ADC measurement and specification

Application Note, V1.1, Nov. 2010
AP32121
A D C m e a su r e m e n t a n d
s p e ci f i ca t i o n
AudoNG, Audo Future and AudoMAX family
Microcontrollers
Edition 2010-12-06
Published by
Infineon Technologies AG
81726 München, Germany
© Infineon Technologies AG 2010.
All Rights Reserved.
LEGAL DISCLAIMER
THE INFORMATION GIVEN IN THIS APPLICATION NOTE IS GIVEN AS A HINT FOR THE
IMPLEMENTATION OF THE INFINEON TECHNOLOGIES COMPONENT ONLY AND SHALL NOT BE
REGARDED AS ANY DESCRIPTION OR WARRANTY OF A CERTAIN FUNCTIONALITY, CONDITION OR
QUALITY OF THE INFINEON TECHNOLOGIES COMPONENT. THE RECIPIENT OF THIS APPLICATION
NOTE MUST VERIFY ANY FUNCTION DESCRIBED HEREIN IN THE REAL APPLICATION. INFINEON
TECHNOLOGIES HEREBY DISCLAIMS ANY AND ALL WARRANTIES AND LIABILITIES OF ANY KIND
(INCLUDING WITHOUT LIMITATION WARRANTIES OF NON-INFRINGEMENT OF INTELLECTUAL
PROPERTY RIGHTS OF ANY THIRD PARTY) WITH RESPECT TO ANY AND ALL INFORMATION GIVEN
IN THIS APPLICATION NOTE.
Information
For further information on technology, delivery terms and conditions and prices please contact your nearest
Infineon Technologies Office (www.infineon.com).
Warnings
Due to technical requirements components may contain dangerous substances. For information on the types
in question please contact your nearest Infineon Technologies Office.
Infineon Technologies Components may only be used in life-support devices or systems with the express
written approval of Infineon Technologies, if a failure of such components can reasonably be expected to
cause the failure of that life-support device or system, or to affect the safety or effectiveness of that device or
system. Life support devices or systems are intended to be implanted in the human body, or to support
and/or maintain and sustain and/or protect human life. If they fail, it is reasonable to assume that the health
of the user or other persons may be endangered.
AP32121
AudoNG, AudoFuture and AudoMAX family
AP32121
Revision History:
2010-11
Previous Version:
V1.0
V1.1
Page
Subjects (major changes since last revision)
1
Titel changed
We Listen to Your Comments
Any information within this document that you feel is wrong, unclear or missing at all?
Your feedback will help us to continuously improve the quality of this document.
Please send your proposal (including a reference to this document) to:
[email protected]
Application Note
3
V1.1, 2010-11
AP32121
AudoNG, AudoFuture and AudoMax family
Table of Contents
Page
1
Introduction ...................................................................................................................................5
2
Terms and Definitions ..................................................................................................................5
3
3.1
3.2
3.3
3.4
Static Error Parameters ................................................................................................................5
Ideal ADC Transfer Curve...............................................................................................................5
Total Unadjusted Error ....................................................................................................................7
Offset and Gain Error ......................................................................................................................8
Linearity Errors ................................................................................................................................9
4
4.1
4.2
Static Ramp Test .........................................................................................................................12
Input Signal ...................................................................................................................................12
Measurement Flow........................................................................................................................12
5
5.1
5.2
Summary and Limitations ..........................................................................................................15
Accuracy and Resolution ..............................................................................................................15
Limitations .....................................................................................................................................16
6
6.1
6.2
6.3
6.4
6.5
6.6
6.7
6.8
Measurement conditions............................................................................................................16
Nominal Conditions .......................................................................................................................16
Supply and temperature range......................................................................................................16
Reference variation .......................................................................................................................17
Frequency variation.......................................................................................................................17
Reference exceeding ....................................................................................................................17
Speed test .....................................................................................................................................17
Selectable reference .....................................................................................................................17
Resolution .....................................................................................................................................18
7
References ...................................................................................................................................18
Application Note
4
V1.1, 2010-11
AP32121
AudoNG, AudoFuture and AudoMax family
Static Error Parameters
1
Introduction
This application note describes the measurement methods used for Analog to Digital Converters in Infineon’s
µ-Controllers. It covers both the definitions of static and dynamic error parameters and the test setup and
algorithms used to specify the ADCs. The test conditions used during the characterization/validation of
Infineon µ-Controller ADCs are also outlined.
