Switching speakers without relays
Mike Green
Posts: 23,101
I have an application where I could use a reed relay to switch an audio source (1-10W) to one of several speakers, but I'd rather use something solid-state that could be switched by a Stamp output. The audio amplifier is a TDA2002 with a 2200uF capacitor in series with the output driving the speaker which is grounded. The signal at this point is AC in relation to ground and the load is very low impedance (<= 8 ohms). Any suggestions?
Comments
I suppose someone has this all available in an IC as a solidstate audio switch.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
"When all think alike, no one is thinking very much.' - Walter Lippmann (1889-1974)
······································································ Warm regards,····· G. Herzog [noparse][[/noparse]·黃鶴 ]·in Taiwan
Goto www.digikey.com type: audio switch
into the search box, click Integrated circuits, Switches (10 items) to see the chips digikey has.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
·1+1=10
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Chris Savage
Parallax Tech Support
csavage@parallax.com
I have two amplifiers with a maximum output of about 10 watts that use a single chip to amplify a headphone output (TDA2002). These have a commonly used configuration with a large capacitor in series with the output of the amplifier since the final stage is biased around 1/2 the supply voltage. The net result is AC at the speaker with a maximum amplitude of twice the supply voltage. The amplifiers are available from KitsRUs.com as their DIY47. In my application, they run off 6-12VDC from a regulated "wall wort". I also have 6 speakers rated for 10-20 watts. The speakers and the amplifiers use a common ground and I'm switching the "hot" side currently using small relays controlled by a Prop-1 at a time when the volume is quiet. For future projects, I'm more interested in eliminating the "clack" of the relays and avoiding the need for the Darlingon outputs than in not generating any clicks or pops in the speakers. My sound sources are conversational voices so, although there may be peaks with sibilants, there's not much average power used.
Since you aren't looking for audiophile quality, consider using paralled N & P type power fets in series with common lead of ea. speaker. Putting them in the high side will require more attention paid to biasing. The gates need to be driven with opposite polarity levels. With both fets turned on they will have an effective resistance in the tens of milliohms for most low cost power fets. Haven't tried this, but it should work. You might even be able to get away with using 1 fet, since fets are resistive devices and not real polarity sensitive in apps like this.
Rick
As you are likely to spend about the same $ for building up a switching circuit for the speaker level signal, why not switch at the low level (-10db) before the amplifier and build up an amp for each speaker? TDA2002 are about $1.70 in the quantity you need. You can then switch the low level audio with CMOS switches (read multiplexers) such as a pair of CD4052, one for each channel, addressed in parallel by the STAMP. You would need to bias the switch inputs with perhaps a small DC bias to keep the signal swing from trying to go below ground. A pair of CD4051 would give you stereo to 8 amplifiers, a 2x8 switching matrix.
If you then wanted to expand to have more sources to the 8 stereo destinations, you should investigate using a pair of 8x8 analog matrix switches. You could get by on just one 8x8 matrix switch IC and still use 4 stereo sources (or 2 stereo and 4 mono or some other combination) to 8 amplifiers and speakers, configured as singles or pairs as you desire. BTW, this IC will also pass RS170A video with proper buffering in and out if you desire to switch video the same way from 8 sources to 8 destinations. Smaller matrix sizes are also available. Find the matrix switch ICs here: http://www.maxim-ic.com/quick_view2.cfm/qv_pk/1931/ln/ and here: http://www.maxim-ic.com/quick_view2.cfm/qv_pk/1974/ln/. I hope this helps.
Rick B (not RickB) just coincidence that we are both replying to you.
It's like Kramer's idea, but compatible with bipolar stuff.
-dave
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
This is not a sig. This is a duck. Quack.
·· I failed to mention this due to my lack of testing to see what effects this would have on an audio signal.· I wouldn't think you would get a linear signal back out, not to mention the possible effects of zero-crossing/bias.· Perhaps I am way off here, but that isn't something I have ever tried, just wouldn't have thought of for the above mentioned reasons.· Perhaps I misunderstand your implementation.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Chris Savage
Parallax Tech Support
csavage@parallax.com
The trick works because of the way that FETs conduct. I dug it out from an old copy of Horowitz & Hill, which has the deep dark dirty reasoning why it works, but the short and mostly correct version is that when you nail a FET into the conduction state, it ceases to care what direction the current flows in, and the back-to-back topology prevents the body diodes from conducting when the pair should remain off (the normal thing people think of when someone uses a FET for bipolar control). There's a little loss because of the series resistance of the FETs, but assuming they're relatively well matched and that you raise the gates to a level sufficiently high that they're ON, not just on, the probelm goes away for signals of sufficient magnitude to swamp the series resistance. Once I get the layout program un-b0oked on my PC, I'll post a picture of the business end of the dimming module schematic I built like this.
-dave
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
This is not a sig. This is a duck. Quack.
