Patrick,
I seen your post about repairing an Audio Research SP-11. I think it's a
hybrid. Anyway, the SP10, I believe, had more tubes and is supposed to be
the best sounding preamp, hands down. What do you think of the preamp you
just fixed as far as circuitry, built, and of course, sound? Did you have a
schematic?
BTW: as an aside... on one of your amp schematics (I can't think of which
one right now), you have a resistor in series with a cap going across the
primary of the OPT. I kept thinking "why did he do that?" Well it's been
eating at me long enough. Can you explain? Thanks so much for answering
these two very different questions.
Cordially,
west

west wrote:

Patrick,
I seen your post about repairing an Audio Research SP-11. I think it's a
hybrid. Anyway, the SP10, I believe, had more tubes and is supposed to be
the best sounding preamp, hands down. What do you think of the preamp you
just fixed as far as circuitry, built, and of course, sound? Did you have a
schematic?

The SP11 sure is a hybrid.

Input comes to the gate of a common source npn j-fet, Q1, with 127k as the RL.
The drain output powers a pnp source follower, Q2.
Q2 source powers the cathode of 1/2 a 6DJ8, V1
The V1 anode is loaded with 100k, and the output goes to the gate of
Q3, an npn mosfet which buffers V1 output.
Q3 source R is 15k plus 3.16k which forms a FB divider feeding V1 grid,
to reduce the gain of V1 from 30 to 6.
Q3 source feeds the grid of V2, another 1/2 6DJ8, with anode RL = 27k
and V2 acts as a phase inverter with a gain of 1.5, with an 18k cathode R.
Anode output from V2 powers the gate of another pnp j-fet source follower, Q4.
Q4 source powers the cathode of V3, another 1/2 6DJ8, with RL = 68k.

The anode output of V3 powers another npn mosfet source follower buffer, Q5.

Output from the phono amp is taken from the Q5 source, via 567 ohms series R.

Global NFB is taken from the Q5 sources and from 1/2 way along the 567 ohms
all the way back to Q1 source, with a seven component RIAA eq network.

So there are 3 triodes and 5 SS devices in the signal path.

In addition to the SS mentioned, another j-fet, 4 diodes, and a zener diode
stop the circuit from self damage of the fets by over voltaging.

Despite the safety voltage limiting, the Q4 in the amp I just fixed
had a leaky gate to source junction, which upset all the
DC voltages of the amp.
It is DC coupled over the 4 stages.

A replacement soon fixed the problem.

The impedance matching swiths was faulty, so I
unwired that from the circuit, although it sits there looking pretty.

I soldered in an MC matching R and finally the amp became usable.

When my customer came around to fetch the amp,
I played Rye Cooder's Jazz album from around
1978, and it sounded clean and fresh, but noise was well above
what I do with a cascode input stage.
Cascode with fet + tube is quieter than cascade with fet + tube.

I didn't spend long comparing my own preamp to the ARC.
The wasn't too much wrong with the sound from ARC.
It was in an otherwise impeccable system, so how much ARC contributed
to the good sound I don't know.

The ARC SP11 is a very complex amp.
My own present preamp has j-fet in cascode with a trioded
6EJ7, then passive RIAA eq, then a 12AT7 bootstrapped follower for more gain,
so 1 fet, and 3 triodes, the last of which is a follower to buffer the phono
stage.
It sure sounds as well as the ARC, and is a little quieter.

The cascode hybrid circuit may perhaps have been used first used by Allen
Wright.
We can trust a j-fet such as the 2SK369, or 2SK147 with a few mv of signal
without any crook thd occuring.

I didn't measure the thd of the ARC, but mine gives 0.1% at 10 vrms output.
This means that at a typical high normal level of 0.5 vrms, there is
less than 0.01% thd, and that's all mainly 2H.

Personally, I'd never buy an ARC amp because despite the
good pcb engineering, there is too much to go wrong,
it is hard to trace the circuit, and removing a switch isn't easy.
And ARC charge a minimum of usd $80 per hour for service.
Probably cheap compared to many other professional fixers.
I like my circuits ever so much simpler.




BTW: as an aside... on one of your amp schematics (I can't think of which
one right now), you have a resistor in series with a cap going across the
primary of the OPT. I kept thinking "why did he do that?" Well it's been
eating at me long enough. Can you explain? Thanks so much for answering
these two very different questions.
Cordially,
west

Typically, I will place a R + C zobel network across each
1/2 of the OPT.

As the F rises, the tube anode circuit will begin to see the leakage inductance

as an increasing Z in series with the load.
Also there is some shunt C from anode to the earthy secondary windings.
The LL and Csh form either a series or parallel resonant circuit
and this excites a sharp dip or peak in the sine wave response, and
when tested with a square wave, there will be some ring, or overshoot,
usually made worse when global NFB is used,
because the LL or Csh causes additional HF phase shift usually
above 80 kHz in a good amp, which we can manage, but a pita
if below 80 kHz as is the case with a poor OPT.

The exact equivalent network LCR model of most OPTs
is a complete pita to have to work out exactly, because the model
must take in the multiple stray C and stray LL between all, the sections
of windings.

But if we place some R across the tranny at HF, these de-facto
LC filter elements can be damped, since damped filters always give a flatter
response than an undamped ones.
Its very like a crossover filter to a speaker.

Say we want to have a bass speaker response tailored to roll off at ultimately
12 dB / octave, we would choose an L and a C which would together form
an LC filter, and these two reactive elements form a low impedance series
resonant
circuit at Fo, which will also be the crossssover point, Fco.
With no speaker connected, the LC circuit is said to have no R element,
and it is an "undamped" circuit.
The sine wave response from the LC junction with no R is perhaps flat
at LF, then as F rises to Fo, the response peaks sharply about 12 dB, before
then
falling rapidly away to end up at attenuation HF by a constant slope of 12
dB/octave.

