The YAHA amp
(Yet Another Hybrid Amp)
Versions
All
information here may change without further notice. I strongly recommend to follow the discussions at the HEADWIZE DIY forum .
First published in May 2005.
Corrections
added in July 2006.
Apologies
English is
not my native language and i would like to apologize
in advance for the humbly grammar, spelling, expression etc..
Credits
I like to
thank John Broskie from TubeCad
and Pete Millett for the inspiration i found when
reading their webpages.
Both had
many words in favour for a portable hybrid amp. Now this little project is an
example how this can come into being.
Thanks go
also to the nice people at the HeadWize forum who
always provided good ideas. The force of so many enthusiasts drove me forward.
Special
thanks go to Paul Reid. His tube knowledge is enormous.
Last but
not least: A big THANK YOU to Chu Moy and his companion webmasters and hosters for giving us a home.
Design Goals
Generally
spoken, the YAHA should combine tube sound with the current capability that is
needed to drive low impedance headphones.
-
Should fit into a small
metal box so that it is really luggable
-
Low parts count
-
Common parts,
especially the tube should be readily available
-
Power consumption
should be optimized to allow for small battery packs and acceptable on-time
-
Should give beginners a
chance to survive first tube experiences
-
Should give advanced
builders a chance for experimentation
-
Last, but not least:
Easy to build.
Circuit
“There is
nothing new under the sun”.
Explanation of the circuit
Basically,
this is a grounded cathode voltage amp followed by an opamp
buffer.
For full
understanding ignore the opamp stage and look at the tubed voltage amp.
Just cut
(in memory) the wire between the tube and the opamp
(+) input.
The triode
is fixed-biased by the grid-charge current that flows through the 1meg
resistor.
To explain
this:
The hot cathode always emitts some electrons
that have enough energy to fly up to the grid.
The grid
resistor closes the current path between grid and cathode so that the grid
discharges across the resistor and
following
Ohm´s law builts up a
negative voltage between grid and cathode = ground.
This way of
biasing works for many small signal tubes like 6DJ8, ECC88, 6922 but also for
12AT7, 12AX7 etc..
It works
for pentodes as well, like DAF96, 1T4.
It won’t
work for power tubes like 2A3 or so because the bias voltage is in the range of -0.3..-0.8VDC only.
Choosing the value of the grid resistor:
It´s written in the books. It is that value named “max. grid resistor” and often is combined with the type of bias:
Automatic or fixed. Choose the value for “fixed”.
Here I have to correct myself:
If the book
says “fixed bias” this is for an external negative voltage to be applied to the
grid.
If the book
says “automatic bias” this is for the bias acquired by a cathode resistor.
Most books
don’t say anything about “grid-leakage bias” !!! If
there is a value, use it.
If not: Use
the value for automatic bias and multiply it by 10.
In my
circuit the R1 value should be up to 10 megOhm .
The circuit
works ok for R1 = 1megOhm but one can be sure that there is a small amount of grid
current flowing.
Choosing a plate resistor:
Normally
you would take the value off the data book, too. But in this case it is not
possible because the plate voltage is so low that no curves exist for most
of the
tubes.
Here is a
procedure how to get a value without curve-tracing.
The plate
resistor value is chosen that it will consume 4..6 V
from the 12
series or
want to use other supply voltages.
Always
adjust the plate resistor to consume 1/2 to 1/3 of the
supply voltage.
If you try
pentodes the plate resistor may consume up to 2/3 of the supply voltage.
Start with
a high value, up to 200kOhm, and adjust to lower while continously
measuring the voltage between plate and B+ until it fits.
Now take
the adjustable resistor off the circuit and measure the resistance. You now may
choose the standard value nearest to the result of the measurement.
An
amplified input signal can now be seen at the opamp
(+) input if we “repair” the wire we cut above.
The opamp in the prototype is an NJM4456 dual opamp. It is able to feed as much as 70mA into a 150 Ohm
load.
The NJM is
not as “hot” as modern opamps with enormous high
gain-bandwidth products. So it is not very likely that you will have problems
with noise or oszillation.
You may
also try other opamps like NE5532 or OPA2134 or so.
Make sure that they are “unity gain stable”. That means they will work
flawlessly at a gain of 1.
That´s the
configuration we need. We wire the output to the negative input and feed the
signal into the positive input. From the DC point of view this input sits at
about 1/2
of the supply voltage, according how exactly you have chosen the plate
resistor. In steady-state the output of the opamp
follows the input (with a gain of 1 here)
and
therefore sits at 1/2 of the supply voltage, too.
Side note:
You could
use a standard buffer like the BUF634, too. It is more expensive, provides even
more current and you need two of those.
The result
is the same: The output follows the input and sits at 1/2 B+.
Ok. Here is another correction:
It was
reported that the BUF634 would be outside a feedback loop and therefore is
nearly incontrollable. It may work or not…
A better
way is to put a MOSFET source follower as a buffer behind the opamp.
We do not
like to feed the headphones with DC, so we need to go with an output cap.
Warning:
Never go without output cap, otherwise your headphones (and probably your ears)
will be ruined immediately.