·· Thanks for the explanation.· Perhaps the OP could make use of this.· Myself, I will probably stick with a mechanical relay.· It won't be switching so much that it will be a problem.· But it is nice to know that you can use the FETs in this manner.· I just figured it would be more involved with support circuitry.· But I have only used FETs in automotive applications for switching 12V lights, so I am not the foremost authority on them.·
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Chris Savage
Parallax Tech Support
csavage@parallax.com
Attached are snippits of the schematic and PCB for the dimmer I mentioned. Patent pending stuff has been obfuscated to protect my own tail.
-dave
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
This is not a sig. This is a duck. Quack.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Chris Savage
Parallax Tech Support
csavage@parallax.com
To the OP (Mike Green), this isn't something you'll ever run into switching speaker loads. I'm working with some pretty crazy stuff to establish the upper performance limits of the system.
The one thing I did forget to mention is that the FETs need to be switched with a voltage on their gates relative to the common sources. In the schematic, you'll see two lines running down the center of the pais of FETs. One is the common (to all parts) gate drive. The other is what I called FETGND, and it's the floating ground that's used as the reference for the gate drive. If you try to ground that center point and drive the gates from logic ground, you'll create a dead short to ground, and probably blow things up. It's required because the turn-on volatge for a FET is measured between gate and source (Vgs), thus we connect all the sources to a single, virtual ground. I'd suggest floating the circuit with a secondary power supply, and making FETGND real ground. Just make sure to use optocouplers for the switch interface to whatever is controlling it, to keep from shorting anything.
-dave
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
This is not a sig. This is a duck. Quack.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
·1+1=10
That's correct, but the levels of flyback required to destroy the IRFP360s is much higher than the pulse generated by the pair of 12" overhung woofers I have. To put it in perspective, the shrapnel I referred to was the result of a load dump disconnect test for an 8kW 3ph purely inductive load. Yes, it was a torture test, and yes, I will be designing in load-dump protection in the next board rev.
-dave
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
This is not a sig. This is a duck. Quack.
Building LARGE tesla coils? .... Just kidding
What precision are the resistors you are using?
Is there any possible way that your "cratered plastic packages" are a result of an imbalanced gate drive?
In a way, thermal runaway?
i.e. If one FET get's turned "on" slightly before the others, then THAT FET experiences the brunt of the demand.
This can cause a chain reaction with all the other FETs.
What would happen if you put all of the FETs in parallel using only 1 resistor per 8 FETs?
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Beau Schwabe
IC Layout Engineer
Parallax, Inc.
I narrowed it down to the complete and total meltdown of the protection module on the load, a COTS unit that the customer specified against our engineers requests. It's potted, so when it went, it went quietly, but shorted around the load, dumping the three phases together into the neutral and making a big mess. I'll see if I can take a picture of the puddle of silicon that formed where the die used to be in one of the parts that didn't vaporize.
In other news, I have a really near Si and Cu sputtered plexi blast shield. I plan on hanging it on the wall of the shop as a reminder.
The gate 'resistors' are left over from an earlier version of the driver circuitry that wasn't capable of directly driving the Qc of the gates at the speed we needed. The way around that is under the missing pixels in the PCB screencap
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
This is not a sig. This is a duck. Quack.
Dave, You're way out of my league. I get really careful anytime I play around with more than a couple of amps at 12-14V.
Mike
I used to have a Vne (Voltage to Never Exceed) of about 12V too. Then I built a 30W amplifier. Then I built a 100W amplifier. Then I built a 1kW light controller. Then I got my degree. Then...
You can see how it snowballs. The reed relays may not be the best solution unfortunately, unless you can mute the amplifier for a half-second or so before you switch the coils. The reason has to do with both the risk of switching a speaker in the middle of a transient (generating a vertical voltage transition) and cooking a tweeter or welding the contacts of the relay. I know it seems unlikely, but I'd use beefier relays if you can. What are your supply rails for the amp? You'd be surprised how much current can flow at the top end of a kick drum thud or a good snare drum thwack.
-dave
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
This is not a sig. This is a duck. Quack.
I'm controlling the output of two MP3 players. I only switch the speakers attached to a channel when the current track has finished playing. The only audio going through the amplifier on that channel is the noise of the idle MP3 processor (essentially inaudible). The amp runs off 6V for now, but I'm going to increase that to 12V since we need a bit more volume without distortion sometimes. Usually the 6V is more than adequate in a quiet room, but, at the opening, there were too many people talking to hear the audio installation piece. There'll be a closing lecture in a couple of weeks and we expect a lot of people again.
Mike
At those power levels, you can probably get away with some 500mA reed relays and not have a problem. Just remember, 12V across 8 ohms gives you a max. current of 1.5A. I'm not saying you'll ever get to that level, but it is possible, especially with higher bit-rate MP3s that (can) have more dynamic range. 4 ohms and 12V gives you 3A peak, which is worthy of some better relays. I could go on and on about peak power requirements for amplifiers and headroom values, since I'm a little crazy about audio (just ask Chris Savage
-dave
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
This is not a sig. This is a duck. Quack.
·· I would say Dave Paton is a little more serious about his audio than even I am, and I spent years building custom amplifiers in a business...Ah the good ol' days...·
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Chris Savage
Parallax Tech Support
csavage@parallax.com