AC/DC Inverters

timrobbins

Joined Aug 29, 2009
318
The ripple current rating for an electrolytic is typ given for 100Hz sine or 10k or 100k . The current being transfered in the inverter is not sine, so some margin is needed, but rms level is a good start for comparison. How close you operate to ripple rating is just a lifetime choice. Need to use a proper specced cap.

The mains frequency ripple current drawn from the DC source needs to be locally bypassed by C3. The higher frequency transient current resulting from deadzone dI/dt needs to be locally bypassed by C3. C3, FET and transformer return need to be a star connection. And 555 has a requirement for C3 loop.

The transformer will have significant leakage inductance that will likely swamp level of L1. An RC snubber across Q2 can provide some level of local bypass of higher frequencies, but probably better to tailor FET transition times.

"No gates are wanted at this point." - 'gates' refers to FET gates. Concept would be something like an asymetric gate drive to each FET such that turn-on is delayed by considerable amount using higher value series resistance. If you make the series resistance an LDR, and use a LED in the output circuit to feedback the output current to the LDR, then the deadzone would reduce with increasing output current - aka quasi square regulation scheme.

Ciao, Tim
 

SgtWookie

Joined Jul 17, 2007
22,230
Here's my idea for preventing shoot-through. Someone else described this same idea in text.

The goal is to make the turn-off fast and delay the turn-on. The diode quickly changes the gate voltage toward the turn-off state for each of the FETs, at least down to around the Vf of the diode close to a Vgs of 0. The resistor delays the turn-on. Adjusting the value of the R changes the turn-on time, in combination with the gate capacitance of the MOSFETs.

The waveform of the output of the push-pull will be different... there will be time when both MOSFETs are nearly off, between each phase.
Well, the big delay will be during the Miller effect. While a MOSFET gate responds more or less like a capacitor, there is a point right at the threshold where it takes a relatively large amount of current to get over/under. It looks like a step in the waveform. At that point, the MOSFET is in a partially conductive state, and will dissipate power as heat. The idea is to control the dead time, but keep the switch times relatively low.

I am really curious about how to make a sine wave AC output from a DC source, with efficiency. I could use a microcontroller and a lookup table and DAC to make an approximate sine wave. That is easy. But then how does one efficiently drive that voltage output at a high power? With a push-pull stage?
This again ... Check out the project this guy did for fun:
http://www.hazmat.com/~mjb/projects/rswg/
Note that to generate the quadrature sine waves, he used 128-byte lookup tables for his outputs, due to the page size limitations in PIC assembly vs speed considerations.

I have an idea that this will whet Bill Marsden's appetite for PICs ;) Really, don't you want a spinning sprocket display on your O-scope? :D
 

tom66

Joined May 9, 2009
2,595
If you're not too worried about cost I can tell you I can generate a sine wave and cosine wave using IEEE754(!) FPU math on a dsPIC33F, at at least 100Hz.
 
Well, the big delay will be during the Miller effect. While a MOSFET gate responds more or less like a capacitor, there is a point right at the threshold where it takes a relatively large amount of current to get over/under. It looks like a step in the waveform. At that point, the MOSFET is in a partially conductive state, and will dissipate power as heat. The idea is to control the dead time, but keep the switch times relatively low.
Sgt, that is a really good point. It's not the ideal solution -- really is much better to use separate signals for each transistor and serious gate drive current to make very sharp transitions, especially if you're dealing with high current flows.

I have used my simple solution and it worked very well for low power devices where the gate drive is simply the output pin of a microcontroller, and the switching frequencies are also slow. It did eliminate shoot-through and therefore power loss. It also simulated well. But it's not ideal for high power stuff, as you say.
 
Still, isn't there a relatively simple analog way to generate a nice sine output with high efficiency? Isn't there some way using L-C resonance to make it happen with a few components? Perhaps not a perfect sine, but reasonable?
 

timrobbins

Joined Aug 29, 2009
318
For at least a decade or two the quasi-square provided inverter manufacturers with a reasonable approximation to a sine, and allowed the rms output to be regulated, although the crest variation was always a problem for some loads. The switching loss of the FETs is then relatively quite low at mains, and can be a tradeoff for reducing emi by slowing down the transitions quite considerably.

Moving to class D pwm to simulate a better a sine, requires much faster switching times to manage the significantly increased switching losses, and hence much closer control of the switched transition of the FETs and much better gate drivers for largish FETs - ie. not a simple circuit solution path, and requiring more attention to emi control.

Class D audio has seen a huge effort into ways to manage such a design issue, using full H bridge and spread spectrum and very specialised gate drive techniques - well worth the read.

Maybe there is a way to use a 556 with one half producing the deadtime control.

