Defining ADC input voltages via input VCD file

[backtail]

I have been unable to find a conclusive answer via official docs, source code, blog posts by enthusiasts, etc. whether I can define ADC input from a VCD file for a simulation or not, so that I can pass it with simavr path/to/elf -i input.vcd:

$timescale 1ms $end
$scope module logic $end
$var wire 16 ! adc0 $end
$upscope $end
$enddefinitions $end
#0
b0 !
#100
b1000 !
#200
b10000 !
#500
b100000 !

I am successfully simulating input pin changes by defining iogPORT_PINNUMBER variables, as suggested in this issue.

When looking at the implementation of the IO pins, I found this definition:

simavr/simavr/sim/avr_ioport.h

Line 46 in 193cba5

#define AVR_IOCTL_IOPORT_GETIRQ(_name) AVR_IOCTL_DEF('i','o','g',(_name))

That would suggest to me, that maybe this definition should make it possible to hook up external VCD changes to the internal ADC interrupts - if I can name the variables correctly.

simavr/simavr/sim/avr_adc.h

Line 57 in 193cba5

#define AVR_IOCTL_ADC_GETIRQ AVR_IOCTL_DEF('a','d','c','0')

However, I don't get it to work with a file like above. That means either I am doing something wrong or it is not possible to do like this in the first place.

[gatk555]

The code is there, so it should work, and may actually be working. By sending input like: "b10000 !" you are providing some possibly unexpected values, maybe even a NaN! It is an analogue input, so the preferred format is something like "r1.2 !", with the voltage in ordinary decimal or scientific format. The "r" form is not in Verilog 2001, but may be in later versions. Icarus Verilog and Ngspice certainly produce it and it is included in the Wikipedia article on VCD.

Edit: I missed a detail: in the file heading you need something like "$var real 32 > adc0_5 $end", with the signal name also specifying the ADC multiplexer port. And "real", not "wire".

[backtail]

Tried that, however could not get it to work. My VCD file looks as follows:

$timescale 1ms $end
$scope module logic $end
$var real 32 a adc0_0 $end
$var real 32 b adc0_1 $end
$var real 32 c adc0_2 $end
$var real 32 d adc0_3 $end
$var real 32 e adc0_4 $end
$var real 32 f adc0_5 $end
$var real 32 g adc0_6 $end
$var real 32 h adc0_7 $end
$upscope $end
$enddefinitions $end
$end
#0 
r0.0 a
r0.0 b
r0.0 c
r0.0 d
r0.0 e
r0.0 f
r0.0 g
r0.0 h
#100
r1000.0 a
r1200.0 b
r2500.0 c
r5000.0 d
r300.0 e
r4100.0 f
r3300.0 g
r5000.0 h

I use the technique that incorporates the simavr tracing within the binary like so:

AVR_MCU(F_CPU, "atmega1281");
AVR_MCU_VCD_FILE("output.vcd", 100000);
AVR_MCU_VOLTAGES(5000, 5000, 5000);

And I read the all the ADC values left adjusted in a periodic manner like this:

uint8_t __adc_get(adc_channel_t channel) {
    ADMUX &= ~ADC_MUX_CH_MASK;
    ADMUX |= channel & ADC_MUX_CH_MASK;

    ADCSRA |= (1 << ADSC);
    while ((ADCSRA & (1 << ADSC))) {}

    return ADCH;
}

I use AVCC as a reference and trying to disable digitial filtering also does not help.

[gatk555]

It works. I modified one of the test programs and your VCD file so that the values change during the loop of conversions:

../simavr/run_avr atmega644_adc_test.axf -i /tmp/t.vcd
Loaded 2090 bytes of Flash data at 0
Loaded 176 bytes of Flash data at 0x82a
Read 8 ADC channels to test interrupts..
Chan 0 got 0..
Chan 1 got 0..
Chan 2 got 0..
Chan 3 got 1023..
Chan 4 got 93..
Chan 5 got 1023..
Chan 6 got 1023..
Chan 7 got 1023..
All done. Now reading the 1.1V value in pooling mode..
Read ADC value 0155 = 1098 mvolts -- ought to be 1098..

Files here: t.zip

One possible source of trouble: the simulator exits at the end of the file. I needed to add a dummy entry with a large time stamp to make it run to completion.