gallery/2 ecpsoftware articles diy audio links contact

MEHA v0.1a

This is the project page for the MEHA v0.1a. A few prototype boards were sent around, but I have not heard whether they worked, whether people liked them, or what the story is. For now, I'll leave this page up as a bit of history which might be useful to someone.

Board Layout


This project is intended to extend the low voltage hybrid headphone amplifier designed by Pete Millett. In that project, Pete used a low voltage triode tube (12AE6A) for voltage gain and DC coupled it to a solid state buffer for current gain to drive headphones. While the project is very innovative, and sounds quite good, it requires a coupling capacitor on the output, and because headphones present a low impedance load, that capacitor must be quite large -- upwards of 470uF for 32ohm headphones.

The changes of this project over the original are several fold. The first, is that the tube is AC coupled  to the buffer. The reason for this is that, hopefully, it allows the use of a high quality film capacitor in between stages instead of the electrolytic on the output. The second change is the replacement of the 12AE6A's with a a dual triode (6GM8/ECC86) which is subjectively of higher quality and, as a dual tube, only requires a single tube. Third, the project uses a buffered ground on the output which has been found, in many cases, to be beneficial for headphone amplifiers. Last, the project is physically smaller than the original, which seems like a good thing.

At this point, the project is in the early prototype stages. You can view the schematic here, or the inital board layout here. Prototype boards have been ordered, and we'll see where it goes from there. Keep an eye here or on and for updates. You can also send me an email at

Why "MEHA"? There are a whole bunch of low voltage hybrid headphone amplifiers floating around. With the exception of Pete's, these all seem to have names like SOHA, or YAHA, etc. The "ME" here stands for Millett-Esque. I am not sure if the "HA" part is Hybrid Amplifier, or Headphone Amplifier, or both, or neither.

Some Gory Details:

In order to pick some of the parts, it is important to understand the basics of how a tube works, and how to pick operating points. However, to get a more thorough understanding, it is worth reading some of the many tutorials on the web. Here are some good links: Anyhow, the basics are that a tube utilizes the thermionic effect. That is, a basic tube has two parts or electrodes. A Cathode and an Anode (usually called the plate). If heat is applied to the cathode, this causes electrons to boil off. If a voltage is then applied to the anode, which is some distrance from the cathode, the electrons are attracted to it, and a current is caused to flow with the electrons jumping across the vacuum. (It is worth noting that current flows from a negative to a positive potential. This is the opposite of how people tend to think in solid state circuits.) This flow of current is constant in a two electrode tube so long as the heat and the voltage applied to the anode are constant.

However, a third electrode can be added. If a grid of wire is placed between the cathode and the anode and is at a negative voltage to the cathode, it will repel some portion of the jumping electrons. By altering the voltage of the grid up and down, the flow from the cathode to the anode can be increased and decreased. And since music signals do alter in voltage, applying that voltage to the grid causes the flow to increase and decrease.

Further, since it is the amount of potential difference between the cathode and the grid (that is, how much the grid's voltage is negative with respect to the cathode) that determinales how much current flows, it is by setting the grid's voltage that a tube's operating point is set. This is generally referred to as biasing the tube. (A note -- on the original Millett amplifier, the bias is measured on the plate, not on the grid. The plate and the grid are proportional to each other, changes in one influencing the other, so measurements can be taken at either. This design has test points at both the plate and the cathode (the grid being connected to ground) so you can convince yourself of this.)

There are several methods that can be employed to bias a tube, but the simplest one, and the one used here, is to place a resistor between the cathode and ground. Since current must flow in a complete circuit, the current flowing across the tube must begin at ground. Since, at a given current, a resistor will decrease voltage by a predictable amount (V=IR) adding a resistor in series between the cathode and ground will drop a certain amount of volts as the current flows resulting in the cathode being at a positive potential (voltage). Further, if one references the grid to ground (by connecting the grid to ground with a large resistor), the cathode is effectively "biased" at a positive voltage with regard to the grid. To determine how much current will flow across a tube in a given setup, one must look at the datasheet.

Below, the red dot on the graph shows an operating point where there are approximatly 16V on the tube's plate, about 2mA flowing through the tube, and by looking at the curve, the grid is biased at -0.75V with respect to the cathode.

This was just selected as an arbitrary point, and you can pick a different one as well. The first thing one must do is to determine what value of cathode bias resistor will yield this operating point. This is a simple calculation, and plugging in for


, one gets

0.75 = .002 * R

or 375 ohms. This is not a standard value for a resistor, but 374 ohm is which is closer than you will usually get. It is also worth mentioning that one should determine how much heat that this resistor will need to dissipate. In small signal tubes this is not usually an issue, but with power tubes it can be. Resistor power is determined by the current multiplied by the voltage, so here we only need to dissipate (0.002)(0.75) or 0.0015W, so any resistor will do.

This flow of current, however, is not that useful on its own. However, if some method is employed to convert the current, and more specifically the changes in current, into a voltage, then the tube becomes a voltage amplifier. The most common method of doing this is to add a resistor to the plate. The resistor resists the flow of current by allowng the voltage to swing up and down. This is just a simple application of the common V=IR where voltage and current are proportional.

Thus the next step is to use this operating point and a chosen power supply to determine the load line and the necessary plate resistor. Note, you can do this in any order you want. That is, if you know your power supply already, or you know what plate load resistor you want to use, you can work from that. The easiest thing to do here is to draw a line through the operating point that contacts the X and Y axis of the graph. In the figure below, this line passes through our operating point, and contacts the X axis at ~23V (which is the plate voltage at 0mA of current) and the Y axis at 6mA (which is the plate current at 0 ohms).

To determine what value of plate loading resistor will cause such a load line, you simply calculate the change in current for the change in voltage, and use this to determine the size of the plate resistor. Here, current change is 6mA and voltage change is 23V. So, again using V=IR, the resistor should be 3833 ohms. 3830 ohms is a standard resistor value, so again we are pretty spot on. Now, as a general rule of thumb, the plate loading resistor should be 5x the tube's internal plate resistance. The datasheet is not clear what the plate resistance is, however, so we'll use this value for now.

There are two things to note here. The first is that the load line contacts the X axis at ~23V. I am assuming that we are using a 24V power supply. The 23V is to allow for the fact that that the cathode is biased up about a volt. That is, the plate voltage is with respect to the cathode voltage, not to ground necessarilly. Second, the signifigance of the load line is that this is where the tube can operate. There is a big graph, but with our power supply, plate load, and cathode resistor, the tube is restrained to positions on our line.

With the above information, you could build this amplifier with a resistor plate load. However, as you may be aware, the design calls for a Constant Current Source (CCS) to be used on the plate. The reason this is used is that it forces the tube to operate in a more linear way (meaning with less distortion). The CCS does two things. The first is that it controls the current flowing through the tube rather than allowng the cathode bias resistor to do so. The second is that is presents an ultra high impedance as a plate load. That is, it holds I constant as well as R such that when converting current to voltage (through V=IR) it will operate more efficiently as only voltage can swing. The long and short of this is less distortion (i.e., cleaner sound) and higher gain.

To see how this works, once again one must consult the plate curves. Here, however, instead of drawing a load line that slopes downward, the load line is nearly horizontal as the current is constant.

With this horizontal load line, adjusting the cathode bias resistor simply moves the operating point left and right across this line. So, for instance, with a CCS operating at 2mA, a 375 ohm resistor will bring us to the same operating point as above. Increasing the resistance will move the point to the right, decreasing it will move to to the left.

One additional important point must be made. Since the CCS holds I constant, it must have room to allow the voltage to swing up and down. That is, the B+ voltage must be higher than the plate voltage or else you will get horrible clipping. And, since this is a hybrid design even without the buffers (the I/V is solid state), that clipping is not nice sounding. Too much voltage, however, and the CCS can burn up by trying to dissipate too much heat.

Parts List
Part Value/Notes Size/#
C1L, C1R These are the coupling caps. There is spacing for a 15mm box type cap (such as a Wima) but there are also extra pads to use other larger caps. Something in the 0.22uF to 0.47uF range is probably sufficient. If you want to use smaller values, you may need to increase the value of the R4's.
15mm or axial
C2L, C2R These are the tube's cathode bypass capacitors. You may be able to leave them out -- doing so increases plate resistance and distortion (mostly 2nd harmonic) and reduces gain, some people find it sounds better in some circuits -- otherwise they should be sized for a decent 3dB point. To determine this, you must consider what resistance the capacitor sees. Generally, you would consider the plate resistance of the tube + the plate load resistor, divided by the mu of the tube (plus 1) all in parallel with the cathode bias resistor. However, since the plate load of the tube is a CCS and is thus, for purposes here, infinite, the math becomes a lot easier. You only need to consider the cathode resistor. To know what that resistor will be, you will need to do a little math, but since we are planning to drop in the range of ~1V at between 1 and 3mA, the worst case is that the resistor will be in the 300R range. A 100uF cap would thus be sufficient. So, use a 220uF cap or larger and forget about it. The formula for figureing out what cap size to use for a given resistance is given here. These caps will only see a couple of volts, so large voltage caps are unnecessary. owever, they are directly in the signal path, so quality counts.
10.5mm case, 5mm pin
C3L, C3R These are film bypasses for the cathode bypass caps. Only use them if you use C2's. 5mm pins
C4L, C4R These are power supply decoupling caps for the tube section. The value is not critical -- 100uF to 220uF is fine, but use what you like. 35V or better is recommended. Unlike solid state circuits, putting huge caps here won't add bass. 10.5mm case, 5mm pins
C5L, C5R Film decoupling caps for C4's. Totally unnecessary -- use them if you like. 5mm pins
C6L, C6R, C6G, C7L, C7R, C7G Decoupling caps for the buffers. use 0.1uF film caps, or ceramics are fine too. 5mm pin spacing
C9 Cap to stabilize the virtual ground splitter. 220uF is typical, a little larger or smaller is fine. 35V is a good idea here. 8mm case, 3.5mm pin
C13, C14 I seem to have skipped C8 on the schematic. These are the main PS decoupling caps. They should be big and at least 35V (technically they can be smaller in voltage, but if the rail splitter goes south, you'll be happy they are over spec'd). However, there are two considerations. One is that the case is only 12mm, so they can only be so big. The other is that they are right by a heatsink that is going to get hot, so 105 degree caps are a good idea. Panasonic makes a 1500uF/35V cap in the FM line that will fit this spot (though they also make one that won't, so check the datasheet.) Also, Nichicon part #647-UDW1V152MHD6 is the same. 12mm case
C10 Input cap for the heater Vreg. The data sheet says to use 0.33uF. 5mm pin
C11 Output cap for the Vreg. The datasheet says to use 0.1uF here. 5mm pin
C12 This spot connects the tube's internal shield to ground. Some people ignore this, some use a jumper, and some use a small ceramic cap. If you use a cap, 0.1uF ceramic is probably good. A jumper is fine too. 1206
U$1 This is the pot that I forgot to rename. An Alps blue or Noble pot will fit as will a Panasonic EVJ. 50K is standard, but anything from 10K to 100K is probably fine. Pot sized
R1L, R1R These are the so-called grid leak resistors. However, in this design the pot already does that. These really offer a connection for the grid to ground in case something goes wrong with the pot. Should be 10X the pot value, so probably 500K is a good value. 1/8W is fine
R2L, R2R These are the grid stoppers that help quell tube oscillations. They may or may not be necessary here. You can probably just jumper them, but if you use them, 100R to 470R is a good range. As high as 1K is fine if that's all you have. In an ideal world, you'd use a carbon film resistor or carbon comp here (AB, Riken) but metal film is fine too. 1/8W is okay
R3L, R3R These are used to set the tube's bias. How to do this is explained elsewhere, but what I suggest is that you set the trimmer to about where you think it should be (by doing the math), then adjust from there. The Bourns 3296W series fits, I would think that 2K would be a decent value to try. Better to be too high here than too low as you can always adjust down. Bourns 3296W
R4L, R4R These resistors set the input impedance to the buffer stage. High is good -- something in the 100K to 500K range is probably good, and what you pick will influence your choice for C1L and C1R. 1/8W
R5L, R5R, R8L, R8R 1K For R8 you might be able to use a ferrite instead. 1/8W
R7L, R7R, R7G These are to set the bandwidth if you try to use BUF634's. However, since it is unlikely that BUF634's will work, these can probably be left out. If you use them, 100-220R or so. 1/8W
CCS1, CCS2 These are CRD's and set the current flowing through the tube and act as I/V converters. Somewhere between 1mA and 3mA is probably a good starting point, but feel free to experiment. The datasheet is unclear on operatig points for the tube, suggesting only 0.9mA at 6.3V on the plate.  Since weare running with considerably higher voltage onthe plate, it is trial and error to find a good spot.  However, it does suggest no more than 0.6W of dissipation from each plate.  If that's too much of a limit, use a 6DJ8 instead.  There is space for either the DO-35 or the TO-92 version. Also, you could use a resistor here. This will require a slightly higher B+ but might be a fun experiment. The suggested buffers can take +/-30V and the TLE2426 can take 40V (so they say) and you'll probably need to upspec the caps, but this will likely offer a tube-ier sound.
BUF1, BUF2, BUF3 These are the output buffers. It is likely that the OPA551 is the best choice here, but try other single opamps or BUF634's to see if you can get the DC offset down. Don't forget the sockets.
U$2 Another part I forgot to rename -- a TLE2426 in the TO-92 package. TO-92
VR1 This is a voltage regulator for the heater. The heater actually wants 6.3V (though anything between 6.0 and 6.6V should be fine), but this is an uncommon Vreg voltage, so use a 6V. Any LM7806 in TO-220 is fine -- make sure it can deliver over 330mA. TO-220
D1 Reverse current protection diode. Some of these Vregs have this built-in, others don't. It's a $0.03 part, so worth including. A 1N4002 or better is fine. DO-41
KK1 This is the heatsink for the Vreg. I don't have a part number as I have piles of these around, so I haven't ordered one in a while. Keep in mind that this will need to dissipate over 6W of heat, so big is good.
Tube The tube this was designed for is the 6GM8. However, a 6DJ8, 6922, or 7308 may work too. Don't forget the socket. 9 pin