Overview

The idea for this pre-amp started with the desire to experiment with directly heated triodes (DHT). DHTs are appealing because of their incredible sound, simplicity, and they all were pretty much made before 1940! After some research I discovered several so called battery tubes that were used in portable radios. More specifically the 01-A, 30, 31, and 112-A tubes. All these tubes are low gain (amplification factors from ~4 to 9), which are perfect for my system. There is nothing wrong with using the full range of the volume control! Also, they all have low plate resistances, which means no output transformers or buffer stages are needed if you only need to drive high input impedance power amps (input impedance > 100K), but could be used if wanted or needed. Thomas Mayer has excellent reviews of these tubes at VinylSavor. I wanted to keep the circuit simple but allow for the use of different tubes to allow for experimentation. In order to accomplish my goal, I needed to be able to change the plate current, grid to cathode voltage, and the filament voltage/current to accommodate the different operating conditions of the various tubes. I also wanted only one tube per channel, so I opted to use a single cathode bias gain stage per channel since it works well and is simple to implement. For more information, read this.

Filament Circuit

Since I needed the ability to adjust the filament operating conditions, I had to design a circuit that would be adjustable, simple, quiet, and reliable. Also, all of the triodes I want to use in the pre-amp were designed only for DC voltage; therefore, I opted for an adjustable regulated DC filament power supply. For the circuit design there was no real need to reinvent the wheel since Pete Millet already has designed a regulated DC filament supply design. I use his regulators in other projects and they perform very well. However, I have several projects in the works that will require a regulator with a low drop-out voltage, so I decided to replace the 3-pin LD1085 regulator with the high current, low dropout 5-pin Micrel MIC29152WT regulator (TO-220 package) for the filament power supply circuit. Other than the regulator the circuits are very similar. This regulator has a dropout voltage of only 220 mV at 750 mA of current and the output noise voltage is specified at 260 uV RMS with a 33 uF output capacitor. Ripple reduction is on the order of ~60dB for the 31 tube and ~68dB for the 112-A tube, which translates to around 2mV. It is rated at 1.5 amps maximum current. In the future, I think I will also try the 5-pin Linear Technology (Analog Devices) LT1963A regulator, which shares the same pinout and features but has a lower output noise voltage at 40 uV RMS with a 10 uF output capacitor. It also has about 60dB of ripple reduction. The low dropout voltage is useful because you can reduce power dissipation by reducing the input-output voltage differential. However, the filament current for the aforementioned tubes ranges from 0.06 amps to 0.25 amps and the filament voltage is either 2.0 volts or 5.0 volts; therefore, power dissipation will be pretty low (<1 watt) if you use a transformer rated with a 6.3 volt secondary, which are commonly available. In other words, you can use the LD1085 regulator for this amp and it will work fine. As far as noise or sound quality, I can not hear a difference between the two regulators. For this amp I used the Hammond 166G6 transformers with a 6.3 volt secondary rated at 0.6 amps. You need a separate regulator and transformer for each DHT. See schematic below.

Here is a picture of the filament regulator in the pre-amp:

Here is the adjustable regulated DC filament power supply schematic. Note: the 500 ohm trimmer resistor and the 121 ohm resistor positions should be reversed. This is because the reference voltage of 1.24V is between the adjust and ground pins. With a 121 ohm resistor there is a fixed load current of 10 mA (1.24V/121 = 10mA). With the current schematic, the load current varies from 3.4mA to 17mA, depending on the filament voltage. The minimum load current specified in the datasheet is 5mA for up to 1.5A. It works fine since the current draw is small for the various tubes.

