Analog pin seems not to be an input

[BrixInMotion]

I'm having problems reading the analog pins on an XMC4700.

A. XMC4700 Relax Kit for 5V Shields
B. Lib Version 1.5.0
C. Arduino IDE Version 1.8.13
D. No extra lib, Hardware: 100k Potentiometer

All analog pins behave the same: When I ramp up the voltage with the potentiometer everything works fine until 1V. Afterwards it looks like in the picture below.

According to the picture, the Serial output prints random values between 0 and 1023.
For me it looks like the pin is defined as output and is not floating. I'm using minimal code.

uint16_t result = 0;  
void setup() {  
   Serial.begin(115200);  
   wiring_analog_init();  
}  
void loop() {  
  result = analogRead(A5);  
  Serial.println(result);  
  delay(500);  
}

[techpaul]

Random values and seeing 25 MHz oscillations, normally means BAD ground (0V) or other connections. It may not be but that is majority of this type of issues.

Code note you do NOT need the wiring_analog_init() call in your code it is ALWAYS done in the core before setup is called.

Going back through datasheets, circuits etc, I can STATE that analog inputs especially A0 to A21 are PHYSICALLY INPUTS only and cannot physically be configured for ouputs. With exception of A13 and A14 which can be made ONLY to DAC outputs.

I can see nothing YET to suggest the problem is in the software. If I get some time I will run up a board and duplicate this.

[BrixInMotion]

Thanks for the fast response. I would eliminate that it's a Hardware problem as I use exactly the same wiring for an Arduino UNO where it works perfectly. The wiring_analog_init() was only for testing, but nice to know that I don't need it for sure.

[techpaul]

Please NOTE I am a contributor and do not work for Infineon

I am sorry I have over 40 years experience in electronics, computers (from filling rooms to tiny almost wearable) and many many different manufacturers and chip types. Just because hardware worked on one system does not mean ALL connections and parameters are identical, i have seen it too many times especially with breadboards and flying wires. Check the SAME hardware still works again on UNO, could be faulty pot.

The only time i have seen oscillations in the past is mis-wiring, intermittent connections or missing grounds (even on scope probes). I have seen things like worn pot internal conductive tracks or even cracks in tracks.

There is a POSSIBILITY that your 100k pot is TOO large a value, I have not yet foun the input impednace for an Analog input, so your pot could be a mismatch. Many new processors have only 30 to 50k input impedance compared to old style using 1M Ohm or greater input impedance. So when your pot reaches 1V level the impedance is roughly 20% of full travel (20k Ohm) and the impedance mismatch along with stray capacitance (or broken ground) is causing inaccuracy AND oscillations.

First try a 10k or 1k pot instead.

My personal candidates are

1. Broken pot
2. Wiring issue
3. Impedance mismatch
4. other hardware problem.

These types of issues are normally (99.9% of the time) hardware issues.

@mhollfelder and @jaenrig-ifx can anyone find any definitive details on the hardware Analog input impedance in VADC mode for an analog input, this is NOT in any datasheet for XMC4700 or any XMC4 series I can find

[BrixInMotion]

Ok so I just did some exact testing. I used a new 1K linear poti, double and tribble checked the wires and used even a second, unopened XMC4700 kit to check. It behaves exactly as described above (on the new and the old XMC) but ONLY on pins A4 and A5, so I'm sorry about what I've written before, Pins A0- A3 are normal as they should behave.
The Poti is connected between 3,3V and Gnd. The exact thresholds, between the noise appears are above 1,6V and below 2,6V.
I don't think that it has something to do with impedance. The poti just works as a voltage divider and provides an absolute voltage on the analog pin. So there should be no difference in using a 100k or 1k poti.

[techpaul]

Right so we know A0 to A3 are fine A4 and A5 have an issue which MIGHT be a software setup issue, that needs checking as this may be what groups are used in VADC.

What is the performance on A6 to A21 ?

For analogue AND digital circuits, input and output impedance is IMPORTANT. in Digital circuits if a gate or output can drive 10 input loads that is because the parallel input impedance of 10 input loads is at LEAST 100 x larger than the output impedance of the gate or output. This is simple fact borne out of the principles of Optimised Power Transfer,
circuit loading. and simple parallel impedance networks, output and inputs are one COMBINED circuit. For good digital and analogue circuits to avoid loading output impedances are VERY low (< 100 Ohm), input impedances are often measured in the Mega, Giga or some cases Terra Ohms. If you had a Voltage Follower circuit between the pot and the ADC your statement would be true. This sort of circuit is often an op-amp in Unity Gain configuration or NPN transistor
in Common Collector configuration. With direct connection of Pot to ADC input it is ONLY true when the input impedance of the ADC is at LEAST 100 times larger than the output impedance of the circuit driving it. This is basic electronics.

