Sweep Tube OTL Monoblock © 2004 Alan Kimmel
Categorie: Alan Kimmel
Gepubliceerd: donderdag 01 september 2016 12:00
Geschreven door Peter
Sweep Tube OTL Monoblock
by Alan Kimmel
Copyright © 2004 Alan Kimmel. All Rights Reserved
Special thanks to Scott Nixon, who encouraged me to design and build this OTL.
We know that tube amplifiers with transformer coupled outputs can sound great, typically better than solid-state amplifiers that need no output transformer. But it is also possible to make tube amps with no output xfmr, as the name OTL (Output TranformerLess) indicates. These can sound great too. For some of you who have inefficent speaker systems that you just can't part with, this is a project you can warm up to.
The earliest commercial OTL amplifier was made by the Stephens Manufacturing Company. Their OTL used 2A3 output tubes. The main drawback of that OTL was that it required a special speaker of 500 ohms impedance. Although that OTL was a commendable early step, it cannot compare with OTLs that can drive 8 ohm loads. Driving 8 ohm loads was the real breakthrough in OTL development. Vacuum tubes were never intended to drive such a low impedance directly; theoretically it should not be possible or practical. But a good OTL can do so nonetheless, and do it well. It is these two characteristics of OTLs that distinguish them from all other amplifiers and make them seem almost magic:
1) Their superb sound potential and
2) Getting vacuum tubes to directly drive 8 ohm loads
The next and most famous OTLs were produced by the brilliant Julius Futterman. He manufactured several different models. His OTLs work fine with standard 8 ohm loads. Later, companies like New York Audio Labs and Fourier Components manufactured Futterman OTLs. There will probably always be companies that will carry the Futterman torch.
Currently the most well known OTLs are the Atma-Sphere OTLs. These OTLs are the first commercially produced OTLs to utilize the Wiggins "Circlotron" output topology. An important advantage of the Circlotron is that it allows the output stage to be fully balanced-- both halves of the output stage are identical. This allows direct coupling to the speaker with no DC current flowing through the speaker.
First I had to determine which is the best and most efficient output tube for the task. Julius Futterman's work led me to feel that sweep tubes would be ideal outputs. At the time the sweep tube of choice was type 6LF6. Also at that time the most available triode candidate was type 6AS7/6080. It was no contest-- the 6AS7 could not compete with the power and efficiency of the mighty 6LF6. I suppose I could have gotten comparable results with enough 6AS7s but I didn't want to use a "zillion" output tubes. (Plus, the 6LF6 requires only 2A of heater current.)
Sweep tubes have low voltage screen grids; 150V is adequate for the screens of most sweep tubes. Julius Futterman obtained the greatest efficiency from the 6LF6 when its screen voltage was raised to approach the maximum screen voltage rating of that tube (which is about 270V).
At one point I was curious to see the effect of using the small 6JN6 tubes instead of type 6LF6. I theorized that I would get about 1/3rd the power and, lo and behold, that's what the 6JN6 tubes delivered (along with a higher output Z).
When I built this OTL several years ago, 6LF6 tubes were available. That tube works great in this OTL but now that tube is too expensive and scarce so I converted the OTL to use type 6KG6 output tubes instead. This type is very similar to type 6LF6, the main difference being that the 6KG6 has a large 9 pin "MAGNOVAL" base rather than 12 pin. The 6KG6 is currently manufactured by two companies-- Svetlana and EI. I am using Svetlana's EL509/6KG6, priced at $33 each. (In Russia, Svetlana's 6KG6 is known as type 6P45S.) EI's 6KG6 is distributed by New Sensor as type EL519/6KG6. I haven't yet tried EI's 6KG6 in this OTL but my previous experience with their 6KG6 suggests that it too is an excellent performer.
My next task was to find the best circuit topology. As always, I wanted something innovative-- why duplicate existing designs? I wanted this OTL to have excellent efficiency and performance; balanced output; ability to accept balanced or unbalanced inputs; and above all, uncompromising fidelity. I wanted an amp that would reach out and grab me with you-are-there realism; this OTL does that.
