Sunday, 13 March 2016

Surround Sound Switch #3: Bulgaria (rotary switches)

Mother Of all Relay Boxes
Rolling your own Commutator
Seems there's quite a lot of work in rolling your own commutator. Might it be useful to look at ways of simulating one, using only reasonably priced standard parts more readily available to today's hobbyist? Why yes, yes it might.

Start with a 7P8T rotary wafer switch and electrically collapse it into a 7-pole commutator, by first connecting the seven input signals to the wiper contacts, then connecting all 56 of its output terminals in a septuply nested helix pattern. Despite containing quite a lot of structural redundancy, in terms of the number of potential interconnections versus the number of configurations we actually intend to use, this design at least preserves this nice commutator property: for each of the eight positions on the dial, every one of the seven circuits contains just a single point of dry metal-to-metal contact. So at least we are still minimising contact resistance.

мартеница (Martenitsa - worn for Grandma March)
The Bulgar Fraction

To ebay now, where we find the RFT Palladium contacts 7P8T rotary wafer switches pictured below. Of German manufacture, priced in American dollars, being offered for sale by a gentleman in beautiful Bulgaria.

As you can see from the metric rulers thoughtfully incorporated into their sexy photoshoot, these beauties are about 6cm (2½") in diameter and 15cm (6") from head to toe. Each circuit or pole has its own dedicated 8-output wafer, and the modular design could quite obviously be adjusted to accommodate various other numbers of inputs and outputs. Such stackable wafer components are readily available from various suppliers, but the sheer novelty of finding such an exact, and so steampunk a match, to my very specific 7P8T configuration requirement, simply obligated me to buy one, even if only for a proof-of-concept prototype. Or actually two, since both were in need of some restorative care.

Materials Science

Palladium contacts were developed during the 1980s to replace the too-expensive gold finishes on the contacts of separable connectors used in burgeoning telecommunications systems. To inhibit corrosion by atmospheric chlorine, retard the formation of insulating frictional polymers on the contact surfaces, and improve wear resistance, an alloy, clad 60 per cent palladium / 40 per cent silver (for improved conductivity), with a small amount of protective gold diffused into the surface, and a new palladium electroplating process using a thin gold overplate, were developed:
Palladium contacts do not sulfidate or oxidize, and so offer extremely low electrical noise levels. They have an electrical life expectancy of approximately 10 times that of fine silver contacts. However, because of relatively poor conductivity properties, load currents are limited to about 5 amperes. Palladium contacts require .006” to .012” overtravel to insure good wiping action. Because of this, they are used primarily on telephone-type relays - that is, relays on which the contact arms are parallel to the length of the coil, and on which such overtravel is easy to obtain. Also, palladium contacts should be bifurcated to help insure circuit continuity on contact closure.
-- Tyco Electronics Corporation Relay Products Application Note [PDF]

The term "extremely low electrical noise levels" is relative to context. Telecommunications relays demand much less than audiophile listening! The "poor conductivity" is also cause for concern in our application, where the active load is only a few (typically 8) Ω. We'll return to this thought right after this...

Brief Note on Switch Soldering

After soldering wires to the terminals of a wafer switch, don't be tempted to use solvents to clean away any flux residue. This applies equally to any other soldered switch and relay types, including relay bases, used in your alternative designs (later designs in this series use mostly crimped and screwed connections, for which this advice is irrelevant). Such a solvent wash will indeed dissolve the resin as advertised, however the resultant dissolved resin can then spread over the switch contact area, depositing a thin, hard, and clearly undesirable film of resin as the solvent evaporates. As Ray Skipp's Workmanship Standards Manual advises, "Much less harm will be done by leaving small amounts of flux residue on the soldered joints."

