Mixed-Signal VLSI Design Course Code: EE719 Department: Electrical - - PowerPoint PPT Presentation

1

Mixed-Signal VLSI Design Course Code: EE719 Department: Electrical Engineering Lecture 20: February 19, 2018

Instructor Name: M. Shojaei Baghini E-Mail ID: mshojaei@ee.iitb.ac.in

2 2

IIT-Bombay Lecture 20 M. Shojaei Baghini

Module 27 Histogram (Code Density Test) for Measuring DNL/INL of ADC

Reference

• Chapter 5, the Data Conversion Handbook,

Analog Devices, 2005.

3 3

IIT-Bombay Lecture 20 M. Shojaei Baghini

Basic Concept

• Collecting many digitized samples for a desirable

shape of input signal

• A slow triangular waveform which exceeds the full

scale range provides equal distribution of the input samples.

• Number of occurrences for each code is a measure
• f DNL for that code. If DNL is zero there will be

equal number of occurrences for each code.

4 4

IIT-Bombay Lecture 20 M. Shojaei Baghini

Measuring DNL and INL – Histogram Method

• Automated measurement
• Almost independent of internal noise of ADC due to

averaging over many samples

• Input waveform (triangular or sinusoid input?)
• Amplitude? Frequency?

Periodic triangular waveform generator (more than FS range)

5 5

IIT-Bombay Lecture 20 M. Shojaei Baghini

• Total number of samples for codes 1 to 2N-2: MT (0 and 2N-1 are

not included.)

• h(n): Measured number of hits for code # n
• Ideal value of h(n) = MT/(2N-2)
• DNL of code n (in LSB unit) = !"#$%& ((*)

,-.%& ((*) − 1

• Condition on the frequency of triangular input
• What about the first and last code, offset and gain error?
• What about monotonicity? (linked to distortion and hence FFT is

also necessary.)

Procedure

Triangular waveform generator (more than FS range) N-bit

6 6

IIT-Bombay Lecture 20 M. Shojaei Baghini

Example Histogram for a Triangular Input

7 7

IIT-Bombay Lecture 20 M. Shojaei Baghini

Derivation of DNL from the Histogram

8 8

IIT-Bombay Lecture 20 M. Shojaei Baghini

Derivation of INL from DNL

9 9

IIT-Bombay Lecture 20 M. Shojaei Baghini

Requirements

• Number of collected samples and duration of

DNL/INL simulation/measurement depends on the desirable measurement accuracy.

• Enough number of analog samples per code for ADC

simulation.

• Slow varying ramp.

M: Number of samples per code ⇒ Total required number of samples excluding overhead = 2N × M

• Measurement time = 2N × M × Ts × K

10 10

IIT-Bombay Lecture 20 M. Shojaei Baghini

Module 28 Limitations of the Triangular Input for Measuring DNL/INL of ADC using Histogram Method

Reference

• Chapter 5, the Data Conversion Handbook,

Analog Devices, 2005.

11 11

IIT-Bombay Lecture 20 M. Shojaei Baghini

Periodic waveform generator (more than FS range)

Histogram Method

12 12

IIT-Bombay Lecture 20 M. Shojaei Baghini

• Generation of the input? A high resolution fast settling

DAC?

• Accuracy of the measurement can be verified by

repeating the measurements several times.

• What about linearity of the triangular input?

Remember such input signal is not even single tone and filtering should not add to the non-linearity.

• Why not to use a single-tone input with high linearity

and high SNR?

Discussion Around the Triangular Input

13 13

IIT-Bombay Lecture 20 M. Shojaei Baghini

Single-tone Input

• Ideal value of h(n)

= p(n) × MT

• DNL of code n

(in LSB unit) =

!"#$%& ((*) ,-.%& ((*) − 1

𝑞 𝑜 = 1 𝜌 sin89 𝑜𝑊𝐺𝑇 2𝑂×𝐵 − sin89 (𝑜 − 1)𝑊𝐺𝑇 2𝑂×𝐵

14 14

IIT-Bombay Lecture 20 M. Shojaei Baghini

End of Lecture 20