Ciao, Tim
 

SgtWookie

Joined Jul 17, 2007
22,230
As timrobbins implied, instead of switching at 50Hz or 60Hz, the class D type is switching in the 30kHz to 70kHz range - some a good bit faster. There are quite a few MOSFETs nowadays that can be switched very quickly, due to their low gate charge - particularly those with low Vdss limits.

Just for an example, take a look at an IRLR8721/IRLU8721 n-ch logic level power MOSFETs; Vdss=30, Rds(on)=8.4m Ohms, Qg=8.5nC, Id=65A. Of course, you have to keep the MOSFET cool to sustain that much current.
Datasheet: http://www.irf.com/product-info/datasheets/data/irlr8721pbf.pdf
 
Tim, do you have any leads on where to read about Class D gate drive techniques?

There is a really nice Wikipedia article on Class D amplifiers.

I am starting to understand the Class D technique. I see that it is really like a buck converter, with the target output voltage following a changing input voltage level. The switching speed must be many times the speed of the frequency of interest in the guiding signal.

Would it be possible to make a reasonable 50 or 60 Hz output using a 10 kHz switching speed?

Could it be done with 1 kHz?

How does that affect the output filter (L-C low-pass filter) component sizing? I'm guessing a bigger C would be required. Since it's on the low voltage side, before the input to a transformer, then perhaps it could use supercaps and be pretty small and inexpensive while still having lots of C.
 

Ghar

Joined Mar 8, 2010
655
I've built a PWM H-bridge inverter with a very nice 60 Hz output but low voltage and low power. My switching frequency was 50 kHz.
I had a simple op-amp oscillator driving the error signal of the PWM chip.

At the time I knew very little about power electronics so there were quite a few problems. The plan was to add a transformer but of course it just ruined everything from saturation.
 

tom66

Joined May 9, 2009
2,595
A few ideas I had... I thought I would share them. This is coming from someone who has never designed an inverter.

I think it would be easier to use two SLA batteries.

First, if 12V batteries are used, then there is 24V to switch. Higher voltage, lower currents, less losses. Maybe(?) a smaller transformer.

Additionally, it makes a switching system easier. A push-pull configuration would be needed. The middle terminal between the batteries is the ground. I am trying to think of a way to implement this; I have an idea using a Half H-bridge IC.

The circuit presented (page 1) originally depended on one battery. The problem with this is you must generate an AC signal to feed into the transformer. The circuit used a capacitor and inductor to produce AC spikes, which would probably not work even in the best of cases.

How feasible is it to generate +325V and -325V dc buses and then use PWM to filter these into an AC signal (230V here)? No transformer: makes it much smaller and cheaper.

The problem I can see is designing the DC-DC converter to work. The DC-DC converter would probably be able to give out high surge currents, so it would not be a circuit for a novice to build.

I have previously tried to design (never build!) a 325V dc power supply. I was somewhat successful, in that it could supply up to 10 amps... that's well over 3 kilowatts. But it had lots of problems, especially with ripple, and inductor selection. Battery current would be massive... around 155 amps at 3 kW (2x12V SLA), assuming 80% efficiency, which is somewhat optimistic. And I could never design the negative rail easily, or find a way to switch the 325V dc.

A ridiculously dangerous design would be to have a dual DC-DC converter, with 325V and 650V rails. The 325V rail would become a virtual ground, and switching between the 0V and 650V rails would produce ac. But I can't help thinking that this could be quite dangerous; this would raise the potential of the neutral 325V relative to the battery, creating a dangerous shock hazard if it were installed in a car or other area where the negative terminal is usually connected to the metal chassis.

Just letting my brain spill everywhere.

I did not fully read the thread... so sorry if any of this has come up before.
 

Ghar

Joined Mar 8, 2010
655
You don't need the two +/- supplies if you use the full H-bridge. With a half bridge you can still get it without the second battery, you can do the mid point voltage with large storage capacitors.

I think stepping up to a very high DC voltage is the general idea used. You make an isolated boost DC-DC converter with a small high frequency transformer then you PWM with a full bridge or half bridge down to the 60Hz 120 Vac rms or whatever.

I've never built a legitimate version but I think that's how they do it. I believe you do the same type of thing in power factor correction circuits.
 

Thread Starter

Wendy

Joined Mar 24, 2008
23,415
Tim, do you have any leads on where to read about Class D gate drive techniques?

There is a really nice Wikipedia article on Class D amplifiers.

I am starting to understand the Class D technique. I see that it is really like a buck converter, with the target output voltage following a changing input voltage level. The switching speed must be many times the speed of the frequency of interest in the guiding signal.

Would it be possible to make a reasonable 50 or 60 Hz output using a 10 kHz switching speed?

Could it be done with 1 kHz?

