Why is this simple Class B amplifier not working?

[colejohnson66]

I'm trying to amplify a voltage from -5~5 V up to -48~48V without using an op amp. The reason being: I need at least amplify a few dozen signals in my project, and the best 100V op amp I could find was the OPA454. However, that is ~$6 a pop in small volume (which is too expensive).

For my requirements, a little distortion is allowed as the voltages don't need to be exact. Also, I am aware of the dead zone, and want to do a Class AB ultimately, but I'd first like to understand how a Class B works. Using LTSpice, I have recreated a schematic I found online (I've attached it here), but when I simulate it, all I get at V(out) is V(in) minus the PN junction drop of ~0.7V. Why?

[alsetalokin4017]

Your original schematic does not have a "V3", but your LTSpice sim does. What happens when you lose the V3 from the sim and just connect this to ground as the original schematic shows?

[hamster_nz]

Isn't that schematic set up to amplify current, not voltage? (i.e. It is a power output stage)

[colejohnson66]

If I connect the collector of Q2 to ground like in the original schematic, V(in) is "clipped" and doesn't go below 0V.

[soldar]

You are not going to get +/- 48V out with a single sided 12 V supply.

If you have a single 12V supply the most you can get is +/- 6V if you use a DC blocking capacitor.

[colejohnson66]

What? I have a 48V supply I can use with some buck converters for ~45 and -45V (close enough to 48V). The problem I have is I have a DAC that outputs -5~5V and I need to amplify that.

[soldar]

I am going by the schematic you posted.

[colejohnson66]

Ah. Well, that's just a schematic I found online. I added a -48V voltage source at the collector of Q2 so it can go negative. However, I'm still confused with what you're saying. Are you saying that this circuit won't amplify a voltage up to the rails?

[edavid]

The schematic you posted has no voltage gain.  You would need to add a gain stage and feedback to make a complete amplifier.  Obviously you will need to use higher voltage parts as well.  Doug Self's power amplifier book will give you a good idea of what's involved (although it sounds like you would be able to simplify his circuits):

http://douglas-self.com/ampins/books/apad.htm

You might find that using HV op amps or audio power amp ICs (LM3886, TDA7294?) ends up being your best option after all.

[magic]

Probably the simplest solution is to use one of the bazillion circuits that boost a low voltage opamp's output to ±50V with a few 100V transistors.
Feedback signal is divided down back to ±5V so the IC doesn't need to handle high voltages. It can be LM324, TL084 or anything.

There was a million threads on that on the forum, maybe somebody can help with links or just look around.

[alsetalokin4017]

Ah, am I confused yet? Isn't Vout effectively looking at the voltage drop across the 10 ohm resistor?

[rstofer]

Look at Q1:  For it to conduct, the base has to get 0.7V above the OUTPUT.  It can only get to 6V so the output can only get to 6 - Vbe or around 5.3V

[schmitt trigger]

You have not mentioned what the output load will be.
One thing is to drive a 1000 ohm load, another is to drive a 4 ohm load @ +/-48 volts.
The output architecture increases significantly.

[Audioguru again]

Like many amplifiers, the power supply for these two transistors has a single polarity. Therefore the input must be DC biased at half the supply voltage so that the input and output of the transistors can swing up and down.
A speaker is damaged by the resulting continuous DC on it so an output coupling capacitor blocks the DC but passes the AC audio.

The transistors are simple "emitter followers" with a gain of 1 but minus two base emitter voltage drops. Lookup the voltage gain of an Emitter Follower in Google.

[colejohnson66]

Thanks for the suggestion @magic! I searched the interwebs and found AN-1593 ("Extending Op Amp Operating Range via Bootstrapping"). I recreated the schematic in LTspice using my -5~5V source and my -45~45V rails (48V from the PSU, but I'll lose some from inverting it (from what I've read)), and it seems to work! The V_cc and V_ee rails are "flying."

For parts, I just chose a random op amp from the list and set it to a gain of 9 (45/5=9), and two random P-channel and N-channel MOSFETs.

I only have one more problem now: the "flying rails" are not going all the way to the -45V and 45V rails. Instead, they only go to roughly -37~37V. This clips the output of the op amp to roughly -35~35V. Now, clipping at the rails is fine, if they were at the PSU rails, but they're not. How can I fix this?

[magic]

AN-1593 figure 1 is captioned "voltage follower" and I can believe that the whole appnote is about it. That is, unity gain.

Indeed your circuit will never work because the high voltage output is your opamp's output and it needs to swing ±48 while the opamp's inputs need to stay at ±5V. No way that both the output and the inputs will stay within the opamp's supply rails, whatever they are.

There are more complex bootstrapping circuits which work with gain, but I don't remember their names. That being said, their complexity probably approaches that of a fully discrete solution, so for completeness, here's one.

