| Electronics > Projects, Designs, and Technical Stuff |
| Inverter (isolated supply) quesions? |
| (1/1) |
| essele:
Hi, I've been considering getting myself an isolated supply as I'm working more and more on mains kit, but ideally a variable one so that I have a bit more flexibility, and as a result I've got a bit distracted looking at Pure Sine Wave Inverter concepts. My initial experiments have been pretty positive, but I have a couple of questions which are probably obvious, but I can't find the answers anywhere, and I haven't had time to fully test the theories yet. I've got a PSoC device creating a nice SPWM and can easily vary the switching frequency and resultant line frequency as well as control the amplitude for regulation and the variability part, so this is great. I've tested it with a couple of off-the-shelf H-Bridge boards to check it works, all is good although I've only used some very basic filtering (available parts!), but the no-load output is great. So my approach was to create a 50kHz spwm signal, and then output this on one output for the first half of the cycle, and then another output for the second half, so both sides of the H-Bridge are working identically just dealing with the different parts of the cycle ... and this all works fine. However a number of the solutions I've found online (and an inverter I've been tearing down) seem to separate the switching from the 50Hz line frequency, with one half of the bridge doing the 50kHz and the other half simply doing the line frequency. (Q1) Is there a reason for this? It feels better to me to keep things symmetrical? At the moment I'm switching a relatively low voltage, I'm experimenting with 12V, but my plan was to use 30v and then put the resulting output through a normal power transformer to step-up to mains voltage. Again initial experiments (no-load) look good. For my experiments I had the filter before the transformer, but for reasonable loads (300VA ish) this is going to need a high current inductor (or large cap) which seem to be difficult to find and expensive, so... (Q2) If I have a 50kHz SPWM output, should I filter this before the step-up transformer, or can I feed this into the transformer without any problems? If I do this presumably I can put the filter on the output, which is a lot less current to worry about? (Q3) Is this all wrong, should I just be switching 350VDC and be done with it? (And presumably then just use a 1:1 isolation transformer to ensure isolation?) (Q4) If I stick with the unfiltered 50kHz signal, could I used a smaller high frequency transformer to get the isolation so I can save on space/cost etc? (like this one? https://www.mouser.co.uk/datasheet/2/445/749196540-539789.pdf) As you can probably tell, this is definitely stretching my understanding, and as much as I like experimenting I am slightly nervous with these kind of voltages, so any guidance would be much appreciated. Thanks, Lee. |
| soldar:
For working on devices you want to isolate from the mains you can't beat the simplicity and ruggedness of a regular mains isolation transformer. That is what I have been doing for years and I can't see much benefit to going the electronic route. |
| T3sl4co1l:
Don't worry about generating PWM. It's what you're doing, but it's not why you're doing it. Why are you doing it? To build current in an inductor. In a flyback converter, that inductor is the transformer. In a forward converter, it's the filter choke, which can be on the input or output side, but traditionally is on the output. The equation you're using is: V = L * dI/dt V is the supply voltage, minus losses. Stepped through the transformer ratio, if applicable. L is supposed to be constant (but it does drop at high current, hence why you want to buy an inductor with saturation current greater than the peak current you're using). dI is a change in current. The inductor remembers its current from cycle to cycle, unless it's been discharged fully (and you aren't using a synchronous converter that forces continuous current flow). dt is the pulse width. This only gives the change in current, dI. The total current is made up from all the dI's added together. When the switch is on, dI is positive and V is supply; when off, dI is negative, and V is the output voltage (which is negative, from the perspective of the inductor, mind). If dI_rising is more than dI_falling, current keeps ramping up. If you don't control current in some way, just setting any random PWM value from the controller, current could rise to dangerous levels. What's worse, if you accidentally allow setting PWM = 100%, switching never occurs, and no voltage is ever delivered to the secondary side! Easy enough to set a hard limit so PWM% never goes too high, but that won't prevent current from ramping up under certain conditions. So what to do? Measure current, and control that first. Then control the current setpoint with a second