| View previous topic :: View next topic |
| Author |
Message |
asmboy
Joined: 20 Nov 2007
Posts: 2128
Location: albany ny
|
| pic hardware may vary - the same code - different results |
 Posted: Wed Sep 12, 2012 11:49 am |
 |
|
I have a code module that does low side PWM, effective "buck" modulation, of a 36v load, using a fast 10a N channel mosfet, with a gate capacitance of about 400pf.
The code has been in use for several years now using a 16F886 @ 8 MHZ on internal clock.
So far so good.
I recently re-purposed it to drive the same load but using a 16HV616 in a 'mini' version of the product.
In the original code, run on the 886, when I wanted to zero out the load I would execute:
and after rechecking the old design, there is no output voltage.
The 16HV616 version had an odd residual 30 mv when shut down,
and on closer inspection the buck storage inductor was being tickled with a ~ 130 ns pulse, and producing just a tiny discontinuous ringing voltage - in spite of the 0 setting.
The cure was simple.
Instead of set_pwm1_duty(0);
I switched to setup_ccp1(CCP_OFF);
Which cures it, but the CCP hardware in the HV616 seems to switch on for 1 cycle of the master Fosc clock (1/4 of a full inst cycle) - and then crisply back off, but not in the 886.
Be it a genuine silicon error or not- I know the ONLY way I am going to
shut off the CCP PWM channel going forward.
Comments ?? |
|
 |
temtronic
Joined: 01 Jul 2010
Posts: 9248
Location: Greensville,Ontario
|
|
| Posted: Wed Sep 12, 2012 2:11 pm |
|
|
my comments..
1) I wonder why you changed processors (and PCBs)? You had a great working unit only to spend $$time in R&D to make a new board that doesn't do the job. Sigh...been there done that.
I find it very very hard to justify 'jumping ship', making a 'mini' when you have a tried and true product.
Will you save bags of $$$ after factoring in R&D,headaches,loss sleep,etc.?
2) changing silicon is a risky prospect, as you've found out. You've found one 'gotcha'...think they'll be others ?
curious like my cat.
jay |
|
|
asmboy
Joined: 20 Nov 2007
Posts: 2128
Location: albany ny
|
|
| Posted: Wed Sep 12, 2012 4:54 pm |
|
|
| Quote: |
I wonder why you changed processors (and PCBs)?
|
The original 886 version supported TWO PWM modulated devices, numerous solenoid valves and analog monitoring circuitry - all under pic control.
A new simple variant of the product was required that had to be smaller and cheaper to build. Closed loop feedback was added for speed control, and the original design's other features made for a larger PCB too.
The CUSTOMER for the design did not need or want all that stuff in the 'mini' product and was willing to source the $$ to do a new one. What is not to like there ?? |
|
|
Ttelmah
Joined: 11 Mar 2010
Posts: 19554
|
|
| Posted: Thu Sep 13, 2012 12:53 am |
|
|
Old story.
If you are using more than 8bits on the PWM at any point, then 'set_pwm1_duty(0)' is _not_ guaranteed to make the output zero. You need 'set_pwm1_duty(0L)'. The effect will depend on what is in the two low bits of the PWM. That it worked on the old chip, was 'luck'.
Sending set_pwm1_duty(0), only updates the high eight bits of the PWM registers. If there is a value in the low two bits you will still get an output...
Best Wishes |
|
|
FvM
Joined: 27 Aug 2008
Posts: 2337
Location: Germany
|
|
| Posted: Thu Sep 13, 2012 9:20 am |
|
|
Unfortunately the specific behaviour of set_pwm_duty() isn't described in the CCS C online help and wrongly described (MSB instead of LSB) in the pdf manual.
| Quote: | | Writes the 10-bit value to the PWM to set the duty. An 8-bit value may be used if the most significant bits are not required. |
IMHO it's a strange way of function overload, apart from missing or incorrect documentation. |
|
|
asmboy
Joined: 20 Nov 2007
Posts: 2128
Location: albany ny
|
|
| Posted: Thu Sep 13, 2012 9:52 am |
|
|
| Code: |
setup_ccp1(CCP_OFF);
|
after careful checking - there seems to be no glitching associated
with this method of shutting off the pwm and is a very simple software trap for a zero argument
actually i KNEW not to use an 8bit value though.