You may use
an electrolytic of 220uF for cans above 32 Ohms, like the PortaPro
or SportaPro. You should use 470uF for lo-Z
headphones.
Some people
put a small film cap in parallel to the electrolytic to have better high
frequency response.
Important: Watch the polarity of the cap. The +
side of the cap goes toward the opamp output !
Essential
are the two resistors after the cap.
The bleeder resistor:
This one,
to ground, will allow the cap to charge at power on.
This helps
to minimize the POP when you connect the headphones.
Side note:
You should
connect / disconnect the headphones always if the amp is up and running.
You won´t notice POPs and this helps
to preserve the sound qualities of the headphones.
The bleeder
resistor should be about 10 times higher than the value of the headphone impedance plus the output resistor value.
The output resistor:
The output
resistor is chosen between 22 and 150 Ohm depending on which final gain the amp
should have.
The output
resistor forms a a voltage
divider together with the headphone impedance.
As a rule
of thumb choose the low value for high-Z headphones (>=250 Ohm) and the high
value for lo-Z headphones (32 Ohm).
Changing
the value of the output resistor has _some_ effect on the sound and the stability
of the opamp stage.
The NJM
likes load impedances of 100 Ohm and higher best.
Feeding the heater:
This is
done by an LM317T voltage regulator wired as constant current source.
You only
need one for two channels because the tube is a double triode, having two
triode systems inside.
Check the
tube data: The 6DJ8 is fired with 300mA constant current or 6.3VDC constant
voltage.
We choose
constant current. This is easy to achieve with the LM317 and an additional
resistor.
The current
flowing is given by the formula: I = 1.25 V / R, or solved for R:
R = 1.25V /
I (desired_heater_current) = 1.25V / 0.3 A = 4.2 Ohm
Construction details
I am the
least talented case builder under the sun so I always prefer readily available
boxes.
This one is
an Altoids “Winter Green” tin. I guess it works with
any of those Altoid tins.
I don´t know if it works with Fisherman´s friends, too. I will test it if someone sends me a
box of Penguin mints (hint, hint).
So here´s the look into the inside:
The tube
socket was a standard bakelite chassis mount type. I
had to “Dremel”-away some material to get it fit into
the box.
All
resistors are standard 1/4 watt metal film types.
The circuit diagram does not show the small
You can
leave it out if you have a volume control with your source.
I use
paralleled resistors to achieve the resistor value for the current source
LM317T. You may use one resistor as well but be sure that it can consume
at least
1 watt. The configuration of my amp is not bullet-proof. Because
I use 1/4 watt resistors in parallel. They get really hot
!
The inside
of the tin is covered with cardboard. This is really necessary because the
LM317T has the output at heatsink and would be
shorted
to ground
if touching bare metal.
The layout
itself is a bit messy, after all. I placed the components according to space
optimization. I would be rather happy if somebody could
provide a
PCB layout.
Addition: DIGI01 from the HeadWize
forum made some. Check for it at HeadWize !
Now here´s a last look:
Building
this amp was fun, but sometimes a drag because of the small enclosure.
Power supply
As you see
12VDC are necessary to power this amp. It consumes a little more than 300mA.
You may use
a batt pack of 10..12 NiCd or NiMh rechargables
giving you about 5 hrs. of continuous listening if you
choose 2000mAh cells.
Or power it
off your car battery if you go for a camping trip.
Use a
wall-wart at home, preferably a regulated and well filtered one.
Try solar
cells if you live in a sunny place.
How does it sound ?
Oh, well.
After you had such a hard job to build this amp it is very unlikely that it
sounds bad.
The sound
is mainly determined by the tube. It works at a very low plate voltage so that
it probably produces a lot of distortion because the curves
are far
from being linear.
There is a
way to measure distortion using a PC-based tool. But I did not try. My PC
soundcard is a lousy onboard chip with ugly noise from the computer
ps. So it really is a task someone else should
perform.
I think the
distortion is of mainly even order, as characteristic for triodes.
I
experienced the sound being very warm, easy to listen to. Not that sterile as
transistor amps and not as aggressive as some opamp
based amps.
I think the
NJM and the 6DJ8 are good companions giving a nice and unfatiguing
listening experience.
This amp
works fine with a large bandwidth of headphones. Maximum usable voltage is 4V(peak) and up to 70mA current which leads to a minimum
load
impedance of 53 Ohm.
This is the
absolute maximum and the amp will distort heavily. But I admit that I did not
listen at that level. My oscilloscope and a dummy load did that for me.
Side note:
A 53 Ohm
load impedance consists of the 22 Ohm output resistor in series with an about
30 Ohm dummy load.
According
to Ohms law this means the dummy only gets about 2 V(peak)
of audio. Imagine this going into your Sennheiser
PX200 and you will know what
LOUD means.
What is next ?
I hope for lots of comments. If the feedback to this circuit is
big enough I will try to convince Chu Moy to make an article for the project
library of HEADWIZE.
My webspace is not big enough to hold more than these few pics and this will not change until my hometown finally is
connected to DSL.
At the
moment there will be no way for me to have step-by-step instructions to answer
all questions.
If my time
allows it I will carefully answer all your questions.
Including
the corrections I made this file was sent to Chu Moy on July, 10th 2006.