Introduction

The indirectly heated 9-pin 12B4-A tube is one of those tubes that you inevitably hear about when looking into building a pre-amp. It was originally designed for vertical deflection applications in televisions, but it found its way into the DIY vacuum tube world many years ago. There are many DIY 12B4-A pre-amp projects out there, so there is a lot of information on how to design your own 12B4-A pre-amp. One look at the data sheet, you can see why this tube would work well in a single triode gain stage circuit. One other thing, there is only triode per tube, which is very old school! Despite being somewhat of a current hog, the low amplification factor (~6.5) and low plate resistance (~1030 ohms) make it well-suited to a pre-amp if you do not need a lot of gain and/or want to use the amp primarily as a volume control. I highly recommend this tube as it is very transparent and the prices are still very reasonable. Out of the many pre-amps I have built, I have kept this one and of course, the DHT pre-amp. Here is the datasheet.

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

The single gain stage uses a clone of the Bottlehead C4S Cascode Constant Current Source (CCS) circuit for the plate load and replaces the traditional pate resistor or choke. The cathode is biased with a Kiwame 2W resistor bypassed with a high-quality Nichicon UKA audio grade electrolytic capacitor. Current is set by the 47.5 ohm resistor and drops  < 1 volt (47.5 x .018 = 0.85 volts). The MJE5731 drops most of the voltage (~ 81 +/- 5 volts, depending on the operating point of the individual tube) and gets a heatsink since it dissipates ~1.5 watts of power. The HLMP-6000 LEDs drop ~1.6 volts each and "turn on" the 2n2907A and MJE5731 transistors. The base-emitter turn-on voltage (saturation voltage) for the MJE 5731 and 2n2907A is ~0.7V, which means that the voltage difference between the emitter and base pins is ~0.7V. The 2N2907A only drops ~1.6V, so no heatsink necessary. The benefits of a CCS include improved Power Supply Rejection Ratio (PSRR), lower distortion, and improved gain. Since the plate resistance is low (~1,030 ohms), there is no need for a buffer or transformer stage if the input impedance of the amp is at least ~15K ohms. The output caps are the Auricaps by Audience, which I find to be transparent and do not "color" the sound in any noticeable way. One of my favorite caps.

There  can be quite a bit of variance in 12B4-A tubes. Since the current is constant through the cathode resistor, the cathode voltage will be constant, but the plate voltage can vary. I have measured plate voltages from 112V-122V, but this is not an issue for performance in this circuit.  One other thing is that the cathode is quite bright and noticeable with the 12B4-A, which is good for "tube glow".

The 1K grid stopper resistor is included in the circuit because it helps prevent parasitic oscillation in the tube caused by high frequencies and it prevents radio frequencies from entering into the gain stage.

See below for a picture of the PCBs with the CCS circuit along with the cathode resistors, cathode bypass capacitors, and the output capacitors. I did not include an output load resistor in the pre-amp since it is used to drive power amps with an input impedance of at least 100K. This pre-amp could probably handle power amps with source impedances down to 15K since the output impedance of the circuit is only ~1030 ohms. Some builders use a series resistor in the output to maintain a linear response for long cables, but I have never had issues not using them and I use cables up to 5 feet.

The Circuit
The Power Supply

See below for a picture of the power supply PCB, which is Glassware Audio Design PS-1 and has a circuit with a high voltage current regulator paired with a low-voltage regulator, which supplies the B+ and a low voltage regulated DC circuit which supplies the tube filaments. You can read more about the PS-1 here. I have used the PS-1 in other projects and I have experimented with different values for the output bypass capacitors and have found that 10 uF caps sound the best. The kit comes stock with 4.7 uF metallized/polypropylene caps that work great, but I recommend using 10uF caps, especially if the PS-1 directly feeds the tube stage, as is the case in this pre-amp. I have tried both the Solen PPE metallized polypropylene and the Obbligato Gold Premium polypropylene caps and I would say that the Obbligato sounds better but at the end of the day, I don't think it matters as long as you use quality 10uF caps. The other stock parts for the kit work well and I would not necessarily recommend replacing them, except maybe the electrolytic capacitors in the high voltage power supply circuit with Nichicon KX  or JJ capacitors, and the high voltage power supply rectifiers with IXYS DSEI-12A. The output voltage is set at 200 volts by two 20K resistors that form a voltage divider (see schematic below). The transformer I used puts out 360v DC with the ~44 mA total current draw of the pre-amp, so there is ~158.5v drop across the IXCY 10M45S dissipating almost 7 watts of power! I have not had any issues, but I would recommend choosing a transformer with a lower secondary voltage rating (you could probably go as low as 180V for the secondary rated for at least 100mA current, assuming you use a full-wave bridge rectifier). Here is a handy design guide to rectifier use. The power supply is reasonable quiet with a calculated output ripple of ~4mV RMS into the CCS. The PSRR of the CCS is good for another 40dB, so the residual ripple in the output is calculated at ~40uV (4mV/100 = .04mV). This is pretty quiet and indeed the only noise you can hear at the speakers is "tube noise" and you have to put your ear right up to the speaker to hear it.