This document is intended to help users of ADCs in µ-Controllers to understand the specification and to get
the most out of the integrated ADCs.
2 Terms and Definitions
Quantization,
Quantization Error
LSB
Quantization is the process where an analog signal is converted into a digital signal
with a finite number of bits.
The quantization error is the inevitable result of this process, even for an ideal ADC.
n
Least Significant Bit, expressed in voltage it is (Vref+ - Vref-)/2 .
Transfer Curve
A Transfer Curve describes the input to output behavior of an ADC.
Adjusted Transfer
Curve
Transition Voltages
In an Adjusted Transfer Curve, both the offset error and the gain error from a
measurement are accounted for.
The Transition Voltages define the switching points from one ADC code to the next
in a transfer curve.
A test is defined to be static when the input signal to the ADC under test is slowly
varied. The output results are the same as for a constant signal being input.
Typically a ramp is chosen as the input signal for such a test.
Static Test
Dynamic Test
Full Scale
(FS)
Full Scale Range
(FSR)
Resolution
A test is defined to be dynamic when the input signal varies in such a way that the
dynamic errors of the ADC are revealed.
Typically a sine wave is chosen as the input signal for a dynamic test.
The term Full Scale is used to define the highest possible digital output code of an
ADC with resolution n.
Full Scale Range is the range defined by the reference inputs of the ADC
If the ADC has only a positive reference input and the negative reference input is
tied to ground FSR = VRef+, otherwise FSR = VRef+ - VRef-.
The number of bits that the ADC is capable of converting an analog signal into.
This is often a power of 2 and is denoted by n throughout this application note.
Throughout this document small letters indicate an array of data whilst capital letters are used for single
values. For example, tue stands for the whole data array while TUEmin and TUEmax are the minimum and
maximum values used for the specification.
3
Static Error Parameters
3.1
Ideal ADC Transfer Curve
There are two kinds of ideal ADC transfer curves. For the correct calculation of the error parameters of a real
ADC it is important to know its corresponding ideal transfer curve. Figure 1 shows the uncompensated ideal
ADC transfer curve of a 3bit AD converter. Here the code width for all codes is the same. This results in a
quantization error from 0 to -1LSB.
Application Note
5
V1.1, 2010-11
AP32121
AudoNG, AudoFuture and AudoMax family
Static Error Parameters
Ideal ADC transfer curve
8
7
6
ADC code
5
4
3
2
1
0
Figure 1
0
0.125
0.25
0.375
0.5
0.625
0.75
normalized analog input voltage
0.875
1
Uncompensated ideal transfer curve and quantization error of a 3 bit AD converter
Another way of definition is where the width of the first code is only ½LSB wide while the width of the last
code is 1½ LSB wide. This type is often referred to as ½LSB compensated transfer curve and is shown in
Figure 2. The resulting quantization error is now ±½ LSB.
The standard single ended ADCs in Infineon µControllers use the ½LSB compensated transfer curve while
the fast differential ADCs use the compensated one.
Note: For the uncompensated transfer curve the midpoint of the analog input range occurs at code transition
n
n
n
from 2 /2-1 to 2 /2 while for the compensated one the midpoint is located in the middle of code 2 /2.
n
n
There are two 2 codes but only 2 -1 code transitions for both transfer curves.
