# DC - 3 phase AC inverter construction

Ok, I'm looking for a way to go from DC to 3 phase AC, in both small and large applications. For now, just to satisfy my curiosity, I'd
like to make some small 3 phase AC motors on a small scale, so I was thinking my DC input would be a bank of AAs or Ds, either 4, 6, or 8, so 6, 9, or 12v. So, very small scale.
I'd also like one that could be used to say, drive about a 100HP motor.
I haven't found much, I found this site http://www.uoguelph.ca/~antoon/circ/555dcac.html which seems fairly straightforward, but gives only single phase AC. How would I go about getting two more phases from here? Would just using a capacitor act to shift a phase? Or to be exactly 120 degrees apart would I want to somehow have one line get a pulse only on the 1 from the timer, the other on the 2, the other on the 3 and then the first on the 4th, etc. I'm just not sure how to go about that.
Correct me if I'm mistaken, the timer is putting out a square wave and the point of the large capacitor and filter are to make this more sinusoidal?
Do you think this diagram would be good for my small scale application, because that last capacitor is looking pretty mighty.
Finally, I bought a 2200uF cap instead of 2700, which I think would be fine, but I just noticed it says 50WVDC, is this a max capacity or would it be very bad to use it with 9V of DC?
Oh, I also couldn't find a 1uH filter, only a 100uH, I realize that's two factors off, but how precise do I need? I'm less familiar with filters (first I've heard of a microHenrie).
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
I hate to respond so quickly to my own post, but after we have a 3 phase system, how can we add the ability to allow for regenerative braking. Where does this come in?
Do most brushless DC motors already have/support regenerative braking? Another important piece in deciding which system to go with.
snipped-for-privacy@gmail.com wrote:

<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>

One way to get three phase from an inverter, for small applications is to create a second inverter that is slaved to the first, but delayed by 90 electrical degrees. This gives you two-phase output. Then use those two phases to power a Scott-T transformer connection to convert the two-phase to three phase.
Look for designs of VVVF (variable voltage / variable frequency) motor drives. A common way to convert the DC to 3-phase is with six SCR's arranged much like a three-phase to DC rectifier (three 'legs' with two SCR's in each, taking a phase from between the SCR's on each leg).
daestrom
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
daestrom wrote:

I think that the OP already has extensive knowledge of bridges..
--
Sue

<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Ok, it looks like an SCR is basically two opposing transistors hooked together in a certain way. I can understand how that would cycle the gates, but it would be a pretty fast Hz and there'd be no way to control it.
More importantly though, how do you control it such that the three generated ACs are 120 degrees out of phase?
I guess how do you slave the second to the first and how do you delay 90 degrees (so specifically).
I'll look up Scott-T, thanks.
daestrom wrote:

<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>

No, although the basic layering of an SCR looks like two transistors an NPN and a PNP tied 'back to back', the operation is much diffent than what you're supposing hear. Think more like a diode but the diode won't conduct in the forward direction until 'triggered'. Then it continues to conduct until the current drops below a certain threshold. After that it stops conducting.

Typically for a three-phase inverter, it's done with a 'ring counter'. a simple counter that cycles from 0 to 5 (six steps) over and over. At each 'step' a different one of the six SCR's is triggered into conduction. Some 'auxilary parts' are used so that when one SCR in a leg is triggered, the other SCR in the same leg is commutated off (otherwise you have a direct short across the DC supply). I'll leave it to you to figure out the 'firing order'.
For the smaller two-phase system using Scott-T, the slave is just that, 'slaved' to the master. In a fixed frequency output application, a fixed time delay from the clocking of the master is used to clock the slave (could be as simple as a one-shot delay circuit). For a variable frequency design, the length of the time delay has to vary precisely with the master's frequency so that it is always 1/4 cycle off. The easiest way to do this might be to run a 'clock' at 4 times the desired frequency and feed a simple 0-3 counter that repeats endlessly. Use counts at 0 and 2 to trigger each half of the master's cycle, and counts 1 and 3 to trigger the slave.
For small power applications, two transistors can be used instead of two SCR's. But SCR's have a much faster 'turn-on' time than power transistor so there is less heat build up inside. When driving a 'real world' load such as a motor, turning off one transistor in a leg before turning on the other one is a bad idea. Any inductance in the load (motor) will create some serious voltage spikes. That means 'snubbers' and such to absorb the spikes. Switching both transistors simultaneously leads to 'totem pole' types of losses (similar to switching losses in TTL logic circuits).
SCR's are used in higher power applications. Shorter rise times, and one of the two SCR's is always 'on'. The trick is to be sure the opposite SCR in a leg will 'commutate' off when triggering each SCR. Center tapped inductors is one common way of doing it, but it requires a certain *minimum* load current.
Another method is to run the inverter at a high frequency all the time, but vary the duty cycle. By controlling how fast the duty cycle varies from min to max, you control the 'frequency' of the output. Think of PWM where the modulation is the desired frequency. This design has the advantage that by controlling exactly how the PWM is varied, you can filter out the high frequency and have something close to a 'true' sine wave output (sometimes referred to as 'modified sine wave' inverters although it's more a 'modified square wave').
good luck
daestrom
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>

Inverter design is very tricky and you must ahve a good fundamental understanding of electronic theory before you start. There is a lot of literature on the inverters: there are books, technical papers and patents. Dig deep and you will find a lot of stuff on ithe subject.
In inverters, sine waves are developed by Pulse Width Modulation or other techniques of adding square waves not by filtering as you propose. Small filters are used to reduce noise. In high power units, switching frequencies of eight to ten kHz are typical. Higher frequencies can be used in low power units. Switches are usually FETs or other high current semiconductors that can be commutated on and off at a high rate. SCR's are usually not used because they are too hard to turn off. Modern inverters usually operate in four quadrants, meaning they can accept power from the load as well as deliver it to the load. This is all done by switch timing and control.
Virtually all inverters use some form of microprocessor to develop the PWM switching signals. Phases are developed by timing not by phase shifting. Precise control of phase angle is accomplished by the processor clock and counters, usually crystal controlled. The sine wave shape is stored digitally in memory and the pulse width values for each phase at each increment of time is pulled out of memory at the required time by the processor. Furthermore, most modern inverters can vary the frequency by controlling the timing allowing motors to operate with variable speed.
You should start you project by learning enough processor coding to understand how PWM signals are developed. Once you have a processor cranking out multy-phase PWM sine waves, then you can concentrate on getting the switches to work. Good Luck.