To model the circuit you have a pot which is equivalent to TWO series resistors, with the one tied to 0V in parallel with the ADC input also going to 0V. Lets call the two pot resistors R1 (tied to supply) and R2 (tied to 0V, and the ADC input impedance Zin, which means Zin is in parallel with R2, and R1 is feeding the parallel combination. So as R1 increases R2 decreases in a matching ratio, also assume the pot has a linear resistance track.

You are setting the voltage by reading a meter or scope on the pot, but with different Zin or total Pot resistance the angular position to achieve the same voltage out is DIFFERENT. You will also note in regions it is easier or harder to achieve small incremental step changes (e.g. 1.25V to 1.26V) The only thing you can say about pots of different values, is the ratio of the resistances of the POT, should be the same for degrees of rotation. The parallel effect of Zin and R2 will be DIFFERENT depending on the value of the pot. Also the same value pot use on two different ADCs with different Zin WILL be different voltage relationship to position of pot rotation.

So if we use your example of 100k pot and for easier calculations the pot has 250 degrees of rotation between ends, then that gives 400 Ohms per degree of rotation. Inversely that is 2.5 degrees per kOhm. So calculating the data for a Zin of the ADC being 30k or 1M gives different responses related to degrees of rotation. See the graph below

Impedance mismatch on ADCs is a common issue I see mainly on newer micros from ANY manufacturer, and is often overlooked by those in a hurry or inexperienced. This will cause wrong readings or apparent incorrect response or accuracy over range.

[BrixInMotion]

Nice explanation, thank you! I checked the behavior with DAVE and have the same issue there, so it might be a bigger problem. In Dave as well only A4 and A5 are affected. I did not test all analog pins so far, but e.g. 14.6 is fine as well. Can I test the other pins with XMC for Arduino as well (A6 to A21)? According to the wiki they're not supported I thought?

[techpaul]

All A0-A21 is supposed to be support for XMC4700 boards extract from pins_arduino.h

    /* 17  */   {XMC_GPIO_PORT14 , 0},  // A0 / ADC Input               P14.0 (INPUT ONLY)
    /* 18  */   {XMC_GPIO_PORT14 , 1},  // A1 / ADC Input               P14.1 (INPUT ONLY)
    /* 19  */   {XMC_GPIO_PORT14 , 2},  // A2 / ADC Input               P14.2 (INPUT ONLY)
    /* 20  */   {XMC_GPIO_PORT14 , 3},  // A3 / ADC Input               P14.3 (INPUT ONLY)
    /* 21  */   {XMC_GPIO_PORT14 , 4},  // A4 / ADC Input               P14.4 (INPUT ONLY)
    /* 22  */   {XMC_GPIO_PORT14 , 5},  // A5 / ADC Input               P14.5 (INPUT ONLY)
    /* 23  */   {XMC_GPIO_PORT3 , 12},  // SPI_SS_2                     P3.12
    /* 24  */   {XMC_GPIO_PORT5 , 9},   // Additional LED1              P5.9
    /* 25  */   {XMC_GPIO_PORT5 , 8},   // Additional LED2              P5.8
    /* 26  */   {XMC_GPIO_PORT15 , 13}, // Additional BUTTON1           P15.13 (INPUT ONLY)
    /* 27  */   {XMC_GPIO_PORT15 , 12}, // Additional BUTTON2           P15.12 (INPUT ONLY)
    /* 28  */   {XMC_GPIO_PORT4 , 1},   // SPI_SS_3 (SD CARD)           P4.1
    /* 29  */   {XMC_GPIO_PORT3 , 5},   // SPI-MOSI (SD CARD)           P3.5
    /* 30  */   {XMC_GPIO_PORT4 , 0},   // SPI-MISO (SD CARD)           P4.0
    /* 31  */   {XMC_GPIO_PORT3 , 6},   // SPI-SCK  (SD CARD)           P3.6
    /* 32  */   {XMC_GPIO_PORT1 , 6},   //                              P1.6
    /* 33  */   {XMC_GPIO_PORT1 , 7},   //                              P1.7
    /* 34  */   {XMC_GPIO_PORT1 , 4},   // USB Debug RX                 P1.4
    /* 35  */   {XMC_GPIO_PORT1 , 5},   // USB Debug TX                 P1.5

