"Jerry Avins" <jya@ieee.org> wrote in message news:CKOdndrqccaXgJDeRVn-og@rcn.net...> Al Clark wrote: >> "Bob" <SkiBoyBob@excite.com> wrote in >> news:3n3p8dF19jontU1@individual.net: I have a question that no one has asked. >> >> Do you need to use a SAR converter? The antialiasing filter problem basically >> goes away with a sigma delta converter since the actual sampling rate is very >> much higher than the effective sample rate. > > It does not. The internal oversampling allows averaging conversions at low > precision (typically, one bit) to achieve a single sample at high precision > PROVIDED that anti-alias filtering for the final sample rate has been applied.If you interpret "basically goes away" as "becomes very simple" then I agree with Al. You still have to be concerned with aliasing, but your transition band is huge compared with the non-oversampled case, so often a simple 2nd or 3rd order analog filter suffices.>> The penalty for using a sigma delta converter is group delay. This is usually >> the only reason a SAR converter is used instead of a sigma delta at these >> frequencies. >> >> You might be able to use a "trick". If you have sufficient signal processing >> power, you can oversample and use digital filters. In the sigma delta case, >> you can decimate with IIR filters to sample at a lower sampling rate with a >> relatively small group delay. Analog antialiasing filters are going to have >> about the same delay as digital IIR filters > > It comes down again to oversampling. The type of converter makes no > difference.True, but a really nice feature of a sigma delta ADC, is that you get the data out at the sample rate you want _and_ get the benefits of oversampling at the same time! If you just run a "normal" ADC at a higher rate, you have to do the digital filtering and decimating yourself, which can be problematic since you are dealing with a much higher sample rate. Essentially, the sigma delta ADC builds the filtering and decimation into the ADC chip.
Anti-Aliasing filter
Started by ●August 23, 2005
Reply by ●August 25, 20052005-08-25
Reply by ●August 25, 20052005-08-25
Hi Naebad,
I somewhat disagree with your assumptions.
To start with why do you need 100dB of attenuation at Fs/2. You
calculation shows that the you assumed the aliasing components around
Fs/2 are as large as 10Volts which never happens in any system.(There
are exceptions like the GSM or WCMDA recievers where your interferers
are 100dB above your desired signal level. But the present application
seems to be baseband audio/voice and hence I don't see how you have
such a large aliasing components.)
Hence if you assume there are no other interferers (I am assuming
your converting the analog data from a microphone to digital) then the
only thing you need to filter is the noise at Fs/2. If you start with
the assumption that the input signal is better than 16bit i.e. the
thermal noise level is around 16bit then this same noise level extends
upto very high frequencies.
In summary if you have say 10dB attenuation at Fs/2 for this kind
of noise, the folded noise will only increase the noise floor by less
than 0.1dB and hence you can use a very simple 2nd or 3rd order
anti-aliasing filter.
Ofcourse you can always use oversampling to relax this requirement
even further as some of the people suggested.
Let me know if you agree or disagree with my calculations.