About Ratings

Due diligence demands we check the operating voltages, currents and powers present in our circuit, and for this we need information about the equipment we're interfacing. On one end of our switch will be found a Yamaha RX-V777 AV Receiver/Amplifier (discontinued), sporting up to 160 watts per satellite or centre channel, into 8Ω loudspeakers. At the other end, a mixture of satellite and centre units from two merged harman/kardon HKTS 16WQ/230 5.1 Channel Home Theatre Speaker Systems (currently unavailable), all of which obligingly present an impedance of 8Ω to the passage of reciprocating electrons. So, the ballpark has these dimensions:-
Vrms = √(P × R) = √(160W × 8Ω) = √1280 V < 36V
Irms = √(P / R) = √(160W / 8Ω) = √20 A < 5A
Actually, these figures should well exceed the voltages and currents produced in practice, possibly by a factor of two or more. This is partly due to manufacturers' traditional systematic exaggeration of the stated output power of their kit. I've taken the quoted wattage to express average power, which is the only figure that can be quoted honestly; anything else is disingenuous and evil! Funnily enough, manufacturers tend to look at this and say well, if that's just the average power, and if the instantaneous power drops to zero on every half-cycle, then the peak music power (a term they invented for purely marketing purposes) must obviously be double that. So they say yeah, double, sounds good. Let's go with that.

Anyway, on the unlikely assumption that the quoted power is the true average power, we do get the above quoted rms voltage and current. In reality, what we probably get is a grotesque overestimate. Which at least errs on the side of caution.

Other than to confirm that arcing won't worry any switch arrangement, the calculated voltage is of little interest. The room rotation switch should only ever be operated while the audio outputs are silent, preferably in fact when the power is off, so when its contacts are switched the instantaneous voltage is zero. That's important; the transient voltages (including inductive kickback) produced by making and breaking contacts, even at comparatively low volumes, can blow or otherwise damage both amplifiers and speakers.

The calculated current value is, on the other hand, of great interest. This is the required switch contact rating in amps, and five is a good target. Unfortunately, and although they perform acceptably at low volume and in the short term, this target proves too high for our Bulgarian rotators.

Don't Throw Out The Baby

Such shortcomings can be addressed in the manifest. Check these 12-way NSF rotary wafers: silver plated brass contacts, with an initial resistance less than 10mΩ, and a continuous current carrying capacity of 5A. Seven of these can be stacked to give us a 7P12T switch, and again we can wire it either helically, to perform our commutator switching function, or any other way, to get an arbitrary set of 12 independent speaker configurations.

These wafers also come in several other pole/contact configurations, e.g. 3P4T on a single wafer, just three of which could provide us with a full (up to nine channel!) 4-direction switch. For the purpose of these prototypes, I'll stick with an eight-direction switch just now.

The standard shaft assembly includes a balanced detent mechanism with a life expectancy of more than 100,000 operations. This component can also be "stopped" in order to reduce our 7P12T design to the original 7P8T logic. Alternatively, we just might be able to find a good use for those extra four switch positions...

Here's what I have in mind. Most 7.1 AV receivers can do a reasonable job of audio upscaling - providing seven outputs (and some bass) from any number of sources, even mono. However, and notwithstanding the soundtrack of the forthcoming Star Wars: The Force Awakens blu-ray, it remains true that mastered, curated, dedicated 7.1 channel content still forms a minority (albeit a growing one) of commercially available home cinema fare today. And audio-only surround sound content in particular tends to be mastered for the 5.x stage.

5.x: Rear speakers lost.
All Quiet In Back

This leads to a problem in our octoroom. When we select "Straight Decode" on the AV receiver, so changing its standing orders away from the direction of "Make it wow!" and more towards "Let me hear exactly what the 5.x mixing engineer intended", then all three rear speakers fall silent, as illustrated by the diagram on the right.

This seems one compromise too far for our octagonal arrangement of speakers. Surely that can't be what the engineer - erm - what's the audio equivalent of envisaged? Didn't Steven Wilson imagine that we could at least be trusted to put the "surround left/right" speakers a little way behind the listener? And what about all that concert hall ambience painstakingly baked into the rear of the mix, it that to stop abruptly at our shoulders now, as though we've just managed to poke our heads through the entrance?

5.x: Rear speakers regained.
The Big Reveal - Mode 5

My solution is to allow an alternative switching pattern for 5.x playback (see left), whereby the surround and surround back speakers are interchanged. Speaker 8 is still silent of course, as we have no provision for a central surround back signal. But speakers 1 and 2 have been swapped, as have 6 and 7. The result is a bit more like a traditional 5.x layout. Let's dub it Mode 5.

