Tiny Daddy's FAQ
A Tube Stereo Amp:
Once upon a time a long time ago (1967) I built an amp
using 807's but it was not satisfactory sounding. I can't entirely blame the
807's because I was using a cast-off output transformer and did not have very
good speakers (2-speaker mono). Then I (and a lot of others) bought a
solid-state amp and for the next 20 years (!) thought the problem was with the
needle/cartridge. But my SS power amp started developing problems and I wanted
something different. So in the summer of 2000 I started building a power amp for
those really nifty-looking 3E29's that I got at a thrift store for 25 cents, and
now use Mullard QQV07/40, one in push-pull in each channel. The Power supply is
fully tube regulated, saved from a dumpster. I can now say that the design is
stable and sounds wonderful, even blissful. Even subsonic wow and rumble from
vinyl is reproduced. Square waves look good too! This amp really transmits the
sound.
So I have been using Mullard QQV07/40's and so far they
have proven indestructible. I run them a bit hotter than AB with 500V on the
plate so this set-up needs plenty of ventilation. I mounted the square ceramic
sockets 1/4 inch below the chassis. I tried a simple round hole out to the
mounting screws but the QQV07/40's have a larger base than 829B and fill it like
a cork. So I cut the hole out farther on four sides, making an odd-shaped hole.
I also cut holes in the chassis bottom and in the rack shelf.
Regarding the output transformers. Of course RCA's plate loading spec of 13.7K ohms is not practical.
I used 6.6K Ohms, 60 Watts Hammond transformers #P-T1650P.
The speakers (Altec 604E) are rated "8 to 16 Ohms" and are attached to
the 8-Ohm tap. I could have used smaller output transformers but the larger iron
can't hurt the low end and the larger primary wire may survive if something goes
wrong.
This schematic shows just one channel, a cathodyne phase
splitter driving a push-pull output stage. Sorry no ultra-linear connection
because the output tube screen is only allowed 200V. Referring to the schematic,
a small transformer at the top is connected across the 6.3V filament supply.
It's center-tapped winding provides for injecting the DC filament bias and saves
running an extra wire to the power supply, or using a pair of 100 ohm resistors.
The other transformer winding provides a back-up bias supply that is active
whenever the tubes have filament voltage, and supplies just enough to keep the
tubes within dissipation limits. This is diode-ORed with the well-regulated
higher bias voltage from the power supply.
Cathodyne Phase Splitter:
For cathodyne phase splitter with equal plate and cathode resistors:
As a starting point, let's say you put 80 volts on the
12AU7 grid by direct coupling it to the plate of it's driver stage. Then the
12AU7 will conduct and the cathode will be at a voltage several volts above 80
because that's the way it works. So let's say there is now 90 volts on the
cathode and therefore 90 volts is dropped across the cathode resistor. Now since
the plate resistor is equal to the cathode resistor, and the same current flows
through both, there will be a 90 volt drop across the plate resistor. The
remaining supply voltage is dropped across the 12AU7, plate to cathode. It
doesn't matter to the 12AU7 what the DC bias is on the next stage (for instance
6BQ5 or 6L6) because it is capacitor coupled. But it does matter what resistors
you use for bias feed because they form a voltage divider (through the coupling
cap) with the driving stage plate resistor so some signal is lost. For output
bias feed, always use the largest value allowed by the datasheet. Notice that
cathode bias allows a much larger bias feed (also called grid 1 circuit
resistance in the 6L6 data) than does fixed bias.
The power supply schematics are partial, showing only the regulators. The power supply is much modified and provides 400-600 VDC, 0-200 VDC, 0-105VDC and 0-10VAC. No chokes were needed due to the active regulation. The control circuitry is referenced to negative 300VDC to enable the outputs to be adjusted to zero. The tube op-amp inputs are diode-protected from voltage surges. Plate and Screen supplies are entirely independent.
The meters are switchable, one for current, one for voltage. There is also a small meter for filament voltage. It is important to monitor plate current and screen voltage since the screen is so sensitive.
The system is well-behaved, meaning no turn-on voltage overshoot, no thumps in the speakers. AC power is switched by a relay to avoid stressing the preamp power switch.
The photos show the original 3E29 tubes, since swapped out
for Mullard QQV07/40's
Speaking of which, there
is a series of tubes I don't see discussed much but I find interesting.
Here is a list of output tubes that would work in this amp:
3E29, 829b, 5894, 832A, QQV07/40, QQV07/50, QQV06/40 and
there may be others.
Advantages of the 829B family of tubes:
They have certainly been cheap and available.
The curves (Mazda & RCA) for the 829B show greater linearity than other beam tubes.
