ADC does not work with external reference

[Ttelmah]

Good.

I have to agree with Temtronic on this. It's so pointless not to have a few of these adapters in stock. I also keep a couple of serial LCD's for the same reason.
I use the boards rather than cables:

<http://www.ebay.co.uk/itm/CH340G-RS232-Upgrade-To-USB-TTL-Auto-Converter-Adapter-STC-Brush-Module/252424791118?_trksid=p2047675.c100005.m1851&_trkparms=aid%3D555018%26algo%3DPL.SIM%26ao%3D2%26asc%3D20140122125356%26meid%3D3ac58bba8c2b4630a24d9914f065cb96%26pid%3D100005%26rk%3D3%26rkt%3D6%26sd%3D252461894405>

These are great since they operate on 3.3v or 5v, and at under £1 each in the UK (including delivery!), you can't even buy half the parts for the price!...
Great thing is that while for full I/O a proper UART is best, for simple 'got here' messages, a software UART on the PIC is easy to send debug messages, and a simple 5pin socket on the board, or a line on a plug in breadboard, makes connection quick... . You only actually need just one pin on the PIC to send this type of message.

[PCM programmer]

Regarding the Vref+ bug in set_adc_channel() for the 16F914, 16F917, and 16F946, which was discussed earlier in this thread:
CCS emailed me and said it will be fixed in the next release, vs. 5.075.

[[email protected]]

I'm compiling for the 18F87J50 and using an external Vref for the 1st time.
My CCS version is 4,3,0,332
The Vref don't work. It appears to be turned off by set_adc_channel(0);

Failing code:
...
setup_adc_ports(san0 | san1 | san2 | san4, vss_vref);
set_adc_channel(0);
...

A look at the .LST file shows:
... set_adc_channel(0);
00234: MOVLW 00
00236: MOVWF 01
00238: MOVF FC2,W
0023A: ANDLW 83 // My comnt: Anding 0x80 turns off bit6 and bit6 allows the external Vref
0023C: IORWF 01,W
0023E: MOVWF FC2
...

The problem goes away when I code this:
...
setup_adc_ports(san0 | san1 | san2 | san4, vss_vref);
set_adc_channel(0);
setup_adc_ports(san0 | san1 | san2 | san4, vss_vref);
...

The second setup restores ADCON0 bit6.
This appears to be a compiler bug. Should AND the ADCON0 with 0xC0.
Bob

[PCM programmer]

You have CCS vs. 4.xxxx. I tested it with vs. 5.071, and it's fixed by
that time. The 0xC3 mask shown below is correct:

[[email protected]]

Thanks PCM Programmer. Yes, I understand that I have an old version.
Bob