A short note about splitters

[R.G.]

I get a continuous dribble of email about splitters. One of the recurrent threads is whether one buffer transistor or opamp can drive all the driven devices or whether one buffer per device is needed.

The short answer is that except for special cases, you only need one buffer to drive a small but reasonable number of "split" inputs.

It's back at that impedance thing. If you want to "split" a signal, you want to send a signal to two or more following inputs in a way that the two inputs do not interfere with each other. For simple loads, such as driving two or three pedals, a single buffer is fine for driving both/all three/all four/etc. in parallel. If the output impedance of the drivng buffer is less than 1/10 of the parallel combination of all the driven input impedances, then none of the driven inputs can affect the others.

This only gets tricky when you try to do fancy stuff. If you want to mute one or more of the driven inputs, then you have to choose series-switch muting, shunt-switch muting, or series/shunt muting. Neither of the series switched mutings will have any effect on the output of a good buffer. It's the shunt muting that may cause issues.

A shunt mute is a switch to ground. A real, no-fooling switch to ground is a lower impedance than any buffer, so a simple switch to ground will mute all the driven inputs by shorting the output of the driving buffer to ground. The answer is simple - put a series resistor from the driving buffer to the shunt-switch. When the shunting switch is closed, the buffer sees the series resistor as a load, and if this resistor is big enough (say, 2K to 10K) then the buffer will not be overloaded and there will be no effect on the other driven inputs. You pick the series resistor value to have no effect on the input being muted and no overload on the buffer.

The series-mute versions are prone to letting the muted effect produce noise or oscillation in some cases. This is because the input is open. The series/shunt muted version is immune to this problem.

[Processaurus]

That's good to have a confirming take on that.  I've slowly started to get more confident with the sturdiness of opamps, that a follower can source a ton (well, a ton in the realm of small battery powered audio widgets) of current to keep its output exactly following its input, and that moderate loads aren't likely to to sag it, or interact with each other. 

The only examples of educated splitting in pedals I've seen are in Boss schematics, where they tirelessly buffer everything discretely (I can't tell if it is to be bulletproof, or it was just part of their education), and Orman's page on FET opamp and discrete FET splitters, where he uses a device per output, but feeds them from a single input, and only biases them once.

But fanning out a single opamp with some small resistors sounds good, and refreshingly minimal.

[dano12]

I hadn't thought about the shunt issue before. Interesting problem.

Here's the schemo of a simple mixer I used on a multi-effects box:

[Top Top]

Thank you for this post. I have wondered about this issue and have asked about it here with confusing answers (probably more my fault than the people answering). This puts it very clearly and practically.

The shunting to ground muting issue is still a little confusing to me. A schemo showing it would really help clarify the issue, but I won't get too mad if I have to just breadboard it and figure it out myself 

[caress]

great post as usual, r.g.

[R.G.]

I hadn't thought about the shunt issue before. Interesting problem.

Here's the schemo of a simple mixer I used on a multi-effects box:

Notice that connecting any output to ground through a mute switch also grounds all the others. That's the problem.

[birt]

but if you use a SPDT relay or switch you can toggle between input (of the effect after the splitter) to ground or input to output splitted signal. its like having a series and shunt mute working together.

[R.G.]

but if you use a SPDT relay or switch you can toggle between input (of the effect after the splitter) to ground or input to output splitted signal. its like having a series and shunt mute working together.

Yes, you can. That is in fact what a series/shunt switch is: two simple one-pole switches, one in series and one in shunt, and operated in opposition so that one is open while the other is closed. They are exactly the same thing. I would reword your post as follows:

(and) ... if you use a SPDT relay or switch you ... have... a series and shunt mute working ... from the same actuator

.

However, the instant you get beyond wiring everything in hard switches and go to electronic switching, the fundamentals get really important again.

[PRR]

> The short answer is that except for special cases, you only need one buffer to drive a small but reasonable number of "split" inputs.

Correct. (Of course.)

"Special cases" are many, and fun (sometimes not).

When a broadcast originates in Chicago, say, but is networked across the nation, (before digital telco) the signal must be split to the local transmitter, to the east coast, to the west coast, etc. Long lines need considerable "push". And long lines are prone to trouble including short-circuits or cross-connections to unwanted conversations. A "Distribution Amp" takes the one signal and makes many isolated outputs. If the flood in California shorts the west line, you still have local and east-line signal.

We've all seen hasty press conferences with dozens of mikes on the podium. And organized press conferences with just 1 or 2 podium mikes, yet dozens of reporters get a "feed". Many reporters show up with funky gear and no tech skills. Here they use a "Press Box", a distribution amp with not so much "push" but very tolerant of shorts and miswires.

