Re: Prototype "tube-like" solid-state amp circuit

On Jan 6, 4:50 am, "Elvis Kabong" <ampscie...@xxxxxxxxxxxx> wrote:
<morris.***...@xxxxxxxxx> wrote in message

news:9214a988-f7ff-42cb-a269-cd0d398a18ce@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx



Warning - amp content!

Ok, I know that calling a transistor circuit "tube like" is both a
provocation and a cliche.  I admit - I'm not trying to emulate a tube
circuit.  I just want something that sounds good.  And right now I'm
trying to design a simple, good-sounding discrete transistor amp, 10W
or so in power - and I want it to sound good.  Don't want it to just
buzz or beep when a note is hit, don't want that Peavey hard-clip
washed-out mess.

Here's the schematic of the prototype that I've been playing with
today:

http://img229.imageshack.us/img229/2820/10wpowerstagemb9.png

I figure that a lot of the nice sound that tube circuits give for
guitar is derived from their simplicity, as opposed to typical
transistor audio circuits with many active devices and boatloads of
feedback to cancel all their many nonlinearities.  On the other hand,
you can't just swap a transistor for a tube.  Transistors don't have
the combination of gain and linearity that tubes do.  So what I
decided to do was use the transistors in Sziklai or Darlington pairs,
with plenty of local degeneration to improve linearity.

JFET Q1 and PNP transistor Q2 are a Sziklai pair with local feedback.
Bias adjustment - the 300 ohm resistor - is a bit sensitive.  And it
will always have a bit of asymmetrical clipping however it's biased.
But it can handle a clean guitar input and sounds nice.  Not perfectly
clean clean, more like jangly rockish (even-order harmonics), but I
like it.  Especially the lack of the need for an input cap/biasing
network to suck out bandwidth of the input signal - this stage
actually can handle input signals that go negative to ground!  Q1 also
gives a nice high input impedance.   Why not just use a simple one-
transistor JFET stage?  They just don't have enough gain and
linearity, not nearly as much as a triode would.  So Q2 just helps Q1
out here.

Q3-Q6 form a class AB output stage using a common-emitter topology
with a swinging, floating power supply in order to obtain voltage
gain.  Darlington pairs Q3-Q5 and Q4-Q6 are used instead of single
power transistors in order to get sufficient gain, as well as to
provide a safety buffer against base-emitter zener breakdown and
consequent transistor destruction.  Resistors R5-R10 form a feedback
network and are used to set gain, linearity, bias current (in
conjunction with emitter resistors R13-R14), output impedance - ok,
they are pretty overworked!  But reasonable compromises are
available.  Capacitors C2 and C3 provide AC bypassing of a portion of
this voltage divider, allowing higher signal gain (at the expense of
some linearity).  They also tend to cause bias shifts when this power
stage is driven hard, which cause some compression - remniscent of
cathode-bias bypass capacitors in tube stages.  Somehow, this goofy
center-tapped floating power supply sort of looks like a push-pull
tube stages output transformer, if I squint at it hard enough . . .

Capacitors C4 and C5 are both filter capacitors and output coupling
caps.  I know some folks don't like capacitor-coupled audio output but
- the power supply caps are ALWAYS in the circuit one way or another.
Might as well use them this way, I figure.  And who wants to worry
about DC offset frying speakers if something goes wrong?  Not me!

The output is somewhat underdamped - output impedance is about 6 ohms,
according to SPICE - but it sounds good when tried in real life.  No
worse than a typical pentode amp without feedback.  And it's not all
that distorted, but overdrives nicely when played hard.  I rather like
the sound.  Pretty sure that this prototype circuit is going to end up
fleshed out into a complete amp.

--------------------------------------------------------------------------------------

I think that this looks pretty atypical for a solid-state guitar amp.
No op-amp front end like Peavey uses.  No simple single-transistor
JFET front end stages either.  And definitely the power amp is
different.  Not at all the typical setup of integrator-driven output
stage with massive feedback and slew-rate limiting for stability, no,
it tries to get by with local feedback.  I wonder - why people don't
do it like this?
--------------------------------------------------------------------------------------

What do you guys think?

To emulate the tube sound, I'd try germanium transistors
and an output transformer to "warm" up the tone.

Yeah, like I said I aren't really trying to emulate any particular
tube sound, I just want it to sound good.

I've thought about how a solid-state amp might be built with an output
transformer, even thought about doing it, but such a device makes no
sense. Transistors are small, everyone says, they're ever so much
smaller than tubes. Why, this current amp project uses two little
TO-220 power transistors each no bigger than a fingernail! Except
that they'd last about 30 seconds ran like that, even at idle. So of
course, they need a heatsink capable of dissipating 20 Watts of heat
or so. Mounted them on a Pentium II type heat sink, with nylon bolts,
thermal stick-pads, and silicone gunk, it hooks to the main breadboard
with a couple 3-pin computer fan connector cables that just slide
neatly over the transistor leads, totally modular. And if you hold
that assembly in one hand, and hold a pair of EL84s in the other, the
transistors & heatsink together are BIGGER and HEAVIER than the
tubes! The ONLY reason why the transistors are 'smaller' is the lack
of the output transformer! And if I have to use a big old transformer
for solid state output, I've just built something that's every bit as
bulky as a tube amp, more complicated, and likely more costly and less
easily maintained.

