Geoff C wrote:
Patrick Turner wrote in
:
But I really can't see too much wrong with using an all SS bridge but
with each SS diode consisting of 3 series 1000piv rated diodes, to be
sure to be sure,
12 diodes in total.
To stop noise, use appropriate seris R and some caps,
to prevent large current pulses charging a cap, or a group of series
caps.
A choke input filter avoids the current pulses of a capacitor input,
but a large well made
choke is needed.
I only use cap input filters and SS rectifiers and I add series R to
limit charge current pulses.
There is no noise from rectifiers on the output.
There is not even any noise in the SS amps I occasionally build with
large caps in the supplies and no series
current limiting resistances, and no snubber caps across the diodes.
But I do try to screen off the rectifiers inside a box for the PS.
What keeps PS noise low is good layout, screening, earth path
rightness, and adequate rail filtering. I never use oil caps, and am
quite happy using el cheapo plain non fast SS diodes and generic
electrolytic caps with LARGE C values,
and good ripple current and low esr ratings.
I tried "fast recovery rectifiers" and got NO magical improvements
that I could detect.
Patrick Turner.
I have found that with no caps on a diode bridge you will get a great
deal of noise generated from diode switching in the range 50-300KHz. It
is easy to attenuate by a good 20dB with a 100n cap across the ac
connections to the bridge.
There have been amps I have cobbled up only to find some little
peaks, or groups of HF rings once for every diode switch off/on.
The 0.01 uF caps have rarely if ever been successful in supressing such PS
spikes in an amp output.
The current surges in the diodes excites the stray C and leakage L of the
power tranny,
and presto, you got a resonant circuit at 50kHz to 300kHz.
Often a 0.05uf 1,000 v straight across the HT secondary, and other larger
caps
across windings where rectifiers are used.
Series diodes so that R = about 6 times the reactance of the cap input
also work, so if a 470 uF cap has 3.4 ohms XC at 100 Hz, use
about 27 ohms in series, and the peak charge current into that cap will be
greatly reduced, and usually so will the mechanical noise in the power
transformer.
Dissipation in the HT winding relates to I squared x R,
so if I is reduced 3 times, the dissipation in the HT winding is reduced.
Its in the series R instead, but I'd rather blow a few 5 watt wire wound R
rather than
blow a power tranny.
I spent considerable time building an LCLCL filter to make the mains clean,
but with inductors on the last led to the amps I was testing to
gain my C-cick accreditation number.
The mains thus has 50 ohms Z at over 50 kHz, and at first I have
a lot od switching ring noise
trying to go back into the mains.
A 0.05uF across the HT reduced it by 30 dB and well below
the 2 mV permissible stipulated by the authorities.
Usually linear PS mounted within all metal enclosures are very easy to have
comply with
authorities emission requirements and its easy to keep the signal clean.
SMPS which charge directly off the mains into a resevoir cap and then
take a square wave signal out at 500 kHz are renowned for causing all sorts
of
bothers.
Not in my PC though, where they seem to have the problems all sussed.
Not in Halcro amps either.
One has to do whatever it takes to keep an amp quiet, and usually its a range
of measures,
not just one.
Having input sockets and wiring or feedback wiring within 100mm of any
other wire mains input wiring or wiring near anything connected to a
rectifier
and thus carrying switching currents is likely to result in noise at the
output.
So screen things off, loctate things properly, try different positions, star
earth,
and use shieled wire for FB loops.
Merely placing caps across diodes don't do much.
The noise is mainly reflected back to the
mains side, and I detect this with a spectrum analyser and a LISN. Even
an electrostatic screen and large iron bulk of the mains tranny won't
save you from this noise. However, the noise is broadband and outside the
audio band, so as long as it does not modulate other equipment it will
not be noticaeable. But it is really easy to eliminate with one cap. Have
a lok at a schematic for a Tektronix CRO and you will see such an
instance.
Patrick Turner.
|