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

[jdb]

To back out the inductance from the scope pictures, you need to know the BEMF provided by the motor at whatever speed it was running as well as the battery voltage. That's somewhat inconvenient at any speed other than zero.

- Can you run the test with a locked rotor?
- Please report the battery voltage being used for the test. To be easier on the controller for the test, use the barest minimum battery voltage that your controller will permit. If it will run on 12V, then feel free to use 12V. This will reduce both the ramp rate, and the peak current, since both are proportional to the input voltage. It will be harder on the controller, but much safer.
- The current signal is barely showing up on the scope. And it will be even harder with the lower voltage supply. Maybe increase the gain by 10x by lengthening the wire? A higher shunt resistance will affect the measurement at higher DC current levels, but it shouldn't be too bad below 50m or so... maybe.
- Where is the shunt connected, exactly?

V = L*dI/dt

When resistance is neglected for a short span at low current,
L = V*delta_t / delta_I

over the rising half of the signal.

That was a start voltage of 82 with 20s lipo it may have sagged a bit but not to much because it is a 10 ah 20-30c and basically new pack.

Its not the effect of the battery sag, which should be pretty much negligible at no-load. Its due to the BEMF. The V in that equation is V_battery - V_bemf - V_resistance, expanding to V_battery - Kv*rpm - I_phase*R_phase. A no-load, but nonzero-speed, Kv*rpm will be large enough that you can't say with much reliability what the voltage across the motor's inductance is.

 

[jdb]

Here is some scope pics it looks to me like the build time is 12uS before the big spikes the total on time was 25uS if anyone has an imput as to what I am seing and if I am wrong about the build time being 12uS pleas let me know. IM in new teratory!

Here's my best guess:
Each cycle is a prompt rise over < 1us, followed by a slow ramp that includes several other big spikes over 25us, followed by another prompt fall and a slow ramp down over 39us. I believe that the prompt jumps are because the shunt is connected to the phase that is "high" during that measurement, and the scope is not rejecting the 82V common-mode noise very well (80% confidence guess). The spikes are not real current spikes, they are voltage spikes caused by transmission-line-like ringing down the phase wires over the cycle (30% confidence guess).

The motor V_bemf can be backed out by assuming that the total current across the motor is exactly zero. Its not quite true because there is some no-load current, but the I*V loss for the motor is so small it can probably be neglected. With that in mind, the BEMF can be backed out from the duty cycle for a continuous-mode buck converter.

On-time (total ramp including the "spikes" in the picture): 25us.
Off-time (nice slow smooth ramp down): 39us.
Total PWM period: 64us
PWM frequency: 15.6 kHz <-- Please double-check the controller's documentation since this can be used as a check on the work

The duty cycle for a continuous-mode buck converter is
D = Vout/Vin ~= V_bemf / V_battery = T_on / T_total = 39%
V_bemf = D*V_battery = 32V
Over the off-cycle, L = V_bemf*delta_t / delta_I = 32V * 39u / 11 = 113 uH

That seems a little high for the motor alone, considering that the stock turnigy 80-100 is known to be 40uH (according to bigmoose's datasheet). Did you have the homemade inductors attached to the motor at all?

 

[rhitee05]

Measuring inductance via the voltage drop is a really painful and unnecessarily difficult way to do it. Since you already know the resistance, calculating the inductance from the time constant is much, much easier. Probably more accurate, too.

It's very easy now that you have your current shunt. It doesn't matter whether you do this at full throttle, partial throttle, full speed, standstill, whatever. You just need to make two measurements of the current. You can do this on either the rising or falling edges, but I'll assume you use the rising edge (works equally well, but the equation is slightly different). The first measurement should be near the beginning of the pulse, as soon as you can get a clean measurement. The second needs to be later in the pulse, but before the current levels out - when the current reaches about halfway between the starting and final levels is a good place. You need to measure the current at both times and the delta-T between them. The time constant is then: tau = - delta-T / ln(1 - I1/I2), where I1 is the first measurement and I2 is the second. Once you know tau, L = tau * R.