AND
i should have been more accurate in my posting.
what i listed was simplified ( i thought) for simplicity:
the ACTUAL code that fails is this:
unsigned int16 pwmval;
.....
pwmval=0;
set_pwm1_duty(pwmval);
ie:
i HAVE been using a long int - that when it becomes ( or is set to ) zero still glitches with the HV616 , but NOT with the 16F886 ( and this morning )
checked with 18F2620 also
using the HV616 - i can nicely resolve the pulse timing difference between an argument of '1' and zero as well.
with my timer prescalar i do get the full 10 bit resolution as expected -
only the HV616 seems to do this cute trick |
|
|
newguy
Joined: 24 Jun 2004
Posts: 1911
|
|
| Posted: Thu Sep 13, 2012 10:45 am |
|
|
| I hope you've reported this to Microchip (I'm assuming they're unaware of this and that it's not mentioned in that chip's errata already). |
|
|
FvM
Joined: 27 Aug 2008
Posts: 2337
Location: Germany
|
|
| Posted: Thu Sep 13, 2012 11:18 pm |
|
|
| Quote: | | using the HV616 - i can nicely resolve the pulse timing difference between an argument of '1' and zero as well. |
You say, the glitch is only a fraction of '1' pulse width? |
|
|
Ttelmah
Joined: 11 Mar 2010
Posts: 19554
|
|
| Posted: Fri Sep 14, 2012 4:38 am |
|
|
I think he is seeing exactly one master clock cycle (125nSec at 8MHz). He says:
"a ~ 130 ns pulse",
which would correspond exactly.
This is why I felt it was the low bit of the register not being cleared.
Things that get possible:
This is an ECCP, not a CCP. It's be interesting to try setting it up as a single output ECCP, rather than using the PWM 'compatibility' mode, and see if this changes things.
It's also be interesting to try manually clearing the two low bits of the PWM duty cycle, to see if there is a code problem here.
If it was staying low, rather than high, I'd suspect that the ECCP dead band delay was somehow getting invoked, but this shouldn't affect compatibility mode, and would give a low, not a high pulse. It would though be interesting to see if clearing this register had an effect.
'No output', is explicitly in the data sheet for both the chips:
"The CCP1 pin is set. (Exception: If the PWM duty
cycle = 0%, the pin will not be set.)".
Best Wishes |
|
|
asmboy
Joined: 20 Nov 2007
Posts: 2128
Location: albany ny
|
|
| Posted: Fri Sep 14, 2012 12:22 pm |
|
|
| Quote: | a single output ECCP, rather than using the PWM 'compatibility' mode, and see if this changes things.
It's also be interesting to try manually clearing the two low bits of the PWM duty cycle, to see if there is a code problem here.
|
Excellent ideas.
I have to be away for a few days but will test these ideas when I am back. |
|
|
asmboy
Joined: 20 Nov 2007
Posts: 2128
Location: albany ny
|
|
| Posted: Tue Sep 18, 2012 7:00 pm |
|
|
as fate would have it - i am having a devil of a time trying to configure
the HV616 in ECCP compat mode with ONLY channel one output active
all other possible EPWM pins on the part are of course used for other stuff
and i am rather paranoid about accidentally activating one of the other EPWM outputs in the extended mode .
so my puzzle is which CCS setup argument actually commands
"Single output; P1A modulated; P1B, P1C, P1D assigned as port pins"
NONE of the hex equivalent arguments for setup in the device file seem to set the CCP1CON register as i 'think' it ought to be for EPWM :
i get 0x0C as the argument to use but of course thats what i have been using in standard mode ....
#define CCP_PWM 0xC
yet the data sheet seems to be saying it IS selecting EPWM already ..??
the CCP_PWM_PLUS_1 does not seem to do it for the HV616 and since i have functional, deliverable code - my interest in potentially damaging a working prototype with errant programming has been a factor in trying to find out more. ..... |
|
|
|
FAQ 
Search
Register
Profile 
Log in to check your private messages 
Log in
