How Do I Make a Distribution Amplifier?

This article started out as part of the FAQ section of my CMoy pocket amp tutorial. I kept getting questions from people who weren’t satisfied with my answer there, which basically said, “No, the CMoy isn’t a distribution amp, go Google if you want some designs.”

Before we move on, please allow me to Google that for you right now. All of the links below come from the first page of results. Perhaps you will find some of the other results enlightening.

I will not in this article give you a complete step-by-step project. I will merely point to the simplest options and comment on them, explaining the choices, warning of the concomitant problems, and extolling the benefits. I will also hint at ways to achieve the more complex features you can get in expensive commercial designs.

At the end of this, I hope you will see why I did not just add a tweak item to the original article or add the links below to its FAQ. The issues are complex and must be considered before you start building. It’s why, until now, I have left this issue to other articles on other websites. (And to a large extent, still do.)

Current-Shared Outputs

The simplest thing that could possibly work is to add another output jack for each set of headphones you need to drive, plus a series current-sharing resistor per headphone driver. If you need to drive 4 headphones, this solution adds just 4 jacks and 8 resistors.

Kevin Ross’ Headhone Distribution article includes a schematic showing the basic idea.

This solution is simple and cheap. I’ll bet you’ve guessed the real cost, though: it brings a lot of problems and restrictions.

The two big restrictions are:

  1. You must use identical headphones for each output.

  2. The volume level for each channel is adjustable only with a soldering iron.

Okay, okay, it’s not that bad. Only exactly that bad in practice.

It is not absolutely true that you have to use identical headphones. They merely have to be very close in performance, or you have to be willing to accept that each channel will have a different volume level, or you have to be willing to fiddle with the current-sharing resistor values to null out the efficiency differences. One drawn to this solution’s simplicity and low cost probably won’t like any of those choices, though, so for the sake of discussion we can continue with the assumption that all the channels use identical headphones.

As for the second restriction, my soldering iron comment probably got you immediately thinking about potentiometers. Again, strictly speaking, this restriction is not absolute, but it is in all practical likelihood true enough that we can take it as fact. The biggest reason for this truth is that audio pots don’t come in values low enough to be useful here. If you’re willing to tolerate the poor volume control behavior of a linear pot, you have a choice of two additional problems. The first bad choice is that a series pot-as-variable-resistor scheme will create a volume interdependency among the channels, because they are parallel resistive circuits; lowering the volume in one channel will increase the volume in the others and vice versa. The obvious alternative, a parallel scheme using the pots-as-pots avoids the interdependency but increases the load on the amplifier, with consequences lined out below. And it still leaves you using linear pots! P’tui.

I didn’t cut this option from the article because in some situations, several identical outputs are acceptable. Museum listening stations, for example.

This solution still has two big problems.

First, the extra required resistance on each output may hurt sonics. Some claim they prefer the sound change caused by adding a little extra resistance here, with some headphones at least. I claim that no headphone designer designs for this case, so you are probably better off with a 0 Ω headphone output if you can manage it. Even the practice’s stauchest supporters will tell you that adding a lot of resistance here will hurt sonics. You might not be able to get away without adding a lot.

Problem 2 is that you can’t do this with the stock CMoy pocket amp design if you use low-impedance headphones like Grados. (32 Ω across the line, as far as I’m aware.) The stock design has enough trouble driving a single pair of such cans as it is. It’ll do it; it’s just not the best deisgn for the job. Doubling the load will probably result in unlistenably bad sound.

There are many solutions to Problem 2, however:

  1. Use high-impedance headphones. Most headphones are probably 32 Ω or under, since lower impedance offers a better tradeoff for portable players and portable use dominates the market for headphones. That said, 64 Ω, 120 Ω and 300 Ω are also popular impedance design targets. In the bad old days, 600 Ω was the most popular choice; some of these design relics may still be commercially available.

