LGT8F328P Nano AD_REF_SOURCE problem

[Don Rosselot]

In using the internal voltage reference on the LGT8Fx , I am getting conflicting results with the different sources.
When I use 1024 or AVCC as the reference, the ADC results are spot on, but when I use the 2048 or 4096 reference, the ADC results are consistently low by about 6%. I tested the same chip on the Arduino IDE and all the reference sources provide consistent results. The ASM indicates the correct assignment is made to REFS2:0, so I don't know what is going on. I have minimized the program to just using ReadAD12() and a serial connection to the terminal, and the errors do not go away.
Any advice would be appreciated!
I have been enjoying GCB since 2007. A big thanks to the whole team for their efforts! I have also found Evans videos most informative!

[Anobium]

Interesting.

Got any ASM from Arduino ? That may give a clue.

[Don Rosselot]

I have tried to get ASM, but apparently there is no way to get that, I have tried. I know it used to be an option, but not anymore.

[Anobium]

Nice.

I can look Sunday morning London time.

Post your program please. I am assuming no external Vref.

[William Roth]

Does Arduino even generate ASM ? I found this but it may be dated.....

https://forum.arduino.cc/t/how-can-one-view-the-assembly-code-output-from-the-arduino-ide/50302

[Anobium]

I disassemble using avr-objdump. I got into this when I wrote the LGT support within GCCode.

[Don Rosselot]

I have 'disassembled' the arduino hex file with avr-objdump (lgt_asm.txt), and also attached the arduino sketch, and the GCB sketch. (I think I attached them correctly)
Good luck!

[Don Rosselot]

I am sending files one at a time, I guess...

[Don Rosselot]

And the disassembled hex file....

[Anobium]

A couple things I have learnt this morning.

For a VREF of 4.096v the VCC must be greater than 4.7v
For a VREF of 2.048v the VCC must be greater than 2.7v

But, I do not think the AVR code is correct for the LGT. I am working a change here. You will have to test.

[Anobium]

OK. Here is a test program.
This uses the vector ADREADPREREADCOMMAND to insert a method called LGTSetVREF ( in argREFS as Byte ) . Its use is very obvious. Once we have this sorted then we can move bac into the main line code.

The issue is that the ADC vref is not correct and the calibration register is missing from the .DAT So, you need to update the .DAT In Line 80 in the C:\GCB@Syn\GreatCowBasic\chipdata\lgt8f328p.dat add the VCALs as follows. Please inspect the datasheet to ensure I got these correct.

VCAL,200
VCAL1,205
VCAL2,206
VCAL3,204
'End of chip specific changes

The code attached is my test GCCode for you. This hooks ADREADPREREADCOMMAND then calls LGTSETVREF ( AD_REF_4096 ) - you must try the other values AD_REF_2048 and AD_REF_1024. Then this sets the registers. Please code review etc. As I said.. if this works then we need to move back into the main line code and fix all the DAT files.

Evan

---

'Evan - trial fixes using ADREADPREREADCOMMAND

#DEFINE ADREADPREREADCOMMAND  LGTSETVREF (  AD_REF_4096 )


Sub LGTSetVREF ( in argREFS as Byte )

    // update calibration as required
  REFS2 = 0

    // set voltage reference
    ADMUX = ADMUX & 0x3f

    if ( argREFS = AD_REF_4096 ) Then
        VCAL = VCAL3    // calibration for 4.096V
        REFS2 = 1       // See register definition of ADCSRD and ADMUX
    REFS0 = 0
    REFS1 = 0
    else if ( argREFS = AD_REF_2048 ) Then
        VCAL = VCAL2    // calibration for 2.048V
    REFS0 = 0
    REFS1 = 1
    else
        VCAL = VCAL1;   // calibration for 1.024V
    REFS0 = 1
    REFS1 = 1
    End if


End Sub

[Anobium]

The root cause is that ADC library DID NOT read the calibration data, and, set the VCAL appropriately. And, VCALs were missing from the DAT file.

Please test and resolve - if you are able/willing please resolve in A--D.H for everyone. I have attached a new version based upon the previous post

Evan

[Don Rosselot]

Success! That fixed the 2.048 and 4.096 reference issues. Accuracy went from 6% to .5%!
And yes, just modifying the chip data file and using the new a-d.h file, all sources work perfectly.
Thank you, Evan for you speedy solution!

[Anobium]

Brilliant. You analysis and your files really help. Thanks to you . it only took a few moments.

I will update all the DATs and update the master a-d.h

Now, go to this url and leave a very positive comment - https://sourceforge.net/projects/gcbasic/reviews/new?stars=5

[stan cartwright]

When will a new gcb release that has this sorted be posted please?

[Anobium]

Great question. You can get the latest build from the Release Candidate - https://sourceforge.net/projects/gcbasic/files/Release%20Candidates/Patches/gcb%40Syn.zip/download

That the fixes and a few more before this issue sprung up. :-)

[stan cartwright]

Thanks, I'll wait for the next release. I use a lgt328nano but not tried a-d and it has
12bit a-d so nice this problem has been identified and sorted.

[Anobium]

I wont be long. :-)