Application Note
6
V1.1, 2010-11
AP32121
AudoNG, AudoFuture and AudoMax family
Static Error Parameters
Ideal ADC transfer curve 1/2 LSB compensated
8
7
6
ADC code
5
4
3
2
1
0
Figure 2
3.2
0
0.125
0.25
0.375
0.5
0.625
0.75
normalized analog input voltage
0.875
1
Compensated ideal transfer curve and quantization error of a 3 bit AD converter
Total Unadjusted Error
The transfer curve of a real ADC deviates from that of an ideal ADC. The difference between the measured
transfer curve and the ideal transfer curve is the total unadjusted error (TUE). An example for a 3 bit ADC is
n
shown in Figure 3. Note that the tue calculation is based on the code transitions, hence there are 2 -1 tue
values in a tue plot:
tuei   real _ code _ transitioni   ideal _ code _ transitioni 
i  0  2n 1
The tue values are normally expressed in terms of LSBs of the converter. The minimum and maximum
values of the tue array are valid for the specification of the ADC:
TUE max  max(tue)
TUE min  min(tue)
As the name implies, the TUE contains all the errors present in an ADC. These are offset error and gain error
and linearity errors.
Application Note
7
V1.1, 2010-11
AP32121
AudoNG, AudoFuture and AudoMax family
Static Error Parameters
Real and ideal ADC transfer curve
8
7
6
ADC code
5
4
3
2
1
0
0
0.125
0.25
Figure 3
Total Unadjusted Error
3.3
Offset and Gain Error
0.375
0.5
0.625
0.75
normalized analog input voltage
0.875
1
For offset and gain error calculations two methods exist, the endpoint method and the best fit method.
Endpoint method
The endpoint method uses only the first and last codes for the calculations. The offset error is defined as the
difference between the measured first code transition and the ideal first code transition. For the gain error a
line through the first and the last code transition is built. The deviation of the slope of this line from the ideal
slope defines the gain error.
Note: For the differential ADCs using the uncompensated transfer curve the offset error is defined in the
middle of the transfer curve. This is because here the ADC should deliver the mid-code when its inputs
are connected together. In contrast the single ended ADC should deliver code 0 when the input is
connected to ground. For that reason the endpoint approach makes no sense for the offset calculation
in the middle of the transfer curve.
Best fit method
The best fit method is based on a best fit line through all codes. For the compensated transfer curve the
offset error is then defined as the deviation from this line to the ideal one at the lowest code. For the
compensated one the deviation in the middle of the transfer curve is the offset error.
The gain error is defined as the deviation of the slope of the best fit line to the ideal slope. An example of a 3
bit ADC using the ½ LSB compensated transfer curve is shown in Figure 4.
All ADCs in Infineon µControllers are specified by the best fit method.
Application Note
8
V1.1, 2010-11
AP32121
AudoNG, AudoFuture and AudoMax family
Static Error Parameters
Real and ideal ADC transfer curve
Gain error
8
7
6
ADC code
5
4
Offset error
3
2
1
0
0
0.125
Figure 4
Offset and Gain Error
3.4
Linearity Errors
0.25
0.375
0.5
0.625
0.75
normalized analog input voltage
0.875
1
There are two kinds of linearity errors, Differential Non Linearity (DNL) and Integral Non Linearity (INL).
Again, there are different definitions used in industry for both the DNL and for the INL. This can potentially
lead to mis-interpretation of the specification.
Differential Non Linearity
The DNL is defined as the deviation of the real code width from the ideal code width (1LSB). The real code
width is defined as real_code_transition[i] – real_code_transition[i-1]. As the first and the last codes do not
have a predecessor respectively a successor, the width of these two codes is undefined and hence are not
included in the DNL error. The DNL can be calculated out of the tue:
dnl i   tuei   tuei  1
i  1 2 n  2
n
n
Note: The DNL array has only 2 -2 values starting from 1 while the tue array has 2 -1 values starting from 0.
The minimum and maximum of this array are valid for the specification:
DNL max  max(dnl )
DNL min  min(dnl )
The lowest code width is “0” and occurs if a code is missing. In this case the DNLmin error is -1LSB. Hence
DNLmin has a lower boundary of -1LSB. DNLmax can exceed +1LSB and has no upper boundary.
Figure 5 shows an example of a 3 bit ADC where the code width of code 2 is greater then 1LSB hence
having a DNL error greater then 0.
Application Note
9
V1.1, 2010-11
AP32121
AudoNG, AudoFuture and AudoMax family
Static Error Parameters
Real and ideal ADC Transfer curve
8
7
6
ADC code
5
4
3
2
1
0
Figure 5
0
0.125
0.25
0.375
0.5
0.625
0.75
normalized analog input voltage
0.875
1
Differential Non Linearity, code transition testing
In the above definition the code transitions are used for the DNL specification.
Another method of DNL calculation is where the code mid-points are compared to the ideal code midpoints. This method may hide certain ADC problems. These problems can easily be seen in the example in
Figure 6. In this example every second code is too wide and in between every second code is too small.
However the code centers are placed exactly on the ideal positions. Hence this definition is not used for the
specification of the DNL in Infineon µ-Controller ADCs.
Real and ideal ADC transfer curve
8
7
6
ADC code
5
4
3
2
1
0
Figure 6
0
0.125
0.25
0.375
0.5
0.625
0.75
normalized analog input voltage
0.875
1
Differential Non Linearity, code mid point testing
Application Note
10
V1.1, 2010-11
AP32121
AudoNG, AudoFuture and AudoMax family
Static Error Parameters
Integral Non Linearity
The INL is defined as the deviation of the measured transfer curve from an adjusted transfer curve. The
adjusted transfer curve is an ideal transfer curve but with the same offset and gain error as the measured
n
one has. Hence the code width of the adjusted transfer curve is the same for all codes from 1 to 2 -2 but is
no longer 1LSB. An example of this is shown in Figure 7. In this case the adjusted code width is smaller than
the ideal code width due to the gain error in the measured transfer curve.
The other way round is to correct the tue for offset and gain:
inl i   tuei   offset  gaini 
i  0  2n 1
The minimum and maximum of the inl array are valid for the specification:
INL max  max(inl )
INL min  min(inl )
Note: Some INL specifications take the ideal transfer curve instead of the adjusted one. This would be the
same as calculating the TUE.
Real, ideal and adjusted ADC Transfer curve
8
7
6
ADC code
5
4
3
2
1
0
Figure 7
0
0.125
0.25
0.375
0.5
0.625
0.75
normalized analog input voltage
0.875
1
Integral Non Linearity
Application Note
11
V1.1, 2010-11
AP32121
AudoNG, AudoFuture and AudoMax family
Static Ramp Test
4
Static Ramp Test
This chapter describes the Static Ramp Test which is also known as the Linear Ramp Histogram Method as
described in [1]. It is used for evaluating the error parameters described in Chapter 3.
4.1
Input Signal
The input signal for the static ramp test is, as the name already implies, a voltage ramp ranging either from
the negative reference voltage to the positive reference voltage of the ADC or exceeding these voltages.
In the first case the ADC under test produces output codes ranging from 0 to full scale, assuming the ADC
has no offset (Figure 8a).
In the case of a voltage ramp which exceeds Vref- and Vref+, the ADC under test produces more of both the
0 code and the full scale code (Figure 8b).
Figure 8
ADC code
anaolg in
ADC code
anaolg in
In order to get all codes, especially the lowest and highest codes, it is recommended to use the second
approach for the input signal.
Input Signal and ADC Response for a Static Ramp Test
The input voltage ramp can be generated either purely analog or by using a DAC with a finite number of
discrete output voltages as determined by the DAC resolution. (In reality, in most signal generators the
output of the DAC will be amplified before it is output).
If an analog voltage ramp (e.g. integrator) is used, this signal must be synchronized with the clock signal
which starts the ADC conversions. This signal is often referred to as the strobe signal. Furthermore, the
conversions must be made in exactly equidistant time steps. If these two conditions are not met, the
measurement can give poor results which are not related to the ADC under test.
Using a DAC as the signal source makes the setup easier as the only requirement here is that the ADC
samples the DAC voltage after it has settled. After the conversion is finished, the DAC can be updated to the
next output voltage. With an appropriate handshaking technique, any sampling jitter of the ADC has no
influence on the measurement result.