Reply by ●August 25, 20052005-08-25
Jerry Avins <jya@ieee.org> wrote in news:CKOdndrqccaXgJDeRVn-og@rcn.net:> Al Clark wrote: >> "Bob" <SkiBoyBob@excite.com> wrote in >> news:3n3p8dF19jontU1@individual.net: >> >> >>>"Jerry Avins" <jya@ieee.org> wrote in message >>>news:o8OdnYAImb2oPJHeRVn-qw@rcn.net... >>> >>>>Bob wrote: >>>> >>>>>"naebad" <minnaebad@yahoo.co.uk> wrote in message >>>>>news:1124826795.554026.7640@g43g2000cwa.googlegroups.com... >>>>> >>> >>>>Tim's suggestion, oversampling, is at the heart of most practical >>>>approaches. >>>> >>>>Jerry >>> >>>True. My suggestions are useful if oversample isn't an option e.g. >>>because the ADC to oversample doesn't exist or is too expensive. >>>11kHz at 16bits is easily oversampled. >>> >>>Bob >>> >>> >>> >> >> >> I have a question that no one has asked. >> >> Do you need to use a SAR converter? The antialiasing filter problem >> basically goes away with a sigma delta converter since the actual >> sampling rate is very much higher than the effective sample rate. > > It does not. The internal oversampling allows averaging conversions at > low precision (typically, one bit) to achieve a single sample at high > precision PROVIDED that anti-alias filtering for the final sample rate > has been applied.I was not trying to imply that aliasing can not happen. Sigma delta converters include digital filters. From a user perspective, the only antialiasing filtering required is at a much higher frequency than the passband and very easy to accommodate. It most cases, a first or second order low pass is all that is required. This is much, much easier that high order filters near the passband.> >> The penalty for using a sigma delta converter is group delay. This is >> usually the only reason a SAR converter is used instead of a sigma >> delta at these frequencies. >> >> You might be able to use a "trick". If you have sufficient signal >> processing power, you can oversample and use digital filters. In the >> sigma delta case, you can decimate with IIR filters to sample at a >> lower sampling rate with a relatively small group delay. Analog >> antialiasing filters are going to have about the same delay as >> digital IIR filters > > It comes down again to oversampling. The type of converter makes no > difference.Certainly it does with typical parts. In the sigma delta case, the converter is probably oversampling by 64 to 512. This is still much easier to protect from aliasing than oversampling with a SAR converter at perhaps 4-8x oversampling. If you can live with the group delay, sigma delta converters are almost always better than SAR converters (and much cheaper for the same performance). Obviously, there are applications where sigma delta converters are not going to be an appropriate choice. -- Al Clark Danville Signal Processing, Inc. -------------------------------------------------------------------- Purveyors of Fine DSP Hardware and other Cool Stuff Available at http://www.danvillesignal.com
Reply by ●August 25, 20052005-08-25
Jon Harris wrote:> "Jerry Avins" <jya@ieee.org> wrote in message > news:CKOdndrqccaXgJDeRVn-og@rcn.net... > >>Al Clark wrote: >> >>>"Bob" <SkiBoyBob@excite.com> wrote in >>>news:3n3p8dF19jontU1@individual.net: I have a question that no one has asked. >>> >>>Do you need to use a SAR converter? The antialiasing filter problem basically >>>goes away with a sigma delta converter since the actual sampling rate is very >>>much higher than the effective sample rate. >> >>It does not. The internal oversampling allows averaging conversions at low >>precision (typically, one bit) to achieve a single sample at high precision >>PROVIDED that anti-alias filtering for the final sample rate has been applied. > > > If you interpret "basically goes away" as "becomes very simple" then I agree > with Al. You still have to be concerned with aliasing, but your transition band > is huge compared with the non-oversampled case, so often a simple 2nd or 3rd > order analog filter suffices.Are you saying that the low-pass cutoffs of the anti-alias filters to be used in front of an SA and a delta-sigma converter sampling the same signal at the same output rate are different? I think they're the same. ... Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
Reply by ●August 25, 20052005-08-25
Al Clark wrote:> Jerry Avins <jya@ieee.org> wrote in news:CKOdndrqccaXgJDeRVn-og@rcn.net: > > >>Al Clark wrote: >> >>>"Bob" <SkiBoyBob@excite.com> wrote in >>>news:3n3p8dF19jontU1@individual.net: >>> >>> >>> >>>>"Jerry Avins" <jya@ieee.org> wrote in message >>>>news:o8OdnYAImb2oPJHeRVn-qw@rcn.net... >>>> >>>> >>>>>Bob wrote: >>>>> >>>>> >>>>>>"naebad" <minnaebad@yahoo.co.uk> wrote in message >>>>>>news:1124826795.554026.7640@g43g2000cwa.googlegroups.com... >>>>>> >>>> >>>>>Tim's suggestion, oversampling, is at the heart of most practical >>>>>approaches. >>>>> >>>>>Jerry >>>> >>>>True. My suggestions are useful if oversample isn't an option e.g. >>>>because the ADC to oversample doesn't exist or is too expensive. >>>>11kHz at 16bits is easily oversampled. >>>> >>>>Bob >>>> >>>> >>>> >>> >>> >>>I