I have heard of good results in triode mode also. But in
pentode mode only 200V is used on the screen so forget ultralinear mode.
The cathode structure is massive compared to other tubes of
similar power rating. This would benefit attack time and peak power levels. The
829B family of tubes has high perveyance, meaning the cathode surface area is
large compared to the spacing of the electrodes. High perveyance tubes include
TV damper tubes and horizontal deflection tubes and other tubes intended for
high-current pulse operation. Therefore the 829B requires a lot of filament
power compared to some other tubes of equal or greater plate dissipation
ratings, but then so does EL34/6CA7.
Each tube is a matched pair, manufactured at the same time,
in the same envelope.
The 5894 and QQV06/40 are unique because of the simple
cantilevered plate structure on either side. It is possible to look clear
through the tube and see between the plate and the other electrodes. Other
"see-through" tubes are triodes 6080 and 6J6.
829B Grid 1 Bias Voltage:
The output tubes have a twin bias supply. The power supply
uses a 0B2 regulator tube operating from the -300V supply for negative bias.
There is a 10-turn pot across the 0B2 for adjustment. Also there is a small
& cheap 120/12.6V transformer connected across the 6.3V filament supply.
The120V winding produces ~50V and is rectified for bias and R/C filtered. The
higher voltage regulated supply is diode ORed with this supply. The result is
that the output tubes have unregulated bias whenever they have filament voltage,
and after the regulators warm up the bias automatically switches to the higher
and smoother regulated voltage.
For 829B plate connections I used AMP 66592-2 connector
contacts for .062" diameter pins. 829B pins are slightly small for this
connector so I bent the contact finger for a tighter fit and covered this with
several layers of heat shrink. You may want to use silicon tubing (available at
aquarium shops as air tubing) for better heat resistance.
The 5894 has larger plate pins. You will have to use a larger connector contact such as Molex 02-09-1119 for .093" pins (not gold plated and I haven't tried it).
Alternately it is possible to smash apart one of those
European set-screw terminal strips and use the contacts.
The 829B and QQV series seem very mature products. Also
perhaps more difficult and expensive to manufacture. I remember in 1968 buying 3E29's for 25 cents each at a Union
Mission store because they had been rejected by the Signal Corps due to a
breakage problem. They were loose in a barrel and I bought the only three that
were not broken. Breakage was at the seal between the top and base glass.
Mullard used a different base
829B Screen (Grid 2) Voltage:
The 829B screen grids share a common Pin 3, therefore no
ultra-linear transformer connection
is possible. But the 829B is already more linear than 6L6 or 807 so perhaps
ultralinear operation is not required. Pin 3 must be connected to a
well-regulated supply because screen current varies considerably with signal
amplitude.
Something to remember about the 829B series of tubes: the sensitive screen voltage requirement of 200 Volts which must be closely regulated. Regulation implies generating a higher voltage and then reducing it to the desired value. But if the regulation circuitry fails, the screen voltage may go up and destroy the tubes. It helps to have meters so you can glance at the screen voltage and plate current meters occasionally (same as having gages in your car). If something should go suddenly wrong and destroy the tubes, it's just time for replacement tubes. Of course the power supply has fuses but these would not prevent damage from a partial failure causing slight but continuous over-dissipation. An option would be a solid-state comparator that monitors the screen voltage and/or plate current and shuts the system down. But perhaps that's unnecessary and easier said than done.
Another thing to remember regarding multi-element tubes: If
screen voltage is lost due to regulator failure, loose connection or whatever,
the disconnected screen will tend to float up in potential to the plate voltage.
This is bad and will destroy the tube. To prevent this from happening, install a
1 Meg screen bleeder resistor to ground at each tube socket.
829B Plate Voltage:
Since the 829B operates class AB there is a lot of plate
current variation. I use a regulated supply but an unregulated supply with large
filter caps should work as well. A look at the characteristic curves shows that
the tube becomes more linear at the higher voltages. The curves are nearly
horizontal because the plate current does not vary much with changes in plate
voltage. I have tried voltages from 200V to 600V with good results. I am using a
fully adjustable supply set to 500 Volts. I could use 600V but there seems no
reason to subject the tubes to the extra plate dissipation.
Tube Rectifiers:
From the 5U4 tube data, 5U4 and every tube rectifier has a maximum cap value. Looking at the RCA data sheet for 5U4, RCA recommends 40uF for a capacitor-input filter, with a note that larger caps may be used if the supply impedance is increased. This means limiting the current by adding a resistor (or choke) from the 5U4 cathode to the first capacitor. The reason is that the capacitor only charges at the peak of the AC waveform when the voltage peak exceeds the voltage on the cap, which had until then been discharging into your circuit. This means that the cap is charged by a current spike near the peak of every half-cycle. This current spike is many times the average output current of your supply. And when the circuit is first switched on with the cap fully discharged, the inrush current is greatest. Unlike silicon diodes, vacuum tube rectifiers are not able to handle large current spikes. When the current exceeds a tube's ratings the tube will arc. This is bad. So the tube manufacturer recommends a maximum safe value of capacitor to use.
You may use larger capacitors on the other side of a choke
(RCA says use 10 Henries) or resistor. The larger capacitor is isolated from the
5U4 by the choke or resistor and may be large without causing excessive current
spikes because the choke evens out the current.
Choke input filtering is easier for the power transformer
because the current is averaged. However there can be voltage spikes that may
cause failure of silicon rectifiers.
Another issue: with a capacitor input filter, the first
capacitor will charge to 1.4 times RMS, minus rectifier drop. But with choke
input the second capacitor will charge to only 0.9 times RMS so your B+ may be
more than 100 volts less than with capacitor input. But on initial power-on,
before the tubes warm up and draw current, the voltage from a choke-input filter
may surge to 1.4X the transformer RMS and your filter caps must be rated for the
higher voltage. This is more of a problem when using SS rectifiers.
Power Transformer selection:
If you are using a transformer that you do not have
specifications for, some testing must be performed. But it's not enough to just
connect a resistive load to the secondary because the rectifier/filter draws
current in large pulses at the peak of the AC waveform. I suggest building a
rectifier circuit with 5U4 and 47uF filter cap (may use two 100uF in series with
220K 2 Watt balancing resistors). Connect the transformer and apply a load,
something like 1.5K, 20 watts. Quickly Measure the voltage and switch off or the
1.5K resistor will smoke. Then calculate the current. The transformer voltage
will decrease with load and rectifier drop will increase so you may want to
graph the results at a few different load points. If you get the voltage/current
you want, you will need larger power resistors, or just build the amp and use it
as the load, plus a parallel resistor and adjust the bias so it draws 250mA.Then
run the transformer and use a thermometer to check for overheating or to be
conservative, use the "rule of thumb": if you can hold your thumb on
it then it's not too hot. Depending on the severity of the overload, a
transformer may overheat immediately or take hours. Also no strange burning
smells are allowed.
The life of organic insulation such as varnish, paper and
tape is related to temperature. Cooler is better. The heat is developed in the
core and windings and may be partially due to insufficient turns of wire on the
primary if for example the primary was intended for 115 volts. Some poorly-made
transformers get too hot even with no load. Since the heat is internally
generated a fan has limited effect. You must use a thermometer tell what is really happening. Using your thumb
is unreliable because of the non-linearity: anything hotter than 98.6F feels
really hot. Most transformers need ventilation and cannot be enclosed in an
unventilated box unless the box is filled with a thermally conductive potting.
Generally, larger = better for your transformer.
Did you ever open a faulty unit and find a hole burned
through the circuit board that you could put your finger through? Must have been
a power surge, or flames or explosion, or something, right? Wrong. I first
noticed this happening in the 60's with phenolic circuit boards that used a pair
of edge-mounted pins for the AC connection. The set, either a television or
radio, used a removable power cord that was retained in the back of the case and
shoved onto power pins on the PC board as the case was installed. The power pins
were too close together. Since then I have seen similar damage to boards in AC
surge protectors, various stereo amplifier boards and battery pack boards. Once
a conductive path forms, due to dust or whatever, current continues to build
until the board material begins to char and burn away. The burning is usually
slow, taking days or weeks. I saw the way the battery pack boards were burning:
there was a blackened area that had become 1/2 inch in diameter before anyone
noticed something was wrong. The batteries (14.4V) were discharging through this
blackened area, causing a small bright pinpoint of light that was slowly but
continuously on the move through the charred fiberglass. This process continued
24 hours a day until a hole was formed as the fiberglass was eaten away.
Capacitor outside foil:
Cylindrical and oval film caps are wound such that one foil ends up on the outside. This was more of a consideration decades ago when the capacitors were huge by today's standards and it was necessary to ground the outside foil to reduce unwanted coupling of signals into adjacent circuitry in crowded radio receiver chassis.
If you are interested in using the grounded outer foil as a shield, or in connecting the outside foil to the lower impedance (tube plate): Use a scope or an audio signal tracer, or even a cable connected to a sensitive amp input. Apply an audio tone through the cap to ground. Bring your probe near the cap. You will detect much less signal from the cap when the outer foil is grounded. Use a marking pen to mark the "outside foil" end of the capacitor.