Amplifying MHz signals

[Infraviolet]

I need to amplify some 3MHz sine wave signals by a factor of around 5, with an amplification method which keeps signal magnitudes proportional (so a 0.1V peak to peak would come out as 0.5Vpp, a 0.3Vpp to 1.5V and so on).

The trouble is the rail to rail op-amps I've got to hand (MCP6024 and MCP6294, GBWP 10MHz) can't keep up with it, their maximum slew rate is around 6V/us, and the signal varies about this fast even when unamplified.

I'm in a situation where I've only got 5V and ground power rails, no negative rails or higher voltages available.

I can drive a few mA of current from the unamplified signal if necessary, so a method for amplifying does not need supr high impedance inputs.

The key things I need to preserve in the amplified signal are the peak voltage of the sine wave (accurately), and whether it is in phase or in anti-phase with another similar sine signal. I would be ok with getting something as distorted as a square wave out so long as peak voltage and phase were preserved.

My assumption is at these frequencies one starts using methods more similar to those for RF amplification rather tha using op amps, unfortunately I know very little about RF, so any suggestions of circuit types to search for would be helpful.

Can anyone suggest how I might do this?
Thank you

[TimFox]

You can use coupling capacitors at MHz frequencies, and keep the DC level midway between your rails.
Fast op amps are ubiquitous.

[Vovk_Z]

You may try low-voltage fast (almost jelly-bean except price) opamp AD8066.

[RJSV]

   To create a virtual ground you simply make a divider between your zero and +5 volts, that could be a 10 k resistor up, to +5, and another 10 k down, to zero.  Put a fairly large cap, like 5 ufd or larger, and you have an ac reference point.
   To obtain more details perhaps try to search on some similar analog application notes, for another type of IC, like maybe LM358.
   I've used circuits with the (quad pack) LM324, although that one can't deliver output all the way up to your + rail...    Still lots of info in various application notes...maybe too much at first.

[tszaboo]

My assumption is at these frequencies one starts using methods more similar to those for RF amplification rather tha using op amps, unfortunately I know very little about RF, so any suggestions of circuit types to search for would be helpful.

Can anyone suggest how I might do this?
Thank you

Nah, you don't need RF stuff for that, regular, but fast opamp will do. Read the datasheet, there is someting called GBW, gain bandwidth, you need minimum of 15 MHz of that, since you have a 3MHz signal and your gain is 5. And then you calculated the slew rate, which is ballpark 1.42 x f x u = 21V/us. I would try the OPA2350, which would be my goto opamp for such a task, but there are a few that you can try. You also want all the resistors in the 1 KOhm range and not significantly higher so your stray capacitances are not turning it into a filter (or oscillator).

[Infraviolet]

TimFox, RJHayward, can you give some links to webpages about those capacitor based amplification methods.
Thanks

EDIT: I might have misinterpreted your responses. I already have a virtual ground rail at 2.5V, that is what the signal is oscillating about, but I need to amplify its peak to peak, not just shift the DC reference level which it oscillates about.

P.S. LM324 and LM358 already have far worse slew rates than the MCP6024 I'm already using, do you mean there are app notes about how to make them work above their slew rate which might be applicable to be trying togo above the slew rate of the chips I've got?

[DavidAlfa]

The specified bandwidth is at -3dB gain (Gain 0.5), so you need something way faster, at least 50-60MHz I'd say.

[TimFox]

TimFox, RJHayward, can you give some links to webpages about those capacitor based amplification methods.
Thanks

EDIT: I might have misinterpreted your responses. I already have a virtual ground rail at 2.5V, that is what the signal is oscillating about, but I need to amplify its peak to peak, not just shift the DC reference level which it oscillates about.

P.S. LM324 and LM358 already have far worse slew rates than the MCP6024 I'm already using, do you mean there are app notes about how to make them work above their slew rate which might be applicable to be trying togo above the slew rate of the chips I've got?

Can you re-phrase that?  What do you mean by "amplify its peak to peak"? 

[bdunham7]

Can anyone suggest how I might do this?

If you want to preserve phase and amplitude (scaled) you simply need a faster op-amp and they are available at operating voltages of +/-2.5V or less.  You'll need to read the datasheets to find a suitable one, in addition to supply voltage and GBWP there will be a minimum stable gain with faster op-amps that are decompensated.  This means that your design needs to be set up with at least that much gain, and since that might be 10 or more, you might need to attenuate the input signal to make all that work.

