The switching transistor in the diagram is a series pass element (as is a linear regulator
Actually not - the transistor/zener is a switching element.
There is a small added power loss across the zener, base resistor and transistor itself, but insignificant compared to the alternative of burning it all off as heat in the regulator.
Unless I am misreading the schematic -- you put a voltage and current into the "switching circuit" and a voltage and current come out. Since the output current and the input current
is the same (ignoring the zener current) the power dissipated is (Vin - Vout) * Iout.
The more power you dissipate in switching and regulator circuits the lower your
efficiency. Raise your input voltage and your efficiency drops.
(* jcl *)
The issue here is efficiency across a wide input voltage range. With a standard series regulator efficency drops proportional to load current multiplied by input/output voltage difference. As long as you control both the input and output voltage, the difference can be kept at a minimum and a series regulator is then the best option.
With a swicthing regulator, power loss is proportional to load current only and virtually unaffected by input voltage level in comparison to a linear regulator. This is a better option when you need to reduce input voltage significantly (e.g. 10:1 as was the case here).
I may be forgetting the original question 
IIRC the original question had a 10:1 input range requirement. With that requirement there is no efficient series pass solution. The NPN transistor shown in the "switching circuit" is a series element since the current that goes in comes out (less the
zener current).
(* jcl *)
You seem to forget that the voltage drop across the transistor (Vce or Vsd for FET's) is fixed and small.
Vin/Vout in this context is the supply voltage versus the regulated output voltage.
In your picture you show 12V in (collector) and 5V out (emitter) which gives a Vce of 7V.
Are we talking about the same circuit?
(* jcl *)
Yes - this is the schematic, but you need to read my post as well to put it into the proper context.
Vce is not (12-5). You will find this value in the datasheet for the transistor of choice. Typically this will be around 0.2V for an NPN. For a FET there is a very small series resistance and likewise a very small voltage drop.
Vce is not 0.2V for a transistor. Vce(sat) is 0.2V. You need to saturate the transistor
to get 0.2V. You can also use the transistor in a linear mode (essentially as a programmable resistance). In a linear mode (which is the way linear regulators work) the current through the transistor is varied to achieve regulation. Changing the current
changes the Vce.
In your post you state the input source is 9-50V (collector) and the output is 7V (emmitter). This yields a Vce range to 2V to 43V.
If you have a 50V source that is low impedance and then saturate that transistor
you will first get a very large current as you charge whatever capacitance is on the
emitter and then after the capacitance is charged you will get 50V - 0.2V not 7V.
(* jcl *)
If we force the transistor off, Vce will climb to Vin. When the transistor is conducting, Vce will drop to Vce(sat) and this is the basis for the efficiency comparison.
In contrast, the linear regulator by itself will have to dispose of the full (Vin-Vout)*Iload as heat. With the switching supply, input to the linear regulator can be maintained close to its minimum dropout voltage and reduce power loss across the regulator to Vdrop (plus a safety margin) *Iload.
For the series regulator by itself we have efficiency:
e=VoutIload / VinIload
For the switching supply we have:
e=VoutIload / (Vce-satIload + (Vout+Vdrop)*Iload )
For the 10:1 case, effiiciency of the linear supply is less than 10% whereas the swicthing supply may get as high as 80% depending on the transistor and regulator we choose.
When the transistor is saturated what is the Vin (which is Vc) what is the Vout (which is Ve)? By definition Vce = Vc - Ve.
If the transistor is saturated and Vc is 50V and Ve = 7V then your circuit is not
functioning since one equation gives Vce of 43V and the transistor should be
at Vce(sat).
(* jcl *)
When the transistor is saturated, current (Ice) is flowing freely and then voltage across ce drops to Vsat. When this happends, Vin (voltage output from the power source) will drop as Ice increases. This is similar to short-circuiting a battery. A power source will maintain its power output (V*I), but at the expense of V.
The capacitor across the regulator input and ground will maintain Ve at the level we set with the zener. When the transistor shuts off, the capacitor will source the regulator and discharge through the load until voltage again drops below the zener level. At this time, the transistor switches back on and the cycle repeats.
Ice is AC and its RMS value will be roughly equal to Iload which is DC.
BenF: your circuit
is NOT a "switching regulator"; it's a traditional zener/transistor linear regulator and has all the power dissipation problems of any other linear regulator. See Linear regulator - Wikipedia, the section labeled "Simple Series Regulator."