Germanium transistors - have thought about them. NTE still has some
in production and I could order a pair or two. They are fairly sweet
devices, actually! Not available in as huge sizes as the biggest of
silicon transistors, but they have 2 very major advantages. First of
all, they have a much higher Vebo breakdown voltage than silicon
transistors do, allowing them to be driven much harder in a common-
emitter or common-base topology. Secondly, they have a much softer
turnon, a much lower junction potential - they can be ran much colder
at idle in a class AB stage with pretty good linearity, even without
global feedback! Yeah, germanium transistors are sweet. I will buy
some someday to play with.

I've seen a schematic in an ancient book - world's simplest transistor
output stage - center-tapped interstage transformer, driving 2
germanium power transistors in common-base mode, the collector circuit
driving an output transformer, zero-biased - given the soft turnon of
germanium, this probably doesn't have too much crossover, it probably
sounds real good! I'm not sure that it would be any cheaper to build
than a tube power stage, but yes, I've seen that in old books, it
probably does sound awesome. Two transformers, two transistors, one
power supply, no resistors or caps or anything, now that's simple!
Might be a really terrific way to do a micropower amp, 1 Watt or so,
or maybe a headphone power stage.

However, even though I design and build tube amps (with some
advice from real engineers once in a while) I am not an
engineer of tube amps and especially not solid state amps.
I just go by what I've observed and deduced.

In fact, many 90's Fender SS amps use NPN and PNP
Darlington pairs for outputs and the amps sound fairly
good for SS amps, but once one ABs one of them with
a tube amp, the difference is quite evident, especially
how the sound projects.

Good luck

A/B testing. Interesting. Take it to the extreme and you have a
modeling amp. Those are pretty nice, I think, provided that you don't
mind having all these 'controls' you can't actually adjust. And if
you don't mind using a power amp and speakers to match, about twice
the power of the amp you happen to be modeling. And if you don't mind
a millisecond or three of latency from the processing chain. These
things do bother me, that's why I don't have a modeling amp. But if
you're not emulating anything in particular, why A/B? I dunno.

So . . . here's my theory. Call it lame, but here is my theory. My
theory is . . .

that the One and Only Standard Solid State Power Amp Design is
completely unsuitable for guitar. That any decent solid-state guitar
amp cannot use OOSSSPAD.

For some reason, there's just one design that seems to have been
hanging around since the 60s that is used for solid state
amplification. One design - implemented as discretes, or as modular
ICs, it's still the same one. It's a bugger of a thing. At it's best
it can sound beautiful and wonderfully clean (like my Yamaha RX
stereo). At worst, it can be utter crap. And for guitar, I think
it's proven completely unsuitable.

Advantages of OOSSSPAD? You can make it big. You can get distortion
figures too small to measure, so long as you don't clip it, and you
won't clip it if you make it big. Slewing it counts as clipping it,
so you get much more power in the bass than in the treble, but that's
okay right, as most audio has most of the power in the bass? You can
make it with low damping factor, flat frequency response, your only
limits are the dissipation and speed ratings of the semiconductors you
have available. Trouble with OOSSSPAD? Clip it and it just dies.
Hard clip, doesn't recover gracefully either.

OOSSSPAD is a big op-amp. Differential pair, one side used for input,
the other side for feedback. Voltage gain stage with dominant pole
compensation - in other words, an integrator. Op amps are, as Barrie
Gilbert once pointed out, actually integrators rather than gain
blocks. Cold AB power stage, highly efficient. Maybe do dominant
pole compensation here too. And a great big load of feedback thrown
back to the original differential pair. Zobel network on the output
just to make sure the silly thing doesn't oscillate. Need less
distortion? Just use more feedback. Need more gain to get more
feedback? Add more stages. Stages add distortion? Use more
feedback.

For guitar use - I think the ideal solid-state power amp stage uses
local feedback and tries for reasonable, not exceptional, linearity,
with distortion that increases monotonically with signal level. So
not OOSSSPAD, but something simpler in idea. Something hearkening
back to the original two-transformer two-germanium transistors power
amp.

That's my theory. It's not especially original nor is it especially
surprising, but it's what I'm working with for now.

So for me it's not about whether I use Darlington's or not, but it's
about how I use them. Lots of OOSSSPAD amps use Darlingtons. And
even Trarlingtons. So much gain. And it's all sent back to the diffy
pair. I'm trying to keep the gain and feedback in little local
lumps. One transistor can't do a tube's job. But maybe two can.
It's a rough theory. Might work out though.
.

Loading