In order to get a sufficient amount of samples for every ADC code, the resolution of the DAC must be greater
than the resolution of the ADC under test. This is directly related to the measurement accuracy and
repeatability as explained in Chapter 5.
4.2
Measurement Flow
A general measurement flow with a DAC as the signal source is shown in Figure 9. The DAC is initialized
with the start voltage of the ramp before the ADC under test makes the first conversion. After the voltage on
the DAC output has stabilized, the ADC converts it to a digital code which is then stored in a memory. The
DAC is updated to the next voltage and the process repeats until the programmed ramp end voltage of the
DAC is reached.
Application Note
12
V1.1, 2010-11
AP32121
AudoNG, AudoFuture and AudoMax family
Static Ramp Test
At the end the memory contains the ADC results corresponding to the ADC input signal but including the
errors of the ADC (offset and linearity errors). This data cannot be used directly for the error calculations as
the resulting transfer curve is not monotonic. In other words the ADC may toggle up and down as it
approaches a code transition. This is due to the nature of all ADCs as the transfer function is a “many-toone” mapping function. For example, all input voltages between Vref- and ½ LSB of the ADC ideally should
be converted to code 0 in case of the ½ LSB compensated transfer curve.
A DAC, by contrast, converts one code into one output voltage, hence has a “one-to-one” mapping function.
For more details refer to [1].
Note: For the error calculations described in Chapter 3 it is necessary to have a monotonic transfer curve to
compare with an ideal ADC transfer curve.
DAC initialization
DAC update
ADC
conversion
NO
store ADC result
in memory
programmed end
voltage reached?
YES
error
calculations
Figure 9
Static Ramp Test Measurement Flow
Figure 10 shows an example of raw measurement data for the first 4 ADC codes. Here an ADC with a
resolution of n bits has been tested using a DAC with a resolution of n+3 bits. The green line is the transfer
curve of an ideal ADC, the red curve represents the measured data.
With a 3 bit resolution difference between the DAC and the ADC, each ADC code width is tested for 8
3
different input voltages. (The LSB of the DAC is 2 times smaller than the LSB of the ADC). Note that for a
characterization test of an ADC this would not be accurate enough, it is just for illustration of the
measurement flow that this number was chosen.
To get a monotonic transfer curve out of which the error parameters can be calculated, the codes are sorted
in ascending order. This process is shown for our example in Figure 11. Now this transfer function can be
directly subtracted from the ideal transfer function and the tue plot can be built. Out of this array all errors can
be calculated.
Application Note
13
V1.1, 2010-11
AP32121
AudoNG, AudoFuture and AudoMax family
code
Static Ramp Test
Figure 10
Example of a raw measurement data
Figure 11
Example of a sorted measurement data
Once the resolution of the ADC increases, it becomes difficult to maintain the difference (ADC resolution –
DAC resolution) > 4 for example. To overcome this problem we can sample every DAC output voltage more
then once by the ADC. Clearly the raw data can no longer be plotted as in Figure 10 as it may happen that
the ADC produces different codes for one and the same input voltage. The solution is to build a histogram
which is simply a mapping of how often every code occurs. A histogram, denoted by H, is a 2 dimensional
n
array with (2 * DACresolution * Oversampling Factor) values. Every value represents the occurrence of that
code. An example of a histogram for a 3 bit ADC is shown in Figure 12a. The total number of samples here
is 64. This could be the result from using a 6 bit DAC as the input signal source or a 5 bit DAC as the input
signal source where every voltage was sampled twice by the ADC.
Note: The process of code sorting as described previously is exactly the same as building a histogram.