have a question that no one has asked. >>> >>>Do you need to use a SAR converter? The antialiasing filter problem >>>basically goes away with a sigma delta converter since the actual >>>sampling rate is very much higher than the effective sample rate. >> >>It does not. The internal oversampling allows averaging conversions at >>low precision (typically, one bit) to achieve a single sample at high >>precision PROVIDED that anti-alias filtering for the final sample rate >>has been applied. > > > I was not trying to imply that aliasing can not happen. Sigma delta > converters include digital filters. From a user perspective, the only > antialiasing filtering required is at a much higher frequency than the > passband and very easy to accommodate. It most cases, a first or second > order low pass is all that is required. This is much, much easier that > high order filters near the passband. > > >>>The penalty for using a sigma delta converter is group delay. This is >>>usually the only reason a SAR converter is used instead of a sigma >>>delta at these frequencies. >>> >>>You might be able to use a "trick". If you have sufficient signal >>>processing power, you can oversample and use digital filters. In the >>>sigma delta case, you can decimate with IIR filters to sample at a >>>lower sampling rate with a relatively small group delay. Analog >>>antialiasing filters are going to have about the same delay as >>>digital IIR filters >> >>It comes down again to oversampling. The type of converter makes no >>difference. > > > Certainly it does with typical parts. In the sigma delta case, the > converter is probably oversampling by 64 to 512. This is still much > easier to protect from aliasing than oversampling with a SAR converter at > perhaps 4-8x oversampling. > > If you can live with the group delay, sigma delta converters are almost > always better than SAR converters (and much cheaper for the same > performance). > > Obviously, there are applications where sigma delta converters are not > going to be an appropriate choice.I need to study this. A one-bit converter oversampling 256:1 allows one to average those 256 samples to produce a 16-bit result. (Noise shaping helps.) What happens to the aliass that are also sampled if they're present? The effective averaging can reduce their effect. Is that all? Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
Reply by ●August 25, 20052005-08-25
"Jerry Avins" <jya@ieee.org> wrote in message news:IKOdnfco4cMmQJDeRVn-pw@rcn.net...> Jon Harris wrote: >> "Jerry Avins" <jya@ieee.org> wrote in message >> news:CKOdndrqccaXgJDeRVn-og@rcn.net... >> >>>Al Clark wrote: >>> >>>>"Bob" <SkiBoyBob@excite.com> wrote in >>>>news:3n3p8dF19jontU1@individual.net: I have a question that no one has >>>>asked. >>>> >>>>Do you need to use a SAR converter? The antialiasing filter problem >>>>basically goes away with a sigma delta converter since the actual sampling >>>>rate is very much higher than the effective sample rate. >>> >>>It does not. The internal oversampling allows averaging conversions at low >>>precision (typically, one bit) to achieve a single sample at high precision >>>PROVIDED that anti-alias filtering for the final sample rate has been >>>applied. >> >> If you interpret "basically goes away" as "becomes very simple" then I agree >> with Al. You still have to be concerned with aliasing, but your transition >> band is huge compared with the non-oversampled case, so often a simple 2nd or >> 3rd order analog filter suffices. > > Are you saying that the low-pass cutoffs of the anti-alias filters to be used > in front of an SA and a delta-sigma converter sampling the same signal at the > same output rate are different? I think they're the same.I say they are different. The delta-sigma digitizes at a higher rate but then internally downsamples to the desired rate. With delta-sigma, you need only remove aliases of the higher internal sample rate. With SA converters, you need to remove aliases of the output sample rate. This is one of the biggest advantages of oversampling converters--removing the need for a expensive, high-precision brick wall filter (which also usually introduces undesirable phase shift in the pass band). We trade off digital filtering for analog, which given Moore's law is a great deal! This is one of the reasons you can get great-sounding converters so cheaply today (and how they can make decent-quality consumer digital audio gear for dirt-cheap). For example, check out this product description: http://www.cirrus.com/en/products/pro/detail/P70.html To quote: "The A/D converters use Delta-Sigma modulation with 128x oversampling followed by digital filtering and decimation, which removes the need for an external anti-alias filter." Note: that although it claims no need for the anti-alias filter, their evaluation board does have one--just a very simple one. Typically, you might see a couple of poles starting to roll off around 40kHz or so (the cut-off is chosen to be high enough so as to have negligible attenuation at 20kHz, e.g. < 1 dB.) See Al's post on this as well.
