12/945,704   filed 11-12-2010


A Rectifier-less Bidirectional AC to DC converter is just a simple

DC to DC being used differently. This invention can convert AC<->DC
without using any diodes.
The schematic above is showing diodes
only as a reminder that shoot thru current needs to be addressed.

The gate driven duty cycle of a DC to DC converter defines

an output voltage. If this output node instead gets connected

across a secondary of an AC transformer, and if the duty cycle gets 

defined to track secondary current, then the secondary will be seeing a 

simulated resistor across it. This simulated resistor will absorb

energy and transfer it to the split supplies. It is based on using

the Energy Harvesting Resistor described here. 

The circuit and equations are shown above. The working breadboard is 

shown below. 



This breadboard takes a 3.2Volt rms 60 Hz input directly off a 

secondary of an AC line transformer, and transforms it 

directly into +/- 6.7 Volt splits supplies to drive 8 LEDS. 

Most of the circuit is for duty cycle modulation experiments. 

And this is all being done using analog circuitry.

The full schematic is included at the end. Everything appears 

to be working as expected.


There appears to be two main modes of operation. The first

mode is to measure the secondary current, and then use it

to modulate the duty cycle of a simple DC to DC converter 

above and below 50%. In the breadboard, this causes the 

secondary to think it is seeing a 22 Ohm resistor across it. 

The modulating duty cycle is multiplexing the 150mA rms input

current to the two supplies. The actual AC waveforms of the secondary

voltage Vac and the Vsense voltage are shown above. Both waveforms

are on the same scale. The current flowing across the 8 Ohm 

sense resistor is the secondary current. 


The bidirectional mode of operation involves monitoring the the 

secondary's voltage instead. In this case the secondary voltage times 

some constant Kgain is used to modulate the duty cycle. 

There will be some value for Kgain where the Vsense voltage

equals the VAC input voltage. Under this condition, little power

should flow. The two AC waveforms off the breadboard are shown 

above. This should raise some questions as to how close can the 

duty cycle get to reflect the voltage that will appear at Vsense.  


Now if this value for Kgain is increased, then the vsense voltage can get 

larger than Vac, and now power is flowing into the 60Hz AC socket. Unlike

normal AC inverters, this is transferring power from the split supplies 

as a negative resistor. The waveforms above show that Vsense is larger

than Vac. An external power source needs to be applied to VCC and VEE

to do this test. It appears that the power drain on the external supply

is what would be expected.


Lower the value of Kgain enough, and now the power is flowing from AC to DC.

This method of voltage driving the duty cycle mode appears to be

ideal for bidirectional AC<->DC power transfer since the circuit goes to an 

open circuit between the power flow directions.

In this breadboard, the efficiency appears to be around 84%. 

It is not obvious how this invention is little more that just using 

a simple DC to DC converter in a new application. If that is so, 

then the same efficiencies for DC to DC converters should be possible. 


This breadboard is actually self starting. When the AC first turns

on, it forward biases the drain bulk junctions of both 

power MOSFETs. This turns on the duty cycle controller which

drives the split supplies to a level higher than the AC input



When the breadboard is doing the job of a full wave rectifier,

it loads the AC line exactly like a resistor. According

to some recent articles in EDN and Power Electronics Technology,

things like having a high power factor is becoming important. 


Apparently the nonlinear loading of full wave rectifier circuits

is starting to create a need to raise the power factor

of AC to DC convertors. 


The figures above are from a Fairchild application Note about Power factor 

correction. The attempt is to be able to draw power off the ac

line like a resistor. Not very many power factor reduction

methods however lack rectifiers in the power signal path.



The breadboard sure does look like it can load the ac line like a

linear resistor. And even as a negative linear resistor. If solar power

becomes more widespread, would it generate a need for a high power factor 

DC to AC convertor?

The power companies are installing smart meters in peoples homes.

This might allow power companies to charge customers at different

rates over the day. Could it be economical someday to store AC power

into home batteries during the time when rates are low, and then

reconvert it back to AC for use when rates are high?


The invention is mainly just a DC to DC converter with differences 

in terms of input/output ports, and in how the duty cycle gets

defined. The easy way to do the duty cycle in analog is to build 

up a triangle waveform, and stick it into one input of a voltage

comparator. An analog signal voltage on the other input will modulate the 

duty cycle. 

The analog signal and its effect on the duty cycle is doing two things.

First it generates a voltage at Vsense which is scaled between the two

supply voltages. Second, the duty cycle multiplexes any current through 

the sense resistor between the two power MOSFETs and their power supplies. 


