While my speaker project is ramping up I thought I would start a parallel task of amplification. I need to come up with about 20 channels of identical circuits, using supplies to control max Pout at 50-100rms. These will not be tube amps, my floor couldn't withstand the weight of the output transformers, never mind my wallet and the accompanying issues with that.
So lets keep the costs down, the bandwidth wide. Need I mention low thd, im? Efficiency counts (20 amplifiers and I think my house wiring is 14 gauge) , but I'll not think about a pure "B". Class "A", while exotic, doesn't meet the efficient condition, so unless someone can come up with a "D" that doesn't broadcast to my neighbor down the street, I'm leaning toward an "AB" with just a hint of q current. As much as I'd love FET's, I think costs will dictate a bipolar design.
So, what do I look at? As I say, I'm way out of touch here. Would like a proven design that I don't need to spend a bunch of bench development time on. If there is something out there that I am dismissing, but should consider, please mention it. I'll spin some numbers around whatever you suggest and see what come out.
And, thank you very much for the guiding hand
So lets keep the costs down, the bandwidth wide. Need I mention low thd, im? Efficiency counts (20 amplifiers and I think my house wiring is 14 gauge) , but I'll not think about a pure "B". Class "A", while exotic, doesn't meet the efficient condition, so unless someone can come up with a "D" that doesn't broadcast to my neighbor down the street, I'm leaning toward an "AB" with just a hint of q current. As much as I'd love FET's, I think costs will dictate a bipolar design.
So, what do I look at? As I say, I'm way out of touch here. Would like a proven design that I don't need to spend a bunch of bench development time on. If there is something out there that I am dismissing, but should consider, please mention it. I'll spin some numbers around whatever you suggest and see what come out.
And, thank you very much for the guiding hand
Member
Joined 2009
Paid Member
My gosh, how did you get the need for 20 channels, that must be either a tri-amped 5.1 system or you're wiring the whole house for sound - this project sounds really impressive !
I could see it being a chore building 20 channels from scratch, if you don't 'need' to build them yourself you should take a look at the myriad of amplifier modules available on ebay these days - most likely you will find something suitable. If you really want to build them all yourself then it will be Class AB, building Class D isn't very easy. For Class AB this forum offers a few options that I'm sure others will suggest - I'd being by looking at the work of "OS" in the MONGREL thread, and the numerous 'DX' amplifiers by Carlos spread over multiple threads here.
I could see it being a chore building 20 channels from scratch, if you don't 'need' to build them yourself you should take a look at the myriad of amplifier modules available on ebay these days - most likely you will find something suitable. If you really want to build them all yourself then it will be Class AB, building Class D isn't very easy. For Class AB this forum offers a few options that I'm sure others will suggest - I'd being by looking at the work of "OS" in the MONGREL thread, and the numerous 'DX' amplifiers by Carlos spread over multiple threads here.
See fig.2
Audio-frequency power amplifier that ... - Google Patent Search
This is class G, you could make it class H too.
Audio-frequency power amplifier that ... - Google Patent Search
This is class G, you could make it class H too.
Reading your post set me thinking. the idea of shoe horning 20 off 100watt amplifiers into any system is no mean feat. Irrespective of efficiency. Wow - even the cabling will be complex and bulky. (imagine looms of cables the size of your wrist!)
The following is honestly out of interest - if the questions seem pointed please take them as me thinking "hey - what about..." and "what would I do if..." ...
There is no doubt you can do this - I reckon it would even be fun and good challenge. Without a hint of knowledge of your application, any comment fro me is speculative...
I am a bit of a discrete design nut - however for this application I would be thinking of an integrated amplifier design, and either find an off the shelf module, or lay one out myself, that allowed tight packing of multiple amplifier modules into a case. There are a lot of chip amp designs out there.
There is no doubt that you could use a discrete design - I use a "Self - blameless" design a lot and have one that I have put multiple channels on one box - but even with a single output pair a discrete implementation is running to 20 pairs of power transistors - will add up in cost fast. How important is distortion? How important is roll your own?
You will need to size the power supply sensibly. In all bar an extreme application, you will be very unlikely to need anything like sinewave output from this system, and thus need nothing like a 2KW+ PSU. You will however need to find the right sweet spot for the size of PSU. So: What level of loading does your application demand?
I think you really need to do a bit of systems engineering before you embark on the detailed design here... which in english (as opposed to engineering speak) means:
- Clearly define to yourself the function of the system
- From this how it will be used, and from that
- What the requirements of the 20 amplifiers are.
This will raise (and then lead you to answer) questions like:
- Does roll your own or "buy a kit" or module matter?
- Do all channels need to (or ever) deliver 50-100W at the same time
- How long will they do this for?
- How big of a case can you live with?
- Can you live with a fan (20 Class AB amplifiers = lots of dissipation) seious driver for cooling and case design.
The following is honestly out of interest - if the questions seem pointed please take them as me thinking "hey - what about..." and "what would I do if..." ...
There is no doubt you can do this - I reckon it would even be fun and good challenge. Without a hint of knowledge of your application, any comment fro me is speculative...
I am a bit of a discrete design nut - however for this application I would be thinking of an integrated amplifier design, and either find an off the shelf module, or lay one out myself, that allowed tight packing of multiple amplifier modules into a case. There are a lot of chip amp designs out there.
