what is the best unity gain buffer to use?

[ug02048]

is there any buffer that provide the best unity gain? for S/H circuit say 20Mhz sampling rate?

 

[XNOX_Rambo]

I cannot say if it is the best, but it might be good enough:
https://www.analog.com/en/prod/0,2877,AD8047,00.html

/Rambo

 

[lianghuamo]

There are about 10mv offset between the input and ouput in unit gain buffer, but how to elimiate it?

 

[lastdance]

Can you do auto-zeroing?

If you are looking for a unit gain buffer, the best way to do so is to do some homework. Go to the manufacturer's web site, and do a parametric search. Eg. Linear, ADI, National...
There's no such thing BEST in analog, as analog talks about trade-off. U just need what's best for your application.

 

[avinash]

first you tell me whether you wank to design it or to buy it through manufacturer

 

[crazyamd]

your question is too genera. pls give more information, i.e., where the buffer is going to be used, specs etc

 

[ug02048]

the buffer is going to use in the analog shift register ie S/H - buffer back to back . at sample at about 20Mhz , ill probably need to design the buffer as well, in 0.35u process how do I ensure the gain is unity?

 

[crazyamd]

to make a long story short, you need to make sure the amplifier can provide enough gain at 20M. say, if you need the unity gain error within 0.1%, the open loop gain of your op amp should be at leasts 1000 at 20mhz. of course, other issues need to be considered, such as noise, power consumption, linearity etc.

 

[lastdance]

1000 is 60dB, at 20Mhz, is it possible? It's tough enough to do more than 60dB at DC, with single stage, what's more with 20MHz. Can you give me some links where I could find some info. on this? I am interested to see how it work.

In sample and hold, the amplifier must settle to within a range which you are comfortable with, say 0.1% within some time, eg.10ns. With 20Mhz, you have 1 sample coming in every 50ns.
For 0.1% accuracy, you need a settling time of about 7 times the time constant of your network of amplifier and capacitors. Assuming a settling time of 14ns is required, divide by 7 gives 2ns. At unity gain, this points to a unity gain bandwidth of 500MHz.
What was mentioned is purely for numerical illustration, you have to look at your system spec. to determine what's best for you.
The type of amplifier architecture you can use to attain this type of performance (>100MHz) is numerous, but they are all SINGLE stage. Look into literatures to compare btw the different architectures and choose accordingly.

Added after 2 minutes:

1000 is 60dB, at 20Mhz, is it possible? It's tough enough to do more than 60dB at DC, with single stage, what's more with 20MHz. Can you give me some links where I could find some info. on this? I am interested to see how it work.

In sample and hold, the amplifier must settle to within a range which you are comfortable with, say 0.1% within some time, eg.10ns. With 20Mhz, you have 1 sample coming in every 50ns.
For 0.1% accuracy, you need a settling time of about 7 times the time constant of your network of amplifier and capacitors. Assuming a settling time of 14ns is required, divide by 7 gives 2ns. At unity gain, this points to a unity gain bandwidth of 500MHz.
What was mentioned is purely for numerical illustration, you have to look at your system spec. to determine what's best for you.
The type of amplifier architecture you can use to attain this type of performance (>100MHz) is numerous, but they are all SINGLE stage. Look into literatures to compare btw the different architectures and choose accordingly.