PAM8403 might well give the cheapest solution : ??????? 2*3W PAM8403??? 2.5?5V?USB?? ????-???
100 of these costs 72RMB (just a whisker over $10).
100 of these costs 72RMB (just a whisker over $10).
PAM8403DR-H Diodes Incorporated | Integrierte Schaltungen (ICs) | DigiKey
bare chip:
1 pcs. 0,72 Euro
500 pcs : 0,385 Euro
1000pcs : 0,308 Euro
bare chip:
1 pcs. 0,72 Euro
500 pcs : 0,385 Euro
1000pcs : 0,308 Euro
Yes, there are 200 discrete audio channels, as fed by a custom-made 200-channel DSP engine. I’m helping the software engineer suss out the hardware side, but I myself know little about either. 
There will actually be, all total, around 600 discrete channels / speakers. So I plan on building three of these 200-channel amps.
Can anyone provide a bit more detail about how these amps would be built?
Thank you!
There will actually be, all total, around 600 discrete channels / speakers. So I plan on building three of these 200-channel amps.
Can anyone provide a bit more detail about how these amps would be built?
Thank you!
1W, but what impedance loads? 8 ohm? What peak voltage is needed (or crest factor)?
I tend to agree class D could be a disaster of interfering oscillators.
Opamp + current buffer seems a plausible scalable low cost approach, especially as you get dual opamps and
NPN + PNP dual transistors in SOIC8 or smaller.
If the output can be PWM already then you just need the MOSFETs though...
I tend to agree class D could be a disaster of interfering oscillators.
Opamp + current buffer seems a plausible scalable low cost approach, especially as you get dual opamps and
NPN + PNP dual transistors in SOIC8 or smaller.
If the output can be PWM already then you just need the MOSFETs though...
Last edited:
If you need the efficiency of class D, you should look at one of the devices that uses a spread-spectrum clocking scheme. You should avoid devices that use a fixed clock frequency, or else you can run into issues from interference between chips and lots of RFI from the wiring going to your transducers. All of the Analog Devices class D amps use a spread-spectrum clock, and TI has a number of newer devices, also.
I used the SSM2518 and SSM3302 amps in my 22-amp boards, and I had no problems with noise. Here is an ADI app note for an 8-channel sound bar amp.
For a second-generation version of the line array amp, I made a very small 4-channel amp board with 2 SSM3582 chips--it's 1.8" by 1.4". The board has a 14-pin connector which will plug into a motherboard, and there shouldn't be any limit on the number of channels. I've got one of the boards partially assembled, but I'm waiting on a motherboard to test the amps (the boards were shipped from China today). You could make the daughter boards even smaller if you used the SSM2518 at 5V instead of the SSM3582 at 12V (smaller chip, plus less decoupling, less ferrite beads on the outputs).
I used the SSM2518 and SSM3302 amps in my 22-amp boards, and I had no problems with noise. Here is an ADI app note for an 8-channel sound bar amp.
For a second-generation version of the line array amp, I made a very small 4-channel amp board with 2 SSM3582 chips--it's 1.8" by 1.4". The board has a 14-pin connector which will plug into a motherboard, and there shouldn't be any limit on the number of channels. I've got one of the boards partially assembled, but I'm waiting on a motherboard to test the amps (the boards were shipped from China today). You could make the daughter boards even smaller if you used the SSM2518 at 5V instead of the SSM3582 at 12V (smaller chip, plus less decoupling, less ferrite beads on the outputs).
200 ‘small’ amplifiers will necessarily be spread out over a fairly large area. Heat sinking will not likely be a big concern, so trying to squeeze efficiency out of it (which could backfire) probably shouldn’t be high on the list of priorities. A whole $-load of TO-220’s in free air would likely do with standard class B efficiency - with any one transistor dissipating less than a watt.
Keeping the power of each element low, and having a whole lot of them means low voltage and a lot of current overall. I wouldn’t try to run them all on a single “power supply”. I would distribute it in banks where the current demand on each is several to maybe 10 amps, keeping the conductors and PCB traces manageable. You may also consider higher impedance drivers than the typical 4 or 8 ohm if you can get them.
Keeping the power of each element low, and having a whole lot of them means low voltage and a lot of current overall. I wouldn’t try to run them all on a single “power supply”. I would distribute it in banks where the current demand on each is several to maybe 10 amps, keeping the conductors and PCB traces manageable. You may also consider higher impedance drivers than the typical 4 or 8 ohm if you can get them.
Anything will be multiplied by 600 so ultra-simple is the key here.
So avoid Discrete; even more Class D.
Best is something like LM386 which requires no external components (except, maybe, a Zobel network), just an output capacitor because it´s single supply and no heatsinking because it´s plain DIP 8 package.
And can just supply 1W if properly fed.
Research the modern equivalent, a bridge out chipamp (so no output cap but keeping single supply) , internally set gain, no Zobel need, such as usen in countless MP3 players and PC speakers.
Remember that adding just 1 resistor or capacitor means adding 600 of each.
Ground and supply rails may be made out of copper strips or aluminum extrusions and part of the "sculpture" side of the project.
Just keeping 600 different signals separate but reaching designated amplifier will be a challenge.
Speaker wiring can be simplified by keeping speaker and amp side by side or nearby.
"In my days" I did build distributed audio systems (PA + background music) with lots of then-new TDA2003, one per speaker or speaker pair, cigarette box sized amps mounted to speaker frames.
Same program to all of them, of course, although I used "area volume controls", lots of 10k pots mounted on small aluminum plates, one at each room or section.
Worked like a charm.
So avoid Discrete; even more Class D.
Best is something like LM386 which requires no external components (except, maybe, a Zobel network), just an output capacitor because it´s single supply and no heatsinking because it´s plain DIP 8 package.
And can just supply 1W if properly fed.
Research the modern equivalent, a bridge out chipamp (so no output cap but keeping single supply) , internally set gain, no Zobel need, such as usen in countless MP3 players and PC speakers.
Remember that adding just 1 resistor or capacitor means adding 600 of each.
Ground and supply rails may be made out of copper strips or aluminum extrusions and part of the "sculpture" side of the project.
Just keeping 600 different signals separate but reaching designated amplifier will be a challenge.
Speaker wiring can be simplified by keeping speaker and amp side by side or nearby.
"In my days" I did build distributed audio systems (PA + background music) with lots of then-new TDA2003, one per speaker or speaker pair, cigarette box sized amps mounted to speaker frames.
Same program to all of them, of course, although I used "area volume controls", lots of 10k pots mounted on small aluminum plates, one at each room or section.
Worked like a charm.
Last edited:
I would be strongly tempted to design a circuit board which contains a long narrow strip of LM386 amplifiers, say 20 of them, and then reproduce it 10 times, and install in a rack mount case with a simple linear power supply.
RCA jacks could get cumbersome for inputs, but would also be the easiest and cheapest. For outputs, I'd probably go with a phoenix contact type terminal strip.
The idea of locating the amplifiers directly on the speakers, and feeding them DC power from a common power supply, and signal voltage is also valid - in this case, just order a whole lot of the cheap Chinese LM386 amplifier boards from eBay. You can't really beat the pricing.
RCA jacks could get cumbersome for inputs, but would also be the easiest and cheapest. For outputs, I'd probably go with a phoenix contact type terminal strip.
The idea of locating the amplifiers directly on the speakers, and feeding them DC power from a common power supply, and signal voltage is also valid - in this case, just order a whole lot of the cheap Chinese LM386 amplifier boards from eBay. You can't really beat the pricing.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.