1. PNOISE Analysis of Pipelined ADCs
Josh Carnes
OSU

2. Sampled Signal Analysis
Over-sampling
Under-sampling * = * =

3. Sampled Signal Analysis
Any spectrum energy outside of the Nyquist range gets folded/aliased into baseband
Full spectrum of noise is folded into the baseband
BW limited noise
Noise Folding due to
ideal sampling +1 -1 -1 +1 +2 -2 -2 +2 +3 -3 -3 +3

4. PNOISE Analysis
The SpectreRF PNOISE simulates the amount of noise in a frequency range for a periodic system, including “folded” noise from modulated bands
Determines total circuit noise or individual contributions as a total integrated noise or PSD
Simulation parameters determine accuracy/speed tradeoff
Requires a preceding PSS analysis

5. PSS Analysis for PNOISE
Calculates the periodic operating point
Only single clock is applied, additional stimulus disabled
Relevant PSS parameters
Beat Frequency – clock frequency
Output harmonics – … not important
maxacfreq – Accuracy parameter

6. PSS Analysis Forms

7. PNoise Analysis for Pipelined ADCs
Want to pick a specific time during period to calculate noise (sampling instant)
Want PSD calculated for impulse sampling
Do Not Want dependence on (sinc) Noise Shaping effect of S/H
Time Domain PNOISE Analysis

8. PNoise Analysis for Pipelined ADCs
2 important mode parameters to consider and set
Noise Type (PNoise) = Time Domain
Reference sideband (PNoise) = 0
3 important accuracy parameters to consider and set
maxsideband (PNoise) =
sweep range (PNoise) = (100 Hz, 100 MHz)
maxacfreq (PSS/options) =

9. Accuracy Parameters maxacfreq, maxsideband will define PNOISE accuracy
Consider noise folded in from nth harmonic:
maxsideband = n
Some noise missed,
maxacfreq too low
for desired
maxsideband

10. Accuracy Parameters Derived for single pole BW limited white noise [4]
Applicable to sampling switch kT/C noise
Identify “significant” sampling network with largest BW
fp = pole frequency [Hz]
fs = sampling frequency [Hz]
%err = allowed rms noise error of noise source [%]

11. PNOISE Analysis Form

12. The Sampling Instant
TD PNOISE wants to calculate PSD at sampling instant
Difficult to sync TD with exact sampling moment
… or you can do something easier …
“save” voltage at sampling instant with an ideal S/H
Can enter any time value during hold operation

13. The Sampling Instant

14. Tstab and S/H Sampling Tstab PSS parameter
Allows for initial conditions to settle
TD Noise sampling of ideal S/H
AVOID CLOCK EDGES!
Run PSS first
Note clock edge times
In PNOISE(TD), set Number of Points = 0
Set Add Specific Point to some point between clock edges
NOTE: all times are relative to PSS time interval

15. TD PNoise: Pipelined ADC
Rvref Rf
MDAC model
Some resistor noise contributes to kT/C of stage
Other resistor noise sampled by next stage caps
Opamp noise modeled with equivalent input thermal resistance
Feedback factor = ½
ugbw = 240 MHz
Expected noise: Φ2 Rs C Φ1 Rn
Φ1p C Φ1
Φ1p Φ1
Φ1p Φ1 D Φ2
Rvref = 200
Rf = 200
Rs = 600
Rn = 100
C = 1p

16. TD PNoise: Pipelined ADC
Time-domain PNoise simulation

17. TD PNoise: Pipelined ADC
Noise summary (input referred) V2
Noise Type: Time-domain
Sim1: N=16 fstop= 50MHz maxacfreq= 1.65GHz
Sim2: N=20 fstop= 50MHz maxacfreq= 2.05GHz
Noise voltage error
Stage
Noise Source
Expected
Sim1
Sim2
Error (sim1,sim2) 1
kT/C
4.14n
3.6n
3.76n
6.7%, 4.7% 2
1.04n
0.91n
0.93n
6.5%, 5.7% 3
0.26n
0.23n
0.235n
5.9%, 5% Rf
0.31n
0.32n
1.6%, 1.6%
0.078n
0.081n
2%, 2%
Rvref
0.078
OpAmp
0.625n
0.646n
1.7%, 1.7%
0.156n
0.163n
2.2%, 2.2%
TOTAL 7n
6.35n
6.54n
4.8%, 3.3%

18. Words of Caution Flicker Noise: how do I set low end of sweep range?
Linear or log sweep?
Linear sweep can severely overestimate noise due to insufficient points in flicker noise region
Accuracy: This can’t be right!!!
Open your mind to noise mechanisms you haven’t considered (particularly correlation) or noise estimations made along the way
Try a sim with very tight accuracy settings
Work your way up from an ideal model

19. Words of Caution Verilog Hidden State
Additional sources of noise in Pipelined ADCs
With no input signal, reference noise will not be excited

20. References
[1] C.A. Gobet, “Spectral Distribution of a Sampled 1st-Order Lowpass Filtered White Noise,” Electronics Letters, vol. 17, pp , Sep
[2] C.A. Gobet, A. Knob, “Noise Analysis of Switched Capacitor Networks,” IEEE Trans. Circuits and Systems, vol. cas-30, pp , Jan
[3] J.H. Fischer, “Noise Sources and Calculation Techniques for Switched Capacitor Filters,” IEEE J. Solid-State Circuits, vol. sc-17, pp , Aug
[4] K. Kundert, “Simulating Switched-Capacitor Filters with SpectreRF,” The Designer’s Guide Community, www
[5] K. Kundert, “An Introduction to Cyclostationary Noise,” The Designer’s Guide Community, www
[6] K. Kundert, “Device Noise Simulation of Delta-Sigma Modulators,” The Designer’s Guide Community, www
[7] SpectreRF manual – found in Cadence tools help menu