Spikes in CMOS Inverter transients

10 replies · 11K views · last reply
reddvoid
Messages
118
Reaction score
1
I am simulating cmos inverter in CADENCE
I am getting a sharp spike when output is going from low to high
and spike became more amplified like when i made rise time and fall time of input rectangular pulse signal very low . . .can somebody explain why this happening ?
 
Last edited:
reddvoid said:
I am simulating cmos inverter in CADENCE
I am getting a sharp spike when output is going from low to high
and spike became more amplified like when i made rise time and fall time of input rectangular pulse signal very low . . .can somebody explain why this happening ?

With the CMOS inverter, isn't there a point during transition where both of the devices are conducting simultaneously? With static input, one or other of the transistors is fully OFF, but during transition from one state to the other, for a moment both transistors are partly ON, hence the rush of current you observe.
 
its a voltage spike
like this
Digital waveform trace with two square pulses, each with a sharp overshoot spike at its rising edge, and a rounded spike at the level transition.

sry, its happening when output is going from high to low
 
Attachments
  • Voltage waveform sketch showing a square wave with overshoot spikes at the rising and falling edges, at the transitions from high to low levels.
    del.JPG
    2.1 KB · Views: 2,078
It would be helpful if you were to indicate 0V and +Vcc levels.

I understand this is for an unbuffered inverter?

Your output's load is currently an open circuit? Try a reasonable load, and see how this improves. Say, 1MΩ for starters. Just guessing, but it could well be feedthrough of the input.
 
Last edited:
snapshots

this is the circuit
Schematic of a CMOS inverter with a pulse voltage source V1 (v1=0, v2=1.8, tr=2n) driving the gates, a PMOS M0 (pch3, l=2u, w=8u) and NMOS M1 (nch3, l=2u, w=nmos_w), supply V0 vdc=1.8, and a load resistor R0 with r=r_load to ground.

simulation
Simulator plot of voltage V (y-axis, -0.5 to 2.5 V) against time (x-axis, 0 to 400) showing two square-wave traces: a red trace switching between 0 and about 1.8 V, and a teal trace that rises to about 1.8 V with sharp spikes near 2.1 V and dips near -0.1 V at each transition.

Screenshot of the Cadence Edit Object Properties dialog for a vpulse source instance V1 in analogLib, showing CDF parameters DC voltage 0 V, Voltage 1 0 V, Voltage 2 1.8 V, Period 100n s, Rise time 2n s, Fall time 2n s, Pulse width 50n s.


Yes output was floating,
then i tried 5fF load
and resistive load from 1k to 1M.
still getting spikes in all the simulations .
Stacked transient simulation plots of six traces over 0 to 400 ns, each labelled with a different c_load value, showing voltage in V against time in ns; the top trace is a square-wave input and the lower traces show output waveforms with increasingly slow edges as load grows.

different cap loads
 
Attachments
  • Schematic of a CMOS inverter with a PMOS transistor M0 (pch3, w=8u, l=2u) and NMOS transistor M1 (nch3, l=2u, w=nmos_w), supplied by V0 vdc=1.8, driven by pulse source V1 with v2=1.8 and tr=2n, output to resistor R0 r=r_load.
    Capture4.PNG
    11.8 KB · Views: 2,711
  • Simulator plot of voltage v (V) against time, 0 to 400, showing a red square wave between 0 and about 1.8 V and a teal waveform with sharp spikes near 2.1 V and dips below 0 V at each transition.
    Capture1.PNG
    9.3 KB · Views: 2,091
  • Simulator transient plot of voltage V(V) versus time with a black background, showing a green waveform with narrow spikes near 2.0 V and dips near 0 V, and a flat grey trace at about 1.8 V.
    Capture3.PNG
    7.2 KB · Views: 947
  • Screenshot of the Edit Object Properties dialog in a circuit editor, showing a vpulse source instance V1 from analogLib with Voltage 1 0 V, Voltage 2 1.8 V, Period 100n s, Rise time 2n s, Fall time 2n s and Pulse width 50n s.
    Capture2.PNG
    21 KB · Views: 1,872
  • Stacked transient simulation plots of voltage in volts against time in ns from 0 to 400, one trace per output with a labelled load capacitance; the top trace is a square wave and the lower traces are rising and falling edges with visible overshoot spikes.
    Capture5.PNG
    17.7 KB · Views: 2,242
Last edited:
These spikes are caused by parasitic capacitance between the inverter input and output. Most of it is the gate-drain capacitance in the MOS transistor. When the output is high and the input transitions from low to high, the voltage across this parasitic capacitance cannot change instantaneously, so some of the input step gets coupled to the output. It is a real effect. The spike is larger on the rising edge of the input because the PMOS G-D capacitiance is larger than the NMOS (I think - you could check this in the SPICE model by looking at the Cgd parameter).
 
You are seeing it because it's meant to be there. If a problem you might see it reduce if you were to increase the risetime of the gate voltage, so it rises less sharply. The height of the spike is probably limited (clamped) by a pn junction to Vcc. Adding load capacitance is not viable.
 
I agree with phyzguy, it is capacitive coupling.

The threshold voltages of the two mosfets must be chosen so there is little through current during a transition.
When the input goes high, the upper mosfet should turn off before the lower mosfet turns on.

During the short period that both mosfets are off, the inverter input is being capacitively coupled directly to the output load by the CGD of both mosfets in parallel.

The ratio of transition spike energy is related to the ratio of mosfet CGD.
The effective CGD is voltage dependent and highest during transition.