Here is one that meets all your criteria and has a minimum gain of 6--and would probably work OK at 5 but it wouldn't be guaranteed.  So you'd need to attenuate your input 6:5 with a divider, or perhaps you can make a gain of 6 work for your application.

https://www.ti.com/lit/ds/symlink/opa607.pdf?HQS=dis-mous-null-mousermode-dsf-pf-null-wwe&ts=1685981084108&ref_url=https%253A%252F%252Fwww.mouser.com%252F

[Infraviolet]

TimFox, I've got a signal of say peak-to-peak of 0.4V centred around 2.5V (min 2.3V, max 2.7V), I want to amplify it, staying centred on the 2.5V rail so it will be peak-to-peak 2V (min 1.5V, max 3.5V). If the input were 2.4V to 2.6V I'd want to be outputting 2V to 3V... The signal is a sine wave, roughly, I need to give it a larger min-max voltage range and must preserve a linear relationship between V_peak_in and V_peak_out such that, within limits, if the amplitude of the sine wave entering doubles so does the amplitude of the output.

Might a single transistor common base amplifier with series coupling caps on the input and output be worth looking in to for these frequencies?

[TimFox]

To clarify my question above:
If all you need is the AC waveform of a signal in the 3 MHz range, the normal approach is "AC coupling", or capacitor coupling.
Take the input through a suitable capacitor to a suitable resistor from your Vcc/2 "virtual ground rail" bias voltage, connecting that to the amplifier input.
Connect the output through similar capacitor to a similar resistor to ground.
You will lose any DC component on your input this way, but will have an output with zero DC value (mean voltage).
The requirement on both C-R networks is
RxC >> 1/(2pi x f_LF)
where  f_LF  is the lowest frequency of interest ("bass response" in an audio amplifier).

If you want the output centered on +2.5 V, then the resistor on the output should go to your bias voltage rail.
Alternatively, if the DC offset of your amplifier circuit is reasonable, and the input is biased to +2.5 V by the C-R input network above, then you do not need the output C, since the amplifier output will still be centered at about +2.5 V.

[Infraviolet]

TimFox, have I badly misunderstood what you've been saying, maybe we've been trying to describe different things. Can that coupling method which eliminates DC do any amplification itself, it doesn't sound like it, or is that just pre-conditioning to do to a signal before feeding to to an op amp?

[TimFox]

A C-R circuit is merely a coupling mechanism that preserves the AC component while rejecting any DC component:  amplification is required to increase the AC component.
The loss of amplitude at low frequencies is governed by the RC product ("time constant").
Fundamental concept:  there is no DC current through a capacitor;  therefore, the mean (DC) voltage after a C-R coupling network is determined by the resistor load (voltage at other end plus any voltage from DC current flowing in the resistor from other places such as amplifier input bias current).
An op amp circuit is convenient, but if you use C-R coupling at both ends of the amplifier, its DC offset is irrelevant and a single active device (traditional AC-coupled amplifier) will work.
If you need the DC component of the input signal (down to zero frequency), you probably want to use split power supplies instead of single-rail, or you need to look at the output with respect to your bias rail.

[Infraviolet]

Thank you, yes I'd misunderstood an earlier post and thought you implied coupling could somehow serve as an amplifier. I was trying to understand what sort of system you were imagining. Now I understand you just meant it as a way of eliminating DC offsets, yes I'm know of them and can add those where needed. Feeding it to an ac coupled common-emitter class-A type single NPN transistor amplifier with an emitter degenration resistor now looks like a possibility for me, that ought to be fesible in the 3MHz or so frequency range so long ad my transistor can handle this frequency? Ought I try to make the R_e and R_c resistors as small as I can (up to the largest current the transistor can handle without heating up) to give this amplifier the best output slew rate?

[TimFox]

In a simple amplifier, the slew rate is determined by the available device current and the capacitance into which it flows.
You must also consider the "Miller effect", determined by the capacitance from output to input of an inverting amplifier and the source impedance driving it.
One reason to use an op amp, is that the relevant parameters are often easier to find in the full data sheet, while if you build an amplifier from discrete parts you need to calculate them yourself.
There is a large literature from days of yore about "video amplifiers" using discrete components in capacitively-coupled stages (transistor or tube), with interesting output networks (including inductances) to optimize bandwidth.
Example from the vacuum-tube days (1959):  https://worldradiohistory.com/BOOKSHELF-ARH/Technology/Rider-Books/Rider-Video-Amplifiers-Alexander-1951-Schure.pdf

[DavidAlfa]

If you want it centered at 2.5V, simply remove C2 in my example.

