Analog 16:1 MUX IC question, 4 vs 1 stage, transmission gate vs nmos

I'm designing an Analog MUX in IC in cadence virtuoso that will be addressed using a 4 bit decoder (so a 16:1 analog MUX). My first idea was to use a transmission gate (nmos + pmos with inverter). However, my adviser thought this was overkill and that just an nmos would work. Here is some info:

• I am using a bipolar power supply (+/- 2.5V) and my signal going into the analog MUX will have DC values around -0.7V to -1V.
• I am very concerned with noise, delay, and gain.
• My signal will be 1 Hz - 5 KHz, so not super high frequency, and on the order of 100's of microvolts to several mV.
• I have a source follower before the MUX with the current source of the source follower after the MUX which is an ideal current source probably around 100 micro amps.

First question: I am simulating the difference between the transmission gate and the nmos, but I'm wondering if there is a general concern/guiding principal concerning analog MUXs and transmission gates vs just an nmos. I know usually transmission gates are used to prevent Vt drops, but since my signal will be DC value -0.7 - 1 V would just the nmos work or are there other benefits to using the transmission gate?


Second question. Additionally, the first analog MUX I designed I used a design that had 4 stages (one for each bit) and looked like so:



However, my adviser said to combine them into one stage and then using it with just 16 nmos gates in a row like so (with proper digital controls accordingly):


This looks wrong to me, but maybe that's because I am just out of IC class and in class we never would design something like that, but my simulations are okay so far.

Second question: What are some possible flaws to a design like this? So far, my simulations have been okay, but I have not simulated it a lot yet. Would the output capacitance be large? Like I said, there is a constant current source on the output of the MUX that is pulling a constant 100 micro amps.

I have worked on PMOS-only muxes, so a NMOS-only is not
entirely insane. However you need to approach this from
the application inward, specifically signal range and supply-set.
This in itself will tell you whether a single-species mux can do
the job.

If you need to go rail-to-rail, in and out, then your NMOS
would need a charge pumped over-the-rail gate drive supply
(and this in turn might violate device ratings at an unfortunate
input, output, supply corner, if your MOSFET was chosen for
supply range alone). At shis corner the PMOS of a CMOS TG
would be taking up the slack using within-the-rails gate bias.

But your numbers up top indicate that NMOS would likely do,
as you've got an "on" overdrive surplus and a negative enough
"off" rail.