wrong running average parameter calculation

Barricade Question

Apr 29, 2013 00:00

Hello,

How can I calculate the dsp data values for running average.

The help file say following for : 1-RMSTC / (10*Fs)

Comparing the calculated value for

Fs = 48000

RMSTC = 5000

the result is 0.9895833, but result from Sigmastudio is 0.024275.

I get the same wrong results for other calculations.

What's the correct formula for RMS TC ?

Report

KJBob Reply

Apr 29, 2013 00:00

Hi,

I checked the wiki but the incorrect formula is there also. Empirically it appears to be approximately:

Parameter = RMSTC / (4.3*Fs)

Bob

Report

Barricade Reply

Apr 29, 2013 00:00

Hi Bob,

thank you for your fast answer. Is it possible to give (later) the correct formula please.

I've done in my micrcontroller code some unittests, that checks the (generated) dsp data words.

Also I've found that the first and last dsp data word (4 Bytes) are in 5.23 format

but the second dsp data word (4 Bytes) has to be send in 2.80 format.

Report

KJBob Reply

May 1, 2013 00:00

Hi Barricade,

JeradL provides the exact formula in the post https://ez.analog.com/message/14502

(this is for the limiter block, but the RMSTC calculation for all the dynamic blocks is the same).

Parameter = abs(1.0 - 10^(RMSTC/(10*fs))) (edited 6/2/2013)

This one works, as shown in the table below:

The exact formula's downside is it involves a 10^x function, which hopefully is available in your micro's C compiler. I honestly don't understand why my empirical formula (with only a simple division) comes close. There must be a theoretical connection between the two, but I'll have to leave that to some curious soul who had decent math professors in college.

As you correctly noted, the middle parameter is in integer (28.0) format. In general, parameter inputs are in whatever form is most convenient tor the runtime DSP algorithm, to save on instructions.

Best regards,

Bob

Message was edited to add brackets to the exact formula