There is no doubt that you could use a discrete design - I use a "Self - blameless" design a lot and have one that I have put multiple channels on one box - but even with a single output pair a discrete implementation is running to 20 pairs of power transistors - will add up in cost fast. How important is distortion? How important is roll your own?
You will need to size the power supply sensibly. In all bar an extreme application, you will be very unlikely to need anything like sinewave output from this system, and thus need nothing like a 2KW+ PSU. You will however need to find the right sweet spot for the size of PSU. So: What level of loading does your application demand?
I think you really need to do a bit of systems engineering before you embark on the detailed design here... which in english (as opposed to engineering speak) means:
- Clearly define to yourself the function of the system
- From this how it will be used, and from that
- What the requirements of the 20 amplifiers are.
This will raise (and then lead you to answer) questions like:
- Does roll your own or "buy a kit" or module matter?
- Do all channels need to (or ever) deliver 50-100W at the same time
- How long will they do this for?
- How big of a case can you live with?
- Can you live with a fan (20 Class AB amplifiers = lots of dissipation) seious driver for cooling and case design.
Sanity...
Sane... well that is something of a relative term.
This is well and truly possible in class AB - but I agree not cheap, small, or necessarily sensible.
I built a subwoofer couch which needed six 250 watt amplifiers. The amplifier was so damn large I built it into a coffee table, and need to use a sack truck to shift it. The heatsink is two 360*300mm units with 40mm fins, the transformer is MASSIVE and in the end I doubled the capacitors in the photos as rail droop was a huge problem. Final rails have something in the order of 150,000uF (cant remember) but driving 4 Ohms peak current from the supply is just shy of 70 amps....
That transformer is huge - about the size of a breeze block.
Class D solves some of this, but not the issue of big bad power supplies that can double as arc welders, nor fist sized looms of wire.
Silliness aside - I think your underlying point of managing heat is absolutely at the centre of the isue, shortly followed by complexity and cost.
Sane... well that is something of a relative term.
This is well and truly possible in class AB - but I agree not cheap, small, or necessarily sensible.
I built a subwoofer couch which needed six 250 watt amplifiers. The amplifier was so damn large I built it into a coffee table, and need to use a sack truck to shift it. The heatsink is two 360*300mm units with 40mm fins, the transformer is MASSIVE and in the end I doubled the capacitors in the photos as rail droop was a huge problem. Final rails have something in the order of 150,000uF (cant remember) but driving 4 Ohms peak current from the supply is just shy of 70 amps....
That transformer is huge - about the size of a breeze block.
Class D solves some of this, but not the issue of big bad power supplies that can double as arc welders, nor fist sized looms of wire.
Silliness aside - I think your underlying point of managing heat is absolutely at the centre of the isue, shortly followed by complexity and cost.
Attachments
Thanks all very much for the comments, keep'm comin'.
The reason for 20 amps?. Well, think about this: Back in the '30's I think it was, AT&T (?) messed around with a digital speaker. Research and simulations so far are quite encouraging. So I'm pursuing. Taking the audio range and breaking into a series of bp's that are an octave wide, just less, actually. Active xover, therefore all the amps. Also, looking into using constant current amps, not constant voltage. Next go 'round maybe.
I have the bottom end cut off @30Hz, and my Xmax will allow me to reach 120 db , 1 meter, with just about 60W continuous. Not too much headroom with a 100W amp, but certainly the load center on my incoming power line will appreciate the difference! Power requirements drop drastically further up, but want the same signal path throughout, so same amplifier, just less power supply. Thus I dictate 20 100W amps, not because I need 2kw, I just need all the same and the bottom end makes that decision.
I must admit that an "AB" is somewhere less than 50% when all things considered, and a "D" will come in around 75%, so the advantage certainly swings to the "D". So I found the Tripath. On first pass, I see it as having several advantages, costs being a major player here. Size another. Kipman725, googlyone (indirectly - fans!) mentioned heatsinks. Ah huh.
Need to dig deeper into the "how well does it work, lets see some data", sort of things. Just found these last night. (Like I said, out of touch) Seems like it has a 25khz switching freq? What's the high end sound like? Must be nasty!? Now "AB"'s that are biased close to "B" are prone toward yuck also, but ..... And what about rf?
The reason for 20 amps?. Well, think about this: Back in the '30's I think it was, AT&T (?) messed around with a digital speaker. Research and simulations so far are quite encouraging. So I'm pursuing. Taking the audio range and breaking into a series of bp's that are an octave wide, just less, actually. Active xover, therefore all the amps. Also, looking into using constant current amps, not constant voltage. Next go 'round maybe.
I have the bottom end cut off @30Hz, and my Xmax will allow me to reach 120 db , 1 meter, with just about 60W continuous. Not too much headroom with a 100W amp, but certainly the load center on my incoming power line will appreciate the difference! Power requirements drop drastically further up, but want the same signal path throughout, so same amplifier, just less power supply. Thus I dictate 20 100W amps, not because I need 2kw, I just need all the same and the bottom end makes that decision.
I must admit that an "AB" is somewhere less than 50% when all things considered, and a "D" will come in around 75%, so the advantage certainly swings to the "D". So I found the Tripath. On first pass, I see it as having several advantages, costs being a major player here. Size another. Kipman725, googlyone (indirectly - fans!) mentioned heatsinks. Ah huh.
Need to dig deeper into the "how well does