Application Note
14
V1.1, 2010-11
AP32121
AudoNG, AudoFuture and AudoMax family
Summary and Limitations
normalized Linear Ramp Histogram
Linear Ramp Histogram
14
1.8
1.6
12
1.4
Code width [LSB]
Code-Density
10
8
6
1.2
1
0.8
0.6
4
0.4
2
0
0.2
0
1
2
3
4
5
6
0
7
Codes
Figure 12
0
1
2
3
4
5
6
7
Codes
Example of a histogram
To convert the histogram back to a monotonic transfer curve it must be normalized to the ADC LSBs. This is
done by dividing each histogram value by the average number of hits per code:
average _ amount _ of _ hits  sum( H ) / 2 n
H
H _ normalized 
average _ amount _ of _ hits
Figure 12b shows the normalized histogram where every histogram value now represents the code width of
the ADC.
As discussed previously, this transfer function can be subtracted directly from the corresponding ideal
transfer function to get the tue array.
Note: In the case of an analog input ramp which exceeds the ADC reference voltages, the number of codes 0
and FS must be reduced by the same factor as the ramp is exceeded.
5
Summary and Limitations
5.1
Accuracy and Resolution
The method described above has some limitations and issues which must be considered when setting up
this test. Offset and gain error are absolute measurements that require a highly accurate analog input signal.
As an example, for an ADC with a reference voltage span of 3.3V and 12 bit resolution, the LSB voltage is
12
3.3V/2 ~ 800µV.
If the generator used has the same absolute accuracy, the errors of the offset error and gain error
measurement are at least ±1LSB. For the linearity errors DNL and INL, only the linearity of the supplied
voltage ramp is of importance.
Furthermore, the resolution of every code width from the normalized histogram is directly related to the
average amount of hits per code. If the average number of hits per code is 8, the resolution of every code
width would be 1/8 of a LSB of the ADC. This is why the resolution of the DAC should be higher than the
resolution of the ADC under test as mentioned before. Once this becomes difficult an oversampling factor
should be used.
Application Note
15
V1.1, 2010-11
AP32121
AudoNG, AudoFuture and AudoMax family
Measurement conditions
5.2
Limitations
Another error which has not been considered until now is monotonicity. According to [2] an ADC is
nonmonotonic if, for a uniformly increasing or decreasing input signal, the output codes can increase and
decrease. However a monotonicity error must not be confused with the random nature the ADC has at each
decision level from when going from one code to another. Therefore [2] defines an ADC to be nonmonotonic
when for any pair of input levels x1 and x2, with x1<x2 the following condition is met:
y ( x1 ) 
y( x2 ) 
3   y ( x1 )
M
3   y ( x2 )
M
 y ( x 2) 
 y ( x1 ) 
3   y ( x2 )
M
3   y ( x1)
M
(non inverting ADC )
(inverting ADC )
Where y (x) is the mean output code value,  is the standard deviation of a static input voltage x and M is
the number of samples taken at each x value.
The linear ramp histogram method produces, in every case, a monotonic transfer function due to the sorting
of the ADC results. Therefore it is not suitable for revealing monotonicity errors.
Hence a sanity check for monotonicity errors is done before an actual ADC validation using the linear ramp
histogram method.
Note: Sometimes an ADC appears to be nonmonotonic if its output toggles rapidly. This may be caused by
noise from the surrounding environment of the ADC and is not a monotonicity error.
6
Measurement conditions
This chapter describes the measurement conditions at which static ramp tests are performed. The specified
error parameters are the result of all the ramp tests, in other words the ADC holds the specification under all
the described measurement conditions. If applicable the conditions applied for the tests go beyond the
specified limits, e.g. maximum ADC speed, in order to see the available margin to the specification.
Dependent on the features of the ADC there can be more specific measurement conditions which are not
outlined here.
6.1
Nominal Conditions
In this test case normally a channel scan of all ADC channels of all ADCs in a µController is performed. The
temperature is kept at room temperature, all supplies are at there nominal value and the ADC runs at the
maximum specified speed. This test can reveal a potential channel dependency of an error parameter. The
channel with the worst behavior is then preferably used for all further investigations.
6.2
Supply and temperature range
This test case covers the variation of all relevant supplies of the ADC at three different temperatures:

Cold temperature on automotive standards: -40°C

Room temperature

High temperature on automotive standards: +125°C
Although the high temperature on automotive standards is defined as +125°C (which results in a junction
temperature of approximately 150°C) the tests are sometimes also carried out at higher temperatures(e.g.
