I want to build an amp selector (two amps a/b, one cab)

[Gtr0]

Still have this on the back burner.... but still want to do it, so I have been looking around in spare time for the correct relays that are available.

How about these ones??
Finder F36119-009A

10A 9V SPDT

I looked up a datasheet and couldn't find the switching type - though I may have missed it, my daughter woke me up early to play Minecraft so I am a bit hazy still.

[HaroldBrooks]

The main thing to be careful about when switching speakers is to avoid intermittent load, which can cause voltage spikes and destroy the output transformer.

Even if the unused amp is switched to a dummy load, the relays will bounce a bit when switching. If there is no other protection, this would be very risky.

A simple protection is to put a resistor across the amp speaker out signal (e.g. 270 ohms/5W) that is always in the circuit. This is enough load to protect the amp, without affecting the tone.

The safety resistor can be added to the amp (I use it on all my builds), or part of the switch box (on the speaker signal input from each amp, before the relay).

I would also recommend adding a circuit to mute the input of the unused amp.

Also I believe an 625MOV 200J Varistor will sufficently quelch any big spikes to the output transformer, and eliminate any worries, including a total speaker meltdown and open condition. I was told by another engineer at Digikey that transformers should be safe well above a 625v clamping voltage for short spikes, and above the clamping voltage there is an effective loop to ground with virtually no resistance, saving the transformer.

Here's the TDK ones I use as per Rob Robinette's recommendation. I also called Mouser directly to see if the below was good for the intended applicaiton, and they seconded the selection based on typical amp voltages, and capacity for current absorption.

https://www.mouser.com/ProductDetail...3Xfj0G66%2Fw==

Easy to install between the center tap and each tube plate for a push pull amp, and gives piece of mind, not perfect but better.

[R.G.]

I missed the first part of this one.

You need to not leave the amps unloaded. Some, but not all, tube amps oscillate and destroy themselves with no speaker load. You can fix this issue by using a "safety resistor" on the amp itself and a dummy load resistor to eat the excess power is the unused amp is ever accidentally fed signal somehow. The safety resistor sustains the non-oscillation during any tiny not-connected periods while output switches are open.

Unless you've done some fancy grounding and/or signal isolation work, you need to leave the amp grounds NOT connected, especially through the speaker "ground". Speaker ground may not be ground on all amps, although it is on most. The easier solution is to arrange the grounds not to be connected. This means you need to switch both signal and return on the speaker output. You'll still have the common signal ground problem to solve with the inputs, but that's a low current version of the same problem.

Then you need relays. The classical solution is electromagnetic hard switches, which introduce the problem of bounce. The heretical new wave way is with MOSFET-output solid state relays. MOSFETs "saturate" to a resistance, and you can have sub-milliohm MOSFET resistances if you want to pay for them. What's probably going through your mind right now is "but wait - signal going through MOSFETs to speakers will ruin the tone".

That's a solid maybe. Bipolar transistors are fancy variations of diodes. MOSFETs are modulated resistors. They go into pinchoff, but when "saturated" with a big gate voltage, they are just the channel resistance. There is some nonlinearity of the channel resistance, but it can only have an effect to the extent that the change in resistance is significant compared to the resistors/impedance it's connected to. To be audible, this is the speaker. With a channel resistance in milliohms, a variation in channel resistance in the tenths of milliohms (this is downright pessimistic - it's more like tens of micro-ohms), any MOSFET distortion will be 60-70db down, probably more. I've tried this. By actual testing, I can't hear the difference between hard-contact and MOSFET relays on speakers. Your ears may be better, but it might be worth a try.

As for sequencing, relay timing and bounce for EM relays will always be a problem. MOSFET relays, less so. They don't bounce at all unless their control signal bounces. They switch in microseconds after their control signal does.

One interesting approach to a pro setup is to use both MOSFET and EM relays in parallel. It will need some care in timing, but the MOSFET makes and un-makes on a sub-millisecond scale. Hard relays make and break in ten of milliseconds or more and commonly bounce for 20mS to 100mS. Do the controls with a "make" to the MOSFETs and EM contacts at the same time. The MOSFET makes, conducts audio, and then some time later the EM relay makes and bounces for a fraction of a second. The MOSFET "shorts" the open condition of the bounce, and the hard relay "shorts" the constant signal of the MOSFET, removing any possibility of MOSFET signal poisoning. For break, you open the hard relay, wait until its laggardly masses have moved, then drop the MOSFET control signal and it does a clean break.

None of this is particularly susceptible to typical "um... use a diode and a ... ah... cap to delay the.. um" kind of circuits work, unfortunately.

