Re: [PATCH v2 5/5] iio: mcp9600: Add support for IIR filter

[Ben Collins <bcollins@kernel.org>]

[-- Attachment #1: Type: text/plain, Size: 2530 bytes --]

On Thu, Aug 14, 2025 at 09:06:39AM -0500, Ben Collins wrote:
> On Wed, Aug 13, 2025 at 05:52:04PM -0500, David Lechner wrote:
> > On 8/13/25 10:15 AM, Ben Collins wrote:
> > > MCP9600 supports an IIR filter with 7 levels. Add IIR attribute
> ... 
> > >  static int mcp9600_read(struct mcp9600_data *data,
> > > @@ -186,6 +189,9 @@ static int mcp9600_read_raw(struct iio_dev *indio_dev,
> > >  	case IIO_CHAN_INFO_THERMOCOUPLE_TYPE:
> > >  		*val = mcp9600_tc_types[data->thermocouple_type];
> > >  		return IIO_VAL_CHAR;
> > > +	case IIO_CHAN_INFO_LOW_PASS_FILTER_3DB_FREQUENCY:
> > > +		*val = data->filter_level;
> > 
> > We can't just pass the raw value through for this. The ABI is defined
> > in Documentation/ABI/testing/sysfs-bus-iio and states that the value
> > is the frequency in Hz.
> ...
> > For example, for 3 Hz sample rate (18-bit samples), I got:
> > 
> >   n  f_3dB (Hz)
> >   1  0.58774
> >   2  0.24939
> >   3  0.12063
> >   4  0.05984
> >   5  0.02986
> >   6  0.01492
> >   7  0.00746
> > 
> > I had to skip n=0 though since that is undefined. Not sure how we
> > handle that since it means no filter. Maybe Jonathan can advise?
> 
> Thanks for notes. If I'm reading for datasheet formula right,
> 
> k = 2 / (2^n + 1)
> 
> So n=0 would be k=1. I did this formula for n=[0-7] and get:
> 
> n	k
> 0	1.00000
> 1	0.66667
> 2	0.40000
> 3	0.22222
> 4	0.11765
> 5	0.06061
> 6	0.03077
> 7	0.01550
> 
> I'm not versed in filter frequency, but would these be the correct
> values to use for the coefficients?

This seems to be what I was looking for:

Got it. For a 1-pole IIR low-pass of the form
y_t = y_{t-1} + k(x_t - y_{t-1}), the –3 dB cutoff is

f_c = \frac{-\ln(1-k)}{2\pi}\,f_s

Using your k=\frac{2}{2^n+1} and a sample rate f_s = 3\ \text{Hz}, the equivalent –3 dB cutoff frequencies are:

n	k		f_c (Hz)	time constant τ (s) = 1/(2π f_c)
0	1.0000000000	∞		0.0000
1	0.6666666667	0.5245487288	0.3034
2	0.4000000000	0.2439012692	0.6525
3	0.2222222222	0.1199938006	1.3264
4	0.1176470588	0.0597609987	2.6632
5	0.0606060606	0.0298512716	5.3316
6	0.0307692308	0.0149219903	10.6658
7	0.0155038760	0.0074605397	21.3329

Notes:
	•	n=0 (k=1) is just “track the input” (no smoothing), so f_c\to\infty.
	•	If you’re using a different sampling rate, multiply the f_c values above by \frac{f_s}{3}.

-- 
 Ben Collins
 https://libjwt.io
 https://github.com/benmcollins
 --
 3EC9 7598 1672 961A 1139  173A 5D5A 57C7 242B 22CF

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]