1
\$\begingroup\$

I am aware this may be a silly question, but I am constantly told by everyone I met more experienced than me in engineering that the only silly questions are the one not asked, so here I am..

About a year ago I learnt at university the basics of how BJTs work, their I-V characteristics, some of their limits and praises. However, when encountering new circuits, I'm constantly struggling to understand at first sight what is actually happening in the circuit. That is, I cannot understand in a few looks what the circuit is doing.

As an example, take an RC or a CR filter, you can see it as building block filtering incoming signals and outputting filtered signals, that is very intuitive after having seen their transfer function and the physics behind it. Same for operational amplifiers, within limits, you can think of them as building blocks outputting a modified version of their input.

I can not do the same thing with transistor, I usually have to start with an hypothesis (ex. for an NPN, that the base is forward biased), check if it holds and then go on from there, as I have been taught in university. This is fine, but it takes time and may not be that easy for non trivial circuits. I wonder if practitioners and experts have more intuitive/quick methods to understand what a BJT is doing (or, more in general if you feel like it, what a non linear component is doing).

For a pratical comparison, take an astable multivibrator circuit. I can get my head around it reasonably quickly (also because I know what it should be doing), but when we step up to a circuit like the Royer oscillator you see here, it takes me some more time to work it out. How would a professional go to understand its workings? enter image description here

As a final note, I would like to add: this question comes partly from the fact that I know I can plug the circuit in a simulator such as LTspice, but I want to know what is happening (or should be happening) in order to be able to tweak the circuit, and partly from the hope that practitioners may have some commonly accepted and used rules of thumb.

\$\endgroup\$
  • 1
    \$\begingroup\$ To understand the above circuit, you have to first understand W4 is part of the transformer T1. It has an AC voltage which is phase-coupled to the voltage at W1. Does this already help? You can see the transistors as controlled switches here, the circuit is simple enough to allow this simplification. \$\endgroup\$ – Janka Dec 26 '16 at 15:47
3
\$\begingroup\$
the only silly questions are the one not asked

That is absolutely false. Of course there are stupid and silly questions. A question is stupid if you should have known better, and silly if it is asked in the wrong place, among other reasons.

The people telling you this are pathetic do-gooders that are actually doing you a disservice. Perhaps they are tolerant of you asking them stupid questions, but the world isn't. The sooner you learn that the better off you will be. Of course you are judged by the quality of question you ask, whether people tell this to your face or not.

That said, what you ask is not a stupid question. It is, however, not a great fit for this site since it's so broad. This question will likely get closed due to being too broad, so I'll be brief.

You can get a lot of analisys done with a simple model of a BJT. The "trick", if there even is one, is to start with a simple model of the circuit elements just to get a feel for what the circuit is doing. Then you can follow up with quantatative analisys if you need the details.

Start by thinking of BJTs as having about 700 mV B-E drop, and that the collector current can be many times the base current. In fact, it can even be helpful to consider the transistors to have infinite gain for the rough first mental pass of what a circuit is doing. Good BJT circuits work with each BJT having from some minimum gain to infinite gain. Not all circuits are "good" in this sense, but it is a useful start.

Note that when the gain (ratio of possible collector current to base current) is infinite, the base current is zero. This simplifies more things. It won't usually get you the right bias point, but details like bias points are for quantitative analisys later. First you want to see what the circuit is doing, then you can figure out over what voltage range, or whatever, it actually works at.

While this method does work most of the time, it isn't foolproof. Sometimes what the circuit does depends on actual component values, although usually not very detailed.

After you do this enough times, you'll build up some experience and the intuition that comes with it. Just keep doing it. There is no shortcut for gaining experience.

Added

I see the question got edited while I was writing the answer above. You are now asking about a specific circuit as example:

Here's how I approach this circuit to try to undestand it:

Hmm. At first glance looks like some kind of power oscillator. Each transistor drives one end of a center-tapped primary, which is connected to power. That means each transistor must turn on alternately. Let's see if that's true, and if so, how.

R1 and R2 provide forward bias for each transistor. There is no feedback here. Ah, there it is. There's a small winding of the transformer connected between the bases. At this level, I'll take it for granted that it's wired with the right polarity to cause oscillation.

