vivkr
|
Hi manodipan,
Nice to know that you were able to fix your problem so easily.
As for bootstrapping, I would recommend it if linearity is critical for you, i.e. you are either working with very low supply voltages, or if you need very good linearity for a very rapidly changing signal. You can check out the JSSC paper I mentioned in my earlier post.
I see the following pros and cons:
Pros: ----- 1. Excellent linearity achievable, especially for very high frequency input signals.
2. Main switch device very small (simple NMOS).
3. Charge-injection etc. are also good, largely signal-independent, one more reason for the excellent linearity.
Cons: ----- 1. Bootstrap capacitance needs to be large. For low VDD, I would say that Cbootstrap ~ 4-5*Cgate is needed, where Cgate is the gate cap of the main MOS switch. This runs into area.
2. Parasitics must be reduced well, especially parasitics from the assisting switches. See Fig. 7 (I think it is Fig. 7 of the paper). The more parasitics you have there, the lower will be the V(G,S) of the main switch. It never will be VDD anyway.
3. You always need a running clock as the bootstrap is dynamic. This is not a problem in most cases, but can be if you are trying to make an ADC which operates in burst-mode, i.e. the ADC lies dormant and may not even receive any clock, but suddenly the CLK is enabled and a CONVERT signal is asserted, following which the ADC must quickly produce an output. Mostly, there is the additional requirement that the ADC "passively" sample a rather rapidly varying input signal while it is dormant. However, as I said, this is happily not often the case.
Other points: ------------- 1. I would advise using as few bootstrapped switches as possible. Delta-sigma modulators are the best place to use these as you usually only have 2 inputs and need 4 bootstrapped switches. Pipelined ADCs actually benefit from these switches a lot, but you might need several of these there, depending on your design.
2. If your input signal varies very fast, make sure that you use a bootstrapped switch on both sides of the cap, the signal side (floating side) but also on the other end, which is usually tied to a common-mode potential, otherwise your linearity will be poor.
3. For the same reason as listed above, do not try to sample a rapidly varying input signal on a cap, the other end of which is tied to an opamp-controlled virtual ground node in the sampling phase.
4. If you happen to have too many signal branches going into your circuit, requiring several bootstrapped switches, it might be worth considering the use of a dedicated sample-and-hold amplifier upfront. You can realize some gain in it also. This way, you save the area you would otherwise be spending for those huge bootstrap caps.
Vivek
|