It was clear that the output stage would have to be some kind of cathode follower (CF) circuit to have the best chance of directly driving a typical loudspeaker impedance such as 8 ohms. I admire the pioneering work of Julius Futterman. His OTL output stage topology is the "totem-pole" type but I chose an output stage named after Mr. A.M. Wiggins called the "Wiggins Circlotron" because it met all my preferences for the output stage. [See "New Amplifier has Bridge-Circuit Output" by D.J.Tomcik and A.M.Wiggins, AUDIO, Nov.1954, p.17] However, all the well known OTL output stage topologies can be made to sound great and perform well.
At first I was going to operate the output tubes in standard pentode mode by supplying the screens with a DC supply that follows the cathodes. But this might have required a somewhat higher DC screen voltage to get good efficiency. Instead, I chose to operate the output tubes in a more efficient manner: the pass device for the screen grid supply is modulated with the same drive signal as the control grids of the output tubes. This boosts efficiency further, transforming the output tubes into Super CFs. You can call it the "Augmented Mode" if you want. The Augmented Mode provides very good efficiency with a relatively low 225VDC on the screens. So I chose pentode CFs which, by the way, is what the original Wiggins Circlotron amplifiers used in their output stage. Thus I ended up with a balanced Wiggins Circlotron output stage using sweep tube CFs operating in this "Augmented" mode.
That, plus the fact that the output tubes are idled at about 1/3 their total plate dissipation rating means that they are operated conservatively, yet they work efficiently. Tube life can be extended further by switching the function switch (S-1) to "MUTE": Besides muting the signal, this switch also increases the negative bias voltage to the output tube control grids, reducing the cathode current to a fraction of the operating current. Therefore "MUTE" is actually a "MUTE/STANDBY". I think all amps should have a Mute/Standby switch. A Standby mode for a SE amp would have to function gradually and very smoothly because SE output stages have no common-mode rejection. [The Standby function could be made an "Auto-Standby" as follows: after there's been no signal for a predetermined length of time, a small amplifier circuit would activate the Standby mode. The manual Mute switch could also activate the Standby mode, as it does in this OTL.
The Mute function is convenient for inserting or removing the input cable to the amp while the amp is on. This saves the amp needless stress in being turned off and on each time input cables are removed or inserted (namely to the unbalanced input). Because it is a shunt switch, S-1 is completely out of the signal path except in the MUTE/STANDBY position, so it has no effect whatsoever on the sound. I think all amps should have at leasta Mute switch.
The next task was to get the best front end and driver circuits. I wanted it to be able to accept balanced or unbalanced inputs. The best input stage would therefore be a Mu Stage that can accept both types of inputs. Almost any dual triode that can handle several mA will work for V1. I recommend type 6SN7 (my prototype has a 12AU7 for V1). Though not required, it would be good to operate V1's filament with regulated DC. For the pentode CF atop V1 I chose type 6U8, a triode-pentode (V2). V2's triode is the phase inverter for the push-pull stages. I wanted a phase inverter whose AC balance never changes because this OTL is somewhat sensitive to its internal AC balance. If the internal AC balance is off, it creates a slight DC output offset. When this problem occurred with other phase inverters I tried, this effect was apparently inaudible but it did push a little extra wasted DC current through the output stage and speaker.
The solution was to use the split-load cathodyne (a.k.a. "Concertina") phase inverter. (Incidentally, this was Julius Futterman's favorite phase inverter.) Its AC balance is determined solely by the ratio of the values of its plate and cathode resistors, and not by tube gain, tube aging, or anything else. Another advantage of this arrangement is that even if a "balanced" input signal is not well balanced, the OTL's internal AC balance is not affected.
Some people think the split-load phase inverter suffers a lack of balance at high frequencies. This is simply not true when the two output load impedances are equal at all times. Think about it: if the current flowing through the plate and cathode circuits is precisely the same (which is true in this phase inverter), the performance of both outputs must also be identical. The reason some have measured a slight imbalance at HF is because shunt load capacitances were not quite equal. For detailed defenses of this phase inverter see "Notes on the Cathodyne Phase-Splitter" by Albert Preisman, AUDIO, April 1960, p.22; and "Choosing The Phase Inverter" (Part 1) by Norman H. Crowhurst, RADIO-ELECTRONICS, August 1957, p.49. About NFB in amplifiers: The combination of an output xfmr with NFB can be problematic due to xfmr phase shifts at both frequency extremes. Without an OT, NFB need not be the bogeyman that some believe it is. Someone will say that solid-state amplifiers have NFB and no OT yet they can sound harsh or unpleasant. To that I would say this: The open-loop gain of a typical solid-state amplifier is enormous and this is much of the reason why many solid-state amps sound as they do. All of these negatives are avoided in OTLs; their open loop gain is much less than that of typical solid-state amplifiers. Unlike with other types of amplifiers, OTLs can do well sonically with NFB, as the Futterman OTLs demonstrate.