	//Additional pins for port X1 starting here
	/* 36  */   {XMC_GPIO_PORT3 , 4},   // PWM42-2                      P3.4
	/* 37  */   {XMC_GPIO_PORT0 , 5},   // I2C_1 SDA  // SPI_4 MOSI     P0.5
	/* 38  */   {XMC_GPIO_PORT0 , 3},   // PWM80-2                      P0.3
	/* 39  */   {XMC_GPIO_PORT0 , 1},   // PWM80-1                      P0.1
	/* 40  */   {XMC_GPIO_PORT0 , 10},  //                              P0.10
	/* 41  */   {XMC_GPIO_PORT3 , 2},   //                              P3.2
	/* 42  */   {XMC_GPIO_PORT3 , 1},   //                              P3.1
	/* 43  */   {XMC_GPIO_PORT15 , 6},  // A6 / ADC Input               P15.6 (INPUT ONLY)
	/* 44  */   {XMC_GPIO_PORT15 , 5},  // A7 / ADC Input               P15.5 (INPUT ONLY)
	/* 45  */   {XMC_GPIO_PORT15 , 3},  // A8 / ADC Input               P15.3 (INPUT ONLY)
	/* 46  */   {XMC_GPIO_PORT14 , 15}, // A9 / ADC Input               P14.15 (INPUT ONLY)
	/* 47  */   {XMC_GPIO_PORT14 , 13}, // A10 / ADC Input              P14.13 (INPUT ONLY)
	/* 48  */   {XMC_GPIO_PORT14 , 7},  // A11 / ADC Input              P14.7 (INPUT ONLY)
	/* 49  */   {XMC_GPIO_PORT15 , 15}, // A12 / ADC Input              P15.15 (INPUT ONLY)
	/* 50  */   {XMC_GPIO_PORT14 , 9},  // DAC0 // A13 / ADC Input      P14.9 (INPUT ONLY)
	/* 51  */   {XMC_GPIO_PORT2 , 13},  //                              P2.13
	/* 52  */   {XMC_GPIO_PORT5 , 10},  //                              P5.10
	/* 53  */   {XMC_GPIO_PORT5 , 11},  // PWM80-0                      P5.11
	/* 54  */   {XMC_GPIO_PORT1 , 14},  //                              P1.14
	/* 55  */   {XMC_GPIO_PORT14 , 8},  // DAC1 // A14 / ADC Input      P14.8 (INPUT ONLY)
	/* 56  */   {XMC_GPIO_PORT15 , 14}, // A15 / ADC Input              P15.14 (INPUT ONLY)
	/* 57  */   {XMC_GPIO_PORT14 , 6},  // A16 / ADC Input              P14.6 (INPUT ONLY)
	/* 58  */   {XMC_GPIO_PORT14 , 12}, // A17 / ADC Input              P14.12 (INPUT ONLY)
	/* 59  */   {XMC_GPIO_PORT14 , 14}, // A18 / ADC Input              P14.14 (INPUT ONLY)
	/* 60  */   {XMC_GPIO_PORT15 , 2},  // A19 / ADC Input              P15.2 (INPUT ONLY)
	/* 61  */   {XMC_GPIO_PORT15 , 4},  // A20 / ADC Input              P15.4 (INPUT ONLY)
	/* 62  */   {XMC_GPIO_PORT15 , 7},  // A21 / ADC Input              P15.7 (INPUT ONLY)

The holes exist for the extra channels on the multi-way connector mountings and are detailed in the schematic for the board. When I get some time XMC4700 pin out needs second page for the extended pins. Recently did a minor update for XMC4400 pinout diagram.

I suspect the issue IF software will be in XMCLIB and more than likely VADC group related, so a scan of which ones work and which ones don't will help give data on what is actually working. This may be a hardware related issue where someone has not documented something properly, having just found one on a different board.

Right back to other paying work and documentation

[BrixInMotion]

All right, I tested all analog pins and can say for sure that the only faulty pins on my both XMC4700 5V kits are A4 and A5. A0 to A3 and A6 to A21 work fine.
I tested a Lite Kit as well, interestingly no pin there seems faulty. So I will check either other 5V kits ore the PCB design, if there are differences between the lite and the 5V. As this is no topic concerning the XMCforArduino Lib anymore, I will close this issue. Thanks for your support @techpaul !

[techpaul]

Glad this is sort of sorted.

