Should I use one or two stage amplifier in an analog buffer?

[surianova]

analog buffer

Hi all,

In designing analog buffer, is it better to use one stage or two stage amplifer ? I think two stage amplifier is better, because it can provide
higher gain and output almost equal to input in unity gain configuration.
Am i right?

Suria

 

[batdin]

Re: analog buffer

Yes,if properly designed,the two-stage buffer has gain almost 1,higher input impedance and lower output impedance than the one-stage buffer.
But the two-stage is more prone to instability,I think.

 

[JPR]

Re: analog buffer

It depends upon what your requirements are.

A properly designed two stage will usually give you higher open loop gain, and thus less closed loop error.

A poorly designed two stage will give you less closed loop bandwidth, more offset, and more phase error.

 

[surianova]

Re: analog buffer

As i think back, buffer should have very high input impedence and low output impendence. If we use two gain stage amplifier,it will provide high output impedence at the second stage which not very good as a buffer. How about the first stage is the diffrential gain stage and follow by source follower which can provide a very low output impendence ? With lower output impedence, the buffer can work better as a volatge source.

 

[JPR]

Re: analog buffer

Once your amplifier is connected with feedback, the output resistance will drop considerably, which may, or may not, be sufficient for your purposes. Having a low gain, low output resistance stage may, or may not help (as illustrated in the following examples)

For a 2-stage amplifier with 2 high gain stages:
If your amplifier has an open loop output resistance of 1MegOhm and an open loop gain of 100dB, the closed loop output resistance (using the amplifier as a buffer) will be 10 ohms.

For a 2-stage amplifier with 1 high gain stage and 1 low gain, low output impedance stage:
Now the amplifier might have an open loop output resistance of 2kohms, and an open loop gain of 40dB. This will give a closed loop output resistance of 20 ohms, which is actually greater than the output resistance with two high gain stages.

For a single stage amplifier with 1 high gain stage:
With an open loop gain of 40dB and an output resistance of 1Megohm, the output resistance will be considerably greater, at 10kohms.

For a 3-stage amplifier with 2 high gain stages and a low impedance low gain stage (caution- this might be more difficult to make stable at a closed loop gain of 1):
Now the open loop gain is 100dB and the output resistance is 2kohms. The closed loop output resistance will ideally be 20 milliohms. At these levels, care must be taken to not have parasitic resistances and inductances play a role, both in impacting the closed loop stability, and the closed loop output resistance.

 

[surianova]

Re: analog buffer

If the input to the buffer is a 3 GHz signal with common mode 1.2v, the output of buffer will connect to RC low pass filter to extract DC value 1.2v, which configuration you sugest to use?
Thanks in advance.

Suria

 

[JPR]

Re: analog buffer

I guess that I am not sure why a buffer is needed? If the signal being fed to the low pass filter is low impedance, adding R-C filter here should be okay?

If the signal is high impedance, can you just add proper C to perform low pass, or is impedance of signal variable?

It seems to me that amplifier adds source of noise and source of error at high bandwidth that is not needed for your application?

What is the filter frequency response? Why not use active filter instead of buffer with passive filter? It seems to me that I must be missing something in your requirements.

If amplifier is needed, you need to determine the specifications for the amplifier, including (but not necessarily limited to):
bandwidth, slew rate, accuracy of gain, offset, power supplies, current consumption, input impedance (vs frequency), output impedance (vs frequency), input signal swing, output signal swing. All of these requirements can drive the amplifier topology that you select, and will help you to evaluate trade-offs when comparing amplifiers.

 

[surianova]

Re: analog buffer

1) the reason i use buffer:-
The main signal 3 GHz will fed to other block for processing but i need
to extract the DC value for control purpose in other block. So, i just
put a buffer between the 3 GHz signal and LPF filter so that the 3
GHz signal will not be loaded by the LPF. The LPF i have is the resistor
connect in series with capacitor. The cufoff frequency for LPF is 30 KHz.
And i am not sure, we need buffer or not in this case?

2) If we want to connect 2 block together, what is the consideration
for input and output impendence ?

 

[JPR]

Re: analog buffer

Yes, it does sound like you will need to keep LPF from impacting signal for other block. Now I see why buffer is needed.

The consideration for input impedance on your buffer is to keep block from impacting the signal to the other block. Keep careful consideration that you need not only high input impedance (in relation to output impedance of signal feeding to buffer) at DC, but also at 3GHz. This could be a tough requirement, depending upon other requirements. If your circuit can handle it, put a resistance to the input of the amplifier, to keep capacitance from impacting signal to the other block. Resistance value depends upon the capacitance of the input of the amplifier. I would recommend that you keep 1/(2*pi*R*C) > 30GHz to keep from impacting phase at 3GHz, with R being your series resistor and C being the input capacitance of the amplifier. This is especially true if input signal is greater than 100Ohms impedance at 3GHz.

It sounds like the buffer does not need especially high gain-bandwidth, since you will be filtering to 30kHz after buffer. I would recommend keeping GBW greater than 300kHz, again to keep phase from impacting the signal.

The offset specification for your amplifier will be determined by the accuracy you need in the output. The input offset will be directly impacting the signal (1mV offset = 1mV offset in your output).

The open loop gain will be determined from your accuracy. The actual gain will be A/(1+A), with A being the open loop gain.

The output impedance of the buffer should be able to be fairly high, since you will be driving the filter. If your output resistance (closed loop) is less than 1/1000th of the filter resistor, you will change the filter time constant by less than 0.1%.

To use this to help you to determine the answer to your initial question, I would look at the necessary accuracy, use that to determine the necessary gain, which will drive the number of stages that you need. Voltage range will drive the topology: If you have a limited voltage range, many topologies will work, but if you need an extreme voltage range, special input and possibly special output stages might be needed.

Depending upon your necessary accuracy and common mode range, I would guess that a single, folded cascode amplifier (1 stage) or a simple 2 stage amplifier would probably suit your needs. Both of these configurations would be dependant upon the input pair to determine the operating range, and if you need wider range than would be available, a special input stage that will have wider input voltage range may be needed.

 

[surianova]

Re: analog buffer

Hi JPR,

1.If the 3GHz signal is the output from transimpedence amplifer which has low output impedence and connect to the low pass filter which has high input impedence ( correct me if am wrong). In this configuration, the buffer is not needed, rite? Because we connect from low output impedence from transimpedence amplifier to high impedence low pass filter.

I would recommend that you keep 1/(2*pi*R*C) > 30GHz to keep from impacting phase at 3GHz, with R being your series resistor and C being the input capacitance of the amplifier. This is especially true if input signal is greater than 100Ohms impedance at 3GHz.

2.why need to have resistor to the input of the buffer? You said it is better > 30 GHz. That mean i need to have very small R and C to get> 30 GHz. Is it its purpose is to lower down the output impedence of the 3GHz signal?

3. Why the output impedence of the buffer should be high to drive the filter?

If your circuit can handle it, put a resistance to the input of the amplifier, to keep capacitance from impacting signal to the other block

4. The capacitance you refer is the filter capacitance or buffer intrinsic capacitance?

Thanks for your answer so far..

Suria

 

[JPR]

Re: analog buffer

To answer your questions:

1- If the addition of the filter does not impact the signal for your other blocks (keep in mind phase change, as well as amplitude), the buffer would not be needed.

2- The reason for the resistance at the input to the buffer is to keep the capacitance of the buffer input from impacting the signal used on other blocks. Capacitance at the input of the buffer will have impedance of 1/(jwC) At high frequencies, this becomes low impedance, and will impact the signal going to other blocks. A series resistance will keep the impedance from dropping too low at high frequency and loading your 3GHz signal.

3- The output impedance does not need to be high, I was just pointing out that it does not need to be extremely low, either, since it will be driving the high impedance of your R-C filter.

4- The capacitance that I was talking about is the input capacitance of the buffer. The same holds true if you use the filter without a buffer.

 

[surianova]

Re: analog buffer

Hi Jpr,

ok, thanks. I am clear and uderstand with your answer given. By the way, is it possible to connect output that is high impedence to high input
impedence for voltage transfer. If not , how to solve it? Put a buffer between it?

Suria

 

[JPR]

Re: analog buffer

The need for a buffer depends upon the levels of impedance and your requirements for accuracy. If you have a 0.1% requirement for accuracy, the input impedance (again, across your frequency of interest) must be greater than 1000x the output impedance of the previous stage. If your requirement is tighter, you will need a larger ratio, and if your requirement is more relaxed, you can have a lower ratio. Since input impedances are usually MUCH higher than output impedances, this is typically not much of an issue.

 

[surianova]

Re: analog buffer

JPR,

Thanks. One more question is, when we connect voltage source to the input of cmos transistor, do we need to consider the source impedence & do you know roughly what is the range of source impedence? Because as i know, the source impedence will form a pole with intrinsic capatance of the transistor that is cgs+ (1+A)cdg. That is why this source impedence is very important to determine the input pole frequency.

Suria

 

[surianova]

Re: analog buffer

The need for a buffer depends upon the levels of impedance and your requirements for accuracy. If you have a 0.1% requirement for accuracy, the input impedance (again, across your frequency of interest) must be greater than 1000x the output impedance of the previous stage. If your requirement is tighter, you will need a larger ratio, and if your requirement is more relaxed, you can have a lower ratio. Since input impedances are usually MUCH higher than output impedances, this is typically not much of an issue.

As i think back,the input impedence of buffer will even higher and output impedence of the buffer will even lower because of negative feedback configuration. If put a buffer between output stage that have high output impedence and the input stage have high input impedence. Then the output impedence of the output stage will be smaller compare to input impedence of the buffer in term of ratio. And olso the output impedence of the buffer will be smaller compare input stage impedence in term of ratio as well. Then this kind of configuration can work better to transfer voltage without much loss.

Suria