conversion from A-weighted to flat RMS - Page 2

tschrama · 12th October 2003, 12:42 PM

Ha

it seems that

DIN IEC 651 = C weighting

back to google...

Steven · 12th October 2003, 03:21 PM

The difference in dB between the results of different noise measuring filters depend on the character of the noise. For instance, if your noise is completely white, then there is a difference between A-weighting and flat (20Hz-20kHz bandlimited) of approximately 3dB (dBA is 3 dB better than flat). CCIR gives far worse results.
If your noise is not white, the difference might become much more, e.g. if you have 50Hz/60Hz hum in your noise the flat value will change a lot, but the A weighted value far less because the A-weighting takes these low frequencies not so much into account. A-weighting is also less critical about the noise above 5kHz.
If you have a circuit that uses FETs or MOSFETs you may have quite an amount of 1/f noise, typical corner frequencies between 100Hz (good FETs) and even 10kHz (MOSFETs). In these cases the difference between A-weighting and 20Hz-20kHz flat will be more than 3dB.
These kind of differences you also get with digital audio circuits. Multibit ADCs and DACs generate mostly white noise (quantisation noise), but bitstream (sigma-delta) type of converters generate more high frequency noise because of the applied noise shaping. Bitstream converters are almost allways measured with A-weighting of the results, then you get (much) better values.

Steven

tschrama · 12th October 2003, 03:34 PM

Thanks Steven..
any idea what a CCIR weigthing curve actually is and why?

Regards,
Thijs

Pjotr · 12th October 2003, 04:15 PM

Hi Thijs,

Here is a site that gives you the pole and zero locations of the various weighting filters:

http://www.ptpart.co.uk/noise.htm

If you are familiar with the basics of filter theory you can build them. BTW searching around the net I found several times that the difference between flat and A-weighting filters for white noise is 10 dB when integrated over the 20 – 20 kHz bandwidth. Is there a mathematician out there to check this?

Also this is informative reading:

http://www.dwelle.de/rtc/infotheque/...ualpar_05.html

Cheers

fmak · 12th October 2003, 04:53 PM

These kind of differences you also get with digital audio circuits. Multibit ADCs and DACs generate mostly white noise (quantisation noise), but bitstream (sigma-delta) type of converters generate more high frequency noise because of the applied noise shaping. Bitstream converters are almost allways measured with A-weighting of the results, then you get (much) better values.

--------------------------------------------------
The specs can cheat; although this is supposedly based on human response to sine waves! I always look at broadband noise first. The A weighting can hide a lot of circuit probelms.

Steven · 12th October 2003, 06:25 PM

Hi Thijs,

Have a look at http://www.rane.com/par-w.html where you can find the subjects A-weighting and C-weighting. It appears that CCIR (now ITU-R) has its source in telephony switching.

Hi Pjotr,

Also here a difference of 10dB between flat and A-weighted figures is mentioned by Dennis Bohn, but more as a warning that the difference can be as big as 10dB in case of 'nasty low-frequency hum components'. Normally for white noise it is my experience (no mathematics) that the difference is some 3dB. In the past, I found this numerous times when measuring noise with a Sennheiser UPM550 and switching between flat 20Hz-20kHz and A-weighted.

Steven

Pjotr · 12th October 2003, 07:15 PM

Hi Steven,

Maybe you are right that the 3 dB figure is more close. Most of the att. is at the lower frequencies for an A-weighting filter and most of the energy of white noise is in the upper frequencies above 1 kHz where the filter att. is much lesser. Also the method of measuring makes sense, is it quasi peak or is it true RMS? For true RMS you can calculate mathematically the difference by integrating the noise content along the filter slope. But I am too lazy to blow the dust from my old schoolbooks

Cheers

tschrama · 12th October 2003, 08:31 PM

Thanks guys,
excelent reading!

Regards,
Thijs

MarcelvdG · 3rd November 2003, 09:47 AM

I once numerically calculated the noise bandwidth of an A filter. Theoretically, with white noise at the input, the RMS noise at the output of an A filter with unity gain at 1kHz should be identical to the noise at the output of a brick-wall filter with unity pass band gain and a 13471Hz bandwidth. That's 1.712dB less than in a 19980Hz bandwidth.

Thread Tools	Search this Thread
Show Printable Version Email this Page	Search this Thread: Advanced Search

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
flat sub for flat TV	Artmaster	Subwoofers	0	17th December 2008 11:15 AM
DIY weighted diaphragm stretcher (pic)	arend-jan	Planars & Exotics	25	3rd April 2008 11:37 AM
How Flat is "Flat"? (XT18WO + 27TBFC-G = MTM)	zenon	Multi-Way	3	21st January 2007 03:15 PM
Flat C drivers	Bidule	Planars & Exotics	3	21st July 2005 12:07 AM
Too flat?	Dave Jones	Multi-Way	5	3rd October 2003 06:06 PM

New To Site?	Need Help?
Register to Participate Search Privacy Statement Contact Us	Frequently Asked Questions Did you forget your password? Mark Forums Read


	Please consider donating to help us continue to serve you. Ads on/off / Custom Title / More PMs / More album space / Advanced printing & mass image saving

12th October 2003, 12:42 PM	#11
tschrama diyAudio Member Join Date: Jul 2001 Location: The Netherlands	Ha it seems that DIN IEC 651 = C weighting back to google...

12th October 2003, 03:21 PM	#12
Steven diyAudio Member Join Date: Oct 2003 Location: The Netherlands	The difference in dB between the results of different noise measuring filters depend on the character of the noise. For instance, if your noise is completely white, then there is a difference between A-weighting and flat (20Hz-20kHz bandlimited) of approximately 3dB (dBA is 3 dB better than flat). CCIR gives far worse results. If your noise is not white, the difference might become much more, e.g. if you have 50Hz/60Hz hum in your noise the flat value will change a lot, but the A weighted value far less because the A-weighting takes these low frequencies not so much into account. A-weighting is also less critical about the noise above 5kHz. If you have a circuit that uses FETs or MOSFETs you may have quite an amount of 1/f noise, typical corner frequencies between 100Hz (good FETs) and even 10kHz (MOSFETs). In these cases the difference between A-weighting and 20Hz-20kHz flat will be more than 3dB. These kind of differences you also get with digital audio circuits. Multibit ADCs and DACs generate mostly white noise (quantisation noise), but bitstream (sigma-delta) type of converters generate more high frequency noise because of the applied noise shaping. Bitstream converters are almost allways measured with A-weighting of the results, then you get (much) better values. Steven

12th October 2003, 03:34 PM	#13
tschrama diyAudio Member Join Date: Jul 2001 Location: The Netherlands	Thanks Steven.. any idea what a CCIR weigthing curve actually is and why? Regards, Thijs

12th October 2003, 04:15 PM	#14
Pjotr diyAudio Member Join Date: Nov 2002 Location: Netherlands	Hi Thijs, Here is a site that gives you the pole and zero locations of the various weighting filters: http://www.ptpart.co.uk/noise.htm If you are familiar with the basics of filter theory you can build them. BTW searching around the net I found several times that the difference between flat and A-weighting filters for white noise is 10 dB when integrated over the 20 – 20 kHz bandwidth. Is there a mathematician out there to check this? Also this is informative reading: http://www.dwelle.de/rtc/infotheque/...ualpar_05.html Cheers

12th October 2003, 04:53 PM	#15
fmak diyAudio Member Join Date: May 2001 Location: London UK	These kind of differences you also get with digital audio circuits. Multibit ADCs and DACs generate mostly white noise (quantisation noise), but bitstream (sigma-delta) type of converters generate more high frequency noise because of the applied noise shaping. Bitstream converters are almost allways measured with A-weighting of the results, then you get (much) better values. -------------------------------------------------- The specs can cheat; although this is supposedly based on human response to sine waves! I always look at broadband noise first. The A weighting can hide a lot of circuit probelms.

12th October 2003, 06:25 PM	#16
Steven diyAudio Member Join Date: Oct 2003 Location: The Netherlands	Hi Thijs, Have a look at http://www.rane.com/par-w.html where you can find the subjects A-weighting and C-weighting. It appears that CCIR (now ITU-R) has its source in telephony switching. Hi Pjotr, Also here a difference of 10dB between flat and A-weighted figures is mentioned by Dennis Bohn, but more as a warning that the difference can be as big as 10dB in case of 'nasty low-frequency hum components'. Normally for white noise it is my experience (no mathematics) that the difference is some 3dB. In the past, I found this numerous times when measuring noise with a Sennheiser UPM550 and switching between flat 20Hz-20kHz and A-weighted. Steven

12th October 2003, 07:15 PM	#17
Pjotr diyAudio Member Join Date: Nov 2002 Location: Netherlands	Hi Steven, Maybe you are right that the 3 dB figure is more close. Most of the att. is at the lower frequencies for an A-weighting filter and most of the energy of white noise is in the upper frequencies above 1 kHz where the filter att. is much lesser. Also the method of measuring makes sense, is it quasi peak or is it true RMS? For true RMS you can calculate mathematically the difference by integrating the noise content along the filter slope. But I am too lazy to blow the dust from my old schoolbooks Cheers

12th October 2003, 08:31 PM	#18
tschrama diyAudio Member Join Date: Jul 2001 Location: The Netherlands	Thanks guys, excelent reading! Regards, Thijs

3rd November 2003, 09:47 AM	#19
MarcelvdG diyAudio Member Join Date: Mar 2003 Location: Haarlem, the Netherlands	I once numerically calculated the noise bandwidth of an A filter. Theoretically, with white noise at the input, the RMS noise at the output of an A filter with unity gain at 1kHz should be identical to the noise at the output of a brick-wall filter with unity pass band gain and a 13471Hz bandwidth. That's 1.712dB less than in a 19980Hz bandwidth.

Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)