Iain Churches wrote:
>
> "Phil Allison" <philallison@[EMAIL PROTECTED]
> wrote in message
> news:681onaF2qvh31U1@[EMAIL PROTECTED]
> >
> Patrick T wrote:
>
> >> I don't know how many tube amps you have designed and or built or
> >> repaired count. Many tube amps will oscillate at HF when connected
> >> to a capacitor load.
> >
> >
> > ** Only a psychotic lunatic would deliberately connect a capacitor
across
> > the output of a tube audio amp.
>
> Quite wrong Phil. 0.22µF is a standard (albeit tough) test
> for tube amp stability. You will find it mentioned in articles
> by Tremaine and Crowhurst.
>
> >> In order to gain unconditional stability ....
> >
> >
> > ** An unnecessary, pedantic, pseudo technical wank.
>
> Unconditional stability is more important now that is was
> in the hey-day of tube amps when many speakers were 15
> Ohms, and formed benevolent loads.
It matters not what the speaker ohms are.
15 ohms ain't necessarily easier to drive or
inclusive of less reactances than any other Z
For a given geometric layout of windings in any OPT,
there may be 2 lots of sec windings which give a match for 16 ohms
when windings are in series, and 4 ohms when in parallel.
The OPT leakage inductance is an imperfection present when implementing
a
load coupling idea such as an OPT.
This leakage is in effect in series with the perfect model of OPT
which has no winding resistance, leakage inductance or shunt
capacitances.
The change in the way the sec windings are arranged to suit either 4
ohms or 16 ohms
makes not the slightest difference to the amount of reflected effective
leakage in series with the primary 'perfect' winding, and as long as ALL
the sec windings are used in either series or parallel arrangements, and
as long as the
current density is the same for all wires in S windings, then the
winding losses
all stay constant along with shunt C when 'looking into' the primary
of the OPT.
A 16 ohm voice coil will have perhaps twice the turns of a 4 ohm voice
coil.
So it suits higher voltage and lower current.
But the amount of inductance is much increased, and an average
16 ohm speaker will have virtually no loading effect on an amp
at 50kHz, allowing output tube gain to go high, and thus allowing
oscillations.
A 0.2 uF has about 40 ohms reactance at 20kHz, but only 16 ohms at
50kHz.
The inductance value of leakage appearing looking into the 16 ohm Sec
winding of the OPT
is much higher than if the S winding effectively had 1/2 the turns as
would happen in the parallel connection for 4 ohms.
So its the OPT with the 4 ohm outlet that will be less affected by tests
with 0.22 uF,
unless of course the maker of the OPT has the 4 ohm outlet from a CT on
the 16 ohm winding,
just like so many lazy dumbass makers do these days and always have, ARC
included.
So, your statement I quote below is far far too simplistic and
and does not indicate any real truth of the amp-speaker interface
situtation either right now, or 50 years ago....
""Unconditional stability is more important now that is was
in the hey-day of tube amps when many speakers were 15
Ohms, and formed benevolent loads.""
Stability has ALWAYS been important, but makers knew 50 years ago as
some (but not all)
know now that unconditional stability AND excellent bandwidth at full
power costs several
beans and maybe a few extra walnuts towards the production costs.
Mainstream amp making has always been dominated by really craparse
designs, like Quad-II amps
and many Leak models. Its SO easy to build something better than these
sample
poorhouse thinking did. But it COSTS MORE, and the amp becomes heavier,
and it won't sell as well as the worser amp which people are
prepared to put up with. Heck, only BBC engineers had the slightest idea
what unconditional stability actually meant at all.
Phil says, about unconditional stability, that it is ""An unnecessary,
pedantic, pseudo technical wank.""
And in some situations he's perfectly correct, especially in a radio AF
output amp
where there is only going to be ONE speaker connected, the one inside
the
cabinet with the radio chassis.
But Phil isn't correct in other situations where any speaker ever made
might
find itself being hooked up to some amp which will oscillate
because of the way the the speaker fails to load the amp at HF due to
the
speaker inductance.
Or the speaker has some shunt C like most ESL which will cause a phase
shift
and cause the sine wave response to become peaked between say 8kHz and
50kHz.