Admittedly this is different from the recommended layout of such a system, where the sidelining of the surround sound sources is intentional. However, specialist studio recordings custom mixed into 5.x demand to be reviewed and analysed using the extra depth provided by this arrangement. And incidentally, this arrangement can also be used to facilitate deep analysis of retro quadrophonic mixes, increasingly today released in 5.x format, which were patently designed for playback in 4-corner configurations.

Now back to that 7P12T rotary wafer switch solution, where we found ourselves with an extra unused 4 switch positions on the dial. With a few more bits of old wire, we could make this 5.x-oriented configuration available, allow it to be rotated so that any of the four walls is front-of-stage, and declare victory!

Depending on the detailed physical placement of your speakers, you might also consider moving the signals from speakers 3 and 5 down to the recently vacated 2 and 6, so that those two can now be repopulated with front presence feeds FPL/FPR, if such are available on your receiver. Obviously we'd need a further two circuit poles to carry these signals, making the switch an impressive 9P12T. The standard shaft assembly takes up to 10 standard wafers (spec; actually looks designed for 20), so we're still squeezing in. But the closer the physical speaker placement is to symmetric octagonal, the less appropriate would be such an enhancement.

8-way switch
Prototype Wiring: Two Flavours

On the right is is the wiring diagram for the basic 8-way switch. The seven ganged 8-way wafers are represented by the green rectangles, and the "comb" of wipers is shown horizontally along the top, in default position 1.

In fact, the prototype eventually constructed for this design included the four extended "Mode 5" patterns described earlier. To show the generality and versatility of the rotary wafer implementation of a commutator, I decided arbitrarily to interleave these with their corresponding 7.x counterparts. So, if you're sitting on one of the four main compass point orientations, switching one step clockwise takes you to that same orientation in Mode 5, while one more click takes you to the next vanilla 7.x point, 45° further on. Hopefully the following 12-step cycle notation is self explanatory:
N → N5 → NE → E → E5 → SE → S → S5 → SW → W → W5 → NW
If you remove the end stop from the shaft assembly, one more click takes you full circle back to N. But I retained the stop on the prototype. It lets you rewind to a fixed known position; useful when you don't intend to add any switch state indication, beyond a couple of scrawled crayon marks.
12-way switch

The 8-way wiring was pretty clear: just eight adjacent helices, appearing as diagonal lines on the flat drawing. The schedule for the 12-way is a little more complex. For clarity, label the seven wafers abcdefg, left to right in the diagram, and number the output terminals on each wafer 1-12, top to bottom. Then the wiring schedule for our eight output circuits, in order of speaker number 1-8, comprises these chains:
1 --- b2 - a1 - b12 - c11 - c10 - d9 - e8 - e7 - f6 - f5 - g4
2 --- a3 - b1 - c12 - d11 - d10 - e9 - f7 - g6
3 --- b5 - a4 - b3 - c2 - c1 - d12 - e11 - e10 - f9 - f8 - g7
4 --- a6 - b4 - c3 - d2 - d1 - e12 - f10 - g9
5 --- b8 - a7 - b6 - c5 - c4 - d3 - e2 - e1 - f12 - f11 - g10
6 --- a9 - b7 - c6 - d5 - d4 - e3 - f1 - g12
7 --- b11 - a10 - b9 - c8 - c7 - d6 - e5 - e4 - f3 - f2 - g1
8 --- a12 - b10 - c9 - d8 - d7 - e6 - f4 - g3
There are four physical wire crossovers between wafers a and b. Despite being at different "heights" above the surface of the switch drum, these need insulation. The other six big diagonal crossovers that are apparent in both 8-way and 12-way diagrams are artefacts of the mapping from cylindrical switch body to flat plane drawing; in reality they connect physically adjacent terminals. Still, they probably need insulation too, because they're so near the wiper connections. The rest of the wiring can just use tinned copper 0.6 mm² link.

Seven wafers shafted and
ready for octuple helixing
Pro / Con

As mentioned before, the rotary wafer switch approach preserves the commutator property of having just one point of dry metal-to-metal contact per channel on the switch. However, it also preserves the property that selecting between settings involves each wiper crossing a variable number of contacts, and this causes wear. A possible mitigation to minimize this sliding contact is to scramble the wiring diagram, such that the most popular settings become adjacent, reducing total wiper travel over time. Obviously, this might make future rewiring desirable, whenever a new setting (compass direction) gains popularity.

