Making a filter that remains the same when loaded

[Ry122]

To make a filter out of a bjt where the poles don't shift when you load it, do you need to have a buffer stage that follows the bjt such as a unity gain mosfet and then put the load after that buffer stage?

 

[NascentOxygen]

To make a filter out of a bjt where the poles don't shift when you load it, do you need to have a buffer stage that follows the bjt such as a unity gain mosfet and then put the load after that buffer stage?

Basically, yes, though it may be useful to have some gain in the buffer stage.

 

[tech99]

To make a filter out of a bjt where the poles don't shift when you load it, do you need to have a buffer stage that follows the bjt such as a unity gain mosfet and then put the load after that buffer stage?

Yes, this is usually the case.

 

[Baluncore]

The form that the output buffer takes will be determined by the signal power, impedance and the characteristics of the load.
You may be able to use an emitter follower that is already part of the filter.

 

[Ry122]

Can't you just use something really simple like this under all circumstances?

 

[meBigGuy]

That will not be linear. Look up "emitter follower" or "source follower". That would be simplest buffer (with no voltage gain). If you need voltage gain, then you essentially need to build an amplifier. The characteristics of the amplifier will depend on what you need to accomplish.

 

[Jeff Rosenbury]

Your circuit seems to be an amplifier rather than a filter. In addition it will give 180º (ish) phase shift. And as meBigGuy pointed out it's non-linear. (The phase shift is 180º for most of its bandwidth, but drifts near the knee frequency.)

What is it you want this stage to do? If it's a filter, what characteristics do you want it to have (bandwidth, etc.). Are you operating at a really high frequency? Is that why you want a BJT?

There are lots of nice op-amp filters online, but they tend to be frequency limited. Push-pull amplifiers are more linear, but have higher output impedances. Current mirrors can provide impedance matching (like an emitter follower) but tend to be the frequency limiting part of the op-amp.

Filter design is an arcane art. There are lots of considerations that mostly don't matter -- except when they do. We need more information about your project.

 

[Baluncore]

Can't you just use something really simple like this under all circumstances?

Every circumstance is different. There is no one solution to all problems.

Please post the schematic of your prototype BJT filter with your required specifications. We can then give you some good solutions to the output impedance problem.

 

[Ry122]

Alright, thanks. Between C1 and R1 is where the output will be coming from.

 

[Ry122]

Your circuit seems to be an amplifier rather than a filter. In addition it will give 180º (ish) phase shift. And as meBigGuy pointed out it's non-linear. (The phase shift is 180º for most of its bandwidth, but drifts near the knee frequency.)

What is it you want this stage to do? If it's a filter, what characteristics do you want it to have (bandwidth, etc.). Are you operating at a really high frequency? Is that why you want a BJT?

There are lots of nice op-amp filters online, but they tend to be frequency limited. Push-pull amplifiers are more linear, but have higher output impedances. Current mirrors can provide impedance matching (like an emitter follower) but tend to be the frequency limiting part of the op-amp.

Filter design is an arcane art. There are lots of considerations that mostly don't matter -- except when they do. We need more information about your project.

The circuit I posted above is a bandpass filter with a pretty reasonable gain in the midband. There's 40db/decade rolloff either side of the midband and the cut off frequencies are at about 100hz and 10khz.

 

[NascentOxygen]

Is this your design for a particular real-world purpose, or is it a circuit you were provided with and asked to investigate as a lab exercise?

 

[Ry122]

real world

 

[Baluncore]

A real world problem.
The Q1 amplifier with primitive bias will be a disaster.
Gain will be signal amplitude dependent.
There will be harmonic generation due to the non-linearity of the Vbe junction.

Attached is the .asc file for those with LTspice.

 

[Ry122]

A real world problem.
The Q1 amplifier with primitive bias will be a disaster.
Gain will be signal amplitude dependent.
There will be harmonic generation due to the non-linearity of the Vbe junction.

Attached is the .asc file for those with LTspice.

You don't need to worry about that. I've ensured that the BJT is staying within the active region throughout the complete cycle of the AC signal by doing some fairly drawn out small signal analysis.

 

[Ry122]

I graphed the voltage drop across the Vce junction in the circuit you uploaded and you can see that it never falls into saturation, so what are your concerns exactly?

 

[Ry122]

Also, I can't see anything wrong with this bode plot.

 

[NascentOxygen]

There are a number of problems with such rudimentary biassing. Okay, you have tweeked it so it works just fine right now, but ...
things will change as temperature drifts, but worse is when that transistor burns out (for some mysterious reason that I cannot always pin down) the transistor you replace it with will be quite different in gain and the circuit will need to be checked and tweeked all over again. That's why a more stable arrangement is always used.

