[patrick-turner]

On 11 Nov, 11:32, "Alex Pogossov" wrote:
SNIP

Alex typrd....

Also I think that stray capacitance noise will not be so much of a
problem -- because it leaks only in 60% of the OPT, which has 36% of
impedance. Besides, GNFB will kill the residue of noise if any.

I've just done more careful measuring of the noise in the Reformed
Jardis amps.

I did mention I thought the Jardis Chokes had L = only 80mH, roughly.
That was based on measuring 100Hz Vr hum across C2 470uF and across
choke. The 100Hz attenuation factor of the LC then is VC/VL and this
was initially 0.6V/0.05V = 1/12, and XL is about 12 x XC = 12 x 3.4r
at 100Hz = 41r then dividing by 628 gives 0.064H, ie, 64mH, even after
I set the air gap closer to get the minimum Vr with 0.15Adc flow. But
the trap everyone is prone to is that the Vac across a second C in a
CLC filter usually has lots of noise content below 100Hz, hence just
measuring across the C2 tells you a lie. So, I use a 10:1 resistance
divider with 3 x 100k and 1 x 33k so that Vdc is reduced fy factor of
1/10 to reduce the VDC measuring to +43Vdc in this case, and then I
have a 0.22uF across the 300k, so that you have a filter reduces DC
and LF but gives no attenuation at 100Hz, as it is -3dB at around
24Hz. This means nearly all the LF noise on B+ rail dissapears, and by
switching the CRO to max sensitivity and usung the scale I have taped
alongside the screen and calibarted in mV rms, I can read low voltages
even with a few mV of trace wobble due to LF noise.
I found 100Hz noise = 4mV, so attenuation factor of the CLC was better
than I thought at 0.6/0.004 = 150, so XL - 150 x 3.4 = 510r, so L =
0.81H, and that's about right considering Rw = 15r, and the size of
the choke, and the way it responded to correct air gap sizing.
Then I found another Jardis caused source of noise. Jardis put the IEC
chassis plug facing down into chassis and beside the PT. There is a 3
connection double cap on IEC plud with wires to A, E and N and they's
have no effect at LF. Then they had a 50mm green-yellow lead to one of
the 6mm bolts holding PT to stainless steel chassis. Now I had to stop
this amp humming by placing 22r between chassis and 0V rail. But today
I measure 0.47Vac across the 22r, and from PT bolt to chassis nearby,
so some serious I flow is generated within very short distances of
iron material, but hey, I could expect that no? bolts are subject to
stray fields from the large PT. So I reach for junk wire box, find
green-yellow wire, connect Earth terminal to a bolt holding a clamp
for 470uF cap. I get that done and remeasure. Bingo!, only 3mV across
22r, although yeah, the PT bolt is still having some voltage induced
in it. The 0V is connected to chassis via 22r at 300mm away at amp
input end, I find all is well. My ancient olf Fluke hand held meter
probably has dirt inside, and it gives a 2mV reading on Vac even with
leads twised together and propds shorted and away from magnetic
fields. Anyway, I can't measure noise at Vo terminals with meter, even
with no GNFB, But my best probe for my CRO with short earth lead to 0V
and short lead from amp 0V to CRO 0V shows 0.5mV without NFB, and
artifacts are various H of mains. This increase a fair bit without
shorting input to 0V with RCA shorting plug. OK, 120k input R does
make a bit of noise. With GNFB, the noise becomes almost a dead flat
line but I estimate 0.2mV. Earlier I noticed some trace wobble with a
regular timing, which telle me someone is switching something on/off
on mains thus making B+ bounce a bit with main F being probably 8Hz,
which is the Fo for 0.8H plus 470uF. Not much I can do about that, but
hey, a coupla mV of noise below 10Hz doesn't matter. So. I am happy
with result with regard for noise.
OLG response is 20Hz to 16kHz, no GNFB, -3dB.
With 9dB GNFB, CLG response with input CR filter shunted = 6Hz to
56kHz, 5 ohm load, 0.5dB peak at 8Hz, before steep roll off below 5Hz.
With HPF connected, LF pole is 8Hz, -8dB but dead flat from 10Hz
upwards. The shelving means GNFB is reduced to 4.5dB, ( halved ) at
15Hz and 20kKz with lessening FB applied outside this band. Sounds
great, no weird behaviours, all looks well. waether fine, level
horizon, will land for dinner soon.
Then to spend a night on working up ther page for the Net about the
mods, so that if anyone wants a good SET recipe, they can have it.
Patrick Turner.