[hpsdr] Demeter progress

Fri Jul 13 07:18:17 PDT 2007

Dear Tommy, and others,

Thank you for your E-mail. I do agree with you that dI/dt is a factor in 
the noise generation. However, it is just a part of the noise-generation 
problem and not the whole problem as you put it. In the mail I was not 
talking about the frequency of the waveform but about the waveform 
itself. The duty cycle (the ratio of the high-time and period of a pulse 
signal) strongly determines the frequency spectrum of a pulse signal. By 
doing Fourier analysis on these waveforms one can clearly see the 
influence of a high dI/dt (or dU/dt) AND the influence of the duty-cycle.
The duty-cycle is parameter which can be influenced, within boundaries 
off course, by adjusting the power-handling-capabillity of the SMPS. An 
overrated SMPS will work with lower duty-cycles compared to a SMPS 
wihich works at almost the top of its power-handling.
Filtering of the SMPS is necessary, but when we can minimize the amount 
of "noise" generated, the filter can be easier to construct.

As mentioned in my previous E-mail: Soft switching would be a promising 
technique for Demeter. Soft switching lowers the dI/dt and thus lowers 
the amount of switching noise generated. The drawback of softswitching 
is the (slightly) lower efficiency of the regulator. The switching 
elements are used in their lineair region, so the loss (heat) will be 
slightly higher.

This weekend I will try to make a setup for the power supply testing. I 
have a kit for a variable load (heavy MOSFET plus some control 
electronics), I assemble the kit this weekend. I have an old switching 
power supply based on the well known UC3842 which will be the DUT. This 
will be AC-coupled to the spectrum analyzer, and hopefully some usefull 
results will come out.

Have a nice weekend, 73 Jeroen PE1RGE

Hollingsworth, Tommy wrote:

> Jeroen:
>I think you need to do a little more research into EMI and signal integrity.
>The noise/EMI generation is not really driven by the frequency of the
>waveform.  It is driven by the rate of rise of the leading edge and trailing
>edge alone.  Think dI/dT and dV/dT.  Putting an analog regulator after the
>switching regulator does not necessarily do anything to filter out the EMI
>generated by primarily the dI/dT in the switch(transistor).  You can be much
>more effective with the appropriate use of a pi filter with very carefully
>selected ferrites.  The ferrites used in the filter have to be chosen for
>the exact frequency or frequency range of the switching regulator.  This in
>combination with a very meticulous board layout to minimize the loop size of
>the paths carrying the highest dI/dT is needed to keep noise and EMI down
>low.  
>
>The use of a linear regulator after a well designed switching supply
>certainly does not hurt, but do not think that alone will give you low noise
>and or low EMI.  
>
>As to your question about how does load impact the spectrum:  Map the dI/dT
>of the design at a nominal level.   Apply the maximum load and map the
>dI/dT.  You will find that the spectrum may or may not shift with load)
>depends on your topology.  With the higher load the dI/dT amplitude will
>rise dramatically and hence radiated and conducted noise will rise the same
>way.  As in a radio it is important to watch the frequency of the radiated
>signal going into an antenna and the current going into the antenna (both
>real and complex).
>
>Your last paragraph is wrong.  Keep reading, it will come to you in time.
>Build a simple switching regulator and perform the test you just described.
>Keep increasing the load until it self destructs.  It will not become more
>stable as load increases until you actually apply a short.  Stability
>usually follows shortly after that. :)  To get a real crude idea of radiated
>noise or EMI, take the integral of the worst case dI/dT.  This is easily
>done graphically.  You need the peak current and the period.  Yes you might
>think the spectrum was worse at lower frequencies; but no, the integral was
>close to zero since the amplitude was very small.  
>
>
>Good luck with your effort.
>73 
>
>
>-----Original Message-----
>From: hpsdr-bounces at hpsdr.org [mailto:hpsdr-bounces at hpsdr.org] On Behalf Of
>Jeroen Bastemeijer
>Sent: Thursday, July 12, 2007 7:23 AM
>To: HPSDR
>Subject: [hpsdr] Demeter progress
>
>***** High Performance Software Defined Radio Discussion List *****
>
>Dear All,
>
>There are a lot of thoughts around about the implementation of Demeter (just
>read all the previous E-mails). As the whole HPSDR-project is based on
>modules, Demeter should be able to adapt (alomost) to each individual HPSDR
>configuration.
>A modular power supply means a power supply with certain "options" (like
>e.g. a status display or a battery charger) but also with a variable power.
>For one configuartion 5V @ 1Amp might be sufficient while another
>configuration might need 5V @ 3.5 Amps. Building a power supply with 5V @ 20
>Amps will certainly be sufficient for all kind of configurations...... but
>I'm sure this is not the solution we are looking for.
>To make a power supply scaleable I have been looking around on the internet,
>and giving my creative thoughts some room...
>
>There are a few things which would fit in the Demeter concept:
>1. Using a PIC for the SMPS controller (you can check the Microchip website
>for AN216) This will certainly add the High Performance to Demeter.
>2. Build a multiphase controller. Just google for "multiphase SMPS" and you
>will find loads of information. This solution will give some extra
>flexibility: People who need less power, can ommit a couple of "phases" 
>as one or two phases can provide enough power for their solution. People who
>need more power can extend the SMPS to four phases. As this technique is
>rather new, I hope we don't get any patent-issues when we apply it for the
>HPSDR-project.
>3. Another nice technique is 'soft-switching'. This will limit the influence
>of big switching spikes.
>
>A general question I'm trying to get an answer on is the following: How does
>the output noise spectrum (please note: not the level of noise but the way
>it is distributed through the spectrum) of  a SMPS depend on the load of the
>SMPS?
>
>The thought on which this question is based is the following: Looking at the
>spectrum of a 50% square wave you only see odd harmonics. When the
>duty-cycle starts to differ from 50% more even harmonics start to appear.
>The regulation principle of some types of switchers depends on a PWM
>switching signal. This means that they will produce more noise at low loads
>(duty cycle < 50%), the lowest noise at half the power rating (duty cycle
>50%) and more noise for higher power levels. If this is true, it would be
>worth to try to match the power supply as close as possible to the power
>which is required for a certain HPSDR-configuration.
>
>That's all folks, for now at least,
>
>73 Jeroen PE1RGE
>
>--
>Ing. Jeroen Bastemeijer
>
>Delft University of Technology
>Department of Electrical Engineering
>Electronic Instrumentation Laboratory
>Mekelweg 4, Room 13.090
>2628 CD Delft
>The Netherlands
>
>Phone: +31.15.27.86542
>Fax: +31.15.27.85755
>E-mail: J.Bastemeijer at TUDelft.nl
>GPS: Lat N52.00002 Lon E4.37157 Alt 46.2m
>
>_______________________________________________
>HPSDR Discussion List
>To post msg: hpsdr at hpsdr.org
>Subscription help: http://lists.hpsdr.org/listinfo.cgi/hpsdr-hpsdr.org
>HPSDR web page: http://hpsdr.org
>Archives: http://lists.hpsdr.org/pipermail/hpsdr-hpsdr.org/
>
>
>
>  
>

-- 
Ing. Jeroen Bastemeijer

Delft University of Technology
Department of Electrical Engineering
Electronic Instrumentation Laboratory
Mekelweg 4, Room 13.090
2628 CD Delft
The Netherlands

Phone: +31.15.27.86542
Fax: +31.15.27.85755
E-mail: J.Bastemeijer at TUDelft.nl
GPS: Lat N52.00002 Lon E4.37157 Alt 46.2m

 1184336297.0