Hi, I want to build a transformerless power supply. (I know it is dangerous, do not touch when powered etc.) For guidance and tips I was reading the included app note from Microchip. However their calculations do not agree with my results. There is what they claim:
For simplicity assume R1 = 0 and V_Z = 0 (both are negligible compared to capacitive reactance and V_RMS). Substituting V_PEAK = sqrt(2) * V_RMS this reduces Equation 4 to
I_IN = V_PEAK * pi * f * C1.
I don't like the AC current calculations because it confuses me so I took another approach. Voltage over C_1 is changing between -V_PEAK to +V_PEAK. In one half-cycle the current flows in forward direction of D1 not providing any current to possible load at output. Only on the other half the load is powered with any excess current shunted by D1. To get maximum possible average current for the load we can use "basic capacitor equation":
\delta V = (I. \delta t) / C
that can be transformed to
I = (V_max - V-min) * f * C.
So my I_IN = 2 * V_PEAK * f * C1 which is less than the Microchip's. Why?
Also at Figure 12 they claim a bridge rectification should provide 141% more current (sqrt(2) times more) than the simple half wave rectification. Why? If the current is provided at both half waves it should be 200%, not only 141%.
Of course I believe I am right and they are wrong but OTOH they are EEs and I am not. Maybe I am missing something fundamental?
EDIT: after rereading and rethinking I think they are calculating RMS current. While RMS current is important for calculating power dissipation of a resistor it is an irrelevant value for calculating dissipation in the diodes or available current for the load. For this a plain average current is needed AFAIK.
The circuit shown is ridiculous! All the series components - resistor, capacitor and diode - should be in the live line, not the neutral. Of course, this still depends on correct identification of the neutral, but allows some slight enhancement of safety.
Paul__B:
The circuit shown is ridiculous! All the series components - resistor, capacitor and diode - should be in the live line, not the neutral. Of course, this still depends on correct identification of the neutral, but allows some slight enhancement of safety.
I debated making a similar comment but decided not to. While I agree with you, I don't think it makes any meaningful difference to the overall safety, I think from a safety point of view the circuit has to be regarded as everything connected to live, even if it's not. If the neutral gets disconnected for any reason then everything is connected to live anyway. I would not trust my life to the idea that a component connected to neutral is somehow safe to touch.
Of course the circuit is only rudimentary "proof of concept". But I believe the equations are wrong!
And I have another question: is it correct to add capacitive reactance and resistance as real numbers? Shouldn't the reactance be considered as an imaginary value and the calculations done with complex numbers?
I want "how much current the load may consume" value. If C2 is reasonably large this should be equal to average current through D2. Average, not RMS. It should be half of the average value of (absolute value of) current through C1 - the other half is "wasted" in D1 forward current.
What I don't understand is why they do all the RMS magic (except for calculating R1 power dissipation). They get clearly wrong results (at least the more I am thinking about it I am more convinced about it).
Of course the maximum voltage over the capacitor is V_PEAK. But minimum voltage is -V_PEAK. Change of voltage in one half period is from -V_PEAK to +V_PEAK so it is equal the peak-to-peak voltage.
There is also this. It is a 5v voltage regulator (10mA) that can run from an up to 450 volt input. It is tiny (sot-89) but also need a big capacitor. https://www.onsemi.com/pub/Collateral/NCP785A-D.PDF
This is talked on the EEV video mentioned by @TomGeorge.
Manufacturer's log has been removed in that image; it's Hilink. They're selling directly as well, you can find them on Taobao and probably also on Aliexpress.
I've seen a rather comprehensive review some time ago of one of their parts, where the reviewer actually opened it up. Pretty decent design. This reviewer also reported that when used at full power it gets quite hot (not surprisingly) but not excessively so.
I used them recently for a project that needed AC outputs, so had to bring AC to the PCB anyway. Then better this part than external adapter what I normally do.
wvmarle:
Manufacturer's log has been removed in that image; it's Hilink. They're selling directly as well, you can find them on Taobao and probably also on Aliexpress.
I've seen a rather comprehensive review some time ago of one of their parts, where the reviewer actually opened it up. Pretty decent design. This reviewer also reported that when used at full power it gets quite hot (not surprisingly) but not excessively so.
I used them recently for a project that needed AC outputs, so had to bring AC to the PCB anyway. Then better this part than external adapter what I normally do.
I didn't notice the missing logo othewise I'd have posted a different image. I supplied also a retailer product link in my post above to TXHANG (AKA Alice).
The (or a ) teardown/review is here: Performance test of Power Mains to 5V 0.6A Hi-Link HLK-PM01 UK
I am aware there is plenty of other solutions. But I want to try to build the transformerless power supply for various reasons:
I want to make it on my own
I want to build it to get experience and learn how it works (and prove I understand it by doing so)
To get connection to the mains to be able to use the mains frequency as a timebase
Luckily Google was able to provide another application note which is included. Authors confirm the Microchip engineers are utterly wrong and provides other equations which agree with mine. They also provides more detailed insight into how the TPS work and possible configurations. (The Microchip application note prom OP has some interesting safety related pictures but the text describing them is dubious.)
