<radams2000@gmail.com> wrote:>Most digital microphones with PDM outputs use a 2nd-order delta-sigma >modulator inside, so the noise is rising by 12 db/octave. Usually the >rising a/d noise "crosses over" the thermal noise of the mic preamp near >the top of the audio band.The spec sheet for thie mic indeed says it contains a sigma- delta modulator. One should easily be able to see the shaped noise in the microphone output.>It is conventional to use sinc filters where the order of the sinc is >one higher than the order of the modulator so that the noise is dropping >as a function of frequency. This is done to avoid aliased noise when >you decimate. So the use of 3 cascaded moving-average filters in this >application is standard, although the fact that they are all different >lengths is a bit unusual.Not sure why sinc filters are a good choice. I'd use something like a Butterworth filter. In what context are sync filters conventional? Steve
How to convert from PDM to PCM ?
Started by ●September 1, 2015
Reply by ●September 20, 20152015-09-20
Reply by ●September 20, 20152015-09-20
Since you are operating at a relatively high sample-rate, it's good to use multiplier-free filter structures. There is a filter structure that combines the sinc filter together with the decimation called the Hogenauer CIC sinc filter. It eliminates the need for the data buffers and has a very cool way of un-wrapping data that has wrapped around in an open-loop cascade of integrators. https://en.m.wikipedia.org/wiki/Cascaded_integrator%E2%80%93comb_filter These tricks apply mostly to custom hardware designs, and any IC that has a dedicated PDM port will use an optimized hardware decimator behind it. It's very inefficient to code this on s general-purpose dsp or micro-controller but it looks like you have no choice.
Reply by ●September 20, 20152015-09-20
<radams2000@gmail.com> wrote:>Since you are operating at a relatively high sample-rate, it's good to >use multiplier-free filter structures. There is a filter structure that >combines the sinc filter together with the decimation called the >Hogenauer CIC sinc filter. It eliminates the need for the data buffers >and has a very cool way of un-wrapping data that has wrapped around in >an open-loop cascade of integrators.>https://en.m.wikipedia.org/wiki/Cascaded_integrator%E2%80%93comb_filter>These tricks apply mostly to custom hardware designs, and any IC that >has a dedicated PDM port will use an optimized hardware decimator behind >it. It's very inefficient to code this on s general-purpose dsp or >micro-controller but it looks like you have no choice.Notes. In the grand scheme of things, 3.125 Msamples/sec is not a very high sample rate, compared to sample rates found say in a baseband chip in a phone... and you're doing way more processing there. I suggest Butterworth because it is a known-good filter for audio (in terms of subjective listening qualities), and this is an audio application. (But if you are worried about complexity, I'd agree with CIC compared to the cascaded Sinc's.) Steve
Reply by ●September 20, 20152015-09-20
Since the OP is detecting a 40 kHz carrier, I doubt this is a traditional audio application, in which case we don't need to worry about audibility. The SNR around 40khz is likely to be poor, so my guess is that a band pass is more appropriate. Some combination of low-pass filtering followed by decimation followed by band pass filtering is probably optimum, but there's lots of ways you could arrange the filtering. While 3mhz is not a very high sample rate, manufacturers are often trying to implement "always-on" audio monitoring on battery-operated equipment, and the rest of the system is usually asleep while this is happening, so power minimization in the microphone path is critical. That way when you mention to your wife that you've run out of paper towels , the Amazon drone can be hovering in front of your window within minutes.
Reply by ●September 20, 20152015-09-20
<radams2000@gmail.com> wrote:>Since the OP is detecting a 40 kHz carrier, I doubt this is a >traditional audio application, in which case we don't need to worry >about audibility.Good point, I had missed that.>While 3mhz is not a very high sample rate, manufacturers are often >trying to implement "always-on" audio monitoring on battery-operated >equipment, and the rest of the system is usually asleep while this is >happening, so power minimization in the microphone path is critical.Another good point. Thanks Steve
Reply by ●September 21, 20152015-09-21
On 9/20/15 3:55 PM, radams2000@gmail.com wrote:> That way when you mention to your wife that you've run out of paper towels , the Amazon drone can be hovering in front of your window within minutes.be sure to duck when the NRA drone armed with a shotgun blows the Amazon drone outa the sky. Drone Wars - the brave new world. -- r b-j rbj@audioimagination.com "Imagination is more important than knowledge."
Reply by ●September 21, 20152015-09-21
On 9/21/2015 6:30 PM, robert bristow-johnson wrote:> On 9/20/15 3:55 PM, radams2000@gmail.com wrote: >> That way when you mention to your wife that you've run out of paper >> towels , the Amazon drone can be hovering in front of your window >> within minutes. > > > be sure to duck when the NRA drone armed with a shotgun blows the Amazon > drone outa the sky. > > Drone Wars - the brave new world.I thought that kid who posted a video of a drone firing a pistol was being investigated? Will the NRA be violating the law? Guns don't kill people, Drones kill people! -- Rick
Reply by ●October 17, 20152015-10-17
Hi Mauritz, Knowles alternately has an I2S MEMS mic, if that would be easier to design in? Part number: SPH0645LM4H-B I can send you some additional documentation on it and/or on PDM conversion. davin@seltech-intl.com Thanks, Davin Seltech --------------------------------------- Posted through http://www.DSPRelated.com
Reply by ●December 17, 20152015-12-17
Hi angrydude, you have quoted in your response that there are some cheap electret mic capsules which give excellent performance at 40 kHz. Can you suggest such microphones, which are small in size. Actually, I need them for the purpose of creating a microphone array of total aperture length of say 40 mm with uniform spacing between each microphone. I intend to use 10 mics in an array. presently I trying to use Knowles Acoustics SPH0641LU4H-1 mics. The statement of my problem is similar to that mentioned by Mauritz Jameson. Just in my case I intend to detect object with the received echoes of the emitted ultrasonic signals. It will be great if you can help me. --------------------------------------- Posted through http://www.DSPRelated.com
Reply by ●December 18, 20152015-12-18
On Thursday, December 17, 2015 at 3:10:52 PM UTC-5, akanksha92 wrote:> Hi angrydude, > > you have quoted in your response that there are some cheap electret mic > capsules which give excellent performance at 40 kHz. Can you suggest such > microphones, which are small in size. Actually, I need them for the > purpose of creating a microphone array of total aperture length of say 40 > mm with uniform spacing between each microphone. I intend to use 10 mics > in an array. presently I trying to use Knowles Acoustics SPH0641LU4H-1 > mics. > > The statement of my problem is similar to that mentioned by Mauritz > Jameson. Just in my case I intend to detect object with the received > echoes of the emitted ultrasonic signals. It will be great if you can help > me. > > > --------------------------------------- > Posted through http://www.DSPRelated.comPanasonic WM-61A electret capsules will give you excellent response at 40 KHz and they are tiny: http://www.wildlife-sound.org/equipment/technote/micdesigns/ultrasonic.html Of course, for those you'll need decent analog pre-amps and AD converters before you can do anything. Also, mic will pick up both audible acoustic noise and ultrasound - you'll need a high-pass filter - before or after AD conversion I myself never used digital MEMS mics - just couldn't find any cheap breakout boards for those plus it's impossible to see what signal you are getting (e.g. using scope) before you do all those digital conversions. Are you soldering those yourself or using breakout boards ? What is this for ? How big are your "objects" to detect ? With 40 KHz your wavelength is about 8mm so this will limit your resolution.