glen herrmannsfeldt <gah@ugcs.caltech.edu> wrote:>I was wondering not so long ago about the state of the art >in A/D conversion. It seems that 24 bit audio is reasonably >common in digital recorders. Is the analog electronics up through >the A/D converter really that good? Amplifiers with 144dB signal >to noise ratio? The A/D converter itself?The main signal path in professional audio equipment often has a maximum signal level of +28 dBm and an equivalent input noise of -130 dBm. (I think that's about equal to 150 ohms of noise over 20 KHz.) Some audio source signals can generate almost that much range -- for example, a piano or percussion instrument miked with the best condensor mikes in a soundproof studio. So, the answer is yes, pretty much. What you can't get is anything mixed down and listenable with that much range, but some of the source material may close to that kind of range. It's also very difficult to get an audio DAC that is more than about 20 actual bits, but maybe it's been done. Steve
notching filter that removes the power line harmonics?
Started by ●October 23, 2008
Reply by ●October 28, 20082008-10-28
Reply by ●October 28, 20082008-10-28
glen herrmannsfeldt wrote:> I was wondering not so long ago about the state of the art > in A/D conversion. It seems that 24 bit audio is reasonably > common in digital recorders. Is the analog electronics up through > the A/D converter really that good? Amplifiers with 144dB signal > to noise ratio? The A/D converter itself?The dynamic range depends on the bandwidth and the sample rate. The very good audio ADC has 20..21 ENOB in the frequency range of 20Hz...20kHz. The attainable SNR of an analog electronics in the same band is about 130..140dB; however the performance is limited by the nonlinearity. Vladimir Vassilevsky DSP and Mixed Signal Design Consultant http://www.abvolt.com
Reply by ●October 28, 20082008-10-28
Steve Pope wrote: (snip about noise A/D converters, and dynamic range)> The main signal path in professional audio equipment often > has a maximum signal level of +28 dBm and an equivalent input > noise of -130 dBm. (I think that's about equal to 150 ohms of noise > over 20 KHz.) Some audio source signals can generate > almost that much range -- for example, a piano or percussion instrument > miked with the best condensor mikes in a soundproof studio.> So, the answer is yes, pretty much. What you can't get is > anything mixed down and listenable with that much range, but > some of the source material may close to that kind of range.> It's also very difficult to get an audio DAC that is more than about > 20 actual bits, but maybe it's been done.20 bits is enough for me. I sometimes record high school concerts with a Roland R-1 in 24bit WAV mode. I figure that allows me to keep the record level a little lower to avoid saturation. I can then adjust it while converting to 16 bits for a CD. With a live audience the noise is fairly high, anyway. I wrote my own 24bit to 16bit WAV converter, allowing me to specify the bits to use and, more recently, add dither. For consumer or low end pro equipment, probably not 24 real bits then. -- glen
Reply by ●October 28, 20082008-10-28
Vladimir Vassilevsky wrote: (snip)> The dynamic range depends on the bandwidth and the sample rate. The very > good audio ADC has 20..21 ENOB in the frequency range of 20Hz...20kHz. > The attainable SNR of an analog electronics in the same band is about > 130..140dB; however the performance is limited by the nonlinearity.Good enough for me. For live concert recordings, the non-linearity is likely less than that of the musical instruments. It would seem that a 20 bit recording format would be useful, but the choices seem to be 16 and 24. -- glen
Reply by ●October 29, 20082008-10-29
-glen, The dynamic range thing is tricky. It is one thing that we want to see uV signals out of 1Vac un-wanted harmonics, it is another thing to analyze such weak signals. When we want to look into the signal characteristics of the uV signals, we need at least another 40~50dB down. James www.go-ci.com
Reply by ●October 30, 20082008-10-30
On Tue, 28 Oct 2008 11:11:03 -0700 (PDT), DigitalSignal <digitalsignal999@yahoo.com> wrote:>Guys, > >Thank you so much for the advice and various throughts. The >applications are related to the very weak signal measurement, usually >in the range of a few uV, that is damaged by the interference of >powline signals. The analog signal contains very large power line >harmonics, say near 1Vac, as well as the weak signals that we are >interested. If we feed all the signals without analog treatment into A/ >D, the dynamic range will be suffered. If we can remove the power line >harmonics in the analog domain first, the A/D dynamic range can be >fully utilized. Therefore I am looking into this direction now. > >James >www.go-ci.comHi James, I'm entering this thread very late. I can't comment on the adaptive filtering suggestions. However, an FIR notch comb doesn't seem like a greater idea to me. That's because such filters don't have sharp transition regions. You can sharpen the transitions by placing filters poles just inside the unit circle at the same freqs as the notch comb's zeros. However, doing so creates a nonlinear phase filter. Thinkin' about your problem, I wonder if some sort of linear phase interpolated FIR (IFIR) filter might be applicable. Just a thought. Good Luck, [-Rick-]
Reply by ●October 30, 20082008-10-30
Reply by ●November 11, 20082008-11-11
On Oct 30, 9:39�pm, DigitalSignal <digitalsignal...@yahoo.com> wrote:> Hi Rick, Thanks for the thoughts! > > Jameswww.go-ci.comGood Morning James, I would like to take the opportunity to mention that we have indeed developed a tracking IIR comb filter using an FPGA and I believe the performance will meet or exceed your requirements. The response from Rick is correct in that you can sharpen the nulls by placing poles inside the unit circle and by doing that you move away from a linear phase filter, but the phase tends toward zero except at the notches, so when you listen to an audio signal (which we have a sample of) passing through the filter with the Q set to more than 10, the effect on background signals is hardly noticable. The Q on our filter can be manually set to any value from 2 to infinity and the filter is automatically tuned to the mains fundamental with a resolution of 0.001 Hz, so all mains harmonics up to and beyond the 100th are attenuated significantly. Please contact me if you require further details.
Reply by ●November 12, 20082008-11-12
On Nov 11, 1:42�am, kmichaeli...@blackrosernd.com wrote:> On Oct 30, 9:39�pm, DigitalSignal <digitalsignal...@yahoo.com> wrote: > > > Hi Rick, Thanks for the thoughts! > > > Jameswww.go-ci.com > > Good Morning James, > > I would like to take the opportunity to mention that we have indeed > developed a tracking IIR comb filter using an FPGA and I believe the > performance will meet or exceed your requirements. > The response from Rick is correct in that you can sharpen the nulls by > placing poles inside the unit circle and by doing that you move away > from a linear phase filter, but the phase tends toward zero except at > the notches, so when you listen to an audio signal (which we have a > sample of) passing through the filter with the Q set to more than 10, > the effect on background signals is hardly noticable. The Q on our > filter can be manually set to any value from 2 to infinity and the > filter is automatically tuned to the mains fundamental with a > resolution of 0.001 Hz, so all mains harmonics up to and beyond the > 100th are attenuated significantly. > Please contact me if you require further details.It looks cool! If you have any information could you send it to Sales@go-ci.com ?
