OK, here is the in depth reply to Mr Pepper's inquiries.

I am an electronics engineer working for a semiconductor company (I'm
not allowed to say which one), I layout integrated circuits, mostly
large high speed digital with sensitive analog circuitry as well. My
specialty is power distribution networks in these chips, providing low
noise power and ground networks to provide very low jitter on the
interface between the digital and analog sections. This jitter needs to
be kept in the single digit ps range, which is not easy. In order to
meet this we need on chip regulators, these regulators are always
linear regulators, switching regulators simply produce too much noise.


At the system level we usually have a switching regulator on the board,
followed by a linear regulator on the board and the linear regulator on
chip. This is required in order to keep noise levels low enough so we
can meet the jitter requirements. Believe me, if we could get away with
a switching regulator on the chip we would do so. 

As to my personal experience, I have designed many switching and linear
regulators, I have only designed a few switching AC mains to DC
supplies, but I have worked closely with those that do. I have measured
many switching supplies, I have a fairly good good idea of what they can
and cannot do. I have designed many linear AC Mains to DC supplies and
measured those as well. 

As to how this relates to audio, that is a long saga. Many years ago I
was building audio DACs for myself and found out that what power supply
I used had a significant impact on the sound. I set out to find out what
differences in these supplies caused the difference. I ran many hundreds
of tests with different supplies, listening and measuring the supplies
and trying to come up with correlations, it was not easy and took a
couple of years. One thing I learned from these tests was that most
supplies feed a fair amount of noise back into the AC mains as well as
the noise feed to the audio circuits. This noise injected back into the
mains turned out to be one of the most important aspect of the tests,
and one which is frequently completely ignored in power supply tests. 

It turned out that frequency range that caused the most impact on sound
quality was the 40KHz to 200KHz range. MHz and up was usually well
controlled and filtered, and frequencies in the audio range were
usually handled very well by the regulators in the circuits, but the
intermediate frequencies were not well dealt with. And it was the
amount of this noise injected back into the mains that had the most
impact on sound quality, presumably by being carried to other
components in the "stereo system" such as power amps and preamps. I
tried this with many different audio systems, from inexpensive consumer
systems, expensive audiophile systems, pro audio systems and all of them
seemed to be susceptible to this incoming noise on their AC. 

Many of these components contained "line filters" designed to block
incoming noise, but most of those only seemed to be effective in the
MHz and up range, having very little affect in the 40KHz to 200KHz
range. I probed around inside these other components and watched what
happened to this noise coming in over the AC mains and found many of
these components actually had resonances in this range which
significantly magnified incoming noise. In quite a few cases even if
their own power supply was not generating noise in this region, the
resonances wound up feeding significant noise into the circuitry when
noise in the right range was on the AC. 

So the summary of all these tests was that the biggest requirement for
a PS was to not send noise in the 40KHz to 200KHz range back down the
AC line, keeping noise out of the DC fed to the DAC also made an
impact, but it actually was not as big a concern as what went to the
mains. It also had to not be susceptible to noise in that range coming
from OTHER components. 

So I set out to design a PS that met these goals. Since this was going
to be a onesie for my own use it was a lot easier to build a linear
supply. I did a LOT of spice simulations trying out different
topologies and components which eventually evolved to what I have
today. I built several and proved that the real world did in fact
closely match the spice results so I was quite confident that doing the
exploration with spice was a reasonable approach. 

The result is very different than most linear supplies. Most linear low
voltage supplies employ a single stage very large value cap after a
bridge regulator. This of course only conducts over a very narrow
portion of the cycle, giving rise to very high current pulses in the
transformer. Common rectifiers produce switching noise in the bad
frequency range, with in conjunction with the current pulses cause the
power transformer to ring like a bell right in the bad range. This
noise goes right through the power transformer and into the mains. Even
if you have a good regulator which blocks this noise from the circuit
beeing fed from this supply, the noise still goes out to the mains. 

My design attempts to alleviate these issues in several ways, first it
uses Schottky rectifiers, second it uses a choke based asymetrical PI
filter after the rectifiers. The first cap is much smaller than the
second cap. The result is a supply which has cunduction over almost the
entire cycle, thus no large current spikes. Having the small first cap
allows the large conduction angle, but it also allows the filter to not
need nearly as much minimum current as a true choke input filter. What
comes out of the filter is almost pure sine wave which is very easy for
the regulator to deal with. In a tgraditional linear supply the sawtooth
output has a lot of high frequencies which are much harder for simple
regulators do deal with. The third part is a damping network across the
power transformer which damps the winding resonance so it will not
resonante from noise coming in from the outside world. 

The result of all this is a design which has extremely low noise
injected back into the AC line and very low noise sent to the audio
circuitry. It's simple, does not contain any custom components, no
special "audiophile" parts, is easy for inexperienced DIYers to build.
Is it the only design possible? of course not. Can a switching supply
be made which can match this performance, almost certainly. But its not
going to be nearly as simple, will probably need custom magnetics and
could very well cost more money to make and will be way out of the
range of a DIYer to build. 

I have tested many commercially available switching supplies (certainly
not all), designed for many different applications, including many that
come with audio devices including some very expensive pro audio
equipment and NONE of them bettered the noise levels of this simple
linear supply. So while it is certainly possible that a switching
supply good do as well or better, it does not seem that this is common
in the market place. It is also true that most linear supplies are also
very bad as well, and many linear supplies are worse than the switching
supplies that people are replacing. I have NEVER said that a linear
supply is always better than a switching supply.

If you would like to test these claims go ahead and build this design
(its easy and not very expensive) and compare it to any of your
switching designs, paying attention to what is injected back into the
AC mains as well as what gets sent to the device, I'd be very
interested to see how it stacks up with your switching desings. Oh
yeah, and while you're at it try hooking them up to a Touch and see if
you hear any differences.  

Here is the schematic of the design mentioned above:

http://johnswenson1.home.comcast.net/stereo/SB_5V.GIF

John S.


-- 
JohnSwenson
------------------------------------------------------------------------
JohnSwenson's Profile: http://forums.slimdevices.com/member.php?userid=5974
View this thread: http://forums.slimdevices.com/showthread.php?t=82648

_______________________________________________
Touch mailing list
[email protected]
http://lists.slimdevices.com/mailman/listinfo/touch

Reply via email to