by Aren van Waarde
For many years, I have used a solid-state headphone amplifier in my shack. It is an op-amp type of circuit which is built with discrete components (BC559 and BC560 transistors, with a BD139/BD140 output pair). It runs in class A and has no output coupling capacitor. The measured specifications are quite good and it sounds decent after it has warmed up for about 20 minutes. However, I started wondering how a tube circuit might sound. In the stereo of my living room, I now use a tube preamp (Curcio Daniel) which sounds excellent, both on CD and MC phono.
Rudy van Stratum has published a schematic for a tube headphone amplifier in the April and September 1995 issues of the Dutch magazine Audio & Techniek. It was just a circuit idea, without any guarantee that it would actually work. The schematic caught my attention for the following reasons:
- Extreme simplicity (only 2 double triodes required for stereo). This is the simplest tube headphone amplifier that I have ever seen!
- Capable of driving low impedance headphones
- Amplifier stages directly coupled
- No global negative feedback
- Single-ended topology
I decided to give the circuit a try, and after prolonged listening tests (more than 3 months, with both CD and analog tape as signal sources) I can report that it works very well.
THE AMPLIFIER
Figure 1
My own (slightly modified) schematic of the headphone amp is shown in Fig. 1. The first stage uses one-half of a E88CC (6922/6DJ8/ECC88) in a common cathode configuration. This is directly coupled to a cathode follower which employs one-half of a 6AS7G. I have added a volume potentiometer and a grid stopper resistor to the original schematic. The size of the output coupling capacitor was also increased (from 100 to 220 uF), simply because I had this value in stock and also because I intended to use 32 and 60 Ohm headphones. With 60 Ohm headphones, the calculated -3 dB cutoff point is now at 12 Hz, whereas with 32 Ohm phones, it is at 22 Hz.
A prototype which I made on a piece of plywood worked immediately and I really liked it. With good recordings there is a life-like quality to the sound. Voices and instruments are pinpointed on the stage, with lots of musical detail and “air”. Cathode followers have the reputation of “muffled” and “boring” sound, but Rudy’s circuit renders dynamic contrasts very well and it grips your attention. Minor details in recordings become audible. One can hear, for example, the difference between different violoncellos, and the fact that different tracks of a CD are recorded in a slightly different recording venue.The solid-state amp sounds “hard”, somewhat “glassy” and “mechanical” in comparison, with less detail and less precise imaging. This surprised me greatly since the tube amp has an output coupling capacitor which the solid-state amp lacks. Apparently, the absence of global feedback and the simplicity of the tube circuit works wonders for the sound. The single-ended topology of the tube circuit may also result in a spectrum of harmonics which is different from that of the solid-state amp which has a push-pull topology.
Since I am quite happy with the sound, I have now built a final version of the circuit on a matt black aluminum chassis (size 4 x 8 x 1 inches). The amp is hard-wired, without any PCB. For my listening tests, I used Sennheiser HD 465 (60 Ohms) and Panasonic EAH-S30 headphones (32 Ohms). However, I suspect that the tube amp would sound even better when 600 Ohm headphones were used such as the Sennheiser HD 580 or the AKG K240.
THE POWER SUPPLY
Figure 2
Modifications of the power supply have a marked effect on the sound quality of this simple amplifier. For the initial listening tests, the amp was fed from the high-voltage power supply on my workbench. Then I tried it with a regulated solid-state power supply (schematic with 2 x BF459 from a book by R. zur Linde). The regulated power supply was not an improvement (as I had expected) but it caused a sonic degradation! The “magical quality” was gone, the amp began to sound like its solid-state counterpart… Subsequently I wanted to try a power supply with a vacuum tube rectifier, but the EZ81 which I intended to use proved hard to obtain. Finally I settled for an inductor-smoothed power supply with solid-state rectification (Fig.2). This simple circuit sounds very well.
