Stewart Pinkerton wrote:
On Sun, 13 Mar 2005 14:01:19 GMT, Patrick Turner
wrote:
At the end of the day, what "score" of music sounds best is the
final decider, surely.
Indeed, although if as claimed 'the first watt is the most important',
I'll lay odds that the various options are sonically indistinguishable
up to that limit.
I have just had a geek at Pinky's provisional SS amp.........
OK, we have a complementary emitter follower pair of class A bjts presumably
biased with a couple of amps idle
current from an RC smoothed supply, so they'll run in class A.
1 amp quiescent current, giving a maximum output into 4 ohms of 8
watts. The voltage limitation also dictates a maximum of about 8 watts
into 8 ohms.
Let us suppose the current gain was 30 between the output current and the base
inputs. Therefore if there is 1 amp rms in the output current load, threre will have to
be 0.033 amps base input current.
It's typically about 200 for the MJL4281A/MJL4302A devices chosen for
this design, hence about 5mA base current.
Then if that is the case, and the device numbers were not
shown on the schematic I got, then the base current will have
6 times less effect than I predicted.
Now the Gm of the output bjts would be about the usual 30A/V,
It is in fact something like 150 A/V at 1 amp for these devices, so
far as can be seen from the Vbe/Ic curves.
This means the open loop gain would be many times greater than what I had guessed.
and open loop gain would be approx Gm x RL, so if the load is 8 ohms,
each output bjt sees approximately 16 ohms, depending on the current sharing
between the two
bjts.
Therefore open loop voltage gain of the output bjts is 30 x 16, or 480.
Actually, about 2500.
But because Pinky has a EF output stage, closed loop voltage gain is 480 / (1 +
480 x beta),
or very close to 1.0.
So there is a gain reduction due to NFB = approximately 1/480 times
So providing the input voltage can be kept linear, thd should also be reduced
from say 10% for common emitter connection to
0.021% for emitter follower connection, and at 8 watts.
I'd say closer to 0.0015%, but let's just agree that it's negligible.
If your gain figures are correct, then indeed the amount of
applied series voltage FB is far greater than I said.
So the amount of series voltage NFB applied in the output stage is 20 log 480,
so about 54 dB of NFB is applied locally in the output stage.
Its not uncommon for the emitter follower connection to automatically
give between 30 and 60 db of series voltage NFB.
We can neglect the 0.22 ohm emitter bias stabilizing resistors for the present.
In this case, about 70 dB of NFB is applied in the emitter follower
output stage.
You can? Without loop NFB? Intriguing..................
The two 0.22 ohm resistors parallel to 0.11 ohms, and these are outside
you local eitter follower FB loop.
So the Rout will at least be 0.11 ohms....
I have neglected the 0.22 ohms because I wanted the analysis to be
simple.
For an 8 volt swing on the output, there is just slightly more than 8v
applied to the output bases, and a 0.033 amp swing,
so the Rin of the output stage is 8 / 0.033 = 242 ohms.
More like 8/0.005 = 1600 ohms.
One has to remember that with BJT output stages, the Rin to an output stage
varies in proportion
to the output load, so that if 4 ohms were connected, Iin would be 0.066 amps,
if the
voltage swing was the same,
and Rin would be 8 / 0.066 = 121 ohms.
Actually, about 800 ohms.
Now the driver transistor will have a much higher Rc than the load, so
the only thing that tends to keep the drive voltage constant is the R5, 150
ohms.
Correct.
Now the open loop gain of the driver stage also = Gm x RL, since it is a
*current*
source, and RL with 8 ohms is 150 // 242 = 93 ohms, and with 4 ohms its 67 ohms
Actually, about 150//1600 = 137 ohms, and about 126 ohms for a 4 ohm
load.
So the gain of the driver with 8 ohms will be nearly 3 dB greater than
when 4 ohms is connected, and a rough calculation
makes me think the output resistance of the amp will be 6 ohms, due to the huge
variation
in driver gain due to reflected load variation..