After the front end I needed a driver stage with adequate open-loop voltage gain plus a large voltage swing capability to deliver to the output stage. I wanted NFB around the output stage itself. The ideal output stage would be a voltage follower so as to minimize output impedance and provide maximum control of the speaker load. This OTL's output stage is actually a balanced pair of buffers. Each is fed by its own diff amp (V3 & V4, respectively). The drive signal is applied to the non-inverting input of each diff amp and 100% NFB is applied to the inverting inputs. The V3/V4 diff amp pair has an additional feature: a modified, so-called "phase compressor" is built into this stage, reducing the output Z of this stage and the overall noise level. (For more on the "phase compressor" see Langford-Smith's Radiotron Designer's Handbook, p.528.)
The B+ for V3B, V4B, V5, & V6 is bootstrapped to approximate a higher B+ voltage so that these stages can provide large voltage swings when called upon to do so. This bootstrapping is accomplished simply by taking the B+ for these stages from the floating supplies.
Next is the pair of CFs, V5 & V6, which drive the output tubes. For maximum output it is necessary to be able to drive the output tubes into grid current. This requires a CF with its output directly coupled to the grids of the output tubes. For this stage you want a pair of triodes that have good current handling ability, plenty of transconductance (gm), and which can tolerate 300V. A pair of 5687s gave excellent performance. A pair of 7119s also gave excellent performance. Actually, a single 5687 (or 7119) is enough. (If you use just one tube here, its filament should be powered by a filament winding other than those of the floating supplies. In this case, either a 6V or 12V winding could be used for this tube.) The plate of this V5/V6 triode CF stage follows its cathode, which is analogous to the screen of a pentode CF following its cathode. This causes a triode CF to perform like a pentode CF.
A footnote about grid current: when grid current will be encountered you don't want large value grid stopper (gs) resistors at the output tube grids, as this could limit their grid current a bit much; for most tubes (except the few so-called "zero-bias" types) I've found 150 ohms to be a good universal value for gs resistors.
is a block diagram of the OTL. V1 and V2 comprise the first block; all the remaining tubes comprise the tandem unity-gain voltage followers.
The bias voltage is derived from both regulated and unregulated sources, yielding a semi- regulated bias voltage. If the bias voltage is 100% regulated, output tube plate current rises and falls as the AC line voltage rises and falls, respectively. If the bias voltage is not regulated at all, the plate current does the opposite-- it falls and rises as the AC line voltage rises and falls, respectively. (One situation or the other probably applies to most tube amps.) I found that a semi-regulated bias voltage keeps the plate current of the output stage fairly constant despite changes in the AC line voltage.
Each monoblock OTL should be built on two chassis-- one for the amplifier circuitry and one for the power supply. This is the best way to build most amps, a technique made famous by classic MacIntosh amps from tube audio's first Golden Age. This two-chassis technique minimizes noise, makes for easier construction and easier handling of the amp. Use connectors to connect the power supply chassis to the amplifier chassis; for safety, use female connectors for the power supply and male connectors for the amplifier. Use a high current connector to supply power to the output stage filaments and plates (i.e., the floating supplies). A lower current connector can feed the driver circuits.
Most of the power supply capacitors should be located in/on the amplifier chassis. The dotted lines on the power supply schematics show this division, i.e., the components to the left of the dotted line should be mounted in/on the power supply chassis while the components to the right of the dotted line should be mounted in/on the amplifier chassis.
The floating supply for the output stage is shown as Figure-3.
Two of these supplies are required per monoblock. Each 130VAC winding is provided by an isolation transformer rated @ 3A (a higher current rating would of course make for a stiffer plate supply and higher power output from the OTL). Or, instead of two separate isolation xfmrs per monoblock I believe single xfmrs exist that have two 130V secondaries; one of these could be used for each monoblock. After rectification and filtering 130VAC yields about 180VDC for the plates of the output tubes. Because the secondaries of the isolation xfmrs track the OTL output in a Circlotron output stage, it would be ideal to use isolation xfmrs that have low capacitance from primary to secondary, i.e., "high isolation" xfmrs. But the low output Z of the amp should enable it to work okay with ordinary isolation xfmrs.
