Designing a Differential Oscilloscope Preamplifier

[laurence]

im designing a preamplifier, but my waveform is sweeping at around 2.5 volts, as the is a preamplifier it should be swinging around 0.0 volts. Does anyone know would to fix this problem? i am only using discrete components no ic's as this is one of the specs.

the specs are as follows:
1. Balanced inputs, so either input may be grounded and the signal applied to the other input.
2. Voltage gain = 100 ± 3%
3. Frequency response to be flat within 1 dB from 40 Hz to 15 MHz.
4. Input (single ended) impedance > 1 Kohms;, input capacitance < 5pF.
5. Output impedance < 100ohms;.
6. The amplifier should be capable of driving a 100ohms; load (external) up to 2 volts peak-peak without any visible sine wave distortion.

so what i have so far is a cascode amplifier connected to a Class AB output stage.

when i place a capacitor in between the amplifier and the class AB a few things happen
1) i get crossover distortion
2) frequency response cuts off too much and
3) the gain is lowered to like 600mV pk-to-pk as i need 2V pk-to-pk

this project is a university project I am doing, so no real fancy work or dead set accuracy is need. the only thing i need to adhere to is the specs... as i indicated before i can get most of the specs right but it wasnt sweeping around zero and when i place a coupling capacitor frequency or gain is compensated. i tried this with the 2N3904 which i can get more frequency with but the same thing happens. I am using multisim to simulate this circuit as i have to produce a pcb for this project and we only get one pcb run at uni as it would apparently cost the department a lot of money. so each students only gets one pcb. and the specs for the pcb is 100mm x 50mm. so back to my original question is there any other way to make it sweep around zero without using a capacitor?

i was going to post wat i have done, however the image was to big so i posted it on a third party site.
the site is http://www.animex.com.au/preamp.jpg

thanks
Laurence

 

[berkeman]

You can subtract out the output offset by adding a voltage drop stage before your output stage. You can make a variable voltage drop with a Vbe multiplier -- connect a resistor C-B and another resistor B-E on an NPN transistor. The overall voltage drop from C-E is Vbe ratioed by the resistors. It's not real stiff over temperature, obviously, but you can tune the ratio to give you a set voltage drop at some temperature. Maybe try dropping that into the circuit and see what happens with your simulations. Be sure to take the bottom of the voltage drop stage through another resistor to the -9V rail...

Is there a reason that you are using the Class AB output stage? Just stick in a couple-three diodes in the output base drive circuit and go full Class A (add output resistors to set the bias current).

As for frequency response, that is hard to get with a discrete amp. You are doing the right thing using a cascode structure, but you may need to use faster transistors and you definitely need to minimize the parasitic capacitance. Check out the BC847 / BC857 family of fast dual transistors, for example.

 

[laurence]

so does that mean that if I've got a dc offset of 2V i should have a ratio of 2:1 on the resistors? which resistor should be higher Rcb or Rbe? from the output of the cascode does it run through a cap or not, and to where of the Vbe multiplier?

I am having trouble with the output stage as well as there is no amplification, but more like de-amplification.

Ive tried using the BC847/BC857 however my gain seems to be smaller. it was 2.19V pk-to-pk (just at the output of tha cascode not connect to anything) then with the BC family it decreased to 350mV pk-to-pk.

i tried using a high-pass filter but that decrease my output to around 392mV pk-to-pk, any suggestion on another stage to amplify it to 2V without compromising the offset or frequency

 

[berkeman]

The Vbe multiplier is just that -- you have one Vbe across the bottom B-E resistor, and the same current (to within 1/Beta) flows through the top C-B resistor. So if the top resistor is the same value as the bottom resistor, you get 2x Vbe. If it is twice the bottom resistor, you get 3x Vbe. You would run the output of the cascode through the Vbe multiplier and down to -9V through another resistor. You feed the midpoint between the Vbe multiplier drop circuit and the resistor into the output stage.

Your output stage is an emitter follower, so the voltage gain is unity. If you are getting a "de-amplification", then there is something wrong with your simulation, or you are using too-small devices for your load.

There's no way that BC847/857 devices have less gain or bandwidth compared to 2N3904 type devices. They are small devices, so again, check your operating points to be sure you are biasing things correctly.

BTW, what equations are you using (and which components) to set the gain of your cascode stage (and hence the overall gain, since the emitter follower voltage gain is unity)?

 

[berkeman]

BTW, a better way to get a stable gain of 100 and a zero output offset is to set the gain of your amp higher and use feedback resistors to set the overall gain to 100. Have you considered that approach? Have you learned about feedback stability analysis yet?