the R w.r.t. the L
makes it possible to use a hall based / simple sensorless controller, this is why you
guys report better results by adding an external L. But it's not the external L that's
doing it, it's the extra R.

It goes wrong with these big motors at higher RPM's, right ? Just as a thought
experiments (this is not the root cause in my view, but just to explain), the
R / 2*pi*f*L impedance ratio decreases for higher rpm.

I can imagine you need some inductance to not overload the output stage during the
PWM cycle, to smooth out the PWM as it were. But (simple) controller algorithm
wise inductance is a big no-no

 

[Arlo1]

Nice, that almost identically matches the 110uH minimum that you came up with as what we need, 55uH per phase.

I think they mean 50uH measure across one phase so 50 total. The 110uH I have been aiming for is for dumb controllers that don't know phase current. And is a number fro CRSIO and a few other controller set ups I have found. But 50uH is possible with colossus I managed to get 54uH with is WYE wound and running the 2 halves of the stator separated. SO two groups of thee phase wires meaning you will need two sets of power stages.

 

[Arlo1]

I think you guys got it totally wrong with the inductance... These controller's you're using, are they
hall sensor based or some sort of simple sensorless (windmilling / back emf sampling) ?

I have the suspicion that by adding an external L you're actually adding more R than L
(so you're increasing the overall R/L ratio of the windings). Increasing the R w.r.t. the L
makes it possible to use a hall based / simple sensorless controller, this is why you
guys report better results by adding an external L. But it's not the external L that's
doing it, it's the extra R.

It goes wrong with these big motors at higher RPM's, right ? Just as a thought
experiments (this is not the root cause in my view, but just to explain), the
R / 2*pi*f*L impedance ratio decreases for higher rpm.

I can imagine you need some inductance to not overload the output stage during the
PWM cycle, to smooth out the PWM as it were. But (simple) controller algorithm
wise inductance is a big no-no

You sir are 100% wrong in my case. And I can prove it.
#1 With the 27uH inductors I have and put one in series with each phase on colossus. I can ride my bike all day with no problems.
Then keeping the wires hooked up as you can see in my Dyno test videos. I take turns off one at a time and the HP comes up and the torque goes down! I did this test and could make 3.5HP with 27uH extra with a 24 mosfet china controller then I went all the way down to three turns on the inductors core with the rest of the inductor wire STILL CONNECTED there for having THE SAME RESISTANCE and I made 6.5HP and less torque then tried going for a ride and the controller blew 3 seconds into a full throttle acceleration
And no the problem is at LOW rpm not hi RPM.
[youtube]Ui0CVEbf5fg[/youtube]

 

[Arlo1]

Lebowski.
Here is a screen shot of the spreadsheet bigmoose gave us.
Motor A is colossus (100% stock)
Motor B would be colossus with 2x the inductance and the same resistance as stock colossus
Motor C would be colossus with 2x the resistance and the same inductance as stock colossus

Once you see this you quit worrying about anything other then getting enough inductance.
I will re-mention I scoped my 24 fet running my bmx and the PWM on-time will vary from 5uS - 65uS so ALL IT TAKES IS 25uS FOR A 24 FET TO BE TOAST WITH COLOSSUS! If the controller makes one mistake and under estimates the current Bam I have a controller full of popcorn again!

PS look at how the resistance being double almost makes no difference!!!!!

 

[John in CR]

+1 on the low or even 0 rpm with the tiniest of throttle pulses. I've only blown 2 controllers above 5mph and both were on a the same moderate hill while not maintaining sufficient speed due to traffic, ie partial throttle. All contollers have run quite warm with those motors, even the factory 15 fet pushing only 25amps peak.

 

[Lebowski]

It all fits though with my view of how motors and controllers work. You guys got big timing issues, something
you especially notice on very low impedance motors. What Arlo reports, bigger hp numbers at lower torque, fits
and is called 'field weakening'. It's also a slippery slope... I know you guys use 1 or 2 sets of halls, maybe with
Burtie's electronic timing change circuit. Problem though is that correct timing depends on everything, motor speed,
motor currents, amount of throttle, motor temperature (due to changing wire resistance). And with low impedance...
a way to think about is that a motor-phase follows a circle, you get a sine wave back-emf for a full circle
rotation

[youtube]HFOPb7Ub1e8[/youtube]
(the black dotted line is the phasor of 1 motor phase following the circle, in red the sine we measure as back-emf.
a motor has 3 rotating phasors 120 degrees apart)

at 48V the circumvence of the circle is 150V, so 2-3 degrees wrong in the timing is 1V, with 0.01 ohm this
is 100 Amps !

If you think about it a bit more, a normal controller with hall sensors divides the circle in 6 parts of 60 degrees, the controller
provides phasors which jump over 60 degrees. The motors back-emf nicely follows the circle. So as the controller
delivers its fixed phasor, the motor first lags 30 degrees (12 Volt !), then has perfect timing, then leads 30 degrees (12V wrong again !)
before the controller switches to the next phasor.

It's a miracle those simple controllers even work. They're helped by typically working at 24 or 36 V with a low power
/ high impedance motor (250-500W, a few 0.1 's ohm impedance per winding) but it's no wonder those motors
heat up like crazy and that a true low impedance motor blows those controllers to kingdom come.

 

[Arlo1]

Leb. I scoped my hall sensors and they are uber close to perfact now! I have 2 sets installed in my rewound motor. But I dont think you are grasping this because you have not had to deal with it. I Have now spent over 1.5 years learning about this. And I got the idea of external inductors from a friend who had a co-worker with a Totaly amazing motor he developed and NO controller in the world could run it! Then they added external inductors and could make it work. Same thing goes with colossus and a china controller I can ride it around with external inductors and with it just stock even with the controller programed VERY LOW it still blows them in seconds if not less!!

PLEASE stop trying to tell me i have a different problem then I have already found. I do like your out of the box thinking but 8uH is very low. My X5304 has 255uH!!! It can run on any controller!!!!

 

[Lebowski]

And I got the idea of external inductors from a friend who had a co-worker with a Totaly amazing motor he developed and NO controller in the world could run it! Then they added external inductors and could make it work.

I know you're not wanting to hear this but,

I conclude this co-worker also doesn't know what's going on



All in good fun, dude

 

[Arlo1]

And I got the idea of external inductors from a friend who had a co-worker with a Totaly amazing motor he developed and NO controller in the world could run it! Then they added external inductors and could make it work.

I know you're not wanting to hear this but,

I conclude this co-worker also doesn't know what's going on



All in good fun, dude

How about lets play this from another angle... You take a motor with 8uH inductance and simular specs to colossus and run it with 84 volts on lipo (something with little sag) and PROVE ME WRONG! If you are so sure of your self then I know marko and Hall can find you a colossus motor for testing!
As for the people mentioned.
Dude they work for someone who is a pretty big player in the electric revalution, but im not alowed to say what they are working on or with they are both members of the forums and as we are all learning all the time Im quite sure they do know whats going on and most likely more then you will ever understand.... I will let them chime in if they feel it is OK.

 

[rhitee05]

Lebowski, you're confusing and conflating two completely separate issues with motor control. Yes, adding inductance does affect the timing of a motor in that it increases the lag of current relative to voltage. But this is an issue only at mid-high speeds. More inductance also reduces the di/dt in the windings and thus reduces the ripple current. This is mostly only an issue at low speeds. These are very different problems, manifest themselves in completely different ways, and have completely different solutions.

Arlo needs more inductance for his motor because the di/dt is too high and it pops FETs starting from standstill. I'm sure he would love to get the motor going fast enough to have a timing problem...

 

[Lebowski]

How about lets play this from another angle... You take a motor with 8uH inductance and simular specs to colossus and run it with 84 volts on lipo (something with little sag) and PROVE ME WRONG!

I would love to actually But I don't have this motor, don't have 84V and don't have a mechanical load to run it under power... ?
Maybe someone in the Zurich area has a setup like this ? Maybe someone who knows how to handle a soldering iron, has
the setup (including power stage), a scope and multimeter and can connect a (pre-programmed) 30F4011 to the output
stage (and preferably has Ubuntu) ?

If only I had access to the ABB lab down the road...

 

[Lebowski]

Lebowski, you're confusing and conflating two completely separate issues with motor control. Yes, adding inductance does affect the timing of a motor in that it increases the lag of current relative to voltage. But this is an issue only at mid-high speeds. More inductance also reduces the di/dt in the windings and thus reduces the ripple current. This is mostly only an issue at low speeds. These are very different problems, manifest themselves in completely different ways, and have completely different solutions.

Arlo needs more inductance for his motor because the di/dt is too high and it pops FETs starting from standstill. I'm sure he would love to get the motor going fast enough to have a timing problem...

As far as this point goes, I'm assuming the controller Arlo is using has some sort of setup like the 33033. This is a hall-sensor
type motor controller IC that can make PWM output for motor speed/torgue control. These type of controllers
work by turning on one half of the bridge full-time while adding PWM to the other half. This is not a very good way
to go about things as you cannot make a decent short PWM pulse. I don't know exact numbers but would be surprised
if a setup like that can make pulses shorter than a few 100 ns.

I use sine-wave outputs on all 3 motor terminals (also with PWM), in my setup all 3 motor phases are powered all
of the time (instead of only 2 out of 3). For zero signal all 3 motor phases get a 50% PWM dutycycle so all 3 motor
terminals are alternating between V_bat and ground. No effective voltage so no current. Nice long 50% dutycycle
pulses, 25 usec on, 25 usec off for a 20 kHz PWM rate.

The PWM generator in the 30F4011 is clocked at a 30 MHz rate. So in the 20 kHz / 50usec period time I have a
'grid' or 'resolution' of 33 nsec. A small signal for the motor is made by turning one of the motor phase on for
25.033 usec and off for 24.967 usec while keeping the others at 25 usec / 25 usec. The (slow) output stage can easely
do this 'cause it's still making nice long pulses, one is just a teeny tiny bit longer than the other 2. The motor
however only sees the difference between the 3 motor phases, it's effectively connected to the battery for only 33 ns.

Using delta_I = delta_t * V / L with 33ns, 84 V and 8uH gives a delta_I of 0.35 Amps.

So the current increases by 0.35 Amps 20000 times a second. In the mean time the 30F4011 is measuring these currents
up to a rate of 80000 times a second, ready to immediately cut off the motor when something gets out of line.

I think it should work just fine

 

[rhitee05]

As far as this point goes...

And this is relevant how?

The controller Arlo has, like most, does not measure phase current directly. It estimates from the battery current, and the measurement is filtered so that it's really an average measurement that doesn't capture ripple. So, Arlo needs to make sure that there is enough inductance in the motor (or add using external inductors) to keep the ripple current at a magnitude the controller can survive.

Do you actually have something to add to the conversation, or are you just dropping in to tell Arlo that he's doing everything wrong (he's not) and your controller is soooo much better?

 

[Lebowski]

Do you actually have something to add to the conversation, or are you just dropping in to tell Arlo that he's doing everything wrong (he's not) and your controller is soooo much better?

No, I'm trying to tell Arlo that there is a different way of inserting PWM into
the motor signals. He doesn't believe running the motor is possible without
external inductors so I'm bringing arguments to the table which show that it
actually is possible. If you read my last post and think about it you'll realise my
arguments have nothing to do with the algorithm but only with using the
6 channel PWM in a different way from the standard china controller.

Man you guys are so serious. Am I the only one who sees this as just a fun hobby ?

 

[circuit]

Lebowski,

No matter how good your SVPWM is, inductance is still needed to integrate PWM pulses. Take 16kHz 50V 50%-duty PWM and feed it to an inductor, and connect it's other end to 25Vdc. Lets say inductor is 10µH.

So you will have a maximum voltage difference of 25V and 31,25µs duty. Lets say winding's resistivity is 0 (usually the lower the better).
Current of and inductor is: I=V*t/L = 25*31,25µ/10µ=78,125 A.

So, current will ramp up to ±78A with no load at all (at max RPM, lets say motor itself draws 0W). These 78 Amps are over the limit for most TO220 transistors, and thats only to spin the wheel. This explains why lots of controllers blow up with those big RC motors even with no load.
Lets say there are 6 mosfets in the system and each one has R_ds_on of 10mOhm. This gives us total resistance of 15 mOhm (I'm familiar with SVPWM, please don't lecture me).
So, if peak current value is 78A, total power dissipation (on 'ideal' transistors, excluding switching losses and assume that deadzone is 0) in transistors is only around 61W. Within limits you might think, but peak current is ±78A and this blows them up.

The ONLY solution for this is decreasing this current. You could do this either by increasing inductivity or frequency. And you can't increase freaquency due to losses in motor's core.
I've came to conclusion that the only way to solve this problem is to design a custom controller running at like 200-300kHz and including three-phase LC filter.

 

[rhitee05]

No matter how good your SVPWM is, inductance is still needed to integrate PWM pulses.

What he said.