This is a fully discrete opamp which should work for this purpose. It's almost rail to rail. Output can source 1mA of current and sink enough to blow up Q3
Modifications are possible to change this behavior if you need.

[GerryR]

Isn't that schematic set up to amplify current, not voltage? (i.e. It is a power output stage)

This!!

[Psi]

I'm not sure of how important signal integrity/linearity is or what the actual data is,
but have you looked into opto isolators.
You can get them with high voltage transistors and they are definitely cheaper than $5

You might find they could work to get your  +/-5V signal up to +/- 50V with some tweaking and DC offsets.

[rstofer]

I only have one more problem now: the "flying rails" are not going all the way to the -45V and 45V rails. Instead, they only go to roughly -37~37V. This clips the output of the op amp to roughly -35~35V. Now, clipping at the rails is fine, if they were at the PSU rails, but they're not. How can I fix this?

Random Op Amps are unlikely to be 'rail-to-rail' on input and or output so the output can't get within a couple of volts of the rails.  Read the datasheet!

For the 741 op amp with +-15V supplies and a 2k load, it can't get within 5 volts of the rail (MIN) or 2 volts (TYP).  Look at "Output Voltage Swing".  A prudent design would use the 5V figure.  In fact, early analog computers used 741s and the anticipated voltage swing for input and output was +-10V.

http://www.ti.com/lit/ds/symlink/lm741.pdf  Page 5

Input Voltage Swing is similarly restricted, you can' put in more than +-12V (MIN) or +-13V (TYP).  Again, a prudent design would use the 12V figure.

When you buy rail-to-rail op amps, you need to realize that 'rail-to-rail' is a marketing term, not a reality.  There will still be limits, they just won't be as severe.

[magic]

Those MOSFETs aren't exactly rail to rail either

That circuit is hopless for the reason I already stated: the opamp can't be bootstrapped to the output voltage because its input needs to stay near ground potential.

A discrete voltage gain stage is needed, like Q3/Q5 on my schematic. There is no escape from that. An opamp, however, could replace Q1/Q2 although some additional trick would be required to translate its output down to the -49.3V potential of Q3 base. Which is why I said that it approaches complexity of a wholly discrete solution. Precision will be better with an opamp, though.

[colejohnson66]

I’ve been thinking about my requirements, and I don’t think I need a “rail-to-rail” op amp (yes, I know it’s just marketing). While my voltage rails will be -45 and 45V (my PSU only has one output, so I’m using buck converters to invert), I only need an output from -40~40V. As for the accuracy, it doesn’t need to be perfect. An output of 5V from the amplifier could have 250mV out and be fine.

I’ve been considering a discrete op amp based system, but I could never find a schematic simple enough. The one @magic posted seems promising, however, I have one question: I can’t find a 100V rated 100 pF capacitor on DigiKey that’s a good price (or any pF magnitude ones at that voltage). Does it need to be 100V rated, or could it be less? As for the transistors, I can easily find 100V surface mount transistors for pennies in the volume I’d need.

[magic]

One end is at 0.7V above the negative rail, the other may swing all the way up to the positive rail. You know what rating you need

I don't believe there are no 100V capacitors. Either MLCC or through hole ceramic disk should be found with ease at any distributor. Try other categories or just type "100pF 100V" into the search box.

And by the way, my circuit actually sucks I thought I'm being clever using that current mirror but I probably should have got a clue when SPICE told me the output current will be 1.5mA. That's because collector voltage on Q5 is much higher than on Q4. And another effect of that will be a serious temperature difference between them and even more current through Q5. I think it's risky business.

I recommend replacing Q4 with two series diodes or an LED and adding a resistor between Q5 emitter and the supply rail to set the current through Q5 and Q3 to 1~2mA or whatever you need. Mind power dissipation in Q5 and Q3, including when the output is at negative rail for prolonged time.

[colejohnson66]

Ah. I was searching in the wrong category for the capacitor. I found one for 3.5¢ or less in volume. And now that I know the difference between a power and a voltage amplifier, what if I added a power stage to the output of the “op amp”? Something like a darlington push-pull buffer?

Also, how do you measure the maximum output current with LTSpice?

[magic]

You need to add some realistic load to simulate the output stage properly. Then you will also know whether it's strong enough to drive your load.
To simulate short circuit protection, short the output to ground or the opposite supply rail with 1mΩ resistor and plot resistor current.

[Audioguru again]

Your first schematic has a +/- 48V supply and a 10 ohms load. If its output swings 90Vp-p then it is 31.8V RMS. Then the maximum undistorted power in the 10 ohms load is (31.8V squared)/10 ohms= 101W.
The power in each output transistor will be almost 40W which is impossible with surface mount transistors.