[vintagekiki]

Late vk come to thread. Chuck H was faster than me.

R = 8 Ohm / 100 - 200W depends on the power output from the amplifier. Resistor is desirable cooling by PC fan.
The relay contacts must withstand the rated current from the audio amplifier.
For 8 Ohm / 200W this is minimum 5 A. It is best contacts for 10 A, to prevent any possible spark when switching.
The relay must be for strong mounting. (Not in socket, not in PCB)

Voltage for the relay perform through an external 12 VDC adapter.
To avoid transient occurrences when guitar switching, inputs amplifiers connect in series (from amp1 - to amp 2)

[SoulFetish]

I missed the first part of this one.

You need to not leave the amps unloaded. Some, but not all, tube amps oscillate and destroy themselves with no speaker load. You can fix this issue by using a "safety resistor" on the amp itself and a dummy load resistor to eat the excess power is the unused amp is ever accidentally fed signal somehow. The safety resistor sustains the non-oscillation during any tiny not-connected periods while output switches are open.

Unless you've done some fancy grounding and/or signal isolation work, you need to leave the amp grounds NOT connected, especially through the speaker "ground". Speaker ground may not be ground on all amps, although it is on most. The easier solution is to arrange the grounds not to be connected. This means you need to switch both signal and return on the speaker output. You'll still have the common signal ground problem to solve with the inputs, but that's a low current version of the same problem.

Then you need relays. The classical solution is electromagnetic hard switches, which introduce the problem of bounce. The heretical new wave way is with MOSFET-output solid state relays. MOSFETs "saturate" to a resistance, and you can have sub-milliohm MOSFET resistances if you want to pay for them. What's probably going through your mind right now is "but wait - signal going through MOSFETs to speakers will ruin the tone".

That's a solid maybe. Bipolar transistors are fancy variations of diodes. MOSFETs are modulated resistors. They go into pinchoff, but when "saturated" with a big gate voltage, they are just the channel resistance. There is some nonlinearity of the channel resistance, but it can only have an effect to the extent that the change in resistance is significant compared to the resistors/impedance it's connected to. To be audible, this is the speaker. With a channel resistance in milliohms, a variation in channel resistance in the tenths of milliohms (this is downright pessimistic - it's more like tens of micro-ohms), any MOSFET distortion will be 60-70db down, probably more. I've tried this. By actual testing, I can't hear the difference between hard-contact and MOSFET relays on speakers. Your ears may be better, but it might be worth a try.

As for sequencing, relay timing and bounce for EM relays will always be a problem. MOSFET relays, less so. They don't bounce at all unless their control signal bounces. They switch in microseconds after their control signal does.

One interesting approach to a pro setup is to use both MOSFET and EM relays in parallel. It will need some care in timing, but the MOSFET makes and un-makes on a sub-millisecond scale. Hard relays make and break in ten of milliseconds or more and commonly bounce for 20mS to 100mS. Do the controls with a "make" to the MOSFETs and EM contacts at the same time. The MOSFET makes, conducts audio, and then some time later the EM relay makes and bounces for a fraction of a second. The MOSFET "shorts" the open condition of the bounce, and the hard relay "shorts" the constant signal of the MOSFET, removing any possibility of MOSFET signal poisoning. For break, you open the hard relay, wait until its laggardly masses have moved, then drop the MOSFET control signal and it does a clean break.

None of this is particularly susceptible to typical "um... use a diode and a ... ah... cap to delay the.. um" kind of circuits work, unfortunately.

I've been thinking about these as well, and I like the idea of SSRs for all the benefits you mentioned.
The downside is because they would be switching reactive loads, they would need to be protected from Back EMF when switching, or they could be destroyed by overvoltages. In addition to switching, transients could be present during normal operation when the output stage is overdriven (I'm not sure how sensitive SSRs might be in this case, I'm only going on material I'm reading) In addition, any short at the output terminals can damage it and cause it to fail.
Not a deal breaker by any means, but it does add to the parts cost and complexity. There are plenty of protections schemes offered in applications notes, and I would think some of the examples for switching motor loads would be a good place get some ideas. That's probably where I would start, anyway.
RG, in the examples you mentioned having experimented in the past, were you using SSRs switching speaker loads? or using mosfets for protection in line between the speaker and amp?

...Then again, I suppose I'm taking for granted that we would be adding some kind of clamp on the coil of a relay to protect the driver as well.