You can double-check the value you get for tau pretty easily. The current should be within about 5% of the final value 3 tau after the start of the pulse. For L, based on what you've said so far I would expect a value less than 100 nH, probably significantly less.

 

[jdb]

Measuring inductance via the voltage drop is a really painful and unnecessarily difficult way to do it. Since you already know the resistance, calculating the inductance from the time constant is much, much easier. Probably more accurate, too.

*snip*

You can double-check the value you get for tau pretty easily. The current should be within about 5% of the final value 3 tau after the start of the pulse. For L, based on what you've said so far I would expect a value less than 100 nH, probably significantly less.

His last post contained shunt voltage measurements, which should be equivalent to current measurements. In principle I would agree with you about using the time constant, except that I think the pulse lengths are too short to get a decent time constant measurement in this range. One tau is going to be in the neighborhood of several hundred us to a few ms for this motor, and his working controller has a maximum pulse width of only 65us. Besides, 95% of the steady-state current at 82V and 10mOhm is going to vaporize his fets with 7800A of current And to do this with a single pulse means a good trigger setup on the 'scope, too.

100nH? Surely you meant 100uH?

 

[Arlo1]

Did you have the homemade inductors attached to the motor at all?

Uhmmmm ... LOL there is 50uH of inductors conected in series with each phase lead! 100uh extra for each phase!

 

[rhitee05]

His last post contained shunt voltage measurements, which should be equivalent to current measurements.

Yes, this should be a sufficient current measurement so long as his shunt is relatively non-inductive. In any case I'm sure it's the best we're going to get.

In principle I would agree with you about using the time constant, except that I think the pulse lengths are too short to get a decent time constant measurement in this range. One tau is going to be in the neighborhood of several hundred us to a few ms for this motor, and his working controller has a maximum pulse width of only 65us.

No, I expect that the time constant will be significantly shorter than the 65 us PWM period. If it was in the range of a few milliseconds, he wouldn't be having problems in the first place! In one of Arlo's earlier posts he spoke about an 8 us rise time for the motor alone. I'm assuming that the time constant will be on this order of magnitude, probably a little less than that. He said his scope has 2 MHz bandwidth, so I think he should be able to get a good measurement if the time constant is on the order of 5-10 us.

100nH? Surely you meant 100uH?

No, I meant 100 nH. I'm fairly convinced based on the data so far that this motor has a very, very small inductance. Again, if it was 100 uH, he wouldn't be having a problem to start with. I posted earlier if the time constant ended up being 8 us that would put the inductance at 75 nH. If the time constant is less than that, as I suspect, the inductance will be even lower.

 

[Arlo1]

In one of Arlo's earlier posts he spoke about an 8 us rise time for the motor alone. I'm assuming that the time constant will be on this order of magnitude, probably a little less than that. He said his scope has 2 MHz bandwidth, so I think he should be able to get a good measurement if the time constant is on the order of 5-10 us.

That was a mesuremen of the drop across the 15uH inductor and only with 1 15uH inductor per phase wire.

 

[jdb]

100nH? Surely you meant 100uH?

No, I meant 100 nH. I'm fairly convinced based on the data so far that this motor has a very, very small inductance. Again, if it was 100 uH, he wouldn't be having a problem to start with. I posted earlier if the time constant ended up being 8 us that would put the inductance at 75 nH. If the time constant is less than that, as I suspect, the inductance will be even lower.

I am spock's incredulously raised eyebrow.

http://www.technick.net/public/code/cp_dpage.php?aiocp_dp=util_inductance_wire_2

2 wires of 1.5mm in diameter^H^H^H edit: radius, 3.5mm apart, have an inductance of > 200nH per meter. His lead wires alone are going to be greater than 100nH.

 

[jdb]

Did you have the homemade inductors attached to the motor at all?