This little winding causes a offset voltage between the two bases so that one transitor is driven on and the other driven off. That should cause the transformer to be driven even harder in the direction to switch the transistors the way they are.

This would cause the circuit to latch if the feedback was DC, but since it's thru a transformer, it can't be DC. Eventually the transformer is driven as hard as it can be in one direction, and the drive doesn't change anymore. This cause the little winding to stop driving the one transistor more than the other. The derivative of this feeds back, causing the state to eventually swing all the way to full drive the other way, etc.

C1 is a bit bothersome. Unless it's rather small and just to reduce emissions, it looks like it shouldn't be there. This is where having the actual value would help. If it's 100 pF, then OK, just trying to reduce RFI. If it's 100 µF, then either I'm not getting its purpose or this is just another bad design found on the internet. Maybe there is some resonsance with the transformer going on. Without real values, it's hard to say.

The only other wrinkle is that the switch is labeled "PWM", and L1 with D1 look like switching is intended. Without D1, I would have thought L1 is just a filter working with C3 to reduce conducted noise. However, L1 seems to work with D1 a bit like a boost converter when the switch opens.

Overall, it looks like PWM is applied to the input power of this circuit to modulate the output voltage. That appears to be the intent, and would be helped by a capacitor to ground on the right side of L1. Perhaps the oscillation frequency is much higher than the PWM frequency, and this is intended to produce bursts of power when the switch is closed.

Again, don't assume everything you find on the internet is well designed. Lack of cap to ground on the center tap is a bit suspicious here.

In any case, this shows how you can look at a circuit like this. Note the high level of the transistors being "on" or "off", and the details of their biasing, actual currents, etc, don't matter at this level. Even if you eventually need a detail analisys, it's useful to start with a high level conceptual analisys like this first.

\$\endgroup\$
  • \$\begingroup\$ I thought they meant that about "informed" technical questions :) I can see the point you are making though, of course there's a time and a place for most things :). Anyway, I'm not looking for "shortcuts", but for rules of thumb as the one you suggested. For most components, from the theory you can build yourself guidelines to understand their behaviour, but I find it harder to do that for transistors. Experience will do the "trick" as you said, I agree. As for now, let's experiment. Thanks for your answer. \$\endgroup\$ – mickkk Dec 26 '16 at 15:57
  • 1
    \$\begingroup\$ This circuit shown here is just a simplified version of a Resonant Royer oscillator widely used for backlights when LCD was using CCFL as a backlight plus a buck convertor for a dimming purpose. And this circuit comes from this note diodes.com/_files/product_app_note_pdfs/zetex/an14.pdf \$\endgroup\$ – G36 Dec 26 '16 at 17:10
  • \$\begingroup\$ @Olin Lathrop > a capacitor to ground on the right side of L1... No, a capacitor there would interfere with operation. It's an ancient RF design, and a good example of subtle functions of certain components! (Heh, sometimes power-RF seems like a Black Art.) L1 is not a supply filter, it's a hi-Z component which floats the CT connection for AC. Vacuum-tube induction furnaces and diathermy oscillators use the same trick, providing a large air-core choke for L1. Similar to inductors on magnetron filament connections. Search other forums for discussion of Baxandall Oscillator class-D. \$\endgroup\$ – wbeaty Dec 31 '16 at 6:29
  • \$\begingroup\$ @wbeaty: This all depends on quantitative information we don't have, like the oscillation frequency, and whether that's much higher than the PWM frequency. If the oscillation frequency is really RF, then I see your point. The PWM is then switching the whole thing on and off, and L1 is high Z at the RF frequency. If not, then L1 is more part of the buck converter driven by the PWM, and a cap is needed to filter PWM pulses and keep the "supply" to the power oscillator stable. \$\endgroup\$ – Olin Lathrop Dec 31 '16 at 13:34
  • \$\begingroup\$ @Olin Lathrop DOH! I totally missed that "pwm." So, very good point. You're right, the L1 "constant-current-izer" which is present in "C1-having" resonant inverters, is also doing double-duty as a lamp-dimming buck converter via D1 (L1 provides the "smoothing.") L1 does triple-duty as lowpass to keep RF out of DC suppl. A beginner has little chance of full grasp (I missed that last bit myself!) so, must find the long online discussions of that particular circuit. PS, for very tiny xfrmrs in laptops/tablets/phones, expect "rf" to be 70KHz or higher. Heh, and C2 is a fluorescent lamp ballast! \$\endgroup\$ – wbeaty Jan 1 '17 at 7:24
1
\$\begingroup\$

I think it would be a mistake to expect a general answer, but the closest I can offer, if you have also studied software engineering, is to think in terms of "design patterns" (named in bold below).