[Chuck H]

Ok... Only vaguely addressed so far is that there's some proximity considerations WRT handling the input switching and the output switching with the same device. I think it should be obvious that having the inputs and outputs managed in the same enclosure would come with instability risks for the same reason we manage layouts in amp chassis to keep both ends of the amp separated as much as practical. You'd probably be best off to have two enclosures, localized with their respective switching functions, and then the foot switch should only power the relays in each switcher. That said...

It's getting more complicated and I can't help thinking that A/B switches are readily available and the circuit shown by myself and vintagekiki could be operated with a Carling snap switch. This would require pressing a whole two buttons though

EDIT: Another consideration might be to use the snap switch to manage the output switching function AND trigger power to a relayed input switcher. That gets us down to one button again, keeps the two circuits remote, removes the stringency of high power relays and avoids any potential for the amplifier currents to hinky any relays (as noted by SoulFetish) if that's a real concern.

I don't know if such a snap switch is available though. It would need to be a 3PDT foot button. Otherwise one might need to fabricate a toggle within a self supporting treadle. Still doable.

[R.G.]

The downside is because they would be switching reactive loads, they would need to be protected from Back EMF when switching, or they could be destroyed by overvoltages. In addition to switching, transients could be present during normal operation when the output stage is overdriven (I'm not sure how sensitive SSRs might be in this case, I'm only going on material I'm reading) In addition, any short at the output terminals can damage it and cause it to fail.
Not a deal breaker by any means, but it does add to the parts cost and complexity. There are plenty of protections schemes offered in applications notes, and I would think some of the examples for switching motor loads would be a good place get some ideas. That's probably where I would start, anyway.

Anything that can switch an AC line motor load probably won't even get its hair mussed up with speaker loads from nominal tube amps. It ought to be fairly easy to protect this. You're right about looking at the schemes for SSRs and motors as a start.

RG, in the examples you mentioned having experimented in the past, were you using SSRs switching speaker loads? or using mosfets for protection in line between the speaker and amp?

...Then again, I suppose I'm taking for granted that we would be adding some kind of clamp on the coil of a relay to protect the driver as well.

I first tried it with signal voltages, and it worked so well that I tinkered with speaker loads. This was just my tinkering, and it's not by any means exhaustive - I didn't go hunt down a Marshall Major for testing or anything. But mostly it just worked, which is always both a good sign and cause for suspicion (Dang! it just worked! ... Wait a minute - that was toooo easy. What does that mean?)

I used an off-the-shelf MOSFET output SSR rated for 240Vac. I wouldn't use the same device in a real setup as it was too expensive (~$15 each) but the principles are portable. If I were doing a similar design as a working unit, I'd probably pick my own MOSFETs and drivers. An AC switch with MOSFETs is just two MOSFETs connected source to source to get the body diodes opposed and some kind of floating gate drive to turn them both on at the same time. A dual LED to photovoltaic works fine, even though it slows the MOSFET switching down a lot.

Ok... Only vaguely addressed so far is that there's some proximity considerations WRT handling the input switching and the output switching with the same device. I think it should be obvious that having the inputs and outputs managed in the same enclosure would come with instability risks for the same reason we manage layouts in amp chassis to keep both ends of the amp separated as much as practical. You'd probably be best off to have two enclosures, localized with their respective switching functions, and then the foot switch should only power the relays in each switcher. That said...

It's getting more complicated and I can't help thinking that A/B switches are readily available and the circuit shown by myself and vintagekiki could be operated with a Carling snap switch. This would require pressing a whole two buttons though

EDIT: Another consideration might be to use the snap switch to manage the output switching function AND trigger power to a relayed input switcher. That gets us down to one button again, keeps the two circuits remote, removes the stringency of high power relays and avoids any potential for the amplifier currents to hinky any relays (as noted by SoulFetish) if that's a real concern.

I don't know if such a snap switch is available though. It would need to be a 3PDT foot button. Otherwise one might need to fabricate a toggle within a self supporting treadle. Still doable.

Good thought. It would make sense if only for the cabling to make this all connect up well to have it be two units. An input unit to switch inputs, an output unit to switch outputs, and a signal cable to make them both switch at the same time.

The solder-fumes-and-wires version is to run the coils of the relays in the two boxes in series. The modern signaling version is to run the same logic signal to both units. The hyper-fancy version is to put a microcontroller in each unit that understands MIDI.

[vintagekiki]

Good thought. It would make sense if only for the cabling to make this all connect up well to have it be two units. An input unit to switch inputs, an output unit to switch outputs, and a signal cable to make them both switch at the same time.
The solder-fumes-and-wires version is to run the coils of the relays in the two boxes in series.