    It is not the case, however, that ten hypothetical 320 Ω headphones will be as easy for a lone amp to drive as a single 32 Ω headphone set. It will be much harder, despite the fact that the overall impedance is the same in each case. First, we’ve been ignoring the effect of the current-sharing resistors; you will need to increase the volume level to counteract their effect; more voltage equals more current since power to the headphones remains constant. Second, high-impedance headphones require more voltage to begin with, as a rule. (Not a hard-and-fast rule; efficiency varies, causing exceptions.)

    These two facts combine to increase the current load on the amplifier as a whole over the simple multiplication you might assume. Headphone amps are all about supplying current to headphones, but each design has its limits; multiplying headphones brings you closer to those design limits, or past them.

  2. Make the CMoy pocket amp more powerful. Since just one component in that design affects this, the op-amp, you might go looking for a more powerful one. Every such candidate I’m aware of, however, is cranky in one respect or another.

  3. Switch to a buffered headphone amp design. Designs such as the PPA v2 and PIMETA v2 can be 10× as powerful as a stock CMoy pocket amplifier, or even more. Again, this does not mean they can drive 10× the number of headphones, due to power lost in the current-sharing resistors. But, more power does certainly help.

  4. Use a power amp instead. You don’t necessarily have to buy or build a dedicated power amp. Perhaps you have a disused integrated amp in the closet? Maybe one that was set aside precisely because it wasn’t powerful enough to drive your floorstanders to wall-shaking volumes? The weakest power amps are more powerful than almost all headphone amplifiers; the few exceptions are small power amplifiers in disguise. Beware that the more powerful the amp, the higher the value you will need for the current-sharing resistors, to avoid blowing out the little headphone amp drivers. An old school 3 W SET amp is a better choice here than a superannuated 150 WPC home theater receiver. The less you care about sonics, the less this consideration matters, because you can always fix any overpower problem with bigger resistors. Do test with headphones you don’t mind blowing up, however.

What is the correct resistor value?

The proper resistor value is very much contingent on the headphone and amp designs.

I can see values between 10 Ω and 1 kΩ being useful, depending on the situation. Maybe higher, but probably not lower. If in doubt, start high and work your way down until your particular configuration stops behaving correctly, then rebound.

I don’t believe the correct value depends on the number of headphones. I can talk myself into believing it does matter in a degenerate case which won’t happen. The argument goes like this. I first realize that these resistors actually do a second duty besides forcing current sharing, that being to damp the parasitic resonant circuit that is part of any headphone design. “Self,” I say, “it is true that the correct value for the resistor for one headphone should be correct for K headphones, where K ∈ ℕ and K > 1.” This argument satisfies for a time. Then I interject, “No, that argument breaks down as R goes to zero. While we can usually get away without an output resistor for a single headphone, surely this cannot continue to be safe as K becomes large! All those Cs and Ls in parallel...oy!” Then my brain comes to its own rescue, observing “That argument is bogus. We need some resistance for current sharing anyway, so R cannot go to 0. Besides, if you need a little extra resistance to keep the amp stable with some headphone designs relative to others, you probably should use a nonzero output R with that headphone in the single case, too.” And scene, then dénouement.

(There is a stage in my head, and players thereon, and they are all me.)

Paralleling the Drive Sections

You can avoid those restrictions and solve the problems by trading them for extra build time and parts cost. Instead of paralleling just the output jacks, you parallel the entire CMoy pocket amp headphone driver circuit instead. This buys you individual volume control for each channel and freedom to mix heapdhone types.

The PAIA 9206K kit is basically the distribution amp version of the CMoy pocket amp. They’ve kindly made the schematic publicly available, so you can entirely DIY it if you like.

The only significant change over a naïve “CMoy distribution amp” implementation is that they’ve cut the power supply divider resistance by 110 — thereby increasing the power wasted here by 10×. PAIA have done this to reduce the effect of the virtual ground shift problem inherent in the CMoy pocket amp’s resistive divider power supply design. This problem is multiplied by the number of channels, so they’ve not cut it by a factor of 10, but by only 106, or about 1.7×.

(Between the lower virtual ground divider resistors, the multiplication of the output driver sections, and the increased load demand by the headphones, you probably don’t want to use batteries for this.)