Constant Current Source and Adjustable Cathode-Grid Voltage Circuits

There is a lot of information on why constant current sources (CCS) are a good way to go for loading the plate of a triode, so I will not go into detail of the theory on how they work and why they perform so well. Read more here. Once again, no reason to reinvent the wheel since there are many great CCS circuits that have been developed. An excellent reference for CCS circuits can be found in the book Valve Amplifiers by Morgan Jones. I opted to use a cascode CCS circuit similar to Bottlehead's C4S design, but I added a trimmer resistor that allows for adjustment of the current through the CCS, as well as a proper heatsink, and reduced the current flow through the LEDs. The 63.4 ohm resistor in series with the trimmer resistor is there just so there is an upper limit to how much current can flow through the CCS. The current is adjustable from around 2.6 mA to 14.8 mA by varying the resistance with the trimmer resistor (the lower the resistance the greater the current). The nice thing about using a trimmer resistor is that you can just dial in the current. The MJE5731 gets a heatsink since it drops most of the voltage, which is from ~50 to 60 volts and can dissipate up to 0.5 watts of power. One thing to consider is the power rating for the LR8 voltage regulator in the TO-92 package , which is only 0.74 watts. I like to keep the power dissipation below 0.4 watts just to keep the regulator well within its operating parameters (see below for suggested operating points for triodes). Be sure to factor in the reference currents used by the LR8 (1mA) and the CCS circuit (0.5mA). More on the LR8 regulator under the Power Supply section below. Companies such as Bottlehead offer highly regarded CCS kits for purchase. A CCS circuit allows one to easily change the operating conditions of a triode and reduces distortion; therefore, it made sense to implement it in the design of this pre-amp. Another benefit of a CCS over an anode load resistor in a cathode bias gain stage is a ~10-fold improvement in PSSR (power supply rejection ratio). See schematic below.

Again, since I wanted to use different DHT triodes, I also needed to adjust the grid to cathode voltage to accommodate the different operating conditions of each of the triodes. Once again, there are circuits that have been developed that work well and are easy to implement. Texas Instruments documents a circuit using the TL431 regulator that I used for this pre-amp here. The TL431 adjustable shunt regulator works very well and you can adjust the bias of the cathode by simply replacing one of the programming resistors with a trimmer resistor. The only concern is that you can not set the voltage below 2.5 V, which is the reference voltage of the TL431. The range of adjustment is ~2.5 V to 15 V. Since the circuit uses the TL431 in the TO-92 package, I like to keep power dissipation less than 0.25 W, which is not an issue for the suggested operating points of the triodes. Noise output from the TL431 device itself is ~10uV, so very low ( I could not measure the noise output of the TL431 circuit). The TL431 has a low dynamic impedance, so no bypass capacitor is needed to realize full gain. As far as noise or sound quality goes, I can not hear any difference between a resistor bypassed with a capacitor and the TL431 regulator. See schematic below.

The other benefit of using the TL431 is that by changing the cathode voltage, the plate voltage changes as well. This makes sense since the operating point of the tube is being changed. In other words, the plate voltage is stabilized with the TL431 circuit.

The output capacitor is 0.22uF and gives optimal frequency response down ~30Hz with a 220K input impedance. I might try a larger value in the future but for now the amp sounds great. Any high quality capacitor should work well.

Here is a picture of the PCB with the CCS and TL431 circuits:

Here is the amplifier schematic for the pre-amp (only one channel shown):

Operating Points for Triodes

Here are the operating points that I have found to work well in this pre-amp:

31: 2.0 V filament voltage; 5.5 mA plate current; 80 V plate voltage; 10.5 V cathode; 0.2W of power dissipated by the LR8

31: 2.0 V filament voltage; 7.0 mA plate current; 100 V plate voltage; 14.5 V cathode; 0.25W of power dissipated by the LR8

I can not hear any difference in background noise or sound quality between the two operating points for the 31 in my system. Not surprising since DHTs are known for their linearity.

01-A: 5.0 V filament voltage; 3.0 mA plate current; 90 V plate voltage; 4.0 V cathode; 0.13W of power dissipated by the LR8

112-A: 5.0V filament voltage; 8.0mA plate current; 100V plate voltage; 5.0V cathode; 0.33W of power dissipated by the LR8

So far, these are the only triodes I have used in the pre-amp. All sound good, but the 01-A has a "lush" sound, which I mostly prefer. The 112-A is very detailed with outstanding bass, but a bit bright. The 31 gives a very neutral musical presentation relative to the other triodes. The 31 has the lowest gain of the bunch at ~3.8. The 112-A has a gain of almost ~8.5 and the 01-A having a gain ~8. I like the 01-A the most followed by the 112-A. At some point, I will give some type 30 tubes a try.

