General > General Technical Chat
Looking at a small SMPS that doesn't have a controller?!
systemloc:
I'm learning about SMPS in detail with a view to build one. I've got a bunch of small PSU's I have pulled out of various junk items, and I'm looking one over on the bench. Don't remember what it came out of. Yokogawa is the manufacturer on the PCB. Very small and simple, has the usual bridge rectifier, common mode choke, and rated cap input. A single power MOSFET and a single transformer, and two output rails, 15V and 36V. There's only one IC on the whole board, looks like an ST LM393 (Marked ST logo and 393 on it). One of the outputs of the comparator has a switching waveform about 270 kHz driving an optocoupler. The optocoupler leads to a small AF 400mW PNP transistor which appears to be driving the MOSFET. There are no other ICs, and there is one other small surface mount transistor. I haven't figured out much of the glue passive components. I'm totally at a loss as to what is generating the switching. Is it possible that there's a simple LC oscillator feeding into the comparator, with a zener going in as well acting as a simple PWM? I also haven't begun to work out the startup logic, but just looking at the above got me confused, as I've not seen anything without a controller.
T3sl4co1l:
Most often these are some variation on the classic blocking oscillator.
It's not a class C self-excited oscillator, at least not usually. It can approach that, when the load has relatively high capacitance (like for a HV converter), and when operating under low bias. More often, it's saturating fully into class D operation: the transistor latches on (due to positive feedback from a drive winding), and so load current grows over time (due to V = L dI/dt, the transformer primary has significant inductance), and eventually turn-off occurs either due to gain dropping off at high current, or something else forcing it.
When turn-off is caused by saturation of an inductor or transformer, and operation is balanced (push-pull), it's a Royer converter.
When half wave (single transistor), it can turn itself off due to limited hFE of a BJT, or with the help of a current-limiting transistor (BJT or MOSFET). (MOSFETs typically can't be used for self-turn-off because they are capable of much higher peak currents than BJTs for the same efficiency (in terms of Vce(sat) or Rds(on) at same load current), while paying no price in terms of drive current.)
Typically, bias current is controlled, varying the time between pulses, and thus average power output. This can be done through a phototransistor type opto. Sometimes, regulation is done by sensing the negative peak voltage on the feedback winding itself, which can use just a zener (and Rs and Cs), saving the opto, though this also gives poorer regulation (the output flyback voltage doesn't quite match the feedback voltage for several reasons).
For your case, the opto might be something more than a phototransistor type. Not sure what a 393 should be doing there -- phototransistors are nowhere near fast enough to communicate gate drive signals. A 393 can be used as a basic switching controller (wired up cleverly; the main downside is the ton of resistors needed to do that wiring, hence proper controllers are preferable), so that's a possibility if so. Or it could just be used as hysteretic regulator, or fault protection. (A common sight in ATX PSUs is a LM339 for overvoltage protection on all the outputs.)
There should also be a voltage reference somewhere, whether a zener diode, or TL431 or a relative. Perhaps this is the small SMT, or the aforementioned current limiting / turn-off transistor is it.
Tim
amyk:
I think you have what is known as a "ringing choke converter". Oddly enough, the Wikipedia only has a Chinese page on it: https://zh.wikipedia.org/wiki/%E6%8C%AF%E8%8D%A1%E7%BA%BF%E5%9C%88%E5%8F%98%E6%8D%A2%E5%99%A8
Here's a detailed explanation in English: https://web.archive.org/web/20110709011600/http://www.deltartp.com/dpel/dpelconferencepapers/S19P6.pdf
systemloc:
Wow, thanks a lot. This was extremely helpful. The IEEE paper pretty much looks like what I've got mapped out so far. I was not aware of this topology. One of the links led to this example from Jim Williams:
https://www.analog.com/media/en/technical-documentation/application-notes/an118fb.pdf
I've also read a lot of Jim Williams' work, and I hadn't seen that one.
T3sl4co1l:
Here's an old example I once made,
This actually works opposite the usual way; the transistor runs under very light bias (1M) and is increased by the opto. Notice the 0.1+1N914 makes a negative supply, which the opto can pull up, shunting current into the base drive circuit, thus increasing power output. One advantage can be lower idle current (not having to drive the opto at full throttle); downsides are poor startup (it's not able to start into a resistive load of rated power; as I recall it was fairly capable though, like, starting up into 2-3 times that resistance) and needing a minimum load to keep the output from floating too high (which is afforded here by the fairly high current consumption of the feedback network..).
A more conventional one, and more or less shows the scheme as it's quite simple:
Some notes: tantalum aren't really great for this sort of application; they were tolerable for this as the current isn't terribly high (100s mA). I only otherwise had electrolytics, at the time, which weren't suitable. Nowdays, ceramics are widely available and perfectly suited to it. D1 is some salvaged diode with seemingly better performance than 1N914/4148; it's important to note as the circuit is sensitive to its capacitance and reverse recovery. Also the powdered iron inductor core isn't suitable, a lower-mu grade or ferrite is preferable, but it at least doesn't get painfully hot.
This uses a somewhat different tack:
The tapped feedback winding supplies immediate positive feedback via the capacitor, and delayed negative feedback via the RC time constant C3*R2. It's an astable multivibrator, in the usual way (just as you'd wire positive and delayed negative feedback paths around a comparator), but done at higher frequency and with a transformer and a single transistor, no inversion logic required. The bias voltage then varies PWM and frequency, allowing control. Downside is, because it's running fairly fast, and it's driving the power transistor, it takes quite a bit of power to do (R2 is 0.5W I think). T1 is ungapped ferrite BTW.
Tim
Navigation
[0] Message Index
[#] Next page
Go to full version