Return-Path: Received: (qmail 26566 invoked from network); 8 Aug 2001 07:55:32 -0000 Received: from unknown (HELO murphys-inbound.services.quay.plus.net) (212.159.14.225) by excalibur.plus.net with SMTP; 8 Aug 2001 07:55:32 -0000 Received: (qmail 14191 invoked from network); 8 Aug 2001 07:55:09 -0000 Received: from unknown (HELO post.thorcom.com) (212.172.148.70) by murphys with SMTP; 8 Aug 2001 07:55:09 -0000 Received: from majordom by post.thorcom.com with local (Exim 3.16 #2) id 15UO0I-0002X7-00 for rsgb_lf_group-outgoing@blacksheep.org; Wed, 08 Aug 2001 08:43:46 +0100 X-Priority: 3 X-MSMail-Priority: Normal Received: from kauha.saunalahti.fi ([195.197.53.227]) by post.thorcom.com with esmtp (Exim 3.16 #2) id 15UNzt-0002X2-00 for rsgb_lf_group@blacksheep.org; Wed, 08 Aug 2001 08:43:22 +0100 Received: from pmk2 (MMMDCXVIII.hdyn.saunalahti.fi [195.197.164.18]) by kauha.saunalahti.fi (8.10.1/8.10.1) with SMTP id f787gXC28662 for ; Wed, 8 Aug 2001 10:42:33 +0300 (EEST) X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2800.1106 From: "Paul Keinanen" To: rsgb_lf_group@blacksheep.org Subject: Re: LF: Harmonic sampling (Was: Decimation) Date: Wed, 08 Aug 2001 10:45:31 +0300 Message-ID: References: <001201c11de7$6051a9a0$916c97d4@oemcomputer> In-reply-to: <001201c11de7$6051a9a0$916c97d4@oemcomputer> X-Mailer: Forte Agent 1.7/32.534 MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii; format=flowed Content-Transfer-Encoding: 8bit Precedence: bulk Reply-To: rsgb_lf_group@blacksheep.org X-Listname: rsgb_lf_group Sender: On Sun, 5 Aug 2001 22:46:14 +0300, "Johan Bodin" wrote: I found an interesting tutorial article about undersampling and ADC parameters at http://archives.e-insite.net/archives/ednmag/reg/1995/080395/16df1.htm Since the article is from 1995, the list of available components is not up to date. On the EDN site, there are also other related articles, if you enter "undersampling" into the search functionality. >Fs=6kHz seems like a good choice. The selected Fs harmonic should be just outside >the band of interest (lowest acceptable beat note) and the preselector BW (-many dB!) >must be less than Fs/2. Pse correct me if I'm wrong. This is true as this Fs/2 bandwidth is exactly at nFs .. (n+0.5)Fs or at (n+0.5)Fs .. (n+1)Fs. At any other frequency bands, the filter should be narrower in order to avoid the nFs and (n+0.5)Fs frequencies. By the way, standard audio ADCs preceded by a sample and hold circuit might also be usable. When running at 48 kHz Fs, the multiples would be at 96 and 144 kHz, thus the 135.7 .. 137.8 kHz band would be converted to 8.3 .. 6.2 kHz (inverted spectrum). With such high sampling rate, the front end filtering could be done directly at signal frequency using ordinary LC filters. The most problematic alias is 150.2 .. 152.3 kHz, which would require a deep notch. The other close alias is at 102.2 .. 104.3 kHz, but apparently this would be sufficiently attenuated by the stop band of any bandpass filter. The other advantage of using standard audio ADCs producing SPDIF is that sound cards with SPDIF inputs could be used to input the digital signal into PC for further processing. The nasty thing about these converters is that they usually require a clock frequency of 64Fs .. 256Fs, which would be quite problematic to generate at very high precision, if 48 kHz sampling rate is used. The converter might be clocked to run at 50 kHz, simplifying the reference frequency generation, but the PC sound card might not lock on to the 50 kHz sample rate. Paul OH3LWR