When a band does a live-in-concert recording, they don't want two mikes everywhere. Each mikes signal is split to both house-sound and recording-rig. Here there is usually no amplifier and nearly no isolation. Modern mike inputs do not load mikes much, the loading of two such inputs causes less degradation and potential trouble than a mess of splitter-amps. Both crews (house and recorder) are professionals who don't bring shorted cables, are experienced and motivated to cooperate to resolve any problem quickly.

Likewise a guitarist can use a Y-cable to feed the inputs of two pedals. The double-loading may be tolerable. If there's a short, he'll find and fix it.

Which way you go depends a LOT on how well you can control the loads. If all loads are inside your own space, and you can design them "light", never heavy or shorted, just split the signal. If the feeds go through hostile turf to strange people, you want to be sure that NOTHING they could possibly do can screw-up the signal to yourself and other users.

"Distribution Amps" take two forms. One big amp usually with separate resistors on every output; and an individual amplifier per load.

I have seen a chest-high rack with 25 tube line-amps: one accepted signal and fed a bus protected inside the rack. The other 24 amps provided two dozen completely separate high-level feeds which were nearly bullet-proof.

But the more usual form is one beefy amplifier with separate resistors to each load. Say you need up to 24 600 ohm loads. If you just paralleled, it could be 600/24= a 25 ohm total load. A lame loudspeaker amp would do this loafing. However a short on any output would kill all outputs plus smoke the amp. So you put in a series resistor. These days you might choose 50 ohms (true 600 is dead). 650/24= 27 ohm "normal" load. Now run a forklift over wire 13, crushed to dead-zero short. But there's still 50 ohms inside the box to the #13 line. The total load is (click click) 18 ohms. Still not a problem for a loudspeaker amp. It is customary to assume that "a third" of loads could wind up shorted. 16 at 600+50 with 8 at 50 is total 5.3 ohm load, and still acceptable to a loudspeaker amp. On good days the amp runs very light and cool, on bad days it runs no harder than it is designed to.

In non-Broadcast work we don't even find 600 ohm loads. More often 10K to "1meg" (though any cable long enough to have two ends will droop below 1Meg at the top of the audio band). The series resistor reduces signal, and we don't generally know the actual load. We assume a minimum load, decide how much droop is tolerable, we get our series resistor. 1K series makes "negligible" droop in 10K or higher load. Put 1K in series with each output.

Now design your "big amp". Say you have 4 outs, each 10K or higher. On a good day that is over 2.7K. If one output is shorted, its 1K series resistor plus the 3 happy loads is 0.78K total load. Most chip opamps will deliver many Volts in 780 ohms. While the THD may rise, the show goes on. The three happy loads work fine.

Another issue is volume/gain controls. dano12's plan has a volume control feeding 4 outputs. There are no series resistors so one short kills the whole signal. As this is probably a musician's tool, the musician will fix his own shorts, OK. But adding four series resistors still won't make this plan short-proof. The 10K collector load and the 100K volume control are series resistors, common to all four outputs, and with equivalent value near 10K-27K. The individual isolation series resistors should be smaller than the load but much less than the source. This plan can't be made short-proof without changes and probably complications (and for dano12's purposes, simplicity probably trumps short-proof).

When you want an individual volume control per-output, it gets more tricky. The brute-force way is to use a BIG amplifier and heavy volume pots. You often find studio headphones driven this way. A dozen 100-ohm phones, 100 ohm pots, and a fat loudspeaker amp. You are pushing Watts in to get milliWatts to each can, but simplicity is nice. Alternatively you find boxes with one input opamp, a bus of volume pots, and an individual chip for each headphone. More parts, less waste.

Observations from other fields:

Analog computers may need to tap several loads from one signal. The classic transistor opamp can use 100K input resistor and will drive 2K load. Therefore one output can drive 50 inputs. However this is not short-protected: if some mis-patch shorts one input, the other 49 loads go dead. This is acceptable because (presumably) all 50 connections are needed to get the right result, so a short "should" produce a very wrong result.

Digital systems often send one output to many inputs. They use the word "fanout". One old TTL output can "fan out" to four TTL inputs. If you needed more, you used a "bus driver" which could drive 10 loads. As the technology matured, fanout improved (slow CMOS has no real fanout limit). Digital systems are not expected to be short-tolerant. However where shorts happen (printer cables), the cost of digital has been low enough to run a separate buffer for every outside load.