Why is this a bogus three phase BLDC drive circuit?

[BradtheRad]

I once overheard the following snippet of conversation between the Senior Engineer (SE) and the SMDE

SE: “Why did you do it like that?”
SMDE: “Because it means that the current more quickly builds up in the motor coils after they are switched in…engineering sense”

Speculation on my part...
He may have intended for constant current regulation to act during the time a coil is switched on. The coil would receive higher voltage immediately, in order to attract a distant magnet as strongly as possible.

Then while current builds in the coil, the magnet gets closer, and voltage is reduced by the current regulator.

A smps, operating at several kHz, would easily be able to do all this while a coil was energized, say 1/400 or 1/800 of a second.

He may have believed this would increase torque, as well as overall efficiency.

I'm not knowledgeable enough to say whether this would or wouldn't work.

For a fleeting instant , I suspected that he was deliberately trying to lead the project astray, perhaps due to his grudge against the company…..he used to openly declare that he had a grudge against the company….he was often heard shouting at managers who asked him to write proving reports for some of his designs.

Again I'm speculating... He may have wanted to keep some things a secret, in order to protect his ideas.

It is common for a 'star performer' to create a successful project, and then watch superiors take credit for it. He may even have an innovative design which is being developed, which is stolen by a superior or a colleague.

This may be their purpose of asking him for proving reports for his design, so his idea can be someone else's.

One cannot be blamed for becoming disgruntled. As a consequence, he has to do some inexplicable things, such as leave out crucial details when explaining his methods. Etc.

 

[mtwieg]

..sorry but I disagree, -with the "normal" method of BLDC drive control, the coils are high frequency PWM'd within the commutations to control the coil current, and this effectively illiminates the inductance of the coil.....just like in a current mode full bridge converter, ...the fact that its a current mode full bridge means you do not get an LC resonance between the output inductor and the output capacitor of the full bridge....because the inductor current is controlled...it is not allowed to resonate with the full bridge output cap............it is the same principle with the "normal" method of control of bldc drives, you do not get a resonance between the dc link cap and the coil precisely because the coil current is being peak current controlled by high frequency pwm.

As I said before, regardless of which circuit you use the parallel LC is still there. Using an inner control loop like CMC can allow you to pretend the resonance is gone with respect to the outer control loop, but it is, in fact, still there. Also your own schematic shows the current sense after the DC link capacitor, so if that branch is really current regulated then it will effectively eliminate that capacitor as seen by the inverter, and the resonance will be "gone."

I note that your call for a "citation" above has not been heeded, as no doubt I believe you expected.....one thing we can say for sure, is that the "bogus" method of bldc control described in the top post, certainly has no advantages whatsoever over the "normal" method of bldc drive control............so at best its a waste of time....and likely is totally bogus.

I'm also skeptical of whether it's worthwhile. At most, having the current biased inverter might mean you don't need individual current sensors on each leg. Also I think that it would only work well if you used FOC rather than trapezoidal control (so that current draw is truly smooth DC).

- - - Updated - - -

Speculation on my part...
He may have intended for constant current regulation to act during the time a coil is switched on. The coil would receive higher voltage immediately, in order to attract a distant magnet as strongly as possible.

Then while current builds in the coil, the magnet gets closer, and voltage is reduced by the current regulator.

A smps, operating at several kHz, would easily be able to do all this while a coil was energized, say 1/400 or 1/800 of a second.

He may have believed this would increase torque, as well as overall efficiency.

Right so if you use trapezoidal drive you might get sharper rise and fall on the current in each leg. I don't see how that is desirable, though. If anything it should make torque ripple worse.

 

[treez]

Also your own schematic shows the current sense after the DC link capacitor, so if that branch is really current regulated then it will effectively eliminate that capacitor as seen by the inverter, and the resonance will be "gone."

I don't think the capacitor is eliminated, ..it would be if the buckboost had infinite feedback loop bandwidth, but in reality, the buckboost has a bandwidth of a few kHz, and the capacitor is very much there...the buckboost bandwidth is not high enough to cancel out the LC circuit.

 

[mtwieg]

I don't think the capacitor is eliminated, ..it would be if the buckboost had infinite feedback loop bandwidth, but in reality, the buckboost has a bandwidth of a few kHz, and the capacitor is very much there...the buckboost bandwidth is not high enough to cancel out the LC circuit.

So you've made two contradicting narratives. First you said you can't drive a parallel LC circuit properly with a current source (which is objectionable on its own). Now you say that the bandwidth of the current control is not high enough to affect the output impedance of the buckboost at frequencies near that resonance. If the second statement is true, then the first statement is irrelevant (since, as the second statement implies, the buckboost is not a current source at the resonant frequency).

Also didn't you say your DC capacitor was 300uF and the motor was 56uH? That gives a resonance at 1.23kHz. If your buckboost is operating at 150kHz then getting a bandwidth higher than that should not be a problem.

For the sake of clarity we should probably consider the issue of the current biased inverter separately from the design of the actual current bias. So for arguments regarding the inverter we presume we have a controllable current source with high bandwidth, and worry about how to actually implement that later.

 

[treez]

For the sake of clarity we should probably consider the issue of the current biased inverter separately from the design of the actual current bias. So for arguments regarding the inverter we presume we have a controllable current source with high bandwidth, and worry about how to actually implement that later.