The tube heaters are fed from a DC supply with a LT1084CP regulator (see Fig. 2). Since this IC dissipates about 10 W of power, it is bolted to the aluminum chassis. The rectifier diodes get hot too, and they are mounted at some distance of the chassis, with sufficient ventilation. The 1k variable resistor (P2) is used to adjust the output voltage (6.3 Volts, loaded).
I have not included details for the primary circuit of the transformers in Fig. 2. Please use the appropriate fuses for your transformers and mains voltage. My power supply has a mains switch (which activates the heater power supply) and a standby switch (which brings +150V to the plates after about 30 seconds of heating). The power supply is hard-wired and built on a separate chassis (12 x 6 x 2 inches).
MEASUREMENTS
Since I am a hobbyist, I have limited possibilities for amplifier measurements. Here is the only data I was able to collect:
Frequency response (-1 dB):
< 10 Hz…. > 100 kHz (0.775 V out in both 60 and 600W) (I therefore suspect that the output capacitors have a larger value than the specified 220 mF. NB My sinewave generator runs only from 10 Hz…100 kHz)
Output voltage:
ca. 28 Vtt in 600 Ohms (onset of clipping) = 10 Veff
ca. 3.7 Vtt in 60 Ohms (onset of clipping) = 1.3 Veff
Max. power output:
170 mW in 600 W
28 mW in 60 W
Voltage gain:
8 x (i.e. 100 mV at the input produces 800 mV output at a 600 Ohm load, volume potentiometer at maximum)
Squarewaves (1 kHz, 10 kHz, 20 kHz) look perfect, at low and very high frequencies (smaller than 100 Hz or greater than 50 kHz) one sees the influence of the output coupling capacitor.
I think these specifications are good, but the best measuring instruments are the human ears.
PARTS LIST (AMPLIFIER)
P1 – Potentiometer 100 k logarithmic stereo (ALPS RK-27112)
R1 – 1M ohms, 1 Watt carbon resistor
R2 – 33 ohms, 0.5 Watt metal film resistor
R3 – 47K ohms, 1 Watt carbon resistor
R4 – 820 ohms, 1 Watt carbon resistor
R5 – 4k7 ohms, 5 Watt wire-wound resistor
R6 – 3k3 ohms, 10 Watt wire-wound resistor
R7 – 10k ohms, 0.5 Watt carbon resistor
C1,C2 – 220 uF, 400 V electrolytic capacitor (Nichicon)
C3 – 220 uF, 100 V electrolytic capacitor (Nichicon)
C4 – 0.22 uF, 250 V MKT (DDR stock)
V1 – E88CC (Brimar)
V2 – 6AS7G (RCA)
1 – noval chassis-type tube socket (ceramic, gold-plated contacts)
1 – octal chassis-type tube socket (I used an octal socket for an Omron relay !)
2 – RCA jacks (gold-plated, insulated)
1 – 6.3 mm stereo jack for headphone plug
1 – knob for the volume potentiometer
1 – aluminum cabinet 4 x 8 x 1 inch (black, Monacor, for Eurocard PCB)
1 meter Prefer microphone cable (for wiring)
1 meter wire red
1 meter wire black
Please note: C1, R5 and C2 are shared by both channels.
The shield between the two halves of the E88CC is grounded.
The heater power supply doesn’t float, but is grounded to avoid the pickup of hum.
With shorted inputs, or a low impedance signal source, the amplifier is completely free of hum and noise, even at full volume. In practice, the volume control is never increased more than half-way.
PARTS LIST (POWER SUPPLY):
P2 – 1k trimpot (Piher)
R8,R9 – 6.8 Ohm, 1 Watt carbon resistor
R10,R11 – 180 Ohm, 0.25 Watt metal film resistor
C5,C6 – 22 nF, 1 kV MKT (DDR stock)
C7,C8 – 100 uF, 450V (F & T)
C9 – 1 uF, 250 V MKT (Philips)
C10 – 22000 uF, 25 V (Sprague Powerlytic)
C11 – 10 uF, 63 V (Philips)
C12 – 100 uF, 35 V (Roederstein)
IC1 – LT1084CP (Linear Technology)
D1,D2 – 1N4007
D3..D6 – P600A (50V, 6A)
T1 – 220:2 x 115 V, 30 VA isolation transformer
T2 – 220:9 V, 50 VA transformer
L1 – Inductor 10 H, 90 mA, 270 Ohm (Triad)
c. 1999, 2001 Aren van Waarde.