The system model above can show all the equations. For this system model,

the analog voltage Vd is being defined to go from 0 to 1, which

corresponds to a 0 to 100% duty cycle. This duty cycle Vd is being modulated

by the Isense current at some gain. The Vd voltage will both define the 

voltage Vsense, and will multiplex any Isense current. Consider what happens

when Vac and Vsense are identical. No Isense current will flow, and zero

current will get multiplexed between the two MOSFETs. 

Because the Vd term, which defines the Vsense voltage, is being scale

by the input current, the Vac voltage source will see a load looking

like a linear resistor. Now the current multiplexing will always be charging

up one of the supplies and discharging the other. But when the total

power of both supplies are summed together, they are receiving

close to the same power that the Vac source is being loaded with. Naturally 

any IR drop in the power path will reduce efficiency.


This invention tends to have a native output format of power

regulation. For instance, if the output loads are LEDs,

and if the voltage across LEDs can drift with temperature,

the output current will auto-adjust itself the maintain the same output


For the sake of tying up any loose ends, one might be

inclined to make the Vd value track the inverse of the supply

voltages. Since the Vsense voltage is being scaled by the 

supply voltages, that means the Vsense magnitude should be 

insensitive to the supply voltages. If so, then the same input

Isense produces the same Vsense. So the simulated load resistance

across Vac will not change. So the power transfer will be independent

of the split supplies voltages.

Suppose both VEE and VCC increase by 10%. The same Isense current

should produce a Vd with a 10% decrease in modulation. But the

Vsense is also scaled by the 10% higher VCC and VEE voltages. So Vsense

comes out the same. But the multiplexing of Isense current has

been reduced by 10%. So the two supplies are at a 10% higher voltage,

but they are receiving 10% less charging up current. 


Now normally the output power is formated in a voltage regulated

mode. There is nothing stopping the voltage to current relationship

to the input resistor from being adjusted in terms of both magnitude 

or asymmetry. Because VCC gets charged up when Vac is positive, and

VEE when negative, it is possible to regulate both VCC and VEE 

independently. The system simulation above shows both VCC and VEE

being regulated to 7 volts. The supply voltages are low pass

filtered since feedback should have a one pole compensation.

The loop gain is set to 10. Different values for G get switched

in depending on Vac polarity.


In this example, VEE will lightly load with constant 50uA load. 

But VCC will vary its load in steps over three orders of magnitude. 


Under heavy load, one would expect VCC to develop some ripple.

But VEE appears to develop ripple at the same time. This is 

due to the fact that whenever VCC is being charged, VEE is

being discharged. The VEE feedback loop needs to recover this

lost charge when Vac is negative. So even when VCC is

the only supply being used, the Vac voltage can not be completely

asymmetrically loaded. 


The asymmetry of the current loading on the AC line is shown

above. So this type of AC to DC converter cannot load the 

AC line like a one diode rectifier. The lowpass filtering

is not at the moment low enough to reduce the looping in the current loading

curve. These low pass filters will affect the voltage regulation

response to load current spikes. The step response to a high load 

current appears to be fast attack, slow decay.

For single supply applications, it may make more sense to build up
a second DC switching network as is shown on figure 9 on this web page 

to transfer charge from one supply to the next. This enables the resistive
loading to be symmetrical.


The full circuit for the breadboard is above. A secondary

sense current node and secondary voltage sense node are

provided, and both have been checked out. A LM13600 is being used

to provide an easily scaleable triangle wave, and it is also 

used as a voltage comparator. A LM6144 is being used to buffer the 

triangle wave, and to measure the current through the 8 Ohm sense resistor. 

A CD4007 is being used to drive the Power MOSFETS. These MOSFETs have 

large capacitances, and adding some diodes and resistors appear to 

reduce the shoot through current. 

Most of the supply current is going to the LM13600. It draws in fact

a little too much supply current. It is mainly being used just

to check out some scaling concepts. The duty cycle controller

mainly needs just a triangle wave source and a comparator. 

The gate resistor/diodes networks certainly could be replaced with

something better in order to operate at higher frequencies. The 

sense resistor is set at 8 ohms in order to make the voltage across it

obvious enough to show things like bidirectional power flow.  

The Triangle wave is running below 20KHz, so the 30uF cap

(two 60uF electrolytics) is chosen just to make the

Vsense signal look clean. 

Other breadboard experiments are in the works. Running spice simulations 

side by side with a working breadboard has some crosschecking advantages.

Some of the components used to build the AC<->DC converter hardware are

shown below. For the moment, the breadboard is mainly being used for
proof of concept.



There appears to be a lot of recent patent activity in the area
of building "bridgeless PFC convertors". The following are
some of the patents.




And here is some more information for those who may be interested.

A bidirectional, sinusoidal, high-frequency inverter

A new structure for bidirectional Power flow
Bi-directional single-phase half-bridge rectifier for power quality
BiDirectional Converter

Synthesis of Input-Rectifierless AC/DC



Don Sauer