How does that affect the output filter (L-C low-pass filter) component sizing? I'm guessing a bigger C would be required. Since it's on the low voltage side, before the input to a transformer, then perhaps it could use supercaps and be pretty small and inexpensive while still having lots of C.
Pulse Width Modulation
 

Thread Starter

Wendy

Joined Mar 24, 2008
23,415
A few ideas I had... I thought I would share them. This is coming from someone who has never designed an inverter.

I think it would be easier to use two SLA batteries.

First, if 12V batteries are used, then there is 24V to switch. Higher voltage, lower currents, less losses. Maybe(?) a smaller transformer.

Additionally, it makes a switching system easier. A push-pull configuration would be needed. The middle terminal between the batteries is the ground. I am trying to think of a way to implement this; I have an idea using a Half H-bridge IC.

The circuit presented (page 1) originally depended on one battery. The problem with this is you must generate an AC signal to feed into the transformer. The circuit used a capacitor and inductor to produce AC spikes, which would probably not work even in the best of cases.

How feasible is it to generate +325V and -325V dc buses and then use PWM to filter these into an AC signal (230V here)? No transformer: makes it much smaller and cheaper.

The problem I can see is designing the DC-DC converter to work. The DC-DC converter would probably be able to give out high surge currents, so it would not be a circuit for a novice to build.

I have previously tried to design (never build!) a 325V dc power supply. I was somewhat successful, in that it could supply up to 10 amps... that's well over 3 kilowatts. But it had lots of problems, especially with ripple, and inductor selection. Battery current would be massive... around 155 amps at 3 kW (2x12V SLA), assuming 80% efficiency, which is somewhat optimistic. And I could never design the negative rail easily, or find a way to switch the 325V dc.

A ridiculously dangerous design would be to have a dual DC-DC converter, with 325V and 650V rails. The 325V rail would become a virtual ground, and switching between the 0V and 650V rails would produce ac. But I can't help thinking that this could be quite dangerous; this would raise the potential of the neutral 325V relative to the battery, creating a dangerous shock hazard if it were installed in a car or other area where the negative terminal is usually connected to the metal chassis.

Just letting my brain spill everywhere.

I did not fully read the thread... so sorry if any of this has come up before.
The 555 has an upper limit of 15VDC, the CMOS 7555 18VDC. We discussed H Bridges in this thread, High Side MOSFET Drivers. I probably need to go back over it, I think this application would be a good one.

I have an idea how to come up with something close to a sine wave, but other ideas are appreciated. If it beats mine in linearity, stability, and parts count I'll probably use it.

I have worked on a house (industrial warehouse) UPS that had 48 12V wheelchair batteries, I don't know if it was a sine wave model or not. Basically it kept a few factory lights on if the power died. Since they were all in series it was something that kept you focused on the job.

My next diagram is for a full H bridge, the one after that is for a class D amp, but I'm guessing this is obvious.
 
Last edited:

crazyengineer

Joined Dec 29, 2010
156
I'm sorry for posting on such an old thread, but I just have the following questions

1) I know you said square wave is not ideal, but can you just directly connect the 555 to the transformer?

2) You recommend using sinusoidal waveform, so can you forgo the 555 timer and build a sine wave oscillator using a couple of omp amps?
 

Thread Starter

Wendy

Joined Mar 24, 2008
23,415
The 555 has a max voltage of 15V, and a max current drive of 200ma (0.2A). If you were to drive it to this max (which is never a good idea) it would be 3W. Without amplification this is the absolute max, and the real numbers are closer to 1W.

You can always connect a waveform to a transformer, but at 3W you're not going to get much power. At 120V (USA standard) you will get 0.025A. A 60W light bulb uses ½A from 120VAC.

The real weakness of inverters are their currents. For a 120W light bulb using a inverter running off of 12VDC you will need 10A. This isn't very practical overall.
 

SgtWookie

Joined Jul 17, 2007
22,230
I'm sorry for posting on such an old thread, but I just have the following questions

1) I know you said square wave is not ideal, but can you just directly connect the 555 to the transformer?
That would be kind of pointless, as a 555 timer can only source or sink up to 200mA. Additionally, a BJT 555's output can't go much higher than its' supply voltage less ~1.3v even with a light load, due to the output Darlington emitter follower configuration.

2) You recommend using sinusoidal waveform, so can you forgo the 555 timer and build a sine wave oscillator using a couple of omp amps?
You should quote who said that, so the context of what you're asking about can be determined without reading through this whole thread.
 

Thread Starter

Wendy

Joined Mar 24, 2008
23,415
I think he was quoting me, though I don't know where or what the context was. Most inverters use a square wave, which will power a lot of devices, but not all of them. There are cases where a square wave is not processed the same as a sine wave in some gadgets. Square waves are also full of harmonics, which means they can be relatively noisey, so a radio or TV might have some trouble with them. Just being next to something like this could cause interferance.
 
Top