Kevin Aylward wrote:


I have *already* explained, that yes, if the feedback is *low*, than it
can sound much worse. However, if one gets into the *total* THD/IMD
figures of 0.01%, then feedback is great. End of story.

And this has been demonstrated in measurements described in two articles in
the
last 20 years which appeared in Wireless World, and Electronics World.
But the articles also concluded that the "sound worse result"
happened where thd to begin with was high, say over 5%,
and also FB was low, say under 12 dB.
Where thd was initially low, and simple 2H, say below 1%, then 10 dB of FB
will reduce the 2H by around 1/3, and only slightly introduce some 3H as a
result
of imd action, but its at a low level.

What has to be kept in mind about thd/imd with regard to triode signal
amps, is that it
starts off at zero at zero vo, then rises to the maximums we are talking
about, say 1%,
at about 10vrms output from a typical SET signal amp, and then rises to
say 5% at 60 vo.

But usually, its rare in preamps to have more than 1vo, so thd is under
0.1% to start with, and if the triodes are selected right, say 6SN7/6CG7,
and loaded with CCS loads, their thd is considerably reduced further to
negligible levels,
and the thd does not tend to increase linearly with vo, but
is slow to increase at first, and only increase more rapidly as vo rises
above where
we want to use the tube.

I have a preamp with bootstrapped follower stages, ( like mu follower
stages ),
and there is a phono amp, passive RIAA eq, tone control stage, when I wanna
use it,
and a gain stage, also deletable, and a cathode follower buffer.
FB is only used with the Baxandal network tone control stage,
and inherent with the CF output buffer.
With 10v output from any stage, thd is less than 0.2%, and at 1vo, thd is
around the 0.02%
mark, depending whether the 2H is allowed to cancel the next stage's 2H,
which happens when the line gain stage is used with say the phono stage.

Noise is worst in the phono amp, but snr is around -75 dB.
The noise of the phono amp is reduced from what you'd get
from an all tube line up by the use of a fet, the 2SK369, as an input
device, in a cascode arrangement with a following triode.
The snr is a lot better for line inputs, because the signal level input is
high compared
to a phono amp's, and yet the input noise of a triode is typically 2 uV.

The use of discrete j-fets on their own in either SE or balanced stages
isn't a pretty picture
once the signal level climbs above 0.1v, because the fet thd is far worse
than a triode's,
like about 20 dB worse. But where only mV are being processed,
the fet does OK.
Bjts are just as bad as the fets, and offer odd order thd as well.
So to linearise any SS devices, FB is required, but not in triode amps,
if you don't want to, and you design for it.
If FB is added to triode signal amps, the thd becomes very
low without much FB applied, because it was low to begin with.

Balanced circuits using triodes reduce their thd by around 20 dB
without using any FB.

Before anyone condemns triodes on the basis they have high N&D,
they should just check the facts.

Patrick Turner.