The raw B+ for the two screen supplies is derived from a full-wave voltage doubler which takes its power from, and is referenced to, each corresponding plate supply. The two screen supplies could be powered independently of the plate supplies but I chose a more economical method.
The pass device of each screen supply is attached to large heatsinks, as it must dissipate a lot of power (one such heatsink is visible in the photo of the prototype and is slightly larger than necessary). I chose an IRF710 as the screen supply pass device because of its low input capacitance (Ciss). [Something in a TO-3P or TO-247 package with low Ciss would be ideal.
I recommend one 7A or 8A SLO-BLO fuse to feed AC to all the power supplies in one monoblock. Another small fuse (3/4A to 1A) is recommended to feed the front end power supply.
The primary windings of each xfmr should be "polarized". This is done as follows:
a. Connect a Hi-Z AC voltmeter between Earth Ground and the p.s. chassis.
b. Connect an AC test cord to one of the xfmrs.
Plug in the line cord and note the voltmeter reading.
c. Reverse the polarity of the test cord and repeat step B.
d. One of the two polarities will provide a lower voltage reading.
That is the preferred polarity for that xfmr. Mark the primary leads of that xfmr accordingly.
e. Repeat steps A thru D for each xfmr. While testing the polarity of each xfmr, the other xfmrs must not be connected to the AC line.
Exercise the utmost care and safety precautions while doing this! The benefits gained by this "primary polarization" are twofold:
1. Improved reliability and safety.
2. Improved sound may also result.
This procedure, first shown to me by Tom Tutay years ago, is a good idea for all amplifiers. After wiring the primaries I connected the power supply chassis to the AC cord's earth ground, and placed a 0.47 Megohm resistor from p.s. chassis to p.s. ground.
To be conservative I call this a 150W amplifier. Though I built the prototype monoblock with 8 output tubes you can use 2,4,6,8,10,12, or 14 depending on how much power you want. Both sides of the output stage should have an equal quantity of tubes of course. When using less output tubes the output Z will be somewhat higher.
I recommend placing the output tube sockets about 3 inches from each other and from everything else for proper cooling. MAGNOVAL tube sockets are required for these output tubes. Note that Magnoval pins are larger than the more common "Novar" pins. Chassis mount Magnoval sockets are sold by Svetlana as their part # SK509 for $2.00 each as of this writing. Antique Electronic Supply also carries this socket as part # P-ST9-509. It is 100% ceramic and requires care in mounting to avoid breaking it. I mounted the sockets on top of the chassis with a thin fiber washer between each screw head and the socket. Panhead screws are recommended.
Billington Export Limited of England is another source of Magnoval sockets. Their address is:
BILLINGTON EXPORT LIMITED
Units E1 & E2
Gillmans Industrial Estate
RH14 9EZ, England
About the plate caps, a.k.a. top caps: The top cap of type 6KG6 is 1/4 inch in diameter. This is the smallest size, as used by 6J7 and 6K7 tubes. Svetlana now has a suitable plate cap, their part # PC509. Billington Export Ltd also has a suitable top cap, their part # TC12E1. Another option is to use the end clips from fuse holders. Some people have been known to simply solder the plate lead to the tube's top cap (though this may void the tube warranty). In any event ensure that the top cap connection is well insulated.
shows the front end power supply schematic. Figure-5
shows an optional indicator which glows when the OTL is in the MUTE/STANDBY mode. A flashing LED reminds the user that he switched the OTL to MUTE; otherwise he might come back to the OTL and wonder why he hears nothing. There are many possible ways the flashing LED can be powered; Figure-5
shows two methods.
It is strongly recommended that you place a fast-blow fuse in series with one lead of the speaker. This OTL prototype has been reliable and trouble-free but the speaker fuse provides an added margin of safety.