[jasonRF]

Thank you, yes I'd misunderstood an earlier post and thought you implied coupling could somehow serve as an amplifier. I was trying to understand what sort of system you were imagining. Now I understand you just meant it as a way of eliminating DC offsets, yes I'm know of them and can add those where needed. Feeding it to an ac coupled common-emitter class-A type single NPN transistor amplifier with an emitter degenration resistor now looks like a possibility for me, that ought to be fesible in the 3MHz or so frequency range so long ad my transistor can handle this frequency? Ought I try to make the R_e and R_c resistors as small as I can (up to the largest current the transistor can handle without heating up) to give this amplifier the best output slew rate?

A simple textbook common-emitter might work fine for this.  When i put together a simple function generator,  I used one to get a voltage gain of about 12 with something like 2.5 MHz -3 dB point, and I did no work to try and maximize bandwidth. I wanted some roll-off anyway to mitigate dds artifacts, and even had a cap in parallel with the collector resistor.  I did it that way because it was easy to keep stable on a solderless breadboard and on perfboard, which don’t tend to be friendly towards high-speed opamps.

If you don’t get enough bandwidth from the common-emitter you might try a cascode.     

Jason

[RJSV]

Yes it's a virtual ground I was mentioning.   The two op amps I had used...all the way back 45 years ! (Whew).  You would simply search on that:  'Op Amp voltage follower,  ' op amp simple audio gain'
   Ideally to learn, say with gain of X5 using negative feedback, one circuit does positive gain, the other with (AC) gain but as inverter.
The DC gain perhaps X1 (unity gain), so DC input at 2.5 volts makes a DC output at 2.5 volts.
.The AC rides on top of that.
   Generally restrict your AC gain to less than 20, or you start having phase problems.  So, for 2X gain the negative feedback is going to be 1/2, for 5X gain the feedback is 1/5 etc. 
Unity gain is often called a follower, and you would see feedback of one....The idea is closely analogous to a servo:  In the case of 5X circuit your output is 'controlling' to get the two differential inputs match.
So, if the 1/5 divided down output drifts upwards, very slightly, the OP amp magnifies that and feeds back to cause the 1/5 X feedback term to be controlled to go back down.
   Same for case when that 1/5 term drifts low: the output very vigorously responds the other way, to cause the input to move up, thus you have an error canceling effect, very much like a simple servo.

   As long as not too high frequency, because at higher frequencies the phase relations get messy.
That open loop gain is very high...ideally infinite but can be 100,000.  Ditto for that little error, between the differential inputs...ideally at 'zero'.but in practice could be a few micro volts.

Confusing, but try doing search terms like Inverting AC amplifier, unity gain amplifier etc.

- - Rick

[Infraviolet]

jasonRF, That textbook common-emitter design is the same as a "class A" layout?

I'm trying to breadboard a test of such an amplifier at lower frequencies, yet can't get it to amplify at all, even copying examples found online with all the same component values.

I've tried this topology:

The only output I get is identical to the input on the base, until the base signal's amplitude gets past a certain point, at which point huge distortion occurs and still nothing like an amplification of the proper signal. I should be getting an amplified and inverted copy at the collector, but don't. Can't understand why.

[Slh]

The bias is wrong on that circuit. Increase R1 to 11k and it will probably work ok.

[Infraviolet]

That was an example of the topology, not the only set of values I tried, but I seemed to have the same problem whatever I tried. I was calculating that the bias was supposed to be sitting at around 0.5*Vcc (i'm using 5V not 12V) with some alterations to account for the typical 0.7V drop, and my calculations (following the method in Horowitz and Hill (Sec 2.2.5 A) ) all suggested having R1 similar to R2.

Given your note I've now tried a larger ratio though and seem to be able to get amplification of the output voltage, never realised the biasing has to be such as to make V_base very low without an incoming AC signal.

Now to try at the full frequency I need.

[gf]

To get a DC collector voltage of about 6V, the base voltage needs to be about 1.2...1.3V (0.6V voltage drop across RE + VBE).

[jasonRF]

That was an example of the topology, not the only set of values I tried, but I seemed to have the same problem whatever I tried. I was calculating that the bias was supposed to be sitting at around 0.5*Vcc (i'm using 5V not 12V) with some alterations to account for the typical 0.7V drop, and my calculations (following the method in Horowitz and Hill (Sec 2.2.5 A) ) all suggested having R1 similar to R2.