Uhmmmm ... LOL there is 50uH of inductors conected in series with each phase lead! 100uh extra for each inductor!

OK, try again with much less external inductance. There are enough ifs, ands, provisos, and approximations that the ESL of the external inductance totally swamps out the motor's inductance measurement with the last run. +/- 10% would be _extremely_ generous just for the measurement accuracy. Add in another 10-20% or more for the external inductors' construction and you don't really have a good measurement quite yet.

Go with zero external inductance and minimum battery voltage; the barest minimum that the controller will support, and try again. Edit: If the controller will run on 5s lipo, then just use 5s. Other than that, the methodology is sound.

Edit: Also, forget about the locked-rotor part earlier. Depending on the controller, that could make the current discontinuous, which would add noise and could make it harder to make the measurement.

 

[rhitee05]

I am spock's incredulously raised eyebrow.

You're right, that would be an absurd number in retrospect. I didn't really do a sanity check on those numbers. Probably a handful of uH is the minimum reasonable value, perhaps. Your point about the lead wires is correct and well taken. Given the extremely low winding resistance, the resistance of the wires needs to be taken into account too. At this point I'm confused enough about the measurements that Arlo is posting that I'm not going to speculate further unless I'm clear what's being measured.

 

[Arlo1]

Ok guys I used a aligator clip jumper wheel to hook the shunt into the system. so the resistance is hi becasuse of the thin wires conected to it.... So tell he how you want me to set up my next test?

 

[jdb]

I think that at no-load the resistance of the leads and phase wires isn't a major source of error, since we are probably observing pulses that are much shorter than the system time constant. If those little spikes on the rising ramp of the current are caused by ringing, then soldering the shunt into place may help a little bit, but I'm really not sure. Alligator clips will be OK for now.

 

[Arlo1]

OK so what I trying to determin is if the 50uH is enough or not. So I mesured the rise time of the inductor and the amout of current remember 1mv = 1 amp with this shunt!

 

[Arlo1]

So I played with motor c in the spread sheet and found at 170uH the amps will build at about the speed we see in my scope so is it safe to say you can subtract the 100uH I added to the 2 phase wires to get a est of 70uH in the motor it self? It will be interesting to see once the meters show up in the mail.

Edit I just realized the test was at 80 volts and not 96 so I changed the spreadsheet and got an approx of 170 total.

 

[Arlo1]

Motor A is a edjucated guess at collossus in stock form. Motor B is likely what I want to achive Motor C is what I just tested (based on the rise time of the amps)

 

[jdb]

The cursors are not on the useful part of the graph. My earlier text description may not have been clear, so here is an annotated scope picture:



The inductor's rise time is captured by the path from point 2 to point 3 in the picture (or from 1 to 4). The fall time is captured from point 4 to point 5 in the picture. The hard jumps from points 1 to 2 and from 3 to 4 are not real changes in current, they are an artifact of a low Common-Mode Rejection Ratio (CMRR) in the 'scope and current sensor design.

First, before taking any measurement at all, you need a capture whose point 1 is equal to point 5, because that means there was no net change in motor current over that PWM pulse.

Second, the rise time by itself is useless. You have to include the effects of the BEMF, winding resistance, and battery voltage. The BEMF can be estimated when the current is constant over the pulse width using the battery voltage and the duty cycle (point 1 == point 5). The winding resistance can be neglected if the motor is at a no-load condition. No-load preferably means not even the drive chain attached.

 

[Arlo1]

First, before taking any measurement at all, you need a capture whose point 1 is equal to point 5, because that means there was no net change in motor current over that PWM pulse.
.

Thanks but how do I do that?

 

[jdb]

You managed it in this post. It's probably luck-of-the-draw to get one near the middle of the commutation cycle. Constant throttle conditions should help.

 

[rhitee05]

What he means is that for his measurement to work, the motor must be operating in steady-state at some speed and constant throttle (doesn't need to be 100%, you'd probably want it much lower than that). If the motor is running at full no-load speed, then the BEMF will be approximately equal to the applied voltage and the average current will be constant over time. Under those conditions you can make the simplified calculation he's describing. The calculation is much more difficult (basically impossible) if the motor is not in steady-state.

I'm a little confused by some of what I'm seeing on those 'scope traces - particularly the spikes which start halfway between points 2 and 3. I don't know how to explain them, other than I don't think they should be there. It would be helpful if you had a 2-channel 'scope and you could show both the current and phase voltage (phase-to-ground, not phase-to-phase) simultaneously. The spike and small ripples at point 2 look like ringing, which I would expect to see, between the inductance and inter-winding capacitance. The later spikes look too consistent and too large to be ringing.

 

[jdb]

Ack: no-load means different things to different people. Ideal conditions would be nothing attached to the rotor, but at 50% of the maximum unloaded speed. Half-throttle for a voltage-source inverter. Barely cracked open throttle for a current-source inverter. The ripple current over a PWM cycle for a buck converter is at its maximum at 50% duty cycle, which makes it easier to observe on this scope. The ripple current is at its minimum at 0% and 100% duty cycle, corresponding to nearly at-rest, and 100% of the unloaded speed for the motor. The speeds Arlo has been using so far are suitable enough.

rhite05, I think you're right about the secondary spikes. I still don't think they are changes in current, and they do look too late to be ringing from the initial on-ward pulse. I'm kindof hand-wavingly hoping they don't matter, because I don't know what to do about them if they do.

 

[rhitee05]

I accidentally used the wrong terminology, I think. When I said "full no-load speed" what I meant was steady-state speed for that particular throttle setting. Not necessarily maximum unloaded speed (corresponding to 100% throttle).

The only explanation I could think of for those spikes would be the current limit circuit cutting in and out. Except, I thought that the current limit for the Infineon designs wasn't nearly that fast. I can't be sure what the period of those is, but I'm assuming on the order of 5-10 us, and I didn't think these designs were capable of limiting on a pulse-by-pulse basis so this probably isn't the right explanation. I wouldn't expect to see any measurement artifacts there, and the down-slope trace of current is nice and smooth. I also wouldn't expect any transmission line-type ringing, but even if there were that should also present a decaying ripple and the spikes I see are of approximately the same amplitude.

The spikes seems to start about halfway into the PWM pulse in the two pulses shown here. That's probably a clue, but I don't really have detailed enough knowledge of the system to tell what it's pointing towards. I'm inclined to think that we're seeing a real effect and not a measurement artifact, just not sure what kind of effect. Seeing the voltage and current traces together would help, then we would know what the switches were doing when those occur. If the switches stay on and the voltage is constant, that would disprove the current limit theory, at least.

 

[Arlo1]

Good news guys. I don't think it will need as much inductance as I originally though. I got a set of 27 uH hybrid air/iron inductors in the mail today from a very generous forum member who sent them to me for the cost of shipping. The most I can lay down is 2.6 hp with them but until I get better measurements and my inductance meters this is where I will stop. The bike is fun but will need at least 6 hp to do what I want. And I would prefer 20 + but for now I will do as big moose says and make small steps. [youtube]z6f11otmdv8[/youtube]

 

[bigmoose]

Arlo CONGRATULATIONS!!! Even two wheelies that I saw. A journey is made of many small steps, I think you are well on your way. Sometimes 60% performance will achieve 95% of our objectives. You can then sneak up on the last 5% at your leisure.

Good job! Best wishes on your continued developments.

 

[Arlo1]

Thanks big moose. This bike has a lot of potential! Im very excited. It handles awesome and just needs an upgraded rear shock and a whole shit pile of lipo!
Today is a great day!

 

[Arlo1]

I just took 2 turns off the inductors and laid down 3.6 hp on the dyno as well I was up to 77.76 ft/lb at the rear wheel. I tried 2 more turns with no more gains but that could be the voltage droping because my charger is painfully slow!