No, in fact it will be more like 0.72dB, and the Zo of the amp will be
less than 1 ohm.
This is a perfectly lousy result, and well below what the 300B is capable of.
Bull****, since your assumptions are incorrect. Check your facts.
Without being told what transistors you are using, I was left to assume whatever
I cared to. When posting schematics, be prepared to be made to look a fool
if you are sloppy with your presentation.
But from what you have now suggested about the chosen transistors,
you are indeed correct, I have no argument.
Add about 0.11 ohms to the Rout you got.
So Rout would still be just less than 1 ohm, and is acceptable
imho, but does not the 70 dB of series voltage FB
render itself rather ineffective because of the "current reflection"
to the input stages?
Also, if any load caused distortion currents are generated, they too
escape the voltage emitter follower FB clean up, because they flow through to
the bases,
and appear in the input wave form largely unsupressed by any NFB
action.
But let us consider the driver stage, a lone transistor, presumably
some sort of T0220 driver type of bjt with a current gain of maybe a 100,
and Gm = 20 A/V, typical of what is available.
It's already been specified as a MJL4281A, with typical current gain
of 200 at the quiescent operating point of 100mA, and even higher Gm
than at 1 amp.
Since its collector load is 93 ohms when 8 ohms is at the output,
then its gain with the 24 ohm emitter load will be about 93/24 = about 4,
so 2 vrms is about what's needed to produce 8vrms at the collector
to drive the output bases..
Nope, your assumptions are incorrect, and it has a voltage gain of
pretty close to 6.
With your full information, the driver has a gain of
about = collector load divided by emitter load, so 6 is about right.
snip of lots of handwaving and false assumptions
BJTs have current linearity, but make poor voltage amps unless
huge amounts of current NFB are used, and as I have explained above,
Pinky needs both lots of current AND voltage NFB to make his 4 stage
amp work far better.
He really needs to have a a pair of darlington connected N and P
devices for the class A output, then perhaps another darlington pair connected
gain stage fro the input. That'd be 6 devices.
Nope, it will work just fine as designed, for the reasons given above.
Get your head out of RDH4 and look at what modern multiple-emitter
poer BJTs can do.
Get you head into RDH4, and see what tubes can do.
Now Anrdre's amp uses a lone 300B to do the business, and as I have explained to
all
before, there is local electrostatic NFB acting at all AF within the tube.
Its about as local and inimtate an application of NFB as one could ever wish
for,
and it was put there by the God Of Triodes when the tube was made,
and lets not argue with a God in charge of so much high voltage.
**** 'Im, if for some deranged reason you want a *real* SET amp, use
an 845 or 211, and play with proper Frankensteinian voltages! :-)
I prefer a 13E1, or perhaps a QE08, far less common, but such tetrodes
make excellent triodes, and run at normal 300B type
voltages.
a single 13E1 can make 25 watts in UL mode, Ea only 375v.
The mosfet SE amp with OPT will be able to easily make 8 watts,
and with only 23 db of series voltage NFB it will have a very
good DF, and low thd, and measure better than the 300B amp.
The Nelson Pass Zen amp pushes things harder by
making a single mosfet dissipate 50 watts from which 17 audio watts are
available,
and then another mosfet is a CCS also dissipating 50 watts,
but producing no power.
There is 12 dB of shunt voltage NFB applied, and the result
in terms of Ro and thd about resembles the triode amp
with no NFB.
A friend got nothing but smoky silence from his Zen attempt,
and after 3 expensive mosfet changes he gave up.
There's no substitute for skill! :-)
He isn't use to have to praying to the GOT
before farnarkling with voltages.
If properly built, the Zen does what it says on the tin, and the magic
blue smoke stays inside the tin.
I don't like the Zen.
Expecting any T03 device to safely dissipate 50 watts is total BS imho.
Agreed, without heroic and costly cooling arrangements.
Its like saying an EL34 is happy dissipating 32 watts.
It will, but sooner rather than later it will fail.
In my AM radio I use in the kitchen, I have an SE EL34 in triode
with a paralleled 12AX7 driver tube and 12 db of global NFB.