Keep in mind also that they’re only using a 12 V supply with those lower divider resistor values, so 14 W resistors are probably still adequate. Wattage ratings go down as heat goes up, and this change increases heat in the resistors, so if ventilation or convection aren’t sufficient to cool the resistors, you will have to step up to larger ones. If you go with a more typical 24 V supply, you should switch to 12 W or 1 W resistors instead.

PAIA recommend the NE5532 op-amp for this, but that change is not necessary. You can also use any of the standard CMoy-friendly op-amps instead.

Chu Moy himself recommended a similar design to solve the same problem, differing mainly in the addition of an extra buffering stage, which looks like a good idea to me. (It prevents the load on the source from shifting as the volume levels change. To an ideal source, this shift wouldn’t matter. I expect some sources are as close to ideal as matters here, but I doubt all are.) The Chu Moy solution is about 34 the way down the linked-to page.

If you go with such a design, you will probably want to share a single power supply for all the amplifiers. The design discussions for the headphone amps on this site don’t talk much about power supply current requirements because they all draw less than even the wimpiest commercial power supply units can deliver. When you gang them up, that stops being true, in some cases even with a simple dual-headphone amp design. With 4 or more outputs, you really have to watch out. As a rule of thumb, a basic CMoy draws 10 mA, a META42 anywhere from 20-50 mA, a PIMETA v2 50-75 mA, and a PPA v2 90 mA+. If you have extra buffering stages, add them in, too. As my math teacher once told me, “Tiny × Big = Large.”

Hybrids

Another popular design didn’t appear in my first page of Google results, but it’s an obvious combination of the above choices. If you are happy with using the same headphone type for each output and you only need a single volume control, you can follow most any headphone amp design with K buffer stages, one for each output. It’s sort of the inverse of the Chu Moy distribution amp design, which has a buffer stage followed by K volume adjustment and amplification stages.

The advantage of this is that the parts cost is lower than for the Chu Moy design (1 volume control instead of K) and it’s easier to build due to the somewhat lower parts count. Its main virtue over the current-shared output plan is that the buffer stages prevent any interdependency problems among the headphone loads.

The main thing to watch out for here is that the buffer stages mask any performance advantage in the volume control and voltage amplification stage. So for example, while it would work to add K×2 naked BUF634s to the output of a PPA v2, it would be foolish to do so. Your voltage amp stage quality shouldn’t be better than that of any following buffer stages.

The inverse is not true. A mediocre amplification stage followed by K high-quality output buffer stages can work out well. While it is true that the buffer stages in such an amp will faithfully replicate any problems the amplification stage has, the buffering will reduce the number of problems it would have if it were driving the headphones directly. A CMoy pocket amp with a decent but unexceptional op-amp followed by a gang of high-spec PIMETA v2s might well sound better than if the PIMETAs had the same op-amp as the CMoys.

A wasteful but easy and functional way to build such a distribution amp is to build/buy K+1 headphone amps, set K of them to a gain of 1, and build an octopous cable between the output of the remaining amp’s output and the others’ inputs. You probably don’t need current sharing resistors in this case, as that is probably included in the input stages of the headphone amps. The waste is in the fact that you have duplicate power supply units, duplicate power handling circuitry, and duplicate cases. One big one of each shared among the amps is cheaper.

That then brings up a second hybrid option, which is basically K+1 headphone amps, each with a volume control, K of them being fed from the “+1,” which acts as the buffer stage. That gives you a master volume control and a per-channel volume control. Some expensive commercial headphone distribution amps use this design, and it’s one reason they’re expensive.

Conclusion

As you can see, it is true that the CMoy pocket amp is not itself a distribution amp. If you just glance at the schematics linked above, you can be fooled into believing that the changes required are simple. Whichever way you go, though, you either buy problems or you buy parts to fix the problems.

I hope this article has helped you choose the least painful problem set for your situation.


This article is copyright © 2016 by Warren Young, all rights reserved.

Updated Sun Jan 18 2015 04:23 MST Go back to Audiologica Go to my home page