A switching regulator needs some sort of energy storage device (inductor or capacitor) to store energy, which is then switched from input to output by the control circuit.
Here's another good switching power supply that can take in up to 60V and gives out either 5V or 6V, up to 2.5A.
http://www.dimensionengineering.com/VHVBEC.htm
If you're not in the US, the one Mike posted on page 1 may be a better option to save a few bux in shipping.
westfw wrote:
is NOT a "switching regulator";
The proposal was using the transistor/zener as a pre-stage to a standard linear regulator with a capacitor on its input side - not as a regulator by itself.
I have a healthy respect for those who master the art of designing good power supplies and trust the products they bring to market will most likely serve us well (including provisions for safety) for what they were designed to do. I am by no means an expert in this area.
The switching regulators come at a premium price however and possibly with features/packaging we can do without. Hence I left the question open as to whether this added transistor/zener pre-stage would suffice.
In the end, the zener/transistor pre-regulator scheme will need the same total size of heatsink as a plain linear regulator chip, though it might be easier to find a power transistor with the appropriate power dissipation characteristics.
Still, real switching regulators are a good thing to look at for power supplies that you want to handle high differences of input and output voltages. And they've gotten rather cheap. The (very common) MC34063A runs about $0.30 from digikey and claims to handle an input voltage of up to 40V. This is the chip used in most of the automotive "cell phone" power supplies, so it's pretty easy to find a sample circuit to test to destruction. I don't know how difficult it is to make it into a complete supply with a very wide input range; a few moments with one of the online regulator calculators might be useful.
Most LM317 datasheets include a sample circuit for a switching regulator, and another for a "tracking pre-regulator", but I don't think I've seen them combined. And the circuit tends to use large inductors and low freqencies (by today's standards.)
In the end, the zener/transistor pre-regulator scheme will need the same total size of heatsink as a plain linear regulator chip
How do you support that?
Another option would be the TI modules -- PTN78000W. These are in DIP packages,
7V-36V input, 1.5A and run at 95% efficiency. They are around ($13(1) from Mouser) and just require two external caps.
Not quite the 10:1 range the OP was requesting.
(* jcl *)
Here's another good switching power supply that can take in up to 60V and gives out either 5V or 6V, up to 2.5A.
VHVBEC Very High Voltage switch mode BEC powers multiple digital servos from high voltage batteries
Thanks, koyaanisqatsi. However, I want to make the board, not buy an existing BEC. I want to combine BEC and Arduino data logging capability on one device. I already have a 180A current sensing prototype to try out.
I'm learning a lot (and eating some popcorn) watching this thread grow! Keep up the discussions, I am reading it all. I hope we come closer to a solution that won't burn the house down.
IIRC the original request was 5-50V in and 3.3V out at low currents. Is that correct?
How much current do you require? In subsequent posts I saw mention of amps.
If you require low currents then a linear regulator solution may be possible.
Are you willing to use SMD components? State-of-the-art components are almost never introduced in TH packages.
(* jcl *)
an efficient but safe voltage regulator that can take anything from 5~50VDC, and supply clean 3.3VDC to a basic low-current Arduino project
Linear regulators are pretty easy for the DIYer. But I have been studying switching topologies for about the last 18 months and they are pretty difficult to get right. The basic principles are simple to understand, but implementing them with a clean output is a little tougher than one might realize. The switcher is basically a voltage conversion via PWM into an inductor. But the best efficiency is achieved at very high frequencies, which requires some serious EMI considerations. And the output is highly dependent on the load impedance. There are some switcher chips that only require a few external components. But I have only found these in SMD packages.
You can easily create a rudimentary switcher, but the output will not be very stable. I built something like this for my bike alternator that doesn't regulate well, but it does keep the output in a range that another, real regulator can use, regardless of the alternator input voltage. It's a switched capacitor 'regulator' that uses zener diodes to define the highest and lowest voltage acceptable. It switches parallel FETs to handle pretty much whatever current load you need.
I have a hand-drawn schematic that I can post this evening once I can snap a photo of it.
Switched capacitor power supplies are a controversial thing since dumping voltage into a capacitor with no series resistance (to speak of) is hard on the components and can dissipate a tremendous amount of energy. But in the case of an alternator input, it seems to work well and is pretty efficient because the output voltage of the alternator drops as soon as the FET turns on. In testing I found less heat is generated when the switching happens at lower frequencies like 1Hz.