low output impedance at high frequencies

[Braski]

How a feedback amplifier with low output impedance at say 150 MHz can be designed? Should it have such a large bandwidth?

 

[FvM]

What means low output impedance for you? 1 ohm, 10 ohm 100 ohm? Do you want feedback action at 150 MHz apart from the involved effect on output impedance?

 

[Braski]

What means low output impedance for you? 1 ohm, 10 ohm 100 ohm? Do you want feedback action at 150 MHz apart from the involved effect on output impedance?

not very low, around 100-1000 Ohm would be low enough for my purposes. Feedback action is not necessary at that frequency, apart from Zout. I know that the Zout becomes equal to Zout of the amplifier at GBW/A where A is the gain of the amplifier. After that frequency, it remains flat, but then i think it will increase due to secondary poles. So i was wondering if there is a technique to obtain a low Zout with a band-limited amplifier...

NOTE: if i have an output stage with high Rout but Miller-compensated, its zout is dominated by the miller capacitance splitted on the output at high frequencies right? This is true only until secondary poles come out, right?

 

[FvM]

100 to 1000 ohm sounds moderate. Generally a source/emitter follower would be an appropriate means to achieve it. Most voltage output OP designs have it anyway. Of course finite transistor fT involves an inductive output impedance component of followers too.

On the other hand, GHz OPs achieve closed loop output impedances in a 10 ohm range by feedback. But you didn't yet talk about technologies and power consumption.

 

[Braski]

100 to 1000 ohm sounds moderate. Generally a source/emitter follower would be an appropriate means to achieve it. Most voltage output OP designs have it anyway. Of course finite transistor fT involves an inductive output impedance component of followers too.

On the other hand, GHz OPs achieve closed loop output impedances in a 10 ohm range by feedback. But you didn't yet talk about technologies and power consumption.

The design only requires low Rout at "open-loop", so GHz OPs are not necessary.

0.18 CMOS with 1.8 V. The problem is, source follower has a low output swing, and i think it won't be enough for my design. I know about flipped voltage follower to increase output swing with low rout, but they imply a local feedback, with its bandwidth.

No particular constraint on power consumption.

You are saying however that if I use a low Rout output stage, say a source follower, with a closed loop response limited in bandwidth, i will be at the end limited for my rout by the high frequency poles of the follower itself, that i can design to be high-frequency, but with limitation induced by the fT of the technology i am using. Is it right?

 

[FvM]

Thanks for giving a clearer picture of the design conditions. I'm not really familiar with low voltage designs, but I think, if source followers don't give sufficient voltage swing, you need to refer to feedback buffers with common source output stage.

 

[Braski]

that is what i am trying to avoid, using a feedback buffer, because of noise and offset. the problem is that i have to drive a fast switched cap ADC. so i have to charge a capacitance at full swing in a very short time and i need a short enough time constant to do it properly.

 

[FvM]

ADC input circuits are mostly using RC filters to decrease the source impedance seen by the sampling switch. You possibly don't have the option for an on-chip ADC buffer. Low driver impedance with sufficient linearity is required anyway. Apart from the voltage swing problem, it should be carefully analyzed if the nonlinearity involved by an open loop buffer is acceptable.

 

[Braski]

can you pls become a bit more specific on the topic of RC? I have tried some kind of RC low pass or series in parallel to rout, but until now, no appreciable results. Is there some good reference on this topic? I have read some applications note, but they did not give good guidelines about it, saying it depends on the situation and telling to use very large capacitances, which is not possible.

I think that if you have a band limited closed loop amplfiier, at very high frequencies, well beyond the bandwidth, the feedback won't react and the rout will be determined by the high frequency singularities, so the output stage will possibly act as open loop.