error amplifier, to regulate output voltage. That PSoC might have a fast enough ADC to do this (you want to measure current about as often as it's switching), or it may have a comparator (also check that it's fast, under a microsecond say) that can be linked to the PWM timer, or output pin. These would be used for average current control, or peak current control, respectively. Typically, you'll choose average current control for higher power converters, with larger inductors that are cheaper, but can't handle as much ripple current (say, dI < 0.2 * I, that is, the ripple fraction at full current output is 20% or less). The lower ripple current also means some savings on capacitors. Conversely, peak current mode control is good for smaller converters, say under 100W, where the economy of scale hasn't kicked in yet, and capacitors and inductors can handle relatively high ripple currents. Both of these controls are easy enough to do without an SoC at all: the UC3842 family is the go-to peak current controller, and TL494 can be used for average current control, or you can build one from op-amps and comparators: https://www.eevblog.com/forum/projects/building-a-simple-switching-circuit/msg1252706/#msg1252706 Indeed, a PSoC is likely to just get in the way, as it brings a whole mess of software issues along with it, any one slip-up which can explode the switching transistor! It's not that a tightly integrated software control isn't possible, it's just that it's a whole hell of a lot harder to pull off, guaranteeing that it will behave itself! If digital control is needed, better to use DACs and ADCs to control and monitor an analog circuit, which is able to run, independently and safely, even if the updates stop, and which can never be forced into a dangerous condition over the range of settings it can be given. Tim |
| David Hess:
--- Quote from: essele on February 01, 2019, 10:14:39 am ---So my approach was to create a 50kHz spwm signal, and then output this on one output for the first half of the cycle, and then another output for the second half, so both sides of the H-Bridge are working identically just dealing with the different parts of the cycle ... and this all works fine. However a number of the solutions I've found online (and an inverter I've been tearing down) seem to separate the switching from the 50Hz line frequency, with one half of the bridge doing the 50kHz and the other half simply doing the line frequency. (Q1) Is there a reason for this? It feels better to me to keep things symmetrical? --- End quote --- If I am understanding your correctly, then I think this has to do with economics. It is just less expensive to do it that way and a lot of effort is made in reducing cost in commercial designs. --- Quote ---(Q2) If I have a 50kHz SPWM output, should I filter this before the step-up transformer, or can I feed this into the transformer without any problems? If I do this presumably I can put the filter on the output, which is a lot less current to worry about? --- End quote --- The line frequency transformer has way to much loss at high frequencies so the switching frequency must be filtered out. --- Quote ---(Q3) Is this all wrong, should I just be switching 350VDC and be done with it? (And presumably then just use a 1:1 isolation transformer to ensure isolation?) --- End quote --- This is how it is commonly done to limit the current requirements of the output filter. Note that galvanic isolation could be provided at an earlier stage (you suggest this below) so no heavy low frequency 1:1 isolation transformer is required. --- Quote ---(Q4) If I stick with the unfiltered 50kHz signal, could I used a smaller high frequency transformer to get the isolation so I can save on space/cost etc? (like this one? https://www.mouser.co.uk/datasheet/2/445/749196540-539789.pdf) --- End quote --- Absolutely and this is the way to go for a good ratio of power to weight because a heavy low frequency isolation transformer is not required. |
| Yansi:
Q1) Symmetrical means less distortion. If you leave half the bridge at 50Hz and other half at 50kHz, you need to transition voltage at the 50kHz side instantaneously from DC+ to DC-. The waveform at the 50kHz side will look like "U∩U∩U∩U∩", resulting at output with SEVERE crossover distortion. Q2) David Hess - not really true. Most if not all modern UPS modulate sinewave to the iron core transformer directly. The flux in that transformer contains only very little 50kHz component, there is no significant loss in the core. The switching gets filtered by the inductance of the transformer itself. It is exactly the same as happens within an induction motor on a VFD. (yes, you need then a secondary side filter, but that is inexpensive). Q4) You can only use such transformer to get DC/DC isolation. You would then need another chopping stage to obtain 50Hz after that transformer |
| Navigation |
| Message Index |