The author’s website: Aren’s Attic.
Addendum
5/28/01: Mike Fieger built the van Waarde amplifier on two chassis, one for the amp and the other for the power supply for performance and space reasons. He writes: I like to see only the amp part sitting on my table, not a big box. The fan noise is not a problem, since the power supply is far away from my ears. This amp seems to be quite exceptional.
I tested the amp by using function generators, and oscilloscope, and a multimeter. Dr. Aren van Waarde has done a great job in objectively measuring the amp’s performance. Results were quite close to those published in the article. The last 80 percent of volume usually resulted in clipping, easily seen on an oscilloscope, on all frequencies. This was true for what ever load was being tested (30 ohms, 60, 150, 300, 600). On the higher loads, the volume would be so high, it would easily make a person deaf. The higher the load, the higher the ouput power. The amp seems to like high impeadance headphones.
Test Equipment
SONY MDR-CD360 – headphones
SONY Receiver STR-D911 – headphone jack
Sennheiser 580 – headphones
Nakamichi Receiver 2 – headphone jack
Pioneer DEH-P600 – cd player
Kenwood DP-97 – cd player
Heathkit IP-32 – power supply
BK Precision model 2120B – 20Mhz oscilloscope
XR2206 – Function generator
load resistors: 30, 60, 150, 300, 600 ohm
With my cheap (32? ohm) headphones, the sound was quite bad. With these headphones (Sony MDR-CD360) full volume on the amp was extremely distorted, and not that loud. However, I borrowed a pair of Sennheiser HD-580’s, and those made this amp sound really good. Sounds very musical, and even half way on the volume was very loud. This was my second tube amp project, and I’m quite impressed with it. The Sennheiser 580-6AS7G combo sounded beter than the 580’s in the other amp’s headphone jack. Quite better. The sound was clearer, and more open. Not muddy, as in the other amp’s headphone jacks. The 6AS7G is by far the most enjoyable source of music used for listening tests. No buzz was heard when using the described power supply, or when using the Heathkit IP-32 power supply.
I constructed the amp using point to point parts placement. With many resistors in series, parallel, or both, this resulted in quite an ugly mess. I used multiple ground points, and just sort of randomly placed parts. No predetermined parts placement was used (hence the mess). I found no mistakes in the circuit during construction. This amp is very easy to build, should be perfect for the novice (like me).
With some exceptions, all of the resistors and capacitors can be ordered from Digikey. All part numbers used in builing the amp have been included in the list. As seen in the pictures, most of the resistors had to be in parallel, series, and or both. Unless the proper value resistors can be found, building the circuit on a circuit board would be highly recommended. IC1, the LT1084CP voltage regulator proved to be difficult to find, so I substituted a NTE970. No circuit modifications were necessary; the NTE970 just dropped right in. The NTE970 must be mounted on a heatsink, and since it gets very hot, I also used a CPU fan.
Download Mike Fieger’s parts list for the Waarde amplifier
The power supply requires the isolation transformer as described in the article. I did a fairly poor joob of choosing parts, and placing them into the amp. This means that the amp can only sound better with better construction techniques. Of course that means using proper and or better quality components, and good construction practices.
12/17/2001: Ying Mingyao wrote: I have finished my first headphone amp according to the single-ended OTL amplifier from Aren van Waarde. The amp is easy to build and it has a true tube sound and very excellent performance. In order to have a true tube sound, I changed some parts of Waarde’s original design, especially in the power supply.
I use two 6Z4 tube recifiers (a Chinese tube, same as 6X4) and a choke. The tube filament supply of 6AS7G is AC instead of DC – there isn’t any noise also. V1 is either a E88CC or 6DJ8. V2 can be a 6AS7G or 6N5P or 6080. The 6AS7G tube is not easy to find here, so I use the Chinese tube 6N5P or the USA 6080 tube; they all give good results. The 6DJ8 filament supply uses a 7806 regulator. In my amp, the 7806 IC doesn’t get hot. If there is room in the chassis, a small heatsink can be used but without it the IC will also be OK and very safe. The 6AS7G filament uses 2.5A. It is not regulated, because there will more heat and power wasted.
My headphones are the AKG K501 and an Aiwa 36-ohm headphone. They all perform well with this tube amp. I also tried a 16-ohms headphone and the result is good. So I can say this tube amp can really drive low impedance headphone very well.
4/14/2002: Kaj Toivola built this Waarde headphone amplifier with dual mono volume controls for the left and right channels (DACT 100K stepped attenuators). The two switches on the left are for switching the heater supply and the high voltage supply separately. Toivola owns many headphones such as the Grado SR80 and SR325, the Sennheiser HD580/HD600/HD265 and the Koss PortaPro. He mostly uses the Grados and Sennheiser HD600 and says that the amp works better with the HD600 than the Grados.
4/18/2002: Alan Buckbee‘s Waarde amplifier (shown below) has a modified power supply, because the original supply ran too hot. He also supplied a list of US-based sources for the amplifier components. He writes:
The Waarde amp is a great design. It is dead quiet, no hum, etc. Vocals (eg. Diana Krall) have such a rich texture, it is hard to believe that it is a CD! It is as if she is in the same room. Piano is incredible. One can plainly hear the decay in the note just as if one was in the same room with the piano. The air around the music is spellbinding. For headphones I primarily use the Sennheiser HD600.
Mr. Aren van Waarde built his amp using parts primarily obtained in Europe. I built this amp sourcing parts from the USA (see list below). As a result I had to make modifications from the original design to accommodate USA-sourced transformers using standard USA voltage. I built the amp in two separate boxes.
The Headphone Amplifier:
Not many modifications to the original amplifier design except that R6 is a 25 watt, wirewound resistor. R6 gets extremely hot using a 10 watt resistor. So to dissipate the heat better, I went to a 25 watt resistor. For capacitor C4, I have experimented with .22uF 600V “Orange Drop” 716Ps and some Angela Fast Caps. No opinion as to which is better at this time. I also used an Alps Blue dual/stereo 100k potentiometer for the volume control.
Power Supply Section:
This section required that I make many changes. The filament supply (6.3 VDC) line got extremely hot. It seems that both D3..D6 and the IC1 voltage regulator ran way to hot for my liking. The initial T2 transformer (Hammond 167N6) 4 amp supply did not have enough headroom to drive both the 6AS7G and the 6922 without getting extremely hot (the 6AS7G draws about 2.45 amps and the 6922 draws about 55 mA). After 1 or 2 minutes, I could not even touch the voltage regulator or the bridge rectifier.
Within 5 minutes, the transformer was very hot to the touch as well. So I split off the 6AS7G from the 6922 tube, and put the 6AS7G on the T2 transformer. I ran the 6922 off of the 6.3 volt 2 amp supply built in to the Hammond 369AX transformer. The new T2 transformer (Hammond 167S6) is rated at 6.3 volts 10 amps. This change along with using a higher rated bridge rectifier (35A vs. 12A) removed most of the heat from the filament line. The 35 amp bridge rectifier is attached to the chassis to aid as a heat sink. The new T2 transformer now runs at a cool 40 degrees C. I had to add R15 and R16 to get the voltage down from 6.8VDC (based on a line voltage of 124 VAC) to 6. 1 VDC at the socket pin. I ran the 6AS7G with as low as a 118 VAC input (using my variable transformer). At this setting, the filament voltage was 5.76 VDC at the tube pin. The headphone sounded just fine even at this lower voltage. Currently I am running the high voltage line at 150 to 151VDC, the 6922 at 6.01 VDC (measured at the socket pin), and the 6AS7G at 6.027 VDC.
I also added R12 to drop the output voltage down to the 150 to 151VDC range. Without this resistor, the high voltage was running at 158 VDC.
I also added the SW1 and SW2 (and their associated LEDs) to the design. It is preferable for longer tube life to apply the filament voltage (SW1) first (for 20 to 30 seconds) prior to turning on the high voltage line (SW2). When powering down turn off the SW2 first then SW1 switch.
For those who are new to tube designs it is absolutely critical that after transformers T1 and T2 have been attached to the chassis, one place the choke L1 on the chassis and either hook up a set of headphones or a volt meter and experiment with the position of the choke on the chassis. The windings of the transformer will interact with the choke windings, causing very noticeable hum. However, with judicious placement of the choke, it is possible to place the choke on a 12″ x 8″ chassis and get an absolutely dead quiet background.
I ended up placing the choke towards the rear of the power supply and at an angle. This was determined by using DVM hooked up to the choke prior to placing it in the circuit. For this test, the choke is NOT connected to the power supply but is placed on the chassis near the transformers. Any induced voltage from the T1 and T2 will still influence the coil in L1. After turning the supply on, I moved the choke around to find a position of least voltage across the coils. The voltage across the choke is extremely low; however the noise caused by T1 and T2 is very apparent. I then hooked up my headphones to the L1 (in place of the DVM) and noted that the position selected was absolutely dead quiet! This amplifier is totally quiet even at full volume! I spent the few extra dollars to get the enclosed choke Hammond 193B verses the open-frame 158M. An enclosed choke provides better shielding and it looks better (matches T1 and T2).
Optional items were the connectors between the two chasses. One could hard wire them together or use a connector. I chose an AMP series 97 connector (unfortunately, they are expensive) as this makes it easier to move the chasses around and to place the power supply well away from the headphone amplifier. There are a few other connectors that are less costly but most are not designed to carry 150+ volts or 2.5 amps.
On my amp, the cord connecting the power supply is five feet long. This allows me to place the power supply out of sight, if so desired. I have compensated for this length by adjusting the voltages to be about 6.01 VDC at the pin for the 6922, 6.027 VDC for the 6AS7G, and 150 VDC for the high voltage line. These are all based on a 123 VAC input voltage.
If anyone that has built this amp and has some suggestions I will be more than happy to hear from you.
The following parts are the ones I used and sourced in the USA.
Waarde Power Supply Parts List
R14 (size accordingly to match brightness of R13)
| P2 | 1.0K OHM ¾ watt ceramic pot | 3009P-102-ND | Digi-Key |
| R8, R9 | 6.8 Ohm 1 watt 5% metal oxide resistor | 6.8W-1-ND | Digi-Key |
| R10, R11 | 180 ohm 1 watt 5% metal oxide resistor | 180W-1-ND | Digi-Key |
| R12 | 75 ohm 2 watt 5% metal oxide resistor | 75W-1-ND | Digi-Key |
| R13 | 1.2 K ohm 1 watt 5% metal oxide resistor | 1.2KW-1-ND | Digi-Key |
| R15, R16 | .36 ohm 10 watt wirewound 5% resistor | 900-1015 | Radioshack.com |
| C5, C6 | .022 uF 1250 v metal polypropylene cap | P10486-ND | Digi-Key |
| C7, C8 | 100 uF 450 V Elect. Panasonic TSHB cap | P10155-ND | Digi-Key |
| C9 | 1.0 uF 250V metal polypropylene cap | PF2105-ND | Digi-Key |
| C10, C13 | 22000 uF 25V Elect. Panasonic TSHA | P6591-ND | Digi-Key |
| C11, C14 | 10uF 63 V Aluminum Elect. M series | P5189-ND | Digi-Key |
| C12 | 100 uF 35 V Elect . FC series | P10294-ND | Digi-Key |
| IC1 | Adjustable 5A Low dropout Voltage reg. | LT1084CP-ND | Digi-Key |
| D1, D2 | Diodes – 1 A 1000V DO-41 | 1N4007GICT-ND | Digi-Key |
| D3..D6 | Bridge rectifier 12A 50V GBPC | GBPC120005-ND | Digi-Key |
| D&..D10 | Bridge rectifier 35A 100V GBPC | GBPC3501GI-ND | Digi-Key |
| T1 | Transformer Hammond 369AX | Hammond 369AX | Angela Instruments |
| T2 | Transformer Hammond 167S6 | Hammond 167S6 | Angela Instruments |
| L1 | Enclosed D.C. filter choke | Hammond 193B | Angela Instruments |
| Aluminum Box – Hammond | HM267-ND | Digi-Key | |
| Aluminum Box cover – Hammond | HM283-ND | Digi-Key | |
| SW1, SW2 | SPST Switch | 910-4719 | Radio Shack |
| LED1, LED2 | Red LED’s | 276-330 | Radio Shack |
| LED Holders | 276-079 | Radio Shack |
In addition to the above one will need soldier terminal strips, wire, rubber feet, 6-32 screws, self tapping metal screws, IEC male AC connector w/fuse holder, fuse, power cord, and heat shrink tubing.
Waarde Headphone Amplifier Parts List (one channel)
| P1 | Alps Blue dual/stereo 100K pot. | Angela Instruments | |
| R1 | 1.0M ohm 1 watt 5% metal oxide resistor | 1.0MW-1-ND | Digi-Key |
| R2 | 33 ohm 1 watt 5% metal oxide resistor | 33W-1-ND | Digi-Key |
| R3 | 47K ohm 1 watt 5% metal oxide resistor | 47KW-1-ND | Digi-Key |
| R4 | 820 ohm 1 watt 5% metal oxide resistor | 820W-1-ND | Digi-Key |
| R5 | 4.7 K ohm 5 watt 5% silicone resistor | 45F4K7-ND | Digi-Key |
| R6 | 3.3 K ohm 25W 5% wirewound resistor | 900-1359 | Radioshack.com |
| R7 | 10K ohm 1 watt 5% metal oxide resistor | 10KW-1-ND | Digi-Key |
| C1, C2 | 220 uF 400 V Elect. Panasonic TSHB cap | P10145-ND | Digi-Key |
| C3 | 220 uF 100V Elect. Panasonic FC | P10780-ND | Digi-Key |
| C4 | .22 uF 600V orange Drops 716P | Angela Instruments | |
| V1 | 6922-E88CC | T-6922_E88CC | Antique Electronic Supply |
| V2 | 6AS7G – triode, Dual Svetlana | T-6AS7G-Svet | Antique Electronic Supply |
| 9 pin ceramic/gold chassis mount socket | Angela Instruments | ||
| Octal ceramic/gold chassis mount socket | Angela Instruments | ||
| RCA Jacks – Vampire | M1F/OFC | Welbourne Labs | |
| 6.3 mm (1/4″) stereo phone jack | SC1107-ND | Digi-Key | |
| Knob for P1 | 274-0424 | Radio Shack | |
| Aluminum Box – Hammond | HM260-ND | Digi-Key | |
| Aluminum Box cover – Hammond | HM277-ND | Digi-Key | |
| Optional Items: | Amphenol 97 series threaded connectors (see text) | ||
| Connector plug 7 position w/pins | 97-3106A-16S-1P-ND | Digi-Key | |
| Conn. Recept. Box mnt 7 pos w/soc | 97-3102A-16S-1S-ND | Digi-Key | |
| Cable clamp w/bushing | 97-3057-1008-1-ND | Digi-Key |
In addition to the above one will need soldier terminal strips, wire, rubber feet, 6-32 screws, self tapping metal screws, power cable (6 conductor) between boxes, and heat shrink tubing.
9/30/2004: James Deaver (aka vaklov in the forums) built a Waarde amp with a glass chimney over the 6AS7G to convection cool the power tube. The amp is named the “Neva.” He upgraded the 3.3K ohm cathode resistor, R6, to 20W (CADDOCK MP820 film power resistors – chassis mounted), because the original 10W got very hot, very quickly.
The power supply was split into separate low and high voltage sections as separate circuits, so that the amp could run in “standby” mode, or for “burn-in” heating of new tubes before high voltage is applied. Each section has independent, switching, fusing, and indicators. He used Hammond power transformers (model 166P10 @ 10V/5A and model 261G6 @ 250VCT/45VA) to get a B+ of 200VDC, adjustable downward (he prefers an “at plate” reading of about 170 – 175VDC on the 6AS7G).
A 6NO60 octal delay tube (Amperite) is connected in series with the HV primary for a 60-second turn-on delay to allow the tubes to stabilize before turning on the high voltage supply. The LT1084 filament supply regulator was running hot (the two filaments draw 2.8A continuous and the delay tube another 0.5A), so he substituted the LT1083, which is rated at 7.5A.
He writes: I am a disabled vet who chose to build Dr. Aren van Waarde’s OTL amp, and although it took me 2 1/2 years the result is well worth it! I am sufficiently proud of my work that I have devoted a section of my website with pictures of the units and some construction details/mods I would like to share with the other DIYers.
The Neva is named for Stephen and Neva Allen, two dear friends now living in Maine (amplifiers are always female). It uses a glass chimney surrounding a ceramic octal base to convection cool the 6AS7G power tube and the circuit components inside the enclosed chassis (via filtered/screened vent holes in bottom). The concept is VERY old and VERY simple: using a cylinder to surround a heat source to draw air in the bottom of the cylinder and forcibly eject it out the top, creating a Venturi Effect or “heat pump”. The chimney in this amp was a glass hurricane lantern chimney/globe.
About the 6NO60 delay tube in the HV supply, delay tubes were standard in the military. EVERY UNIT WITHOUT EXCEPTION that I worked on while in the U.S.A.F. used plate power delay designs! They were mostly 15- and 30- second delays switched by thermal contact relays in a tube format. Delay tubes VASTLY extend the operational life cycle of power tubes ESPECIALLY when very high plate voltages are used ( >350VDC ). In the realm of audio equipment, I believe they also minimize the annoying “turn-on THUMP” effect.
In the Neva, I connected the filaments of the delay tube with the other circuit filaments, and wired the contact pins in series with the HV transformer’s primary. I recommend “bridging” the contact pins with a small disk capacitor to minimize contact arcing. You occasionally find these on Ebay under “Time Delay Relay” or they are available from Antique Electronic Supply for about $15 US.
The second picture (top view) shows the two one-third arc openings, into the chassis interior (note the wiring for the 6AS7G socket visible in the “upper” of the two). The octal socket is top mounted for better weight loading (probably unnecessary), suspended in middle of the 2 1/2″ opening, and is not visible with tube and chimney in place.
The third picture (bottom view) shows brass screened vents to permit room air to be drawn into the chassis interior by the HOT air (155° F) exhausting from the chimney. Try to position them under the hottest circuit elements, ie. tubes, cathode load resistors, etc.
Here’s the Neva in its “dedicated” hutch. The headphone stand is made from two pieces of “scrap” American Black Walnut left over from fabricating the wood sides for the amplifier module. Add a three inch piece of “freeze proof”, black foam, water pipe insulation my plumber left behind ( with the bill! ). The vibration pad under the amplifier is two 12″ sq. ceramic tiles (surplus from a tileing job), painted flat black, with a 3″ thick block of graphite impregnated, acoustically “dead” foam, left over from unpacking an IBM 6300 Series Winchester disk drive, sandwiched in between. Result: Zero microphonics from the JAN 6922 <:-]
For more construction information, see the James A. & Kathleen C. Deaver Homepage.
HeadWize Memorial 