Reply by ●August 25, 20052005-08-25
"Jon Harris" <jon99_harris7@hotmail.com> wrote in news:Y6mPe.18642$g47.6392@trnddc07:> "Jerry Avins" <jya@ieee.org> wrote in message > news:IKOdnfco4cMmQJDeRVn-pw@rcn.net... >> Jon Harris wrote: >>> "Jerry Avins" <jya@ieee.org> wrote in message >>> news:CKOdndrqccaXgJDeRVn-og@rcn.net... >>> >>>>Al Clark wrote: >>>> >>>>>"Bob" <SkiBoyBob@excite.com> wrote in >>>>>news:3n3p8dF19jontU1@individual.net: I have a question that no one >>>>>has asked. >>>>> >>>>>Do you need to use a SAR converter? The antialiasing filter problem >>>>>basically goes away with a sigma delta converter since the actual >>>>>sampling rate is very much higher than the effective sample rate. >>>> >>>>It does not. The internal oversampling allows averaging conversions >>>>at low precision (typically, one bit) to achieve a single sample at >>>>high precision PROVIDED that anti-alias filtering for the final >>>>sample rate has been applied. >>> >>> If you interpret "basically goes away" as "becomes very simple" then >>> I agree with Al. You still have to be concerned with aliasing, but >>> your transition band is huge compared with the non-oversampled case, >>> so often a simple 2nd or 3rd order analog filter suffices. >> >> Are you saying that the low-pass cutoffs of the anti-alias filters to >> be used in front of an SA and a delta-sigma converter sampling the >> same signal at the same output rate are different? I think they're >> the same. > > I say they are different. The delta-sigma digitizes at a higher rate > but then internally downsamples to the desired rate. With > delta-sigma, you need only remove aliases of the higher internal > sample rate. With SA converters, you need to remove aliases of the > output sample rate. This is one of the biggest advantages of > oversampling converters--removing the need for a expensive, > high-precision brick wall filter (which also usually introduces > undesirable phase shift in the pass band). We trade off digital > filtering for analog, which given Moore's law is a great deal! This > is one of the reasons you can get great-sounding converters so cheaply > today (and how they can make decent-quality consumer digital audio > gear for dirt-cheap). > > For example, check out this product description: > http://www.cirrus.com/en/products/pro/detail/P70.html > To quote: "The A/D converters use Delta-Sigma modulation with 128x > oversampling followed by digital filtering and decimation, which > removes the need for an external anti-alias filter." > > Note: that although it claims no need for the anti-alias filter, their > evaluation board does have one--just a very simple one. Typically, > you might see a couple of poles starting to roll off around 40kHz or > so (the cut-off is chosen to be high enough so as to have negligible > attenuation at 20kHz, e.g. < 1 dB.) > See Al's post on this as well. > >I was going to respond to Jerry's specifics, but I would just be echoing Jon. There are a couple of points that I will add: 1. Usually the ADC antialiasing filter is set higher than 40k for a 20k passband. Since the oversampling is still very far away, there is no need for it to affect the passband at all in a significant way (1dB atten at 20k for example). This has an added benefit that the phase matching of identical converters will tend to be very good since most of the filtering that affects the passband is digital (and therefore identical). The aliasing signals are going to be near the analog sample rate which is typically in the megahertz region for an audio frequency converter. This means that switching power supply and similar noise sources might be aliased but it is very unlikely that noise from a sensor will be an issue (such as a microphone). 