Do you need this filter to continue to work reliably for any length of time?

 

[Ry122]

This is what real world op-amps consist of, you just aren't accustomed to seeing this type of thing because everyone just buys IC op-amps these days.

 

[Baluncore]

This is what real world op-amps consist of, you just aren't accustomed to seeing this type of thing because everyone just buys IC op-amps these days.

That is incorrect. You are getting advice here from experienced professional electronics design engineers.

Also, I can't see anything wrong with this bode plot.

What is the vertical scale dB/div ?

 

[Ry122]

fixed the graph

 

[meBigGuy]

Ry122:

I agree with the others that your circuit is a disaster. It will have high distortion and the gain will vary with temperature, age, and from transistor to transistor.

It is nothing like what is used inside an opamp.

A typical discrete circuit with stable gain and bias would be

You will find no common emitter 1 transistor amplifier circuit without two bias resistors and an Re

Opamps use current mirrors and much more sophisticated structures.

http://electronicsecg1.blogspot.com/2008/01/op-amp.html

 

[NascentOxygen]

We should give Ry122 due credit for devising a filter that does what he wants, and confirming this by simulation.

It sounds to me that you may have come up with this through a good deal of your own effort, so you have good reason to be proud of it. But the reality is that filters can be precision constructs. Your design looks like it may be accommodating the input impedance of the BJT in setting the response, and this impedance will change a little with temperature, and a lot when/if you need to swap in a replacement device. There are improved circuits that seek to minimize this variation, quite separately to avoiding transistor saturation.

 

[meBigGuy]

Maybe calling it a disaster was a bit harsh. I was over-reacting to his charge of ignorance, which was way out of line:
"This is what real world op-amps consist of, you just aren't accustomed to seeing this type of thing because everyone just buys IC op-amps these days."
not realizing that the people answering him could easily design the internals of those opamps.

 

[Ry122]

How come you should use two bias resistors? Also, isn't having an Re only beneficial if you're going to have a non stable/unclean VCC?

 

[LvW]

How come you should use two bias resistors? Also, isn't having an Re only beneficial if you're going to have a non stable/unclean VCC?

An emitter resistor Re (providing negative feedback) has the following advantages:
* DC bias point more stable against temperature changes and BJT parameter tolerances (beta)
* Lowering of signal gain (advantage?)
* Signal gain less sensitive to BJT parameter tolerances
* Reduction of signal distortions (lower THD)
* Drastic increase of input resistance
* Bandwidth increase

Finally, Re provides negative VOLTAGE feedback - therefore, this feedback scheme works best if the input biasing is realized using a voltage divider that can deliver a "stiff" voltage (as stiff as possible). However, due to some other constraints (power consumption, input resistance) the voltage divider must not be too low-resistive. As acommon trade-off the current through the base divider is set to a value of app. 10*Ibase.

 

[Ry122]

Alright, I changed the circuit again to account for the problems that you mentioned. Do you see any problem with implementing this type of filter near the source like this?

 

[LvW]

Because you have ac-shorted the emitter resistance in both stages the input resustance at the base node as well as the signal gain depends considerably on transistor parameters. Bad design.
( I didn`t check the resistor values!)

Question: The first step for designing a filter is to specify the filter parameters (type, bandwidth, gain, attenuation requirements). What is your specification?

 

[Ry122]

Oh, okay. That's something I copied from the image at the top of this page (posted by meBigGuy).
My requirements are that its a band pass filter with a mid band between 100hz-10khz, and 40db/decade attenuation. Gain in the mid band is 400.
The filter part near AC1 is okay though right? That filters out the high frequencies.

 

[LvW]

Oh, okay. That's something I copied from the image at the top of this page (posted by meBigGuy).

OK - but this circuit represents just a gain stage. Of course, due to coupling capacitors and falling current gain for rising frequencies this gain stage has any bandpass characteristics. But - does it fit to your needs? I don`t think so.
My recommensation (in case you are not allowed to use an opamp): Use to separate passive stages for a first-order lowpass (R-C) and a first-order highpass (C-R) - and a buffer (common-collector) between the 2 stages. For the required gain, use another separate gain stage.

Comment: Sorry - I forgot that you did require a 40dB/dec roll-off. Is this really important?

 

[Ry122]

I reverse google searched that image and it seems to be for a microphone which is what I'm also making my amplifier for. I wonder why they want the AC to bypass the emitter resistor.]

If I remove those emitter capacitors then my gain drops from 60db to -20db in the pass band.

Also, why should you put a buffer between the two passive stages? So that you don't lose a substantial amount of the signal due to voltage drop between stages?