# Easiest way

#### thatoneguy

Joined Feb 19, 2009
6,359
Think of a way to make an inverter ring oscillator, out of six of them, you should find two that are 90 degrees out of phase, which you could then tweak into a sinewave or MSW, run through a transformer and come out with a decent sinewave, after a bit of loss.

#### shortbus

Joined Sep 30, 2009
9,326
can you do a little writeup on the motor rewinding once you're done please?
No problem, I found a couple of great books that have both winding diagrams and charts on wire size and other rewinding properties. The charts are set up as nomographs, so it makes things pretty easy.

#### shortbus

Joined Sep 30, 2009
9,326
Ok, for the non-motor experts: Will this work the same way as for a 3-phase AC motor VFD, i.e. the voltage increases with frequency (Sine wave voltage not PWM) and there is an IR compensation at low frequencies to provide some boost?

Btw, the 90 phase shift at changing frequencies makes the circuit a little bit more complex. If one only could use a uC...
To my knowledge (meager as it is) VFDs don't increase or decrease voltage as the frequency is changed. The incoming AC is rectified and then chopped into a PWM replication of a sine wave. But could be wrong as I am known to be.

Yeah, most of the VFDs do use a uC to generate the switching.

Why is the 90° phase shift harder than 120° ? A bipolar stepper driver IC does it without a uC though. Just cant find one of the older discrete logic schematics yet, still looking though. The stepper circuits do it with the clocking frequency changing.

#### shortbus

Joined Sep 30, 2009
9,326
Think of a way to make an inverter ring oscillator, out of six of them, you should find two that are 90 degrees out of phase, which you could then tweak into a sinewave or MSW, run through a transformer and come out with a decent sinewave, after a bit of loss.
Not quite sure what you mean here? Six of what?

#### thatoneguy

Joined Feb 19, 2009
6,359
Not quite sure what you mean here? Six of what?
6 inverters, each with an equal R C to ground, connected output to input, with the final output going back to the input.

I suppose you could use 4 of them and come up with 90 degrees, you'll need to experiment.

Search for "Inverter ring oscillator" for formulas, the topic was just here last week-ish, so it should be easy to find.

Why no uC solution? The Magic Sinewave PWM algorithm would give you a perfect sinewave on the other side of a transformer.

#### praondevou

Joined Jul 9, 2011
2,942
To my knowledge (meager as it is) VFDs don't increase or decrease voltage as the frequency is changed. The incoming AC is rectified and then chopped into a PWM replication of a sine wave. But could be wrong as I am known to be.
I worked a few years in a company that manufactures VFDs. However it was always for 3phase asynchronous AC motors, so that's the only one I can say something about.
Power supply voltage is rectified to DC then passes through a 3 leg IGBT bridge.
This bridge is controlled by 3 PWM signals. The voltage of the sine wave that is being transformed into a PWM increases with frequency, yes. There is also an increase at very low frequencies, they called it IR compensation.

Read the second post here : http://www.eng-tips.com/viewthread.cfm?qid=143590 Maybe they can explain it better.

Why is the 90° phase shift harder than 120° ? A bipolar stepper driver IC does it without a uC though. Just cant find one of the older discrete logic schematics yet, still looking though.
No it's not harder, it's of equal complexity.

#### strantor

Joined Oct 3, 2010
5,910
To my knowledge (meager as it is) VFDs don't increase or decrease voltage as the frequency is changed. The incoming AC is rectified and then chopped into a PWM replication of a sine wave. But could be wrong as I am known to be.
we, they do, in a roundabout way. the PWM pulses are always full voltage DC, but the motor windings average them out (the current through the windings that is). If you read the voltage phase to phase with a meter that has a low pass filter like a Fluke 289, you will read significantly lower voltages at reduced frequency. If you read any drive manual, about sensorless vector mode, it will talk about the "volts/hertz" lookup table. as the frequency increases, so does the voltage.
Why is the 90° phase shift harder than 120° ? A bipolar stepper driver IC does it without a uC though. Just cant find one of the older discrete logic schematics yet, still looking though. The stepper circuits do it with the clocking frequency changing.

#### praondevou

Joined Jul 9, 2011
2,942
@shortbus

I don't see an easy way so I tried thinking of something less easy.

IMO what you need are:

An oscillator, a phase shifter, a square wave to sine wave former (at different frequencies, that's the hard part), a frequency to voltage converter, a adjustable gain amplifier, a PWM frequency generator, square wave to triangle former, then the circuit to compare triangle with sine wave. Then you have a true sinewave modulated PWM with adjustable frequency and voltage.

I drew an oscillator though any adjustable square wave 50% duty cycle oscillator will do the job. This one is adjustable between 2 Hz and 100Hz.
It is then passed through a FF so it's frequency is divided by two . The XOR at the output gives the 90° shifted signal. (the other XOR is only there for a propagation delay) The VFD output frequency will therefore be between 1 and 50Hz. (Q1 of the FF is one output, the XOR output the other)

The next step is missing, square wave to sinewave.

The two sine waves are then fed into the transconductance amplifiers. Their output voltage is controlled by a voltage (actually current) which has to be proportional to the sine wave frequency, this is where the F/V converter comes in. I imagine it could be a LM331. Never used them though.
The LM13700s output voltage will therefore be two sinewaves (90° shifted) whose voltage depends on oscillator frequency.

The rest is easy, Generate triangle, compare to sine wave, feed the PWM to two H-Bridges then to the motor...

Is this to complicated, i.e. a dead end or do you think it's worth continuing without microcontroller?

Maybe someone else comes up with a simpler non-PIC solution, but IMO the fact that the reference sinewave needs to be adjustable in both F and V makes it more complex.

#### Attachments

• 44.5 KB Views: 148

#### shortbus

Joined Sep 30, 2009
9,326
Thank you to every one especially prandevou. I'm going to need some time to digest all of this, and do more reading. When I hit the wall again I'll be asking more questions. Again THANK YOU ALL!

#### praondevou

Joined Jul 9, 2011
2,942
I was still thinking about the easiest way to do it.
I wonder if a quadrature oscillator with an Opamp could work, they already have two output sine wave signals that are 90 degree shifted. But I guess they will not be stable or predictible enough in amplitude...

Another possibility would be instead of using an EPROM lookup table (where the sinewave is stored) to create a digital sinewave with counter ICs. Like the MC14017. But for a less distorted sine wave that doesn't need much filtering you would need two counters --- for 1 sinewave, so 4 counters in all.

They could be synchronized to the 90° shifted signals I posted before. And they would be stable in amplitude if frequency changes...

Also, I was thinking any additional phase shift caused by filters would be bad, but actually it isn't, because both sine waves will suffer the same phase shift, i.e. it'll still be 90°.

Either way this circuit will not be small. But it's doable.

Good luck

#### THE_RB

Joined Feb 11, 2008
5,438
I was still thinking about the easiest way to do it.
...
Another possibility would be instead of using an EPROM lookup table (where the sinewave is stored) to create a digital sinewave with counter ICs.
...
Either way this circuit will not be small. But it's doable.
...
Yep it will be small and easy, just copy my project "Accurate xtal-locked Sinewave inverter!" from this page (just past half way down the page);
http://www.romanblack.com/one_sec.htm

It uses a cheap 8pin PIC to generate complimentary sine waves in PWM for use in inverters etc. The PWM steps are sequenced in the interrupt so if you change the interrupt timing you can change the frequency.

#### praondevou

Joined Jul 9, 2011
2,942
It uses a cheap 8pin PIC to generate complimentary sine waves in PWM for use in inverters etc.
The OP doesn't want to use a microcontroller That's the problem.
He wants two 90° phase shifted sinewaves whose voltage increases with frequency, from e.g. 2Hz to 50Hz. All with discrete circuitry.

#### shortbus

Joined Sep 30, 2009
9,326
I guess maybe I've been using the wrong term. Been saying "replicated" sine wave when should have been saying "modified" sine wave. Most modified sine wave inverters only use a three step PWM wave I am after a increased amount of steps.

The frequency to go between ~25 to 120Hz. And be 90° electrically shifted. I did put a DMM on the out put of a VFD on my surface grinder and between those frequencies the voltage doesn't change at all.

#### strantor

Joined Oct 3, 2010
5,910
I guess maybe I've been using the wrong term. Been saying "replicated" sine wave when should have been saying "modified" sine wave. Most modified sine wave inverters only use a three step PWM wave I am after a increased amount of steps.

The frequency to go between ~25 to 120Hz. And be 90° electrically shifted. I did put a DMM on the out put of a VFD on my surface grinder and between those frequencies the voltage doesn't change at all.
Your dmm most likely doesn't have a filter. It's probably trying to measure the rms value of Pwm, which doesn't work well

#### praondevou

Joined Jul 9, 2011
2,942
Your dmm most likely doesn't have a filter. It's probably trying to measure the rms value of Pwm, which doesn't work well
I agree. Measuring PWM with a DMM is probably not going to work.

@Strantor, since you are more involved with VFD, do you remember what the Volts/Hertz curve looks like after the maximum specified motor frequency, e.g. motor rated for 400VAC/60Hz. Does it go flat? I think it goes flat, say with increasing frequency it doesn't increase voltage anymore.

The maximum voltage is limited anyway by the DC bus voltage.

@shortbus, As I said I know nothing about two-phase motor control. But the original 3phase motor you want to modify is usually controlled by a VFD whose output voltage changes with frequency (the resulting sinewave)
Have a read HERE under "operation".

Are you going to increase the number of turns? What DC-bus voltages are we talking about?

A fixed voltage sinewave would make things easier, since your minimum frequency is 25Hz it could work. And since you don't want a pure sinewave would the following signal fit your needs?

Can be done with this circuit or counters as I stated before.

#### strantor

Joined Oct 3, 2010
5,910
@Strantor, since you are more involved with VFD, do you remember what the Volts/Hertz curve looks like after the maximum specified motor frequency, e.g. motor rated for 400VAC/60Hz. Does it go flat? I think it goes flat, say with increasing frequency it doesn't increase voltage anymore.

The maximum voltage is limited anyway by the DC bus voltage.
I've never tried to run a 60hz motor higher than 60hz so I can't say for certain. I believe though, that the VFD will map the voltage 0-100% to the frequency 0-100%.
So for example if I tell that dive that max freq is 60hz and max voltage is 480V, then at 30hz, voltage will be 240V and at 60hz, voltage will be 480V. If I then change the max freq in the VFD to 120Hz, then at 30hz the voltage will be 120V, at 60hz it will be 240V, and at 120hz it will be 480V. This is only for sensorless vector mode. With encoder feedback, it will adjust the voltage up or down to compensate for torque to maintain the speed for a given hz. I'm about 80% sure of that answer.

I don't really think that the manufacturer's use of "volts/hz" terminology is accurate. The really aren't controlling the voltage per se, as in a DC voltage level. They are controlling the current through the winding by varying the duty cycle of voltage applied to it.
see this picture:

the grey lines are the PWM, the black line is the current. The current is actually where the sine wave is. your meter won't read those hideous DC pulses. I believe you can make a RC filter to read the true AC value, but not sure how. my super fluke does it for me.

#### praondevou

Joined Jul 9, 2011
2,942
I don't really think that the manufacturer's use of "volts/hz" terminology is accurate. The really aren't controlling the voltage per se, as in a DC voltage level. They are controlling the current through the winding by varying the duty cycle of voltage applied to it.
Yes, of course. That is, since he doesn't want a PIC then if he wants to transform a reference sinewave into PWM he will need to control the amplitude of that sinewave...

I was just thinking, for this case it may also be possible to change the amplitude of the triangular wave , for a limited range of frequencies that could also work.... It may be easier than to control the amplitude of two sinewaves... Time to do some simulations work.

#### strantor

Joined Oct 3, 2010
5,910
Yes, of course. That is, since he doesn't want a PIC then if he wants to transform a reference sinewave into PWM he will need to control the amplitude of that sinewave...

I was just thinking, for this case it may also be possible to change the amplitude of the triangular wave , for a limited range of frequencies that could also work.... It may be easier than to control the amplitude of two sinewaves... Time to do some simulations work.
ah, ok now I get it. This whole time I was thinking he wanted to feed a sine wave to the gate of a MOSFET. I didn't get it; was waiting for it to become clear to me. thanks.

#### praondevou

Joined Jul 9, 2011
2,942
Not sure if someone already stated it but I guess a stepper motor control with microstepping feature could be used too.
Like the A3977 for about $6 to$10.
Not as complicated as learning to program microcontrollers but it's a long datasheet too. The output stages would need to be connected to some drivers and two H-bridges since the DC bus voltage for shortbus' motor will be much higher than the max rating of this chip though.