When we connect the speaker after the L and acoss the C to 0V,
we introduce R to "damp" the series resonant circuit.
If the R = 1.414 x the recatance value in ohms of either L or C at Fo,
then the LC circuit is said to be maximally flat, with no peaking
to the sine wave response, and a rapid change to 12 dB/octave attenuation rate.

90 degrees of phase lag occurs at the pole, the - 3 dB point, and this
increases
ultimately to -180 degrees.

You should try setting up a sample circuit with L, C and R, and dual
trace CRO, to see what really happens with such basic and elementary circuits.

The zobel places a mainly resistive load across the resonant circuits,
and the sine wave response flattens, as well as make the impedance looking
into the LCR network into one more easy to drive, ie, without
a dip in Z due to an undamped resonance.

With a bass speaker crossover with LC low pass filter, the Zin
would be perhaps 0.5 ohms at Fo with no speaker,
although Zin a few octaves away fronn Fo the Zin would be very high.

Where RL = 1/414 x ZC or ZL at Fo,
the Zin to the LC filter with R is R ohms well below Fo,
then it dips to 0.707 R at Fo, then rises again as the L impedance increases.

Its not nice to have a series resonant circuit in the anode circuit
of a tube amp and resistive damping reduces the
severe current shunting with undamped LC.
Parallel resonances form high impedance circuits at their Fo, and
also make it harder to stabilise the amp, since
stability worsens if gain is allowed to rise too much and phase shift is
allowed
to increase too rapidly, and gain will certainly rise
with a pentode or UL amp if RL rises.

But regardless of resonances, the finite value of leakage L
may still prevent transfer of HF power to the load, just like the
crossover inductor does in the bass speaker case.

Many speakers present an inductive loading for an amp as F rises
above 20 khz, and into the zone where the gain and phase shift
can cause oscillations.

A zobel RC with R = 2.2k and R = 0.0022 uF will
have a Z RC 3.11k at 32.9 kHz.
At 100 kHz, the Z RC has fallen to close to 2.2 k and from the anode's point of
view
it appears as a mainly resistive load, and one not able to cause much phase
shift.

At below 20 kHz, the effect of the RC zobel on gain is minimal,
if the rated load also appears, since the 0.0022 uF becomes a high Z at low F,
so it has no effect, and prevents the 2.2k loading the amp at the
audio F we want to use.

The zobel RC thus only loads the amp at HF, and only because we want
the mrgin of stability to be as high as possible.

Zobel RC are also often placed across the secondary for the same kind of
reasons associated with damping.

Once placed and optimised, the zobel will help an amp to
give a flatter response into an ESL speaker, and most
importantly, prevent RF oscillations from all to easily occurring,
with catastrophic results, and at least a bad effect on the sound,
because such oscillations can such power from an amp,
and cause serious imd by such RF over loading.

Amps with really wide badwidth OPTs are easier to stabilise, and the
zobel networks designed to act at a higher F than one closer to the AF band.
Amps with global NFB loops benefit far more with judicious
use of zobel networks compared to triode amps with no global NFB.
Sometimes one has to place a zobel across the
load of V1 input triode to curb its gain at HF and to reduce the phase
lag at critical F where oscillations might otherwise occur.

Stabilizing amplifiers when global FB is used isn't for the foolhardy
or the ignoramus, and methinks one reason why some enthusiasts deplore NFB is
that
they lack the ability to measure and get it right, since their knowledge
is based only on subjective listenings, with little regard
to measurements.
Sure the amp must sound well, but it also should be stable.
Well stabilised amps with NFB seem to sound OK to me.
But it is deleterious to over damp the amplifier.


Patrick Turner

"Patrick Turner" wrote in message
...


west wrote:
BTW: as an aside... on one of your amp schematics (I can't think of which
one right now), you have a resistor in series with a cap going across the
primary of the OPT. I kept thinking "why did he do that?" Well it's been
eating at me long enough. Can you explain? Thanks so much for answering
these two very different questions.
Cordially,
west

Typically, I will place a R + C zobel network across each
1/2 of the OPT.

I am interested to know how this network can be physically
implemented. The anode windings of the primary go to the
valve bases of the pp pair, while the primary centre tap goes
to the psu reservoir cap. So these are physically quite a distance
apart.

Iain

Iain M Churches wrote:

"Patrick Turner" wrote in message
...


west wrote:
BTW: as an aside... on one of your amp schematics (I can't think of which
one right now), you have a resistor in series with a cap going across the
primary of the OPT. I kept thinking "why did he do that?" Well it's been
eating at me long enough. Can you explain? Thanks so much for answering
these two very different questions.
Cordially,
west

Typically, I will place a R + C zobel network across each
1/2 of the OPT.

I am interested to know how this network can be physically
implemented. The anode windings of the primary go to the
valve bases of the pp pair, while the primary centre tap goes
to the psu reservoir cap. So these are physically quite a distance
apart.

Iain

I sometimes place the two zobel networks on the OPT,
or between the anodes, with a 0.47 uF from the centre of the zobels to
to the 0V bus.

The leads lengths involved do not make any difference.
To test the zobel networks i often temporarily set them up
using 300mm leads with aligator clips, I get the same results
when I finally place them in the amp in someplace where they
belong.

Same goes for the zobel across V1 anode to 0V.
The leads can be 300mm long without causing any problems.
Just be sensible and don't allow such leads to be near grid circuits during the
testing.

Patrick Turner.