David, If you are a real purist why not sample at RF? I know the cost is about 4x the cost of the Elektor kit but the result must be worth it. When I looked at recent reviews in RadCom the direct sampling boards came out realy well.
Dave -----Original Message----- From: [email protected] [mailto:[EMAIL PROTECTED] On Behalf Of drmail377 Sent: 03 April 2008 16:47 To: [email protected] Subject: [soft_radio] Re: Elector Boards Hi Dave G4UGM, OK, this is top-down typing, please correct any obvious (and likely) glaring mistakes. I'm short on time... You raise excellent questions. You are right, theoretically the divide by 4 for the quadrature LO will reduce phase noise by 20*log(4) or 12 dB, but not really. Even at room temperature, and depending on the sequential quad-divide part thresholds, power supply noise, etc, we might want to at-least halve that. I'm looking into this further, especially considering the Silicon Labs Si570 20ppm LVDS part and Si571 voltage controlled ref input part (wrap the 571 around another PLL loop that is GPS disciplined perhaps?) Let's stick with 12 dB for now. I feel -135dBc/Hz at 10kHz is reasonable for a relatively high performance HF VFO. Let's work back from the example extrapolated from the Elektor's jitter to PN in my previous post assuming a -20 db difference from 1kHz to 10kHz, the center of interest for a 24 kHz BW. -135-12-20=-103 dBc/Hz at 1kHz, 18dB better than the extrapolated Elektor PN from the Cypress jitter spec. Granted, there is a lot of "slop" in this analysis, but one might argue that an error margin of 20dB is not likely methinks. Also, the Cypress part may preform better at certain "select" frequencies and lower frequencies as well (Cypress doesn't disclose), yes it is a PLL device, and as such can trade close-in vs. outer "hump" phase noise performance vs. loop BW. But we're interested in close-in performance for HF reception. BTW my Peppermint "replacement" for the "Blueberry" now seemingly discontinued demo board for the device used in the Elektor LO is not performing anywhere near these numbers. A different animal entirely and seemingly no non EEPROM register to write to in order to set the frequency (EEPROM wear-out problem). But I am feeling better about the Elektor board after you raised these questions Dave. Given that the Si570/671 from Silicon Labs wasn't available at the time the Elektor board was designed, the Cypress part may have been the best choice. I'm sure the Elektor board is doing fine by most if not all users. However, I'm somewhat of a purist and the options for parts are changing rapidly. I just wish the Elektor board was more affordable. I'm a victim of the declining USD. 73''s David --- In [EMAIL PROTECTED] <mailto:soft_radio%40yahoogroups.com> ups.com, "Dave Wade" <[EMAIL PROTECTED]> wrote: > > 1) Whats the phase jitter on a typical PC sound card? > > 2) Doesn't the divide by 4 reduce jitter. The chip is not used to > generate the IQ directly . > > 3) The ready built price to the UK is only marginally more than I can > buy the bits for. I can't find a ready built general coverage board in > Europe for anything like the price. > > > > -----Original Message----- > From: [EMAIL PROTECTED] <mailto:soft_radio%40yahoogroups.com> ups.com [mailto:[EMAIL PROTECTED] <mailto:soft_radio%40yahoogroups.com> ups.com] On > Behalf Of drmail377 > Sent: 28 March 2008 04:53 > To: [EMAIL PROTECTED] <mailto:soft_radio%40yahoogroups.com> ups.com > Subject: [soft_radio] Re: Elector Boards > > > > Hi Jack, > > You might want to rethink building that Elektor board, at least the > way it is "out of the box". I took a good look at it and found that it > uses a Cypress Semiconductor CY27EE16 PLL Clock Generator as a > synthesized local oscillator. The CY27EE16 has a typical jitter > specification of 250ps rms. This is pretty bad. By comparison, a good > crystal oscillator should present less than 1ps rms jitter. To give > you an idea how this translates to phase noise: > > 0.4psrms jitter (Isotemp OCXO134-10 10MHz VC-OCXO): > 10Hz = -105 dBc/Hz > 100Hz = -125 dBc/Hz > 1000Hz = -140 dBc/Hz > > 230psrms jitter (for comparison): > 10Hz = -50 dBc/Hz > 100Hz = -70 dBc/Hz > 1000Hz = -85 dBc/Hz > > Not only that, I get the feeling that the CY27EE16ZE may be set for > deprecation. At least the Blueberry demo board for the part has been > discontinued and replaced with a Peppermint demo board. The Peppermint > demo board uses a CY22393 three PLL clock synthesizer part. The > CY22393 has even worse specified jitter, a whopping 400psrms! But > there are allusions in the data sheet that the jitter may be > significantly less for some "settings", but it doesn't elaborate. I've > just received a couple of Peppermint boards to test. > > Anyway, the seemingly poor choice of L.O. on the Elektor radio, along > with the outrageously high price put me off. If you do build one of > these, you might want to replace the CY27EE16ZE with an Si570. The > Si570's jitter is specified at around 0.35psrms. Mike, KF4BQ has > measured the phase noise of the CMOS version of Si570 to be -130dBc/Hz > at 3-7kHz offset. His report is in the Files section. > > If you do replace the CY27EE16ZE with an Si570 (or something else), > remember the G8JCF SDR software that accompanies the Elektor radio > most likely will not be able to control the L.O. frequency. There's a > group here for the G8JCF software with some discussion on the Elektor > board as well - but not a lot of activity: > > http://groups. <http://groups. <http://groups.yahoo.com/group/G8JCFSDR/> yahoo.com/group/G8JCFSDR/> > yahoo.com/group/G8JCFSDR/ > > 73's David (WB4ONA) > > --- In [EMAIL PROTECTED] <mailto:soft_radio%40yahoogroups.com> ups.com, > "artus1947" <artus@> wrote: > > > > Hi folks, > > > > I wonder if anyone knows of any other sources for the Elector Boards -- > > the shipping to the States (25 euros) is horrible. I tried etching the > > boards myself, but I can't get a good registration between layers. > > Thanks. > > > > 73, > > Jack (W0FNQ) > > >