...Thankyou, yes, you are correct, there is such thing as a "current source inverter" and this is entirely genuine. It involves an upstream inductor controlled basically in a kind of "hysteretic mode" so that its peak and troughs are defined, and thus its average level is defined (half way between the peak and trough).......

however, as discussed in the post 1, this is not the situation of this bogus drive methodology.

We have already agreed that the method of post 1 has no supporting documentation literally anywhere in the world, from any book or website whatsoever, -you yourself have also suggested that at best, its certainly no better than other more "normal" drive methods. FvM has delivered comment to indicate that FvM has considerable doubts about the method of post 1.

Post 1 , as you know, is not a "current source inverter", and I am declaring it bogus.

Also didn't you say your DC capacitor was 300uF and the motor was 56uH? That gives a resonance at 1.23kHz. If your buckboost is operating at 150kHz then getting a bandwidth higher than that should not be a problem.

...you are touching on something here that I was investigating some months ago whilst I was actually working at the place, -I was considering whether it would be possible if the buckboost bandwidth was high enough. -At the time, there was no reponse to this. (by the way, it was 200uF and 56uH and thus 1.5kHZ)...We couldn't fit 300uF onto the small PCB.

So you are saying that the method of post 1 is ok if the buckboost bandwith is significantly greater than the LC resonant frequency? (dc link cap and motor coil)

BTW: I am surprised that no-one has brought up the fact that the "bogus" method of post 1 , if it could be workable , has the advantage of less switching losses in the igbt's (since they only commutation switch, and don't high frequency pwm switch the motor coils). Also, in theory, the reduction in high frequency pwm'ing of the motor coils, could (?) in theory mean a reduction in the radio interference chucked out from the motor (?).

 

[FvM]

I think, the thread could be retitled "How to throw the baby out with the bath water".

You have 1. a design problem: Drive a BLDC motor from a 18 to 40 VDC supply that partly requires boost operation. Without boost requirement you would most likely refer to the simpler pwm operation of a basic three phase bridge.

You have 2. the idea of a hardware topology, a buckboost DC/DC and a (fixed duty cycle) BLDC driver. At this point current or voltage control loops aren't yet fixed.

You have 3. an implementation of the hardware topology combined with a specific control concept ("output current regulation") which apparently failed to drive the motor as intended.

At this point a control system expert would analyze the properties of the involved design blocks and overall system behaviour to find out if approriate tuning of the control loops is sufficient to achieve stable operation or if the control loop structure needs a revision.

As an example, the buck-boost converter has a voltage rather than current output by nature (at least in CCM where the duty cycle sets a fixed Vout/Vin ratio). It can operate "current regulated" by means of a control loop, but the transfer function H(s) of this control loop and the resulting converter output impedance are widely variable.

Instead of detail circuits we would look a block diagrams and transfer functions in this design step (or possibly step back to it if the first implementation approach didn't work). Lacking this general view on design problems can certainly be a reason for project failure.

 

[treez]

You have 2. the idea of a hardware topology, a buckboost DC/DC and a (fixed duty cycle) BLDC driver. At this point current or voltage control loops aren't yet fixed.

Thankyou, by the way, I am not sure what you meant by "fixed duty cycle", but the inverter was not giving a fixed duty cycle for the high frequency PWM'ing of the coils, as the IGBT's were not high frequency pwm'ing the coils, -once the igbts had commutated to the next coil, they just stayed on all the time until the next coil was due to be commutated in.

 

[FvM]

I was looking at the BLDC inverter from a control system viewpoint, it's not relevant in this regard if it's using pwm or not, just that it doesn't change the ratio between dc input and motor voltage. I translate no pwm as 100 % duty cycle.

 

[godfreyl]

I know it is incorrect, but I need the official explanation to take to the boss so we can get the project scrapped.

Why?
You mentioned elsewhere in this thread that the project has already been scrapped and that the people who had been working on it were sacked. You also mentioned that you left that company as well.

I don't understand why you want to go back to a company you used to work for and explain to the boss there why he should scrap a project that he's already scrapped. Seems like a waste of time.

 

[chuckey]

An alternative view might be , if the rise time of the current in the motor coils is excessive. Then the mean power in the motor has fallen and during the time the current is still rising in the coils, the motor has a lower power input. The obvious thing to do is to increase the current, until the mean currents back to where it should be.
Just putting some numbers in, suppose the current pulse is 30% rising with 70% full amplitude. If we consider the 30% as triangular (for simplicity), then the current over the whole period would be 30% X 50% + 70% X 100% or 85%, so we increase the current so it comes back to 100%, increasing the Vcc of the commutating switch by 100/85 would give the same mean current.
Looks a bit complex, but it will work if the Vcc is high to start off with and the mark/ space ratio of the commutation is reduced for low motor speeds. So as the motor current decreases, the mark/space ratio is increased to the manufacturers recommended value at the motors maximum speed.
Frank

 

[treez]

I don't understand why you want to go back to a company you used to work for and explain to the boss there why he should scrap a project that he's already scrapped. Seems like a waste of time.

...thanks, I was cutting a long story short, I want to be sure future companies I may (or may not!) work for don't go down the same road to nowhere.