Contact again if need assistance

[BrixInMotion]

We found the reason for this issue: there are two traces too many on the board, that connect P14.5 (A5) with P0.13(SCL) and P14.4(A4) with P3.15(SDA). This is a unnecessary limitation of the board, as the XMC4700 would have separate pins for ADC and I2C unlike the ATmega328.
A solution here, in order to use both analog pins and I2C is to take a knife and carefully cut the two traces near R104 and R105 shown in the picture .

[mhollfelder]

Hi BrixInMotion,

This connection is in place as for the Arduino Uno Rev3 pin-out, these pins are connected and shared I2C/Analog pins as well:

https://content.arduino.cc/assets/UNO-TH_Rev3e_sch.pdf

The I2C pins on the top are connected to the pins on the bottom of the board itself.
You can also see this from the pin-out as they are using the same signals:

https://content.arduino.cc/assets/Pinout-UNOrev3_latest.png

As some shields rely on this connection, this is implemented here by having this on-board pin connection.
You can either cut it or you set the I2C pins to tri-state to not influence the analog measurements (which blocks using the I2C, of course).

I hope I could give some background why this connection is in place here.

Thanks and best regards,

Manuel

[techpaul]

Hi BrixInMotion,

This connection is in place as for the Arduino Uno Rev3 pin-out, these pins are connected and shared I2C/Analog pins as well:

https://content.arduino.cc/assets/UNO-TH_Rev3e_sch.pdf

This is where there is a Yes and No answer, or as is sometimes said "Comparing Oranges and Apples".

On AVR UNO this is like some of the XMC1xxx boards an INTERNAL alternate pin configuration, on the XMC4700 it is linking pins externally, and at best should have been onboard jumper or individual 0R resistors that could be removed.
Otherwise we are saying you cannot have these two analog inputs (see later). Not checked whether XMC1400 Arduino Kit is Alternate pin function or pins externally connected. This has OTHER issues NOT DOCUMENTED.

The I2C pins on the top are connected to the pins on the bottom of the board itself.
You can also see this from the pin-out as they are using the same signals:

https://content.arduino.cc/assets/Pinout-UNOrev3_latest.png

As some shields rely on this connection, this is implemented here by having this on-board pin connection.
You can either cut it or you set the I2C pins to tri-state to not influence the analog measurements (which blocks using the I2C, of course).

Setting pins to Tri-state is NOT DIRECTLY supported in XMC-for-Arduino, setting as input will add a load and leakage to the analogue signal.

The NEAREST setting is setting the pins to OUTPUT_OPENDRAIN and writing a 1 (HIGH) to the port pin, to turn output off.
The length of tracks may well add a lot more noise to analogue readings, as well as any noise leakage from pin(s).

The files like pins_arduino.h do NOT properly document this, nor does

• Pin out diagram(s) (including XMC1400 Arduino Kit, XMC4400, possibly others)
• Wiki
• Similarly AREF pin is NOT properly documented best use is leave OPEN do not touch

No mention is made of fact that you cannot use I2C and these analogue inputs at the same time in any part of the documents or Wiki.

This needs a documentation overview WITH notes on track cuts where pins are externally connected as well as how to set I2C/GPIO pins to avoid interacting with analogue pins.

This artefact does NOT affect the following boards at current revisions

• Any of XMC1100 boards (XMC1100 Boot Kit is the same format)
• XMC1300 Boot Kit

This artefact DOES affect

• XMC4700 boards
• XMC4400 Platform 2 Go
• AMC1400 Arduino Kit (XMC1400 Boot kit is discontinued and not checked)

[mhollfelder]

I agree to techpaul's comments, ideally it would be an alternate multiplexed pin function which is properly documented.
For the Relax Kit, I cannot fully judge what was the reasoning behind making such a choice of connection.

I can only comment on why the connection is in place, not why the connection was chosen as it is.
Ideally, of course you should have a 0 Ohm resistor or a switch as the best solution.

XMC1400 Arduino Kit has the SAME pins P2.0 and P2.1 connected on the board between I2C and ADC pins.
So it is an alternative pin function and I2C and ADC functions can be activated via these functions.

Tristate can be set via pinMode(<pinNumber>, XMC_GPIO_MODE_INPUT_TRISTATE);, the regular definitions can be used here.
Of course, this has to be documented properly, better would be to include the tristate as an optional mode if required.

Fully agree, it has to be properly documented as the only source where you can find it are the user manuals.

I added two chapters to the XMC1100 Boot Kit and XMC4700 Relax Kit Wiki for documentation.
I will check for the XMC4400 Platform 2 Go and also document it (Edit: DONE).

Thanks for investigating this and sorry that it caused such troubles.

Amendment: The XMC4400 board has to be cut here (blue routes on the back side of the PCB), fyi: