Advanced DIYers - I hope you'll enjoy this page. Newbie? Please read
Disclaimer.
Feb 15, 2003: a new page - OTL Projects. First instalment:
*
Two New Complete OTL
Power Amp Projects! *
Some OTL speculations and
calculations
1 Single End CF Class A into 8 Ohms (Vp 75V, Vg1 ~-12V, Vg2 150V, Ik 500mA/EL509)
No. of EL509s/channel: 1 pc 2 pcs 3 pcs 4 pcs 5 pcs 6 pcs
Pdiss tot (HT + heater, 6-38V across CS3): 56W 119W 188W 263W 344W 432W
Pout continuous @ clip: 1W 4W 9W 16W 25W 36W
Efficiency (of o-p stage incl. heater): 1.8% 3.4% 4.8% 6.1% 7.3% 8.3%
Who ever said SE Class A OTLs were efficient?!
Figures are for "generic" EL509/6KG6s. The 40-volt/ 0.3-ampere ‘PL’ and 40KG6 versions draw 0.6W less per valve.
A Tentative Single-Ended Circuit to Try
Each EL509 grid has its own 1uF/100k combo, bias circuit is common. Heaters are in series from GND to B–. Don’t forget grid- and gate-stoppers (not shown for clarity): 100-300R for small valves and 10-100k for ‘509s and MOSFETs (I'd use something like the IRF710 for the screen reg.). For all versions, adjust NFB to exactly 10x/20dB, except 1W amp 11.8x/21.5dB. Sorry, I haven't calculated the feedback voltage divider for any version yet - just keep them in the range of a few k Ohms. The 22k resistor is 2W min., the 1 Ohm 3W. FETs can have 1-1.5 megohm divider resistors. The power supply (see below) must be substantial, but need not be regulated. The current sources/sinks need substantial heat sinks, especially CS3, but I leave the calculations to you (The formula [0C/W spec.]=25/[U*I] gives a theoretical sink temp. of 500C - it's always higher in practice...).
For details of
DC
Servo and Output Protection Circuit, see Headphone Amp
Project. Use a powerful relay. They're NOT optional here, as you don't want
C3 to pull 0.5-3A (that's 4-24V!!!) from GND through your speaker at
start-up or if a failure occurs! Unless the heaters are pre-heated, the speaker will blow, that's almost a promise!
Power Supply (one channel)
Front-end heater PS is not included above - see Tips and Tricks, section 3 for advice. Tip if you use the 40V PL509: strap the heaters in series pairs across the 80V B+ instead. More pairs in parallel. A bigger xfmr is needed, but the B- xfmr can be smaller to compensate. It is a very good idea to start up heaters a couple of minutes before applying B+.
Current sources
See Tips
and Tricks, section 4! Use the cascode variety for plate loads and the
MOSFET for cathodes. For CS3, a MOSFET or an IC voltage regulator hooked up as a
current reg. could be used for the smaller amps. For the bigger, one or more big MOSFETs like IRFP150
are also necessary. Cascoding may be a good idea here, as big MOSFETs have big output
capacitance. Maybe like this? (Note that an extra few B- volts may be necessary for
the 1-9W amps. A series resistor for the heater string may be needed, but this
is a good thing.)
[Perhaps the ideal CS3 current source would be another EL509.
Note that this would have to work at a much higher voltage than the MOSFET, so
the best solution would be to use a bipolar +/- supply with equal rail voltages
(+/- 80-120V or so). The valve-CS3 screen grid could be fed off the positive
supply rail via a dropping resistor, or better, series regulator, and decoupled
to B-. This variant should give far less offset at startup. A nice way of
soft-starting the amp could be to have a pot or special circuit that after a
delay for heating slowly brings up the screen-reg. voltages simultaneously.]
Suggested Configurations
CS1: 10mA CS2: 20mA CS3: 500mA B–: 6.3V
Sensitivity: 0.3Vrms Zout: 2.9 Ohm
CS1: 10mA CS2: 20mA CS3: 1A B–: 12.6V
Sensitivity: 0.42Vrms Zout: 1.7 Ohm
CS1: 10mA CS2: 20mA CS3: 1.5A B–: 18.9V
Sensitivity: 0.55Vrms Zout: 1.1 Ohm
CS1: 20mA CS2: 40mA CS3: 2.0A B–: 25.2V
Sensitivity: 1.0Vrms Zout: 0.85 Ohm
CS1: 20mA CS2: 40mA CS3: 2.5A B–: 31.5V
Sensitivity: 1.2Vrms Zout: 0.7 Ohm
CS1: 20mA CS2: 40mA CS3: 3.0A B–: 37.8V (hint: a 0.2oC/W at 100W heat sink is required for CS3!)
Sensitivity: 1.35Vrms Zout: 0.55 Ohm
*Alternatives for input valve: 6BU6, 7E6, or 6C4; 6C5, 6J5
or 7A4; ½12AU7 or ½7AF7; ½6FQ/CG7, ½6SN7 or ½7N7.
2.1 Push Pull CF Class AB into 8 Ohms (Vp 150V, Vg1 -35V, Vg2 190V, Ik 220mA)
No. of EL509 pairs/channel: 1 pair 2 pairs 3 pairs 4 pairs
Pdiss tot (incl. heater): 92W 184W 276W 368W
Pout Class AB: 12W 49W 110W 196W
[Pout Class A: 0.8W 3.0W 7.0W 12.4W]
Efficiency Class AB: 13% 26.6% 39.8% 53.3%
2.2 Push Pull CF Class AB into 8 Ohms (Vp 110V, Vg1 –33V?, Vg2 205V, Ik 330mA; g2 feed-forward driven, see below)
No. of EL509 pairs/channel: 1 pair 2 pairs 3 pairs
Pdiss tot (incl. heater): 110W 220W 330W
Pout Class AB: 14.5W 57.8W 130W
[Pout Class A: 1.74W 7.0W 15.7W]
Efficiency Class AB: 13.2% 26.3% 39.4%
2.3 Push Pull CF Class AB Triode mode into 8 Ohms (Vp+g2 170V, Vg1 –30V, Ik 230mA)
No. of EL509 pairs/channel: 1 pair 2 pairs 3 pairs
Pdiss tot (incl. heater): 103.5W 207W 310.5W
Pout Class AB: 9.6W 38.5W 86.5W
[Pout Class A: 0.85W 3.4W 7.6W]
Efficiency Class AB: 9.3% 18.6% 27.9%
A Tentative Push-Pull Circuit to Try
This is a so-called Circlotron circuit. The main supplies (B+ no 1 and 2) must be connected as shown and not touch ground anywhere. Similarly, neither speaker terminal must be grounded! Each EL509 grid has its own 1uF/100k combo, bias circuit is common. Don’t forget grid- and gate-stoppers: 100-300R for small valves and 10-100k for ‘509s and MOSFETs (I'd use something like IRF710). For all versions, adjust NFB to exactly 10x/20dB. Keep the feedback voltage divider in the few-hundred-Ohms range. Calculate resistor power dissipation carefully, and double or triple to be safe (22k, 100 Ohm, and 1 Ohm are 3W). The 100 Ohms are just there to keep the amp happy if the series FB resistor is removed for open-loop measuring. The FET voltage divider can have 1 megohm resistors. Current sources as in the above SE amp. The power supply (see below) must be substantial, but need not be regulated.
For a Futterman-type alternative, see Tube CAD Journal, Jan 2000, pp 17-20.
Screen Connection Details
A quieter Triode mode can be accomplished by connecting up as in 2 but
grounding the lower ends of the parallel resistor and capacitor instead of tying
them to the output. Don't forget to heat-sink the FETs!
Other Details
Suggested Configurations
For all: V2 = 5965; B+ no 3 ~300V; CS1 14.5mA, CS2 10mA, CS3 20mA
1 pair Rout: ~2.5 Ohms Sensitivity: ~0.5V B -: min. -25V
2 pairs Rout ~1.2 Ohms Sensitivity: ~0.7V B -: min. -40V
3 pairs Rout ~0.8 Ohms Sensitivity: ~0.9V B -: min. -55V
4 pairs Rout ~0.6 Ohms Sensitivity: ~1.2V B -: min. -65V
Set quiescent for each '509 to I=33/Vp-k maximum. E.g., 33/150V=0.22A (0.22V
across 1-Ohm plate resistor). Less is safer and may prolong valve life. Also
check that screen dissipation is not exceeded (Imax=7/Vg2-k). E.g.,
7/190=36.8mA, i.e. 3.68V max across 100-Ohm screen resistor.
Comment on all the above amp suggestions
Power output estimates above are theoretical;
they may be a bit optimistic, yet shouldn’t be absurdly wrong (as this isn’t a
product advertising campaign – oops, sorry!). Quiescent current is pretty much
"maxed out" in the examples in order to extract as much Class A power as possible
– AB output power will be the about same if you turn quiescent down, but distortion
may increase a bit. It might not be necessary (or even advisable) to set the
current this high in the push-pull amps, especially if multiple parallel PP
pairs are used; idle consumption will be great and so will cost, as one may need to change output valves more
frequently. One might like to make quiescent adjustable and experiment with
different settings to see where it sounds best. A more scientific approach is
to measure for lowest distortion open-loop with an 8 Ohm load resistor as you
tweak the bias. The amps should perform well open-loop but will of course have
very high sensitivity, be noisy, and suffer a bit from uneven frequency
response in this state with most real speakers connected due to the typically
varying speaker load. Always test with a junk
speaker for a good while before connecting up your precious ones!
A note on EL509/519
EL509 and EL519 are virtually identical in electrical
characteristics. The 519 has a slightly higher plate (40W) and screen (9W)
rating than the 509 (35 and 7W, resp.). The PL509/40KG6 is a 40-volt heater
version of the EL509. These are often cheaper, so if you don't mind using a 40V
transformer for the heaters (and why should you?), you might just as well use
the PL version. No sonic differences whatsoever.
Whether you use new or used samples, you could make sure they are OK by pulse testing
them at Vp = 100, Vg2 = 190, Vg1 = -40V. Apply Vg1 = 0V for a second or
so, to get a rough reading, or you may destroy the valve. They should
preferably pass ca 1400 mA. A stiff PSU is needed
for this test, so the best place to test may be your finished amplifier.
There is little point in ordering matched valves, IMHO. The valves may
need several hours of run-in to settle down, so it's a better idea to acquire
more valves than you need, and match by swapping the valves around in the
amp after running it for a bit, until you have a fair balance of cathode
currents.
Addenda
26/4 2002:
I have recently experimented with toroid autotransformers between my speakers
and my various power amps. This is a tapped single-coil choke/transformer hybrid that
"steps up" the speaker's impedance in order to better fit a valve
output stage by improving the power transfer (voltage/ current ratio). This
affords much increased output power or fewer output valves at the expense of a
somewhat greater voltage swing ability (which means that some of the above amp
examples, esp. the SEs, would need higher voltages to realise increased power).
An "autoformer" of this kind has a wider bandwidth and seems to be more
transparent than a
typical OPT transformer in the same price range, probably due to its low conversion
ratio, a single winding, no idle current through the winding (the output must be
at ground potential at idle), and the fact that it is
driven by a potent, low-impedence CF. A somewhat shocking finding was that
my no-NFB transistor amp benefited from the autotransformers as well!
The devices I am using, called ZEROs and devised by Paul Speltz, are
described on this page. They have an
impressive bandwidth of 2Hz-1MHz or 3Hz-2MHz, depending on which taps are used.
[It is
possible that ordinary mains toroids could be used with satisfactory results in this application, but I
haven't investigated this possibility.]
18/5 2002:
I've added a table showing the numerous hookup possibilities of the ZERO
Autoformers to the Valvulations section.
23/7 2002:
My Circlotron OTL plans are now finally coming to a stage where I might begin
thinking of the practical construction work, but it's going to be a while yet!
Watch this space for further info - I might just upload the tentative schematics
later this year.
28/1 2004: It's been a while! The OTLs are now under construction, but things are going a bit slow due to the considerable expense of the parts, and also due to other obligations taking time.
- Morgan L.