Power Supply

The power supply is a fairly straight forward two-stage RC (resistor-capacitor) filter design followed by a voltage regulator. I opted to use the GlassWare Audio Design PS-7 Power Supply for the first stage since it perfectly met the design needs of the power supply. I have used GlassWare products in the past and they are well designed, perform well, and are reliable. The second stage is a RC filter followed by a LR8 voltage regulator for each channel, both of which are mounted on the same PCB as the CCS and TL431 circuits. I highly recommend you read Matt Renaud's write-up on power supply design for vacuum tube amplifiers. I calculated the ripple voltage input to the first capacitor to be ~ 0.6 V peak to peak (based on ~15 mA of total current draw through the power supply). This will vary depending on the type of tube as well as the operating point of the tube. However, it should not vary that much, maybe ~0.35 V peak to peak to ~0.75 V peak to peak since the total current draw for the pre-amp is fairly low regardless of which tube you use. The ripple reduction factor for the first RC stage is ~17, so the calculated ripple coming from the common power supply to each channel is 0.6 V /17 or 0.035 V peak to peak (measured at 32mV peak to peak). The ripple reduction factor for the second RC stage is also ~17, so the calculated ripple input to the LR8 regulator for each channel is 0.035 V/17 or 0.002 V (2 mV) peak to peak (measured at 1.5mV peak to peak). The LR8 regulator provides another 60 dB of ripple rejection per the datasheet, so the output ripple voltage into the CCS circuit is calculated at 2 mV/1000 or 0.002 mV (2 uV) peak to peak (there is no ripple visible on my oscilloscope). However, there is ~2mV broadband noise. A 4.7uF capacitor connected from the voltage divider to ground knocks this down to <1mV peak to peak, which is a modification I am going to make to the circuit. Factoring in the PSRR of the triode gain stage (~40 dB), the final ripple voltage appearing in the output is ~0.02 uV! This is an extremely quiet power supply which is perfect for a pre-amp. I opted to use the LR8 regulator for this pre-amp because of the favorable results that Bruce Heran of Oddwatt Audio has reported for his projects that use this regulator. Read more here. The power supply design works well in this pre-amp as evident from just how quiet the pre-amp is. There is some "white noise" or tube rush that can be heard if you press your ear right up against the speakers, but no hum or other types of noise that intrude on the listening experience. One other thing, you might be wondering how the output voltage is 180V coming from a 120V secondary? The calculated no load voltage is ~170V. The Hammond 262 series transformers are rated with a 115V primary and since my wall voltage is anywhere around 120 -122V, the actual secondary voltage is higher, resulting in a higher rectified voltage. Additionally, I am sure the regulation on the transformer is probably around ~15%. The power supply schematic is below.

Here is a picture of the PS-7 Power Supply in the pre-amp:

Here is the power supply schematic:

Here is a overview picture of the inside of the pre-amp. The stepped attenuator volume control and the selector switch are made by Goldpoint Level Controls. The phono (RCA) jacks are Neutrik. The pre-amp also uses a GlassWare Audio Design House Ground, which provides for some isolation between the signal and chassis ground, which reduces hum. The coupling capacitors are Russian made K42Y-2 500 V paper in oil types. I am also experimenting with some shielding, but the jury is still out on whether it is necessary since the amp is pretty darn quiet without the shielding. The AC power entry module is made by Schaffner (FN261-2-06).

Here is the pre-amp in action with some Tung-Sol 31 tubes. The 17" x 10" walnut chassis is made by Hammond and is readily available. The brass rings on the tubes are made by Mapleshade and are the Standard Grounded Brass Tube Halos. Without them, there is audible hum from the speakers (I measured a 33mV 60Hz peak to peak signal at the plate). With them, there is no audible hum (output noise is ~3mV peak to peak). With the 112-A tubes the hum is audible during quiet passages. My guess is that the tubes are picking up the mains frequency from the power transformer even though it is enclosed in a Hammond aluminum instrument enclosure. Actually, this is not surprising since aluminum does not do a good job at attenuating magnetic fields, especially with low frequency signals. Steel is much better. The 31, 112-A, and 01-A tubes all seem to pick up hum if they are not grounded. I think I will reconfigure the chassis layout so that the tubes/transformers are farther apart or orientated differently to eliminate the 60Hz hum and use a steel housing. This is the best pre-amp I have made so far and I love the fact that I can change tube types with relative ease. It is quiet and it sounds great! It just needs some fine tuning. Stay tuned for the Mark II version of the DHT pre-amp!

