Professional Documents
Culture Documents
Voltage Follower Part II Circuit Intuitions
Voltage Follower Part II Circuit Intuitions
Ali Sheikholeslami
0.6
Volts
0.6
PARAMETERS USED IN 140 mV vout(SVF) 1
SIMULATIONS OF THE SVF 0.4 0.4 1
AND FVF. 0.2 SVF
vout(FVF) 0.2 0.7
PARAMETER NMOS (M1 & M2) 0 1
0 FVF
0 5 10 15 20 25
W/L 400 nm/45 nm Time (ns) 0 0.2 0.4 0.6 0.8 1
| VT | 0.43 V Vin (V)
Vout/Vin (dB)
Another intuition for smaller out- –10
1.2 GHz FVF
put resistance can be provided with –20 SVF
the help of negative feedback. If the –30
output voltage increases for any rea- –40
son (say from an increase in the load –50
resistance), it will reduce the v gs of 108 109 1010 1011
M1 and, in turn, the current that M1 Frequency (Hz)
feeds to the output node, lowering the
output voltage. This negative feed- FIGURE 6: Voltage gains of FVF (in red) and SVF (in blue) as functions of frequency.
back effectively “resists” a changing
output voltage or, equivalently, pro- note that both designs exhibit a simi- tively, one should note that the volt-
vides a low output resistance. lar voltage gain of g m /g me, which is age transfer function of the FVF is a
We now multiply the expressions around 0.7 V/V. product of its short circuit transad-
for G m and R out to determine the Figure 5 compares the dc charac- mittance (having a pole at the gate
voltage gain of the FVF teristics of the two followers against of M2) and its output impedance
that of an ideal voltage follower. We (having a pole at the output node).
Vout gm
. . observe that the outputs of I n s u m m a r y, w e
Vin g me
the SVF and the FVF follow There are further have shown that an
Interestingly, this voltage gain is their respective input only variations and FVF offers the same
identical to what we have found [1] for for input voltages larger improvements voltage gain as that of
the voltage gain of an SVF. However, than about 0.6 V. For this that can be made an SVF; however, it does
we note that in the case of an FVF, range, any voltage incre- to the FVF circuit, so by providing a much
this same voltage gain is obtained by ment at the input appears and we will cover larger short circuit trans-
the product of a much higher trans- as a voltage increment some of them conductance and much
conductance (by a factor of g m ro) and at the output albeit in Part 3 of this lower output resistance.
a much lower output resistance (by a with a factor of around series. There are further variations
factor of 1/g m ro). This is interesting 0.7 V/V. and improvements that can
as it shows how the FVF achieves our Finally, given a reduction in the be made to the FVF circuit, and we
final goal of having a voltage gain of output resistance of the FVF com- will cover some of them in Part 3 of
around one while reducing the out- pared to that of the SVF, by a factor this series.
put resistance substantially. of g m ro, we expect the FVF to have
We have used the parameters an increased bandwidth by the same Acknowledgment
shown in Table 1 with a nominal factor. This is indeed consistent Many thanks to Jeremy Cosson-Mar-
bias current of I B = 50 nA and a load with the simulation results shown in tin for his assistance with the simu-
capacitance of C L = 100 fF to simu- Figure 6 where the FVF bandwidth of lation results in this article.
late both the SVF and the FVF circuits ~6.2 GHz is five times the SVF band-
and to compare their dc charac- width of ~1.2 GHz. The simulation References
[1] A. Sheikholeslami, “Voltage Follower, Part 1
teristics and frequency responses. results of Figure 6 further reveal a [Circuit Intuitions],” IEEE Solid-State Circuits
Figure 4 shows that, for the param- 40-dB/dec drop in the voltage gain Mag., vol. 14, no. 3, pp. 13–15, Summer
2022, doi: 10.1109/MSSC.2022.3182791.
eters used, both SVF and FVF shift of the FVF, compared to a 20-dB/dec [2] R. G. Carvajal et al., “The flipped volt-
down the input voltage waveform by drop for the SVF. The readers are age follower: A useful cell for low-voltage
low-power circuit design,” IEEE Trans.
around the same dc value, although encouraged to explain why this is Circuits Syst. I, Reg. Papers, vol. 52, no. 7,
with a slightly higher shift in the case the case before reading further. pp. 1276–1291, Jul. 2005, doi: 10.1109/
TCSI.2005.851387.
of the FVF: 580 mV for the FVF ver- A short answer is that the signal [3] J. Ramirez-Angulo et al., “Comparison of
sus 520 mV for the SVF. This larger path from the input to output of the conventional and new flipped voltage struc-
tures with increased input/output signal
shift in the FVF is due to a small volt- SVF has a single pole, correspond- swing and current sourcing/sinking ca-
age drop across a nonideal current ing to the output node, assuming pabilities,” in Proc. 48th Midwest Symp.
Circuits Syst., 2005, vol. 2, pp. 1151–1154,
source we have used in our simula- that we drive the input node with an doi: 10.1109/MWSCAS.2005.1594310.
tions. This voltage drop effectively ideal voltage source. In contrast, the [4] A. Sheikholeslami, “Looking into a node
[Circuit Intuitions],” IEEE Solid-State Circuits
reduces VDS1 in the FVF, causing a signal path of the FVF has two poles: Mag., vol. 6, no. 2, pp. 8–10, Spring 2014,
larger VGS1 to accommodate the same one at the output node and one at doi: 10.1109/MSSC.2014.2315062.
nominal current through M1. Finally, the gate of M2. To see this intui-