I have a sensitive speaker, and about all I use is around
1/2 a watt max average power.
The sound is utterly blameless.
Aside from the speaker, of course.......... :-)
The speaker measures and sounds very well.
I use microscopic amounts of power most days.
I don't expect the radio to reproduce the effect of standing in the middle
of the LSO at full bore.
The rule for simple *low power* power amps is that
the average levels should be 1/10 the maximum levels for clipping,
so if an amp clips at 5 watts, then only expect a nice average 0.5 watts and no
more.
Except of course properly conceived amps like the 1969 Linsley Hood
10-watter, which is just fine all the way from half-power downwards.
A williamson also is fine on the same basis.
And that basis is for a sine wave to R load.
And evan at full 16 watts, the W makes only 0.1% thd.
Average power for music is quite different.
1/10 full power means the output voltage is about
1/3 the full clipping voltage.
Much pop music from radio stations has an average voltage level
only 6 db below the maximum voltage level transmitted.
If you monitor a radio stations's signal just have a dummy load and take the music up to
where it
begins to clip as shown on a CRO with a signal which is rap, bebop, rock or whatever,
and you will measure an RMS voltage about 1/2 the sine wave voltage for full power and
clipping.
But one has to allow for much more headroom for classical or un compressed music,
and a 10 dB voltage headroom is a minimum; more headroom is better.
0.5 watts into a speaker with 95 db efficiency with 1 watt means you
get 92 dB, and that's way too loud for me, so
0.05 watts for 82 db is about all I need.
5 watts allows a 102 db ceiling, quite enough in the kitchen.
Shame that there are *very* few decent 95dB/w/m speakers, and they
tend to cost many thoiusands of dollars.
Its easier to use 4 tubes instead of 1 and use 89 db speakers than be
compelled to use 95 dB speakers with one tube imho,
unless the tube is a biggun.
Andre's requirement for 3? watts is a bit low, but he intends for horns, so its
OK.
But a simple amp is still simple if it merely means that to quadruple the power
we just use 4 parallel output devices and use a more current able driver device,
or use a much larger output tube.
AFAIK, there are not many cheap easily available gigantic transistors or mosfets
rated
for safe continuous dissipation of 200 watts.
As you point out, not a problem if you just parallel them up.
Alternatively, grow a brain and build a 100 watt Class AB BJT design
with negligible distortion at *any* level up to 3/4 of full power.
Yes but I have customers who have such amps, and they still prefer the tubes.
They sometimes go back to the SS just for an hour every few months.
Last month one told me that when he tried this for a couple of hours,
he thought the musicians were all playing separately.
Then when he goes back to the tubes he says they all sounded like they were together.
I don't argue with them.
I build amps that Andre and many others don't like.
Too much power, vanishingly low thd, even the SE35 amps with a quad of
6CA7 make 10 watts into any load between 3 and 12 ohms at less than 0.1% thd.
the ceiling is 35 watts.
Many other SE amps are up there with well over 1%.
I have built/restored 3 watt amps, and tried to use test them on my
89 dB speakers, and they just don't have the power.
I have a 2A3 project under way for someone who does have
some 95 db speakers. He does not use much volume.
Most people with excellent and undamaged hearing don't need
high volume and huge powers.
But if only 1/3 of a watt is the maximum average power, then a
300B is very hard to beat, and the NFB within it is sufficient.
A colleage of mine used 8 x 300B for 80 watts max
for a cane farmer's amp ( 16 tubes for two channels ).
At a watt there was not a lot of measurable problems,
and the sound was OK.
You could ask why he didn't buy an 80 watt SS amp
for 1/10 of the price and weight, but some folks just ain't interested.
They could always buy a cheap car, cheap shoes, cheap house,
marry a cheap lady, but they ain't interested.
I'd still wanna build an SE mosfet amp rather than yours.
You snipped out the juicy bits from my post, so
it wasn't worth your while to comment.
Patrick Turner.
--
Stewart Pinkerton | Music is Art - Audio is Engineering
|