Don Pearce wrote:
On Sun, 25 Nov 2007 01:44:55 GMT, Patrick Turner
wrote:
How true, but I have designed semiconductor audio stuff for years
(actually decades). Now I want to design/build some (pure) tube gear.
Filtering is just as important in SS circuits....
Patrick Turner.
No it isn't. SS circuits are generally built fully balanced, using
current sources instead of collector load resistors etc, which gives
them many dBs more common mode and power supply rejection than you see
in a valve circuit. You can get away with much poorer supply filtering
in SS.
You're not quite all correct in your comments.
It depends what sort of SS design one means.
With all discrete bjts or j-fets, and operating SE,
filtering of collector rails is very important if R loading is used.
But even with the traditional Linn basic power amp circuit
where there is a balanced input LTP stage and followed by a gain or VAS
stage
with CCS loaded gain bjt, and followed by emitter follower output
stages,
hum is quite horrendous without the massive amount of global NFB
applied.
In nearly all the SS amps I have played with I have found hum to be so
massive
when testing without GNFB connected that the amp response cannot be
measured
because the hum level is so huge.
Good filtering of all rails including rails for LTP and VAS stages
makes the open loop hum far lower, and then the open loop response can
be plotted,
and the effect of the GNFB also plotted. The noise is LOW BEFORE GNFB
has been applied,
so when the GNFB is applied, the noise is a lot lower than without the
filtering.
I filter all rails well. Its good practice, and I don't rely on the NFB
to remove all the noise.
We all know what the bean counters try to get away with.
I am not a bean counter.
Now as for valves, don't just slap in big capacitors; think. This is a
filter you are designing, for a specific purpose. So design it. Decide
how many dB rejection you need at 100 (120) Hz and do the sums. That
will give you a set of resistor and capacitor values you can use as
your minimums, Check that the spec is still met when you include the
parasitics of internal resistance in a cap; there is no point trying
to cure hum by making a cap bigger when it is the ESR causing the poor
performance.
The esr is not the cause of the problems with noise and instability.
I suggest the esr plus the resistance between one LARGE cap and the next
LARGE cap
simply sum to form a total R, and the divider action along the line
of CRCRC will reduce hum very effectively and without worrying about ESR
at all.
AND, if one places say a 0.47uF plastic cap across the last electro in
the CRCRC line up
then any RF trying to get along the resistance line ups and esr shunts
then it is shunted by the 0.47u.
The presence of ESR in large value caps these days is not noticeable
compared to ancient and poorly made old caps. This is because they
design electros now
for use in SMPS and the cap must be able to dump its charge fast enough
into a chopper circuit using
a 500kHz square wave. The efficiency of the SMPS depends on the RF waves
being
nice and square, and the SS switching devices being either fully ON or
OFF, so that the product
of voltage across the devices and current flow through the devices is
near zero.
SMPS would overheat if they used sine waves.
In ALL of my experience, hum in rails is reduced in proportion to cap
size.
So where there might be 330 ohms plus 470uF, the attenuation factor is
about 1/100 at 100Hz.
if C becomes 1,000 uF, expect the aten factor = 1/200.
Its that simple.
And where 3 similar RC sections exist, the atten factor = the cube of
the atten factor of one,
and doubling this overall atten factor means the atten factor becomes 8
times greater.
However, this is the case for LF only, and as F rises the
large value electro becomes increasingly inductive and fails to bypass
anything at say 10MHz.
In my SS 2x300W amps I have two 100,000 uF rail caps, one for +70V, one
for -70V,
and without extra other caps also bypassing the rails with short paths,
the bypassing
is poor at HF when the amp is tested with a square wave; the waves show
up in the rail where they shouldn't be.
The large caps have thick leads about 250mm long from drain connections
and 0V connections and its the
inductance and RF resistance of such long leads that causes bypassing
problems at HF.
ESR would play a part, but not a large one it seems.
So the I have 1,000 uF electros close to the drain to 0V star point,
then 2 uF, then 0.01uF.
I found this cured the problem of the appearance of rail signals due to
poor rail filtering.
The rails of the driver stages are regulated and filtered, so noise on
the output rails are
entirely prevented from appearing in drive amp rails.
Some amps run the driver stages straight off the output rails.
This is terribly poor workmanship.
And BTW, the 2x300 can be turned off with music running, and it runs on
for
20 seconds before the signal clips due to slow rail voltage decline, and
then
switched back on again and not the slightest noise is heard or any
indication that the
PS has been turned on or off.
In the 300 watters, less than 0.25mV of hum appears in the output
which is a typical SNR figure of -103dB relative to 300W, 8 ohms,
unweighted.
Without passive rail filtering, its difficult to get this performance.
Ideally, hum in the open loop amp should be less than the
open loop THD which is under 5% at clipping.
So when 60 dB GNFB is applied, THD and noise will all be below 0.006%.
Sometimes hum and noise and distortion is induced into
leads/tracks/paths
of NFB and poor N&D figures suggest the NFB isn't well thought out
as well as filtering.
Patrick Turner.
d
--
Pearce Consulting
http://www.pearce.uk.com
Anode Starvation
Hybrid Mode
