Author Topic: Class A MOSFET amp with rapidly varying voltage supply ?  (Read 1781 times)

0 Members and 1 Guest are viewing this topic.

Offline TheBaconWizardTopic starter

  • Regular Contributor
  • *
  • Posts: 53
  • Country: gb
Class A MOSFET amp with rapidly varying voltage supply ?
« on: July 04, 2018, 09:01:50 pm »
This is likely not the efficient way to do this, but I would like to consider it with the others..

I am providing a top-rail voltage to a MOSFET-based voltage amplifier and then a push-pull. The voltage is provided by a charge-pump. If I omit the final smoothing-cap, there will be significant variation in the voltage, at the 5Mhz i’m Using to drive it.

I actually WANT that in the final waveform, and while yes, I could use a summing op-amp to combine the 5Mhz signal with the 500khz signal the final amp will have on its base, i’d still Like to investigate this option if only for education.

Could I bias the MOSFET in such a way as to set the bias point at Q no matter what Vcc happens to be at the time? feedback resistor perhaps? A very carefully chosen bootstrapping arrangement? These latter 2 options are new to me btw, I am still finding my feet.
 

Offline Cerebus

  • Super Contributor
  • ***
  • Posts: 10576
  • Country: gb
Re: Class A MOSFET amp with rapidly varying voltage supply ?
« Reply #1 on: July 04, 2018, 09:39:43 pm »
1) I think you might have missed part of the explanation as
Quote
combine the 5Mhz signal with the 500khz signal
suddenly springs out with no introduction.

2) I take it you mean "push-pull" in the class-B/totem-pole sense, not the differential pair amplifier sense?
Anybody got a syringe I can use to squeeze the magic smoke back into this?
 

Offline TheBaconWizardTopic starter

  • Regular Contributor
  • *
  • Posts: 53
  • Country: gb
Re: Class A MOSFET amp with rapidly varying voltage supply ?
« Reply #2 on: July 05, 2018, 12:51:33 am »
One signal (5Mhz) is driving the voltage pump which provides the top rail voltage, with a huge 5Mhz ripple if we deliberately omit the final smoothing cap.

The mosfet amplifier stages are  intended to be supplied from that moving voltage. They will of course have something to amplify. That something happens to be a 500khz sine wave. But as far as I know that's irrelevant: the question is, can I bias mosfets such that the bias follows the top-rail's voltage and thus continue to perform as an amplifier according to an input signal of whatever kind. It would of course vary in voltage at 5Mhz as well as following the input waveform. This is all assuming the mosfets are fast enough.

Yes, it's class B for the final output.
« Last Edit: July 05, 2018, 12:59:36 am by TheBaconWizard »
 

Offline Cerebus

  • Super Contributor
  • ***
  • Posts: 10576
  • Country: gb
Re: Class A MOSFET amp with rapidly varying voltage supply ?
« Reply #3 on: July 05, 2018, 01:19:24 am »
Hmm, definitely beard scratching time. Normally one tries to design in as much PSRR to an amplifier as possible, so we're looking at chucking out all the usual design habits.

So, some more questions. What does the rail voltage look like? DC + some 5MHz signal on it, obviously, but how much of each? Do you have a sketch or something of what you're hoping your final output waveform looks like?

You'll have to forgo explicit and implicit negative feedback (e.g. degeneration) because that will raise the PSRR, yet stabilising bias in the face of wildly fluctuating supply voltages would seem to demand lots of negative feedback somewhere along the road. Let's kick the ball around for a bit, by all means, but I have a suspicion that a practical solution is going to be elusive. Doesn't hurt to try, it's always good to try and figure out novel solutions. (Plus a bit of proper analogue electronics is always welcome in the face of the preponderance of bit-twiddlers around here.  :))

Early thoughts are perhaps something more in the way of a common base/gate topology might be more do-able.
Anybody got a syringe I can use to squeeze the magic smoke back into this?
 

Offline JS

  • Frequent Contributor
  • **
  • Posts: 947
  • Country: ar
Re: Class A MOSFET amp with rapidly varying voltage supply ?
« Reply #4 on: July 05, 2018, 04:37:57 am »
The two inout signals, at least an examplenof them, and the desired output should be clear, if you can grab a napkin, a lipstick and a camera and post it here would do.

I imagine a pretty known rail waveform, I imagine dc+squarewave, at 5MHz the bandwidth would probably be noticeably limited but let's say square wave. So its an elevated square, with base voltage half the value of the peaks. This is my first guess.
The second signal, the input signal let's call it a 500kHz sine, but would be whatever 500kHz BW limited signal.