With practice you can start to identify basic building blocks even when their components aren't all there - for example, an R-C circuit when there is no R component, but you know the previous stage has a moderate or high output impedance, which provides the R.

In this case, you can see elements of a multivibrator.

However the transformer should attract most attention : recognise that its primary is centre tapped and connected (as far as DC is concerned) to V+, and each end of the winding is connected to ground via an active device (BJT here).

This design pattern is about 85 years old, and known as a "push pull amplifier" where as you pull one end of the primary down to 0V, the other end (by autotransformer action) rises to twice V+. You can replace the BJTs with triode valves and it'll do the same thing. Note that you can view Q1,Q2 etc as part of either some sort of multivibrator, or as part of a push pull amplifier (operating in saturation), or as switches, which is why I think of them as design patterns rather than purely as circuit blocks.

Also note (the schematic doesn't make this very clear) the winding W4 connecting the transistor bases, is part of the same transformer, and the dots indicate it's in antiphase. So, as you turn Q1 on, the black dots are pulled low, and the white dots are pulled high - providing positive feedback to turn Q1 on harder, and turning Q2 completely off.

Note also, C1 connected directly across the primary - thus the primary is also an L-C resonant circuit, and the L and C determine the resonant frequency.

Beyond that, I'm a little curious myself how the PWM switch and L1 interact - without knowing the ratios of the L-C resonant frequency and the PWM frequency, I'm reluctant to comment on that aspect. But one thing's clear : disconnecting L1 generates inductive spikes which would destroy the switch if D1 wasn't there...

\$\endgroup\$
  • \$\begingroup\$ I'm a little currious Think of L1 as infinite inductance. It's not supposed to interact. This is a common technique in power RF, where the collector resistor is often a pure inductance, so a short circuit for DC, forming a constant-current power supply. The goal is to provide a constant DC supply to the transistor(s), while presenting the RF with an open circuit. For lots of discussion, search: Baxandall. Also called SSTC, solid state Tesla coil, or push-pull power oscillator, or Royer oscillator. (Also, D1 is lacking on most of these I've seen. It's on the wrong side for protecting!) \$\endgroup\$ – wbeaty Dec 26 '16 at 21:05
  • 1
    \$\begingroup\$ And in a single-ended design, a high impedance for L1 would be important. But to a first order approximation, the load here is approximately constant (as either Q1 or Q2 is conducting). So here, with the PWM switch, L1 is more akin to a buck converter (assuming it is relatively large). And as Olin notes, the lack of a decoupling cap at the centre tap here is ... odd. \$\endgroup\$ – Brian Drummond Dec 26 '16 at 21:13
  • 1
    \$\begingroup\$ In a Royer Oscillator the transformer inductances resonates with C1. One end of C1 is grounded through one of the transistors. At every half a cycle the transistors switch the opposite end of C1 to ground. So what will be observed at the collector of each transistor is a half cycle of oscillation with a full sine wave across the transformer. It has the virtue of a sine wave output and zero-volt switching of the transistors. L1 can be visioned as a constant current source that ideally loses no energy, it does double duty here as it also acts as the filter for the buck converter. \$\endgroup\$ – Kevin White Dec 26 '16 at 22:20
  • \$\begingroup\$ @Brian Drummond > L1 is more akin... No need to speculate about common oscillator types, instead read refs. This one is Baxandall, also called Royer, and L1 is discussed here: sophia-electronica.com/… We have bunches of these things here ($1 from All Electronics, cheap kilovolt source) and their L1 measures 700uH. \$\endgroup\$ – wbeaty Dec 29 '16 at 21:39

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.