Given your note I've now tried a larger ratio though and seem to be able to get amplification of the output voltage, never realised the biasing has to be such as to make V_base very low without an incoming AC signal.

Now to try at the full frequency I need.

You can tell that your posted values had problems by assuming it is biased properly then working out the voltages across each element.  Which means that (neglecting the base current) the voltage at the base of the BJT should be about 5.7 volts, which means that there is about 5 Volts across Re from the bias current.  But that same bias current flows through Rc, which is 10x Re so there would be 50 Volts across Rc.  Given your 12 volt supply this makes no sense.   

[Infraviolet]

I've got a solution to this, it seems.

With a 10K and 47K biasing potential divider and a 470 ohm and 4.7K R_emitter and R_collector (found I ended up needing 10 factor gain as the circuit draws a bit of current from the input which lowers its voltage) I'm able to amplify my 3MHz signal. I also added a 10nF from the emitter to ground in parallel with the 470 ohms to allow the output to swing further.

I'll upload a schematic diagram soon.

Can I just check, is this type of class A type transistor amplifier one of those devices which is highly suspectible to individual peculiarties in components? With an op amp type circuit you know that if one of them works then so long as as all the components used in other are within tolerances any copy will work, is this true for single transistor amplifiers? I understand this type is not supposed to depend on things like the transistor's beta property, which varies wildly between two transistors of the same model and batch, but is it dependent on anything else highly variable? I have found that using a BC337 for this transistor gives a rather weaker amplification than if a 2n3904 is used, and the actual amplification is somewhat below the expected -R_c/R_e ratio , does this mean this circuit is probably not something I can reliably reproduce later? or should it be about as reproducible as op amp based circuits are?
Thanks

[jasonRF]

I've got a solution to this, it seems.

With a 10K and 47K biasing potential divider and a 470 ohm and 4.7K R_emitter and R_collector (found I ended up needing 10 factor gain as the circuit draws a bit of current from the input which lowers its voltage) I'm able to amplify my 3MHz signal. I also added a 10nF from the emitter to ground in parallel with the 470 ohms to allow the output to swing further.

I'll upload a schematic diagram soon.

Can I just check, is this type of class A type transistor amplifier one of those devices which is highly suspectible to individual peculiarties in components? With an op amp type circuit you know that if one of them works then so long as as all the components used in other are within tolerances any copy will work, is this true for single transistor amplifiers? I understand this type is not supposed to depend on things like the transistor's beta property, which varies wildly between two transistors of the same model and batch, but is it dependent on anything else highly variable? I have found that using a BC337 for this transistor gives a rather weaker amplification than if a 2n3904 is used, and the actual amplification is somewhat below the expected -R_c/R_e ratio , does this mean this circuit is probably not something I can reliably reproduce later? or should it be about as reproducible as op amp based circuits are?
Thanks

Glad you found a solution!

When I simulate your choice of resistor values, with the bc337-40 LTSpice predicts the -3 dB point is somewhere near 4.3 MHz, so you are already down ~1.7 dB at 3 MHz; the 2n3904 gives a -3 dB point of about 12 MHz so the response has not rolled off at 3 MHz.   Perhaps I will breadboard it tomorrow to see if measurements agree.  The issue is that your bias current is quite low; if you increase the bias you will increase the bandwidth with both of these devices and they probably will be more similar at 3 MHz. 

I breadboarded a 5V, x5 gain version with a bc337-40 biased at 17 mA that was only down 0.2 dB at 20 MHz (the max I could measure with the setup I was using), so you can get a lot of bandwidth if you want it.  When I swapped different transistors (tried 2n2222, 2n3904, bc549c) I don't think the gain at 3 MHz changed by more than perhaps 0.2 dB, but I didn't record any numbers as I was just goofing around. 

I think you forgot about re (intrinsic resistance) in your gain formula. At your bias current of about 0.5 mA, re = 26/0.5 = 52 Ohms, so gain is -Rc/(Re+re) =  -9.   Is that close to what you are getting?

Regarding dependence on the device - yes, the circuit will have some dependence on device parameters.  You can calculate the dependence (I would think most "Microelectronics" type of books will show how) and verify your circuit design is robust enough for your needs.   Biasing is the key, and for starters you really want R1||R2 << beta*Re.

jason