Another margin of safety is provided by the small 150 ohm 1/4 W film resistors that go to the suppressor grid (grid #3) of each output tube. The suppressor grid resistor's only function is to act as a fuse in the unlikely event of a short from the suppressor grid to the plate of the tube. As a group, sweep tubes were more susceptible to this type of short than other tubes. I call these disaster-prevention resistors "short-stoppers" (ss). Do not substitute higher power or non-film resistors for them. A more elegant solution would be to use self-resetting PTC (positive temperature coefficient) thermistors in place of the ss resistors. Such devices are also called "self-resetting fuses". Though the excellent construction of currently manufactured 6KG6 tubes makes an internal short unlikely, it doesn't hurt to have the ss resistors (or PTC thermistors).
Because there is only one bias adjustment it is necessary to use a matched octet of output tubes per monoblock. If you cannot find a matched octet, the next best thing would be two matched quartets.
Adjustments and Operation
If it ever becomes necessary to change the bias voltage range, change the value of the bias range resistor "Rbr", shown as 62K in Fig.2: a larger value gives a larger bias voltage, and vice-versa. Situations that would require a different bias voltage range are:
1. Using triodes other than type 5687 for V5 & V6.
2. Using different brands of output tubes.
Set the BIAS pot to maximum resistance, and the DC and AC BAL pots to center position. Do not connect a load to the OTL yet. With the function switch turned to UNBAL, set the bias so that an average of about 77 mA flows through each output tube plate, producing about 77 mVDC across each 1 ohm plate resistor. (You can install pin jacks for the 1 ohm resistors. I installed a simple meter circuit in the prototype to monitor the drop across the 1 ohm resistors.)
Next, adjust DC BAL for minimum output offset. Next, connect a dummy load to the OTL. Then feed a 1 kHz sine wave into the OTL giving an output of about 10Vrms and adjust AC BAL for minimum DC output offset. Then re-check BIAS and DC BAL. When the output tubes or V5 and V6 are changed, the BIAS and DC BAL must be re-adjusted.
Caution: When bench testing any OTL, especially with steady-state power output tests, do not run a sine wave or other repetitious waveforms through the OTL any longer than absolutely necessary to minimize wear on the output tubes.
Four ohm loads are not the best load for any OTL. (In fact, 4 ohm loads can be hard on ANY amplifier.) This OTL will drive a 4 ohm load but it's not an ideal match. Driving 8 ohm loads directly is quite an accomplishment for tubes but asking tubes to directly drive 4 ohm loads is pushing your luck.
I recommend soft-starting devices for most amplifiers (including this one), especially for amps that have an expensive output stage.
It is good practice to have the OTL in the MUTE/STANDBY mode while it is warming up.
It is also a good idea to install separate AC lines for power amplifiers, especially if the amplifiers can put out some power.
Audiophiles report that listening to music through this OTL reveals a clean, tight, effortless, and above all, very musical amplifier across the entire sound spectrum.
A Few Specs:
Measurements made from the prototype @ 1kHz and with 8 W load except where shown otherwise.
Performance with 8 output tubes:
Frequency Response (referred to 10W @ 1kHz): 16Hz to 26kHz within 0.3dB Continuous Power output:
230W rms into 16W
210W rms into 8W
140W rms into 4W (Power Supply Limited)
*Note: When I made the maximum power output measurements the plate supplies of the output stage were pulled down quite a bit. The screen supplies (which take their power from the plate supplies) were also pulled down somewhat.
0.08% @ 5W rms
0.30% @ 50W rms
0.35% @ 150W rms
Output Z: ~ 0.5 W
Unbalanced Input: 1.7 Vrms input for full output
Balanced Input: 2 * 0.85Vrms input for full output
Combined total idle dissipation of all the tubes including heaters ~ 240 W. (Even less in Standby Mode.)
Performance with 6 output tubes:
Continuous Power Output ~ 160W rms into 8W
THD: 5W ~ 0.20%
50W ~ 0.37%
100W ~ 0.40%
Z-Out ~ 0.6 W
Performance with 4 output tubes:
Continuous Power Output ~ 115W rms into 8W
THD: 5W ~ 0.20%
50W ~ 0.55%
100W ~ 0.80%
Z-Out ~ 0.8 W
Notice: Private individuals may build this project for their own personal use. If you want to use this proprietary design for commercial purposes contact Alan Kimmel for licensing. Licensing includes consultation, improvements and updates, and more.
Contact information can be obtained using the contact form on this website.
Reprinted with Permission
Issue 8 Fall/Winter
Vacuum Tube Valley
1095 E. Duane Ave., Suite 106
Sunnyvale, CA 94086 USA