Update

I finally finished the Mark II version of the DHT pre-amp. The new amp essentially uses the same circuit but with the following changes (see photos below):

Updated Filament Power Supply. The new PCB accepts the 5-pin Linear Technology (Analog Devices) LT1963A regulator, the new Triad Magnetics CMF Series dual-mode chokes (they provide both common-mode noise suppression, and their stray inductance is effective in suppressing differential mode noise - see the datasheet here), has larger capacitance on the output side of the regulator, and a LED indicator circuit. The ripple and noise are not detectable with the new filament supply (see schematic below and photo of the output on an oscilloscope).
New High-Voltage Power Supply PCB. The new power supply still employs a RC filter design, but can accept the larger snap-in style capacitors, has an integrated ground-lift switch, and a LED indicator circuit (see schematic below).
Updated CCS-VR PCB. The new PCB now accepts a voltage divider capacitor for the LR8 voltage regulator, which eliminated the noise output from the regulator, as well as space for a larger bypass capacitor for the second RC power supply stage (see schematic below).

Everything is essentially the same as the original DHT pre-amp. I did switch to the Auricap XO by Audience, which is one of my favorite output capacitors. I also switched to a toroidal transformer made by Antek (AS-05T160). I am at the point in my amp building hobby that I mostly prefer toroidal power transformers because they are super quiet. Other changes are documented in the updated schematics (see below). I positioned the tubes farther apart as well as farther away from the power transformer, resulting in no stray fields finding their way into the tubes. Now, the only noise on the output is the noise from the tubes themselves (~3mV peak to peak). There is no evidence of the 60 Hz mains frequency on the output (no hum!). Also, there is no audible buzzing on the output caused by 120Hz ripple. Just tube noise! I still use the Mapleshade Brass Tube Halos because they help with microphonics and hum induced by my hand getting close to the tubes when I adjust the volume. I know they look clunky!

Here is the oscilloscope trace at the output of the updated Filament Power Supply. The volts/division is 2mV, which is the smallest division on the Tektronix 2213A. No visible ripple or noise! The original Filament Power Supply produced ~2mV of ripple with some modulated noise on the output (see Testing/Specifications page) Not bad!

Here is the updated schematic for the Filament Power Supply. I added an LED indicator circuit that is optional. I just wanted a way to know if the filament circuit is working properly. The LM317 is configured as an adjustable current source and the voltage for the LED is provided by the positive voltage rail before the regulator. The BCY59 NPN device is turned on when voltage from the output is applied to the base, thereby connecting the LED to ground and completing the circuit. If there is no output voltage from the regulator then the LED will not light up. The trimmer resistor is to allow for the adjustment of the base current since different output voltages are used for the different DHT tubes.

The circuit is essentially the same as the original except with larger output capacitors and the replacement of the Triad CMT common mode inductor with the their new CMF dual mode choke. I chose the model with the highest inductance values but with the lowest current rating. This is not a problem with the current draw of the 01-A, 30, 31, and 112-A tubes. It is pretty clever to intentionally build a common mode choke with stray inductance to reduce differential noise. I feel like the use of the CMF model combined with larger capacitance knocked down the ripple to a negligible level. Additionally, the noise level of the LT1963A is negligible as well. I think I will only use this regulator over the MICREL for my future projects. The DHT Filament Regulator will be available for purchase soon. Click here

Here is the updated schematic for the High Voltage Power Supply. The LED indicator circuit is optional. The LR8 is configured as an adjustable current source with the high voltage from the power supply dropped by the 33K resistor to 18V. If the high voltage drops below ~178V, the LR8 will shutdown since it needs at least a 12V input-output differential to regulate. The new PCB accepts snap-in type electrolytic capacitors, such as the Nichicon KX series, and larger film bypass capacitors, such as the WIMA MKP 1uF.

Here is updated schematic for DHT Pre-amp. The major differences are:

The addition of a 4.7uF capacitor from the voltage divider of the LR8 to ground to reduce output noise from the regulator.
I also upped the current running through the HLMP6000 LEDs because a little bit more current is better for stability.
I changed the output cap to 1uF, but I can not tell a difference in sound from the original 0.22uF cap.
I decided to eliminate the 120 ohm grid stopper resistor since it really is not necessary.
I added a 220K grid to ground resistor so that the grid is always at ground potential when adjusting the volume. Technically, the Gold Point attenuator uses make-before-break switch contacts, which eliminates pops or clicks when adjusting the volume. However, I find that the DHT tubes I am using are very sensitive to the make-before-break switch contacts, so for stability I added the ground to grid resistor. The load for the input signal is now 100K and 220K paralleled or 69K, which is still sufficient for low-impedance sources.
The TL431circuit is attached to the 0V filament pin versus the positive filament pin in the original amp. Some amp builders say there is a difference in sound, but in this amp in my system there is no audible difference between connection methods.

Van Morrison

Play Video