[Konstantin]

Hello all.
I need to understand the chopper oscilator frequency selection.
What one must take in account to set the frequency?

Konstantin.

[ToyMaker]

I think the biggest issues are cost and size. If your PWM power supply operates at 60 Hz it will be as big as a tower computer case. Going up in frequency reduces component size so that at ~100 kHz you have a box the size of a computer power supply. At MHz frequencies (and smaller sizes) component cost becomes an issue. Very small, high power parts cost big bucks.

robotic regards,

Tom
= = = = =
"People need a reason to change, but some people need to change before they can reason."
- - Tom Oliver

[ESjaavik]

Chopper frequency of what? ToyMaker is right when it regards a PSU because it contains magnetics. When not like in most hobby grade stepper drives, frequency doesn't affect size much. But you cannot hear the audible noise when you pass beyond 20KHz. So that's where most are placed. When you get higher up, the switching losses becomes dominant. Which as mentioned can be counterweighed by more expensive components. So 20K or somewhat above is a good compromise for a stepper drive.

[Mariss Freimanis]

Interesting question because the choice is a comprimise between conflicting goals and effects.

Increasing freq Pro:

1) Increasing the switching frequency decreases motor winding ripple current. Ripple current causes eddy-current and hysterisis losses in the motor laminations, causing them to get hot. The frequency is high enough once peak-to-peak ripple currents are less than 10% of the total current.

2) Increasing the switching frequency past 15 to 20 kHz makes the switching frequency inaudible. Windings and laminations vibrate at the switching frequency. If the frequency is in the audible range, the motor will squeal or whistle. This can be incredibly annoying.

Increasing freq Con:

1) MOSFETs and bipolar transistors in particular cannot switch instantly from on to off. They spend some time switching from one state to the other. During this time there is both voltage across and current thru the transistor, causing a large pulse of power to be dissipated as heat, called switching loss. The more switching loss pulses per second, the hotter the transistor.

2) Motors and cable have significant capacitance. This capacitance needs significant current pulses to charge and discharge. These current pulses heat the transistors they pass thru. The more pulses per second, the hotter the transistors.

3) Rapidly changing voltages and currents launch electromagnetic radiation which can interfere with other equipment. Increasing switching frequencies increase electromagnetic interference (EMI).

4) Motors and cable are simultaneously capacitive (C) and inductive (L). Circuits having both L and C are resonant, meaning they "ring" at a resonant frequency determined by L and C. Resonant voltages and currents can be destructively large. Increasing switching frequencies "pumps" the resonant circuit (motor and cable) more frequently. A swing pushed on every cycle will swing much higher than one pushed say, every tenth cycle.

5) Practical motor circuits require sensing either load voltage or current. This cannot be accurately done while the motor is resonating. Normally a period of time is allowed to pass before sensing the motor in order to allow the resonant "ringing" to die out. This elapsed time is called "blanking" time. The higher the switching frequency, the less time there is to control the motor based on the sensed input.

6) Higher switching frequencies require faster switching times causing greater di/dt and dv/dt rates which places burdens on printed circuit layout design. Often inexpensive double-sided PCBs won't do; 4-layer boards may be required to keep circuit trace inductances and circulating current paths in check.

The "Con" list is longer than the "Pro" list as you can see. It is a losing proposition to increase switching frequency past what satisfies Pro 1 and 2.

Mariss

[ESjaavik]

@Mariss:
Your posting should be copied to the articles section! That was a very well founded writeup!

I would add to your Con-1 that the switching have to include a delay to avoid shooot-through. And this is a non-productive interval that will take up an increasing (with frequency) part of the time. Assuming the same power devices of course. Move to faster devices to lower this interval, and the cost increases significantly.

Your argument Pro-1 actually didn't occur to me before you mentioned it here. But it does make a lot of sense. I had trouble many years ago with US equipment (60Hz) not having enough iron core to cope with our (50Hz) mains. So even if the winding number and ratio was right, it ran hot at our lower frequency. This is the same effect I suppose?

Sheesh! I need to read through your post again tomorrow. That was the most knowledge expressed in a few words I've seen in a long time.

[Mariss Freimanis]

[QUOTE=ESjaavik]@Mariss:

(60Hz) not having enough iron core to cope with our (50Hz) mains.

Thanks for the kind words. Though frequency related, the cause is quite different because it is iron saturation that is the culprit here.

Uloaded transformers have very high primary inductance (>> Henry) and are designed to operate near magnetic saturation at their rated primary voltage and frequency. The minimum amount of "iron" (size and weight of the core steel laminations) is used for reasons of economy and size.

Magnetization currents are inversely proportional to Hz and proportional to primary voltage. At 50Hz the primary current is 120% of the 60Hz value. This can be enough to saturate marginal transformers.

When saturated, a transformer's inductance drops dramatically and magnetizing currents rise equally dramatically (Ohm's Law). Transformer heating is entirely due to the I^2 R losses due to this "out of sight" primary current.

Good transformers are rated for "50Hz/60Hz" operation. This means their "iron" weight is 20% more than it has to be at 60Hz. This is good insurance for those of us lucky enough to have 60Hz mains. They will tolerate 20% higher primary voltages as well without burning up.

Mariss