No two boxes. Everything is in one box, located near the cab.
In the box is only one relay and four female 6.3 mm jack.
Switching the amplifier (amp1 / amp2) is by with ordinary foot sw. Voltage for relay is any 12 VDC cheap adapter.

No switch inputs. Guitar connect to amplifier so inputs from amplifiers connect in series (from amp1 - to amp 2) or connect via guitar Y-cable.

Everything is simple and efficient without processors and microcontrollers, and most importantly it works reliably.

[Chuck H]

No switch inputs. Guitar connect to amplifier so inputs from amplifiers connect in series (from amp1 - to amp 2)...

??? I'm having trouble picturing how to plug two amplifier inputs in series.

...or connect via guitar Y-cable.

Ick, for the added capacitance on the guitar signal. And I don't think it's best to have continuous amplifier operation when switching to a dummy load. We sure don't want two amplifier outputs connected together, so the switches have to be non shorting. There's a moment of no load. In that circumstance it would be best to minimize any possibility of drive at the amps input.

[vintagekiki]

??? I'm having trouble picturing how to plug two amplifier inputs in series.

Correction
I'm wrong expressed. Amplifiers are connected in parallel, not in series.

Ick, for the added capacitance on the guitar signal. And I don't think it's best to have continuous amplifier operation when switching to a dummy load. We sure don't want two amplifier outputs connected together, so the switches have to be non shorting. There's a moment of no load. In that circumstance it would be best to minimize any possibility of drive at the amps input.

The Y cable is very short, so not capacitively affected on to amplifier
Ultimately, the guitar can be switched with a relay. Schematics follows soon.¹⁾

1)

[R.G.]

Two boxes may make more sense than one. It depends entirely on what the person can deal with in terms of running cables and not getting speaker cables and coaxial shielded cables mixed up. It is possible to get into difficult situations routing high impedance input cables through a box that contains speaker cables as well.

Switching with microcontrollers can be just as reliable as hard switches - if you know what you're doing.

[R.G.]

Correction
I'm wrong expressed. Amplifiers are connected in parallel, not in series.
[...]
The Y cable is very short, so not capacitively affected on to amplifier
Ultimately, the guitar can be switched with a relay.

This is the issue I was referring to. If you use a Y cable to connect the inputs of two amps in parallel, you guarantee that any issues with ground leakages in the two amps will cause hum problems if the grounding isn't immaculate on both amps. On setups with two-wire amps (no third-wire safety ground) hum is almost guaranteed.

Third wire grounding of both amps is really, really helpful.

Y cables on inputs is its own separate problem, which complicates amp and cab switching as well.

[Helmholtz]

The Y cable is very short, so not capacitively affected on to amplifier

It is often overlooked that amplifiers not only have input resistance but also input capacitance. A typical value for tube amp inputs is around 150pF.
So paralleling amp inputs will not only load the guitar by the parallel input resistances but also by the adding input capacitances.

Being an experimental physicist I am always keen to verify theory by experiment. So to see the influence of the Miller effect, I measured the input capacitance of my 80s Super Champ with the amp turned off and on. Results: Coff = 30pF, Con = 130pF. The difference of 100pF can be attributed to the Miller effect.

[Chuck H]

It is often overlooked that amplifiers not only have input resistance but also input capacitance. A typical value for tube amp inputs is around 150pF.
So paralleling amp inputs will not only load the guitar by the parallel input resistances but also by the adding input capacitances.

Being an experimental physicist I am always keen to verify theory by experiment. So to see the influence of the Miller effect, I measured the input capacitance of my 80s Super Champ with the amp turned off and on. Results: Coff = 30pF, Con = 130pF. The difference of 100pF can be attributed to the Miller effect.

So really it's a small capacitance that would be easily countered by using shorter cables or some fancy low capacitance cables. The Y connector has some capacitance of it's own too. If the player typically uses a 10' cord then it's going to be tricky. But if the player typically uses a standard 20' cord then three 10' $$$ low capacitance cables would probably be alright.

And if it's YOU then there's no trouble at all because you use something like a 2000pf cable, right?

Thanks for the real world investigation on the matter

[Helmholtz]

And if it's YOU then there's no trouble at all because you use something like a 2000pf cable, right?

No, for me the optimal cable capacitance with my vintage type PUs is 1000pF +/- 10% . Not mainstream, I know.

My point was that paralleling amp inputs will add extra capacitance, like 200pF/amp including short extra cable. So 4 amps in parallel would add some 600pF and reduce effective input resistance to 250k.
But with a buffered pedal this won't matter.

Paralleling inputs may cause ground loop problems - even more so if additionally the amp output grounds are connected.