The filaments of the 12B4-A can be wired in series at 12.6V (pins 4 and 5 connected to + and - of the filament regulator output, respectively) at 0.3 amps or in parallel at 6.3V at 0.6 amps (pin 4 and 5 connected to the + output, and pin 3 connected to the - output). For this pre-amp, I wired the filaments in series for a couple of reasons. The first is that I pretty much use DC for filament heating in all my projects to eliminate or minimize hum, so generally some kind of regulation is in order. This leads to the second reason; if you use a regulator such as the LD1085, there is a minimum input-output voltage difference required for it to regulate, in this case it is 1.3V. The transformer I used has a 6.3V secondary winding rated at 2.2 amps with no center-tap, which necessitates the use a full-wave bridge rectifier. However, the voltage input at the regulator would be ~8.0V. Since we need an output of 6.3V,  the input-output voltage difference is only 1.7V, which is very close to the 1.3V drop-out voltage of the LD1085. It might work, but this is where a low dropout (LDO) regulator would be useful. The other option is to use a full-wave voltage doubler, which essentially "doubles" the output voltage (output voltage = secondary voltage x 2.8) but with a large penalty for rated output current ( output current =  rated current of the secondary/3.6). Since the 12B4-A only uses 0.3A at 12.6V, we would only need 0.6A from the secondary, which is just barely meet (2.2A/3.6A = 0.6A). The voltage output at the regulator is ~16V, leaving ~3.4V of headroom and the power dissipation through  the regulator is a reasonable ~2 watts. An option with the PS-1 board is to bias the filament output with a voltage divider from the B+, which is highly recommended since this tends to reduce noise. For this pre-amp, I set the bias at ~37V (the heater-cathode voltage difference cannot exceed 100V DC). Noise output is ~ 38mV RMS, which is fine for an indirectly heated cathode. The schematic is below. Also, see the Rebuilding Ideas section below for future ideas on redesigning the filament power supply of this pre-amp, not that it needs it!

Here is the schematic for the regulated DC filament power supply (voltage doubler circuit):

Here is the schematic for the power supply:

Here is the picture of the PS-1 PCB:

Here is an overview picture of the inside of the pre-amp. The stepped attenuator volume control is from Goldpoint Level Controls. The 13" x 10" walnut chassis is made by Hammond and is readily available. The AC power entry module is made by Schaffner.

Here are some external photos of the pre-amp. The transformer is an Amplimo Type 3N604 toroidal transformer and is extraordinarily quiet. The tube dampers are made by Herbie's Audio and are the Ultrasonic SS-9 models, which are highly recommended. Chassis isolation is via Tenderfoot footers by Herbie's Audio. The tube sockets are from Azuma of Japan and are available from Parts Connexion in Canada.

Rebuilding Ideas

There really isn't anything wrong with the current regulated DC filament power supply circuit, but if I had to start from scratch I might consider a new design. Below is a revised DC regulated filament power supply circuit using the Micrel 29152WT LDO regulator. The dropout voltage at 1.2A is ~0.3V, which will give plenty of headroom with the 6.3V secondary since the input voltage at the regulator is ~8.0V. The regulator would still dissipate ~2 watts of power, which is what the regulator in the current circuit dissipates, but the nice thing about the new circuit is that it simple and compact and would probably fit onto a 2.8" x 2.0" PCB. Output noise would be ~260uV RMS, which is much better than the LD1085. The Glassware Audio filament circuit PCB is 3.6" x 2.2" and Pete Millet's PCB is 3.5" x 2.0" but uses the LD1085 regulator.

If I started from scratch, another thing I would consider is adding the LR8 voltage regulator to the power supply circuit for each channel. The LR8 is good for ~60dB of ripple rejection, which will make for an even quieter pre-amp (I calculate the residual ripple at ~0.04uV!). Generally, you want at least a 25V input-output differential and ~1mA for the reference current to get the sweet spot for the LR8. Since there would be 20mA of current flowing through the LR8 it will dissipate 0.5 watts. Therefore,  I would recommend using the SOT-89 or TO-252 package, which will get you a power dissipation rating of 1.6W or 2.5W, respectively. You can fine tune the input and output voltages of the LR8 to suit your needs, but keep the output voltage at least 50V above the plate voltage to allow for adequate headroom. Below is a revised circuit (one channel shown). A PCB for this circuit would be  about 3" x 3".

Also, I would use a different transformer with a lower voltage rating for the secondary to reduce the amount of power dissipated by the IXCY 10M45S but I would keep the PS-1 power supply.