Now, you want the sine to be chopped by the square, but keep the relation, not just clipped when the square is low and sine is high.

Is al that the case?

Then you need your output stage to behave as a constant resistance, variable with the sinewave input signal. The more controllable way would be to add the rail voltage in some way to the input, I'm thinking something like a conatant resistance dummy load right now, recently discussed around here. I don't know about that freq range with this loop but I guess a known load would help a lot compared to a test equipment where load could be almost anything, like an indian cow on new year...

If you want a more open loop approach and still get some kind of predictable, precise output it can get really tricky, as would probably be dependent on mosfet parameter, changing a lot from device to device, temperature and moon alignment.

JS

If I don't know how it works, I prefer not to turn it on.
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 17426
  • Country: us
  • DavidH
Re: Class A MOSFET amp with rapidly varying voltage supply ?
« Reply #5 on: July 05, 2018, 12:10:11 pm »
Could I bias the MOSFET in such a way as to set the bias point at Q no matter what Vcc happens to be at the time? feedback resistor perhaps? A very carefully chosen bootstrapping arrangement? These latter 2 options are new to me btw, I am still finding my feet.

Sure this can be done within the limits of the power supply rejection of the bias circuit.

For instance connect a shunt reference to the positive supply and draw a current to the negative supply.  Or since a MOSFET has a high input impedance at DC, draw a constant current to the negative supply through a resistor connected to the positive supply.  Now the DC bias can be adjusted from the negative supply by adjusting the current.
 

Offline C

  • Super Contributor
  • ***
  • Posts: 1346
  • Country: us
Re: Class A MOSFET amp with rapidly varying voltage supply ?
« Reply #6 on: July 05, 2018, 01:25:19 pm »
As David Hess stated bias is not a problem.

Where you may have problems is with how you look at a circuit.

First remove Ground from your mind, this is a mess.

You have a V+ & a V-. When looked at it like this, when you have positive ripple on one, you have negative ripple on the other, both the same size. The two can cancel out if you do not mess this up.

One way to keep this balance is to always have one signal going positive while a second is going negative. In place of one class A amp, you use two with second having inverted signal.
This can also make other things easer.
With just one Class A amp bjt transistors you need to have matching NPN & PNP transistors. When you switch to a Dual amp, you want two matching NPN's and two PNP's. The output of the two amps might not look as great separate but difference will be good as you are always using one PNP & one NPN.

When you shift from BJT's to MOSFET, you find the same problem that a N channel MOSFET is created different from a P channel MOSFET. But the dual class A amps  makes this work.

Make your design using point to point methods. Keep watch that you have a signal that changes positive matched by a signal the changes negative for both voltage & current change. Signal is the signal you want and also the Power used to create signal.

You can have a pair of signals between the power rails or one each near the rails and shift back and forth between these.

Old timers worked this way all the time.

C
« Last Edit: July 05, 2018, 01:28:53 pm by C »
 

Offline TheUnnamedNewbie

  • Super Contributor
  • ***
  • Posts: 1211
  • Country: 00
  • mmwave RFIC/antenna designer
Re: Class A MOSFET amp with rapidly varying voltage supply ?
« Reply #7 on: July 05, 2018, 02:23:59 pm »
This sounds somewhat like something that is in fact done, at least in research, to get very high power efficiency from RF amplifiers - Polar modulation amplification. The key is that in those amplifiers, the input signal is a phase/frequency modulated and has constant amplitude, and what we pretty much do is create a switching amplifier (class E/F) and then modulate the amplitude information onto the supply rail. As a switching amplifier is far more efficient than a linear class A/AB/B amplifier, this can give us very high efficiency (provided we can actually generate the supply voltage efficiently, that is).

Here is an example of a block diagram (source: https://www.semanticscholar.org/paper/A-1.75-GHz-polar-modulated-CMOS-RF-power-amplifier-Reynaert-Steyaert/e9b1f1c9d0b2c7975992a71f1b403f2b928ed5eb):


The idea here is to throw away the concept of linearly biasing your amplifier, and using it as a switch. However, this requires you to not have any amplitude information on your input signal (the 200 KHz one).

The key to this being possible/usefull/how depends on the signals in question. What exactly are you trying to do? This "500 KHz signal", what is it? Does it have a lot of dynamic range? You say:


I actually WANT that in the final waveform


What is this "that" that you want so bad? Do you just need some ripple? Do you want full-on modulation of the 500 KHz signal onto a 5 MHz carrier? etc...

The best part about magic is when it stops being magic and becomes science instead

"There was no road, but the people walked on it, and the road came to be, and the people followed it, for the road took the path of least resistance"
 

Offline TheBaconWizardTopic starter

  • Regular Contributor
  • *
  • Posts: 53
  • Country: gb
Re: Class A MOSFET amp with rapidly varying voltage supply ?
« Reply #8 on: July 05, 2018, 03:30:53 pm »
Hmmm, I am wrapping my head around some of the advice given so-far. Interesting stuff, all of it.

Some of you have asked for an idea of the final waveform.

Both waveforms are a sine. Yes, they start-off as square wave from a 556 or an Atiny85 or something, but after an RC filter, we get a sine to work with, let's assume.

To re-iterate. I'm aware that a more obvious way might be simply to put both signals through a summing op-amp and then amplify it by conventional means. But how about if my final amplifiers or indeed the Op-amps are low bandwidth? I was thinking that having the supply voltage ripple, would be a work around.

Here's a visualization, not to scale.



 

Offline Cerebus

  • Super Contributor
  • ***
  • Posts: 10576
  • Country: gb
Re: Class A MOSFET amp with rapidly varying voltage supply ?
« Reply #9 on: July 05, 2018, 03:57:57 pm »
As David Hess stated bias is not a problem.

Where you may have problems is with how you look at a circuit.

First remove Ground from your mind, this is a mess.

You have a V+ & a V-. When looked at it like this, when you have positive ripple on one, you have negative ripple on the other, both the same size. The two can cancel out if you do not mess this up.

One way to keep this balance is to always have one signal going positive while a second is going negative. In place of one class A amp, you use two with second having inverted signal.
This can also make other things easer.
With just one Class A amp bjt transistors you need to have matching NPN & PNP transistors. When you switch to a Dual amp, you want two matching NPN's and two PNP's. The output of the two amps might not look as great separate but difference will be good as you are always using one PNP & one NPN.

When you shift from BJT's to MOSFET, you find the same problem that a N channel MOSFET is created different from a P channel MOSFET. But the dual class A amps  makes this work.

Make your design using point to point methods. Keep watch that you have a signal that changes positive matched by a signal the changes negative for both voltage & current change. Signal is the signal you want and also the Power used to create signal.

You can have a pair of signals between the power rails or one each near the rails and shift back and forth between these.

Old timers worked this way all the time.

C

As I've often noted with posts from you, this, rather confused, stream-of-consciousness post has little to nothing to do with what the OP is asking about. I know you're trying to be helpful but just throwing random "stuff you know" into a thread, that is only loosely or tangentially connected with the topic in hand, just tends to cloud rather than clarify the issue.
Anybody got a syringe I can use to squeeze the magic smoke back into this?
 

Offline JS

  • Frequent Contributor
  • **
  • Posts: 947
  • Country: ar
Re: Class A MOSFET amp with rapidly varying voltage supply ?
« Reply #10 on: July 06, 2018, 01:31:47 am »
  Opamp's PSRR at high freq gets worse, and once out of opamp's bandwidth there's not much the opamp can do to reject the PS noise. You might build the thing as you are thinking and get it working close to what you expect, so why not to try it... this is one of the cases where simulator probably won't help much (unless really good opamp models) but prototype can solve the riddle. You could pick an opamp with the right bandwidth you need or try one with external compensation terminals so you can tweak the BW after the fact.

JS
If I don't know how it works, I prefer not to turn it on.
 
The following users thanked this post: TheBaconWizard

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 22436
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: Class A MOSFET amp with rapidly varying voltage supply ?
« Reply #11 on: July 06, 2018, 01:19:37 pm »
It sounds like you're making a class D amp, and therefore just moving the losses back a stage.  To control the output amplitude, the ~5MHz charge pump needs to be modulated.  But charge pumps can't be modulated*, they are driven with resistive switches and you can only burn the difference as heat in said resistance.

*With a great many stages (impractical for a discrete design, but I'm sure it's been used in ICs before), you can get ratios spaced closely enough to deliver high efficiency.  If the ratios still aren't close enough (e.g., you need continuously variable output, not just a steppy "power DAC"), you only need one stage to burn the difference; which, because the ratios are closely spaced, the difference is small, so the efficiency hit is small.

So you could in turn control the voltage into the charge pump, say with an inductive SMPS.  But then you could remove the charge pump entirely, and drive the output from the SMPS.  Which, really, isn't a bad idea, when it works out (in this case, probably not at 500kHz).

Whatever the application is, it still screams of X-Y problem, which I may've felt (or hinted at?) in a previous thread... :)

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf