What is Space Vector Modulation? (Episode 10)

Ever wondered "what is space vector modulation"? Well, this week we look at optimizing our voltage modulation method via space vector modulation (AKA SVM), as well as brushing up on some basic control ideas. Like, subscribe, share, tell Carol what you really think about her baby (reference to the video), but only if you feel like it.
Note on this video, I'm only presenting the alternate reverse SVM pattern. I feel like its the most broadly applicable, but if you have special cases, it may be worth looking into different SVM wave forms. I also feel like its easiest to show the parallels between it and sinusoidal modulation for understandings sake. Anyway, have a good one.
footnote:
You don't necessarily need an external hi res sensor for this, it is possible to run this system looking at back EMF for location, or running basically an observer that uses velocity estimations and hall effects.

[@kuzain73]

Just discovered this channel, love the videos!

[@adnaneakk1068]

That moment when you are speachless to say just thanks ... wow man no words can describe how much grateful iam

[@ulquiorra3504]

This is absolutely golden. Thanks

[@tzushonchen4195]

Thanks for explaining the space vector modulation and how is transformed from the sine modulation!

[@daantruijen1170]

Unbelievable video's, keep it up!

[@mauriceupp9381]

I can tell you I understand part of this and I was going to give up because it wasn't making complete sense but it is so damn interesting I can't get rid of the damn videos I can't go somewhere else

[@salehmohamed1981]

Is it possible to control 20 BLDC drivers with variable speed as well as anti-reverse with two push-buttons

[@damyandimov5920]

Hello. First of all great videos, watched them all and it helped me tremendously with my project. Second I want to ask what program did you use to draw and animate the functions of the vectors, the sin waves etc.

[@niklasleh9409]

amazing content! Haven't seen SVPWM as an addition with a triangle wave before. Maybe you could also reference third harmonic injection for another video?

[@sudheerkumar5966]

Very good sir

[@tahabandrawala]

ty

[@michaelbraverman2001]

Hi, should either one of the switches of each half bridge be active at all times? If so, how again each pwm percent value is interpreted? -70% for instance, does that mean that 70% of the time the lower switch is active and the rest 30% the upper switch is active?

[@jtlee1108]

Great question! Your interpretation is exactly right. I know that notation can be confusing, as we tend to think of individual MOSFETs being sent a given PWM signal, and this is more of a PWM signal to a leg of the bridge. So, as you said, its more of a percent of the time connected to high vs ground; as opposed to (in the typical way we think about it) the percent time that a single MOSFET gate is written high vs low.

[@michaelbraverman2001]

@@jtlee1108 Thanks! I still don't understand though... I scale down the desired voltage on a phase to the range (-1...1). Lets say I've got -0.7V, should I generate the same duty cycle as if I got +0.3V? What happens with 0?

[@jtlee1108]

@@michaelbraverman2001 another great question! So I think I understand the confusion now, and I think its due to my presentation. Yes, within my perhaps confusing notation, -70% duty cycle and 30% duty cycle are equivalent. However, its important to remember that 0 V is +-50% duty cycle. So, instead of thinking of 0.3 as 30% up from 0, we have to think of it as 65% up from -1 (since -1 is our 'ground' in this case). Likewise, -0.7 would be 0.3, or 15% up from -1 (so it would be 15%, or -85% duty cycle). I did not do a great job of explaining that clearly in the video, so I apologize. I hope this clears it up some.

[@Timsturbs]

why do we apply triangle wave? because its simpler than sine? it should be 1/3 in amplitude? its basically a third harmonic?

[@thorlancaster5641]

Even if you just have hall sensors for determining angle you can interpolate rotor position at all but the lowest speeds. Actual position = (start of sector angle) + (time since last hall change / how often hall reading is changing) * 60 degrees.

[@jtlee1108]

Absolutely! it won't work quite as well, especially not for (as you said) low speeds, or even stall torques. Some people will also use the value of the BEMF to estimate speed as well (which you could, potentially use in an observer, with the velocity via halls you were talking about) and potentially make it a little bit more accurate.

[@thorlancaster5641]

At low and near-stall speeds, I think I'll just go with 6-step block commutation. Given the amount of inertia in a large hub motor (what I'm building a controller for), the angle interpolation will probably work quite well at low speeds, down to about 2ish MPH. Edit: Did try this, worked well. Below a certain speed I had to fall back to 6-step but 30% torque ripple isn't bad at all, especially when it's only a problem for half a second at startup.

[@stefanvanmorgen4106]

So when you use above SV Modulation, you effectively control the voltage in each leg. What about the current? Does the current follow the voltage meaning it has the same non-sinus shape as the voltage? Or is the current still sinusoidal?

[@jtlee1108]

good question! So the short answer is that the Current will be sinusoidal. This answer assumes you have a constant command (IE your duty cycle command, as discussed in the next episode is fixed), not doing any sort of servo control. This is because the current is forced by voltage differential, not voltage. And since the harmonics are added into all three voltage signals, the differential is still the same as it would be with a sinusoidal voltage. Hope that helps!

[@stefanvanmorgen4106]

@@jtlee1108 Thank you! I was hoping for some crazy current pattern which might help reducing switching losses especially in the low-current regime, but well, I guess there is no way around those ;)

[@FREZADTYUI]

Amazing video ! You helped me understand a lot, especially why we don't use the full differential voltage. However, I can't find mathematically why or how to find this 86.6%. Do you have any materials I could learn from? THanks

[@Concretizer]

I think that number comes from the "Range Used" graph shown at 7:50. I would define it as max(a(t), b(t), c(t)) - min(a(t), b(t), c(t)), where a(t), b(t) and c(t) are the 3 sinusoidal phases with amplitude 1. The maximum value of the graph is approximately 1.732, while the maximum that we want to achieve is 2. Then, 1.732/2 = 0.866

[@denisleo9069]

Hi, Your videos are awesome ! Just 47k for this. it largely merits to have more followers than Beyonce ! Also, do you have some books to recommand us ? Especially about motors dynamic/transfer function controls ?

[@boonsfrc]

What if we made a ship that ran of electromagnetic power then we could harness the electric Soundwave and then travel on the sound , I guess since light is traveling with the electric Soundwave 🤔🤷maybe we would need to build a ship out of crystal at least the engine and pistons out of crystal, I feel the whole ship should be made out of crystal this could handle space and traveling at the speed of light