Bulgin PX0551, PX0956/S, PX0957/S
and PX0959/P. Not shown: PX0552.
Choosing the Connectors

This is the first in a series of interchangeable concrete prototypes, which I'll be wanting to swap in and out of use in the context of a live home cinema setup. The system I'll be using to facilitate quick swaps is based on this range of Bulgin 8-way plugs and sockets. These are rated 6A at mains voltage (250Vac), which is a little hefty for this application, but they're just what I have available.

Slightly more appropriate might have been the Neutrik Speakon, shown below right. This range is actually designed for multi-speaker connections, and does feature a subset of 8-pole connectors, both cable and panel mounting. On the other hand, these are designed specifically to handle four channels, and to take two wires per channel in a standard pin connection pattern. I'm using eight channels, one wire per channel, and ignoring the common connections. All the '-' outputs are joined together, and to all the speaker '-'s in the wall plate, by a single uninterrupted wire. These are already connected in common inside my AV receiver - I've obtained the circuit diagrams from Yamaha to confirm this. I'm reluctant to use the Speakon solution with nonstandard wiring, but using it correctly would double the number of plugs and sockets involved. It's a first world hard choice.

Neutrik Speakon connectors.
That reminds me, did I tell you not to use any of my prototypes in your car surround sound system? Yeah, those amps tend not to have a speaker common, but drive both ends of the coil in opposition to get the max oomph out of your 12V battery. Don't go commoning up them speakers Cletus, it'll go badly for you.

I'm not going to worry about scare stories of induced currents on unbalanced wires - electrical suppression is of high quality nowadays, and anyway these are relatively high current, low impedance connections.

Making the Connections

Whatever connector family is used, we just need to interrupt the positive amplifier output feeds and insert a cable mount plug/socket pair, with the socket on the amplifier end and the plug going to the speakers. The cable mount adapters are the upper pair in the Bulgin picture. I'm connecting audio outputs 1 through 7 to the corresponding pin numbers on the socket, and the audio amplifier common output to pin 8, just in case the particular switch prototype needs a reference to that voltage level.

If there's no prototype switch in situ, we just connect plug to socket, and everything's as it was last week. But then, each prototype will be built into an ABS project box, fitted with panel mount equivalents of the plug and socket. The panel mount adapters are the lower pair in the Bulgin picture. Take care to connect the panel mount plug to the input (amplifier) side of the prototype switch, and the panel mount socket to the output (loudspeaker) side. Now to use a particular switch prototype, just separate the continuity plug/socket pair, and connect them to the switch box. These adapter polarities will ensure the connections always go the right way.

Lumberg KV81, SV81, SFV80, KFV80.
Once the prototyping's done and I've settled on a final design for the switch, I might convert to 8-pin DIN EN 60529 IP40 connectors (left). These have brass contacts plated gold over silver, resistance < 5mΩ, rated 5A at 60Vac. They should scrub up nicely, albeit fiddly to solder given the girth of my speaker wires. At least the cable mounted plug & socket are available with a 6mm entry option (standard is 3mm), which should help a bit.

I say might convert to these, because I've a doubt about their mutual compatibility. As seen in this illustration from Wikipedia (right), there are two different pin geometries for 8-pin DIN plugs and sockets. And while RS Components assures that these four Lumbergs play nicely together, close inspection of the photos fails to convince. Update: I bought a set of four from RS Components, and it turns out that the cable mount plug and socket use the left pattern, while the panel mount ones use the right. So, yeah, no; they need to be either all 80s, or all 81s, a fact which had been 404'd out of their web catalog.

Post Script

While installing the Bulgin connectors, it occurs to me - not for the first time! - that a similar plug and socket, but with eight poles arranged in a circle (instead of seven with a central one), and with all of its keying Dremeled away, could obviate the switch requirement entirely. Just pull them apart, twist, and rejoin for a new sound stage orientation. But where's the fun in that?

Next time we'll start examining an alternative approach, using toggle switches, relays, and permutation group theory.

No comments:

Post a Comment