View Single Post
  #10   Report Post  
Patrick Turner
 
Posts: n/a
Default



Rich Sherman wrote:

Hello Patrick:

I have always wondered about this question but from another perspective.

Let's say I connect a my 8 Ohm speakers on a Dynaco ST-70 to the proper 8
Ohm tap. Doesn't the feedback loop now intruduce a series inductance
produced by the connection of the feedback to the 16 Ohm tap?


Fair question indeed.

In my experience, the response at the speaker sags while that at the 16 ohm tap
looks better.

For best correction of the sag in speaker response the FB should be taken from
the speaker terminals, not via the extra winding which must introduce
some effective extra series L.
The model for the OPT at HF gets rather complex, and the "unused"
portion of the sec winding in this case remains magnetically locked to the rest
of the
tranny, but nevertheless the response sags.
One might expect the NFB take off from the end of the sec might make the
response correction best, but it doesn't, much in the same way
as taking an anode signal back to some previous stage
for FB.
This last option is sometimes used, (EAR509), to avoid the
phase shift caused by the leakage inductance, since its not included in the
NFB loop.
Careful OPT design is required to minimise these concerns,
and a tertiary dedicated FB winding wound close to the sec
is perhaps the best option, so that the speaker sec is slightly isolated from
the
FB network.




Doesn't this now become a series RL feedback loop?


Slightly, yes.



Or is the inductance of the left over winding negligible?


Its a small L, and the series R of the NFB and this series L
have a very high F pole.
Phase shift occurs before the pole, and as I said,
exactly what the model is for a given OPT for the actual
L and C between each section is a very complex model.




If one uses the 4 Ohm tap then the value of L increases even further, would
this change the feedback any by causing the loop to open-circuit at high
frequencies?


It would open circuit, but by the time one gets to the F pole the
amp open loop gain has already sagged, usually so that the phase shift caused
doesn't
make the amp oscillate due to this cause.
Other things might cause the amp to play up though.



Just curious....


Curiosity never killed too many people's brain cells, and the more you
allow yourself to figure out OPTs, the more questions are raised than can be
answered.

Leak had a range of ways to connect the 4 secondary windings on their
TL12 amps. Each impedance match also required different NFB R and compensation
cap.
The ordinary man in the street could never ever have guessed how to alter the Z
match
without a tech.

And some arrangements of secs in a Leak give quite different
stability margins. Usually the 4 or 16 ohms are the lowest loss,
best response, highest stability compared to the 8 ohm match.
But in 1955, many speakers were 16 ohms.

Patrick Turner.


"Patrick Turner" wrote in message
...


Jim Gregory wrote:

I notice the valved/tubed amps' negative feedback source node is usually

at
the live leg of a secondary designed for feeding a 16-Ohm load.
But I also found that none of the various valve/tube schematics in these

RAT
topics involved an obvious Zobel R+C network across their final o/p

point.
These frequency-conscious shunts are often implemented in audio

buffers/line
drivers coupled to o/p xformers.
When made prudently, its adoption restricts super upper bandwidth to

give
reliable HF stability, thwarting probable misbehaviour arising with high
freq resonance in xformers.
In such power amps, assuming the load is always connected, was it

omitted
because the loudspeaker coils + crossover network/s would band-restrict

any
funnies beyond 22kHz anyway?


The NFB resistor, and its compensation phase advancing cap do not have any
effect
on the open loop gain of the amp at HF, which should be limited by the

zobel
across the whole of the sec winding, so an R load is is effectively

connected at
HF
above 20 kHz where instability will occur.
Most speakers are inductive with rising impedance as F rises, so
without a zobel, virtually no load is connected to the amp at 100 kHz,
where excessive output tube gain may cause instability due to rapid phase

shifts

due to poor OPT quality with large shunt C and leakage inductance.



Or do the NFB components, effectively in parallel with the secondary
winding, also react like a Zobel R+C out of band?


No.

But then there is not
usually a low-value resistor in the NFB loop.


Yes.



Incidentally, why don't any of these o/p transformers ever have two
secondaries for either parallel or series config to cater for 4 or 16

Ohm
loads (you might also then need 2 Zobel R+C shunts), delivering

virtually
the same peak Power?


Mnay OPTs do have more than one winding for various load matches.
See my pages about OPTs at
http://www.turneraudio.com.au/htmlwe...utputtrans.htm



Though 8 Ohms would become the odd-one-out impedance in that event, you
would do away with 1/2 way tapping and its associated half-wasted energy
whose whole isn't a lot of Watts to start with, by today's standards.
Jim


It is indeed harder to cater for 4,8 and 16 ohms.

But the best amp will have no wasted windings on OPTs, and
current densities in each wire of the secondaries is equal when each

impedance
selection is made, and leakage inductance remains constant for all load

matches
when referred to the primary.

Hardly any amps are configured like this, but its *the* right way to go.

Patrick Turner.