2. Its useful to think of sigma delta as 1 bit converters but very few commercial devices actually are. They are usually a few bits instead. There are a variety of good reasons to do this (especially for very wide dynamic range). Sigma delta converters are more than just oversampling & averaging converters. Noise shaping is the big deal. Without noise shaping, the oversample amount becomes huge for high resolution. I'm not sure, but I think its double the sample rate for each extra bit of resolution. 3. As I mentioned before, the main cost of sigma delta is group delay. A typical good ADC might have a group delay of 20/fs. In general, the better the internal filtering, the more the delay. This is why sigma delta is often not well suited for feedback applications where the group delay might be a signicant problem. 4. The industry is divided on whether these converters should be called sigma delta or delta sigma. You can call them whatever you want, they refer to the same thing. -- Al Clark Danville Signal Processing, Inc. -------------------------------------------------------------------- Purveyors of Fine DSP Hardware and other Cool Stuff Available at http://www.danvillesignal.com
Reply by ●August 25, 20052005-08-25
Al Clark <dsp@danvillesignal.com> writes:> [...] > 4. The industry is divided on whether these converters should be called > sigma delta or delta sigma. You can call them whatever you want, they > refer to the same thing.The correct name has been clearly acknowledged for some 14 years: http://groups.google.com/group/rec.audio/browse_thread/thread/add0c2ff2091f1e4/21927da449403dfa?lnk=st&q=candy+group:comp.dsp&rnum=4&hl=en#21927da449403dfa -- % Randy Yates % "Bird, on the wing, %% Fuquay-Varina, NC % goes floating by %%% 919-577-9882 % but there's a teardrop in his eye..." %%%% <yates@ieee.org> % 'One Summer Dream', *Face The Music*, ELO http://home.earthlink.net/~yatescr
Reply by ●August 25, 20052005-08-25
Randy Yates <yates@ieee.org> wrote in news:64ttva7o.fsf@ieee.org:> http://groups.google.com/group/rec.audio/browse_thread/thread/add0c2ff2 > 091f1e4/21927da449403dfa?lnk=st&q=candy+group:comp.dsp&rnum=4&hl=en#219 > 27da449403dfaWhether there is a "proper" definition or not, Here are what these manufactures call it: Sigma Delta Camp: Analog Devices Wolfson Micro Sigmatel Delta Sigma Camp: Cirrus Logic AKM Linear Technology Can't make up our minds camp: Texas Instruments So call it whatever you like..... -- Al Clark Danville Signal Processing, Inc. -------------------------------------------------------------------- Purveyors of Fine DSP Hardware and other Cool Stuff Available at http://www.danvillesignal.com
Reply by ●August 25, 20052005-08-25
"Al Clark" <dsp@danvillesignal.com> wrote in message news:Xns96BD7D4083951aclarkdanvillesignal@66.133.129.71...> "Jon Harris" <jon99_harris7@hotmail.com> wrote in > news:Y6mPe.18642$g47.6392@trnddc07: > > > 3. As I mentioned before, the main cost of sigma delta is group delay. A > typical good ADC might have a group delay of 20/fs. In general, the > better the internal filtering, the more the delay. This is why sigma > delta is often not well suited for feedback applications where the group > delay might be a signicant problem.That is one area where there is some advantage to 96kHz (or even 192kHz) sampling rate. Since the "fs" is half as long, the group delay is cut in half.