150°C ambient temperature) if it is required.
Application Note
16
V1.1, 2010-11
AP32121
AudoNG, AudoFuture and AudoMax family
Measurement conditions
6.3
Reference variation
This test case covers the specified FSR of the ADC at full ADC speed and nominal supply and temperature.
This is done by variation of the positive and negative reference inputs (if available) of the ADC.
Reducing the FSR means also a reduction of the absolute value of the LSB voltage:
Example:
FSR = 5V, resolution=12bit,
→ 1LSB = FSR/4096 = 1.22mV
FSR = 3.3V, resolution = 12bit → 1LSB = FSR/4096 = 0.805mV
As the absolute accuracy (in mV) of the ADC keeps constant, the relative accuracy specified in LSB has to
be adapted to the FSR decrease. Therefore the specification is relaxed for a smaller full scale range by
specifying a k-factor which defines the increase of the error in LSB by decreasing the FSR.
Example:
FSR = 5V, Offset error = ±4 LSB
FSR = 3V, k=1.51 → Offset error = 6.04LSB
6.4
Frequency variation
An integrated ADC in a µController may be influenced by noise and activity of the surrounding environment.
This test checks for possible frequency dependent errors caused by different CPU-system frequencies.
Therefore the ADC is kept at a constant frequency while the CPU and the digital interfaces operate at
different selected frequencies.
Note: If port activities are involved, the ADC performance may be further reduced. But here also board
effects can play a major rule, e.g.: board layout, supply blocking. For this reason this test case does not
cover influences caused by port activities.
6.5
Reference exceeding
If the positive reference voltage is increased above the ADC supply a parasitic diode starts to conduct. This
is also the case if the negative reference is lower then VSS. This affects the ADC performance, especially at
hot temperature. This test case checks if the error parameters are within specification for the allowed limit of
VREF+ - VDD at room temperature and hot temperature.
6.6
Speed test
This test checks if the ADC can operate within the specified minimum and maximum ADC clock speed
without violation of the maximum allowed error parameters. The decisive factor is the clock at which the ADC
converts, e.g. denoted with internal ADC clock fADCI or analog clock fANA. The low speed limit is a result of
loss of charge during the conversion phase. The maximum speed is limited either by the sampling phase or
the conversion phase or a combination of both.
This test is carried out at room temperature and nominal supplies.
6.7
Selectable reference
Some ADCs have the possibility to select the reference from an analog input channel, e.g. from channel 0.
This introduces a further resistance in the reference path and limits again the maximum ADC speed.
The worst case condition for the maximum ADC speed is normally under the following conditions:

Positive reference input selected from an analog input channel

Supply is at the lower limit

Hot temperature
Application Note
17
V1.1, 2010-11
AP32121
AudoNG, AudoFuture and AudoMax family
References
This test checks if even under this conditions the error parameters are within specification for the maximum
ADC speed.
6.8
Resolution
Beside the maximum possible ADC resolution there is often the possibility to convert at lower resolutions,
e.g. 8 or 10-bit resolution instead of 12-bit. This can be used to reduce the conversion time and to increase
the possible sample rate of the ADC.
This test is a pure functionality check of the possible resolution settings because the ADC simply stops the
conversion earlier at lower resolution. Hence the specified error parameters can be scaled according to the
following formula:
Errormbit resolution  Errorn bit resolution 
1
2
nm
For example an offset error of ± 4LSB12 can be multiplied with 0,25 for conversions with 10-bit resolution.
7 References
[1]
Mark Burns and Gordon W. Roberts An Introduction to Mixed-Signal IC Test and Measurement, Oxford
University Press 2001
[2]
IEEE. 2000. 1241:2000. IEEE Standard for Terminology and Test Methods for Analog-to-Digital
Converters. New York: IEEE.
Application Note
18
V1.1, 2010-11
http://www. inf ineon.com
Published by Infineon Technologies AG