AM modulation by pulse width controle / Tesla Coils / Forums

Forums

4hv.org :: Forums :: Tesla Coils

« Previous topic | Next topic »

AM modulation by pulse width controle

1 2 next

Move Thread

LAN_403

teravolt

Tue Nov 17 2009, 06:07AM

Registered Member #195 Joined: Fri Feb 17 2006, 08:27PM
Location: Berkeley, ca.
Posts: 1111

I am toying with the idia of a tesla with AM modulation. the way I want to do it is with changing the pulse width to the gates of fets. I picked up 6 FD600R17ZKF6C_B2, thease devices have a body diode and a seperate 600A fred diode. thease monsters are designed for chopping.

GeordieBoy had this to say in a previouse tread>

GeordieBoy in your opinion if I wanted to controle a full bridge power out by controling gate pulse width would the mur1660 be nesisary? The object would to amplitude modulate a tesla.

Yes, when you decrease the duty ratio of the gate-drive signals you introduce dead-time where no MOSFETs are conducting. The resonant sinusoidal load current doesn't stop however, it circulates through the free-wheel diodes during the times when all MOSFETs are off. It is when you next try to turn on a MOSFET that the problems can start. It is particularly the commutation of current from a slow free-wheel diode to a fast switch on the opposite side of the bridge leg that causes the problem.

With excessive dead-time added to reduce the power and implement audio modulation, you are almost guaranteed that the MOSFETs will turn on well after the current has gone through zero and commutated to the free-wheel diode across the opposite device. This is exactly the situation where the MOSFET body-drain diodes are too slow, and need to be bypassed by fast recovery external devices like the MUR1660.

However, I would say that if your ultimate aim is to achieve amplitude modulation you might want to look at implementing it another way. You can keep the duty ratio fixed at full, and modulate the frequency of the drive instead. This Frequency Modulation causes indirect AM of the current drawn by the resonator and still achieves audio modulation of the spark. The neat thing is that if you do this by detuning the driver above the natural resonant frequency of the coil, it essentially works up and down the high-side slope of the resonant peak where the load always appears inductive. This type of load is much easier on a MOSFET inverter, and you can surely get away with using the internal body-diodes in this case.

In short, I would say that the external bypass diodes are required if you want to build a MOSFET driver that is robust enough to withstand any manual mistuning, or errors in automatic tuning without blowing up. If the load seen by the inverter is always net inductive the body-diodes should be sufficient as they never see forced-reverse recovery in this mode. (That is why most power electronics applications don't require the body-diodes to be bypasses by external discrete diodes. The most common exception being DC motor control where forced-reverse recovery is a problem.)

Fast antiparallel diodes are not normally required in a DRSSTC that uses IGBTs. This is because most IGBTs are co-packaged with a second die comprising of a fast-recovery diode. The switchign properties of this free-wheel diode are usually chosen to match and complement those of the main IGBT. The important point here is that IGBTs contain a proper fast diode that has been chosen specifically to do the free-wheel task well - Conversely MOSFETs contain a crappy slow diode that you get for free, that is occasionally just adequate to do free-wheel duty in some applications!

-Richie,

I want to controle the gate time from Zero during each half cycle and not a widening of dead time. I have tried this on a small scale a cuple of years ago and I could vary power with a SSTC but never persude it. I am hopefully going to try this with a bridge of cm600s. The gates are driven with fiber optics and drivers as an expieriment and this is one of the things I'm going to look at. what do you think does my plan have merit.

GeordieBoy

Tue Nov 17 2009, 01:11PM

Registered Member #1232 Joined: Wed Jan 16 2008, 10:53PM
Location: Doon tha Toon!
Posts: 881

Duty ratio control is a well established and proven power control method used throughout the power electronics industry, so it will definitely work. There is no doubt about that.

In SMPSUs the designer is normally trying to control some DC output voltage (or current) by varying the duty ratio of the chopping, and then following this up with an LC filter to convert the chopped DC into smooth DC again. What is actually happening here is that you are chopping up the DC which forms a whole frequency spectrum including the switching frequency, harmonics and a variable component at DC. As you change the duty ratio from 0 to 100% the DC component goes from nothing to maximum. In a switch-mode power supply the output filter blocks all of the AC components and just allows the DC component to pass to the output of the supply. So in summary you vary the duty ratio of a pulse train to vary it's average DC component then you filter out all of the AC stuff to just leave the DC component. That's how it works for SMPSUs the world over, but SSTC's are a little different.

With an SSTC the resonantor doesn't select the DC component out of the resulting frequency spectrum, it selects an AC component at its resonant frequency. Usually the TC resonator is chosen to resonate at the fundamental switching frequency of the inverter so it more-or-less draws a sinusoidal current at the switching frequency. As you vary the duty ratio of the switching from 0% to 100% it turns out that the spectral energy in this fundamental switching frequency goes from zero to maximum and back down to zero again. Maximum power is in the fundamental when the duty ratio is 50% and the waveform is perfectly square. The voltage contribution of the fundamental part of a PWM pulse-train is actually proportional to the sine of the pi times the duty ratio.

This means that unlike our SMPS which has a linear transfer function, the PWM modulated SSTC has a curved transfer function. It is the first quadrant of a sinewave (0 degrees to 90 degrees) so it starts off rising linearly and then flattens out. If you keep the modulation depth low by keeping the volume low the sound quality will be reasonable. As you turn up the volume the modulation depth will increase and the "saturation" of the sine-shaped transfer function will cause more and more audible distortion.

So there is a little info about exactly why PWM causes AM modulation in SSTCs and why it doesn't sound too good!

-Richie,

teravolt

Wed Nov 18 2009, 04:19AM

Registered Member #195 Joined: Fri Feb 17 2006, 08:27PM
Location: Berkeley, ca.
Posts: 1111

thanks Richie, I had to think about this for a bit. when I have done any sstc work there was the harmonic frequencys that I have noticed. so if I get this right than when the width is adusted to a quarter wave length or 180 deg. of the harmonic it will become dominant verses the fundimental even if the beating frequency is the fundimental. so if I had a DRSSTC with out the secondary oscillating and I changed the width that it would oscillate at a harmonic depending on the width that I limit it to.

I happen to own a tektronix spectum analyzer and I want to set up a expieriment where I could see the frequency splitting in a DRSSTC like in spice simulations. just monitering it whyle it running it I don't think will work because a DRSSTC is running around its lower pole. if I pulsed the primary or frequency swept it or swept the pulse width maby I could get it to reviel the frequencies in a spectral form. I am looking for some idias. thanks N.B.

Dr. Dark Current

Wed Nov 18 2009, 09:49AM

Registered Member #152 Joined: Sun Feb 12 2006, 03:36PM
Location: Czech Rep.
Posts: 3384

What about dead time modulation?

Proud Mary

Wed Nov 18 2009, 09:37PM

Registered Member #543 Joined: Tue Feb 20 2007, 04:26PM
Location: UK
Posts: 4992

Aren't A3 and PWM quite different sorts of fish?

GeordieBoy

Thu Nov 19 2009, 12:23AM

Registered Member #1232 Joined: Wed Jan 16 2008, 10:53PM
Location: Doon tha Toon!
Posts: 881

> Aren't A3 and PWM quite different sorts of fish?

They kind of are, and they aren't. Let me explain...

In Switch Mode Power Supplies we normally modulate the width of a train of pulses in order to control the average level (the DC part.) However, when you modulate the width of a pulse train at say 200kHz you also modulate the intensity of the fundamental frequency component at 200kHz, and also all of the harmonics too! We just don't normally think about this because all of the HF stuff is usually filtered out to just leave the wanted DC.

In theory you could build an A3E (amplitude modulated double-sideband) transmitter by using a Class-D switching amplifier and just modulating the width of the switching pulses. Then follow it up with a narrow-band filter to make sure only the frequency component that you want to transmit goes to the antenna. I've tried it and it works, but...

From a comms point of view there is a catch. If you do the PWM using a traditional sawtooth ramp comparison, you only modulate one of either the rising edge or falling edge of the pulse train. This causes modulation of the phase of the fundamental frequency component too. So the RF engineers would say that not only have we achieved AM, but we have introduced undesirable incidental PM (phase modulation) too. In the frequency domain this messes up the nice symmeterical sidebands that an A3E DSB transmission should have. It makes it take up more bandwidth, causes interference, and ultimately limits audio fidelity too.

In theory this incidental PM can be overcome by implementing double-edge PWM using a triangular reference waveform instead of a ramp. However, for practical reasons it is hard to make the triangular waveform's slopes precisely identical and some incidental phase modulation still occurs. (If you are familiar with RCA's Ampliphase technique, this PWM thing can be compared to their method of phase modulating two RF carriers and then combining them to achieve AM by constructinve/destructive intereference of the two carriers. Invariably some PM always remains!)

So for these reasons high power AM broadcast is not usually done via the carrier-frequency PWM method. The most common methods are either high-level modulation of one masive RF power amplifier with a seperate PWM controlled series pass modulator, _or_ discrete switching of multiple smaller RF power amps who's outputs are then combined together and filtered to remove the stepping. The later method also providing excellent fault redundancy and carrier level control properties.

For SSTCs though the incidental Phase Modulation caused by single-edge PWM is probably not a problem. After all we want to modulate the average current through the arc to hear audio, not transmit a clean radio signal to distant lands.

-Richie,

Proud Mary

Thu Nov 19 2009, 01:39AM

Registered Member #543 Joined: Tue Feb 20 2007, 04:26PM
Location: UK
Posts: 4992

What a very thorough reply, Richie! Thank you for making things clear.

My own intuitive understanding seems to have stopped at Class C. A quick glance at a circuit diagram and I can identify A, B, or C and see their phase angles almost visible before my mind.

I shall get a more recent primer on modulation for Christmas, and so perhaps escape the fate of the living fossil,

Stella X

teravolt

Thu Nov 19 2009, 06:07AM

Registered Member #195 Joined: Fri Feb 17 2006, 08:27PM
Location: Berkeley, ca.
Posts: 1111

perhaps modulation could be achieved by doing class D on the dublier circuit that supplies the energy for the bridge. alot of thease super base amps that push 2kw at the lower frequencies are class D.

Russell Haley

Mon Nov 23 2009, 09:54AM

Registered Member #2478 Joined: Mon Nov 23 2009, 03:24AM
Location: Texas A&M University
Posts: 47

I think it might be possible to linearize the response to duty cycle with feedback from the primary current.

Most obviously, you could rectify and lowpass filter the output of a current transformer on the primary, and then get an error signal from the difference of the average current and you audio reference.

Another method would be to use something like the cosine-intercept scheme for generating firing signals on phase controlled rectifiers.

Integrate and invert the instantaneous primary current. (If primary current is sine, this is cosine.) Then send the cosine to two comparators. The first comparator would compare cosine(primary current) to (reference - audio signal) while the second would compare cosine to (reference + audio signal). AND the outputs of the two comparators, and you have your gate drive waveform with symmetrical dead time. The gate drive for the other half of the cycle could be generated in a similar fashion.

The second method uses a whole lot of analog signal processing, but it would probably have much faster transient response than average current feedback control.

GeordieBoy

Mon Nov 23 2009, 02:43PM

Registered Member #1232 Joined: Wed Jan 16 2008, 10:53PM
Location: Doon tha Toon!
Posts: 881

Yes, there are many solutions. Mostly based around one or both of the following techniques:

1. Pre-distortion of the audio to linearise the open-loop system,
2. Closed-loop negative feedback to linearise the system.

The only problem with taking primary current as the feedback parameter is that it is not a good indicator of resonant current flow in the TC resonator. The primary current is heavily contaminated with the triangular magnetising current of the primary winding itself. This represents a large reactive current flow that doesn't contribute real power flow.

If you want to use negative feedback for linearisation of modulation, you are much better either feeding back the magnitude of the secondary base current, or preferably doing synchronous detection on the base-current to extract only the magnitude of the in-phase part.

Both techniques are good in their own right, but the combination of negative feedback around a system that has been substantially linearised with pre-distortion should give superb linearity for audio.

-Richie,

1 2 next

Moderator(s): Chris Russell, Noelle, Alex, Tesladownunder, Dave Marshall, Dave Billington, Bjørn, Steve Conner, Wolfram, Kizmo, Mads Barnkob