The critical damping network should not only stop
oscillations when no load is present, but also stop oscillations when
ANY value of pure C is used as a TEST LOAD.
( Are you listening Phil? I never said you expect any speaker to ever be
a pure C.)
If you connect say 2uF across the output of many amps, ( and including
some SS )
there will be a peaked sine wave response at low levels where the music
is
listened to, maybe 1/20 of the clipping output power.
If the amp is tested at full PO of the 1kHz level into an R load,
and has 2 uF connected, then by 10kHz, the amp will have a pure 8 ohm
reactance
and the response will begin to sag, and by 20kHz, there will be only 4
ohms reactive loading
and the waves will be severely distorted due to over loading and vain
attempts
by the NFB network to overcome a declining RL value.
So NEVER test amps with pure C loads above about 1/4 of the maximum
clipping VOLTAGE
level of mid F performances into rated R loads.
So, with output levels low, you will see the peaked response caused by
the
2 uF cap loading. A typical peaking occurs at 32kHz, maybe the peak is
+6dB above the
reference voltage output at 1 kHz.
So despite the lowering of C load value as F rises, more output voltage
is produced across this load than at LF.
Some portion of the rise in response above 8kHz towards a peak may be
well within
the audio band, which is not wanted.
Hence the need to keep leakage L low, which means more interleaving,
thicker interwinding insulation layers, and this always means
less units per hour can be produced in a factory, and that what is made
has to be done by
the best and highest paid skilled workers.
((((( Bean Counters was a guy who carried a pistol during his factory
tours.
When he spoke to unionised workers about the skills and time they needed
to wind what the
engineer told them to, he shot the worker on the spot.
Then he shot the engineer, and used their bodies in the factory boilers
which kept the workplace just above 13C most winter days.
Back at Cumpunny Headquarters, he told the design team who shivered in
fear
to make the PO 20% higher and weight and size 20% lower,
and told the marketing guys to create a line of utter bullshit to
promote the product.
This is how profitable ventures are realized. Never give anyone what is
ideal.
Give the bastards just enough to keep 'em alive. No need to
tempt them to get fat.)))))
The phase shift caused by Csh makes the phase
relationship of Vin at V1 grid to VFB at the V1 cathode less phase
coherent,
so much so that the NFB becomes LESS effective than when no cap is
connected.
AND the leakage inductance and C load get together to make a series
resonant circuit.
So, Peter Walker and others deliberately made sure their ESL speakers
have SOME series R looking into the input of their speakers.
Its extremely unusual to find any speaker ever made that
presents itself as a pure C load to an amp.
ESL are the ones that have much C in their make up, but because there
is always a step up transformer, there is always some series input
resistance,
and possibly some shunt R as well.
The model I use to test amps for their suitability for use with SEL
is a 2uF + 1.5 ohm in series, plus 15 ohms in shunt across the R&C.
Many tube and other amps with a 2uF load give a very typical peaked
response at 32kHz, maybe +6dB
and maybe +3dB at 20kHz. But with an added series 1.5 ohms, the
peaked response is reduced to maybe 2dB max at 32 kHz, and
maybe only +0.5dB at 20kHz, which is quite acceptable.
The added 1.5 ohms is the DAMPING resistance needed in what is an
effective series resonant
LC circuit so that the Q of the resonance is much lowered.
The leakage L + C form a second order low pass filter.
For an unpeaked response from such a filter you MUST have some shunt R
across
either L or C or both, or some series R .
This is very basic LCR theory, and all speaker and amp makers need to be
very aware
of the need for critical damping of LCR networks if they wish to
see a flat response before or after a pole where the beginning of
attenuation has been planned to occur.
Some peaking is allowable with pure C loads.
It is a normal function of an LC filter with insufficient damping
But we do NOT NEED a perfectly flat response above 20hHz, so
damping brought by added R to the LC circuit does not need to be greater
than what is needed to stop the amp oscillating.
Some ringing will always occur in amps with a C load,
and its not necessary to prevent it all.
The F of the ringing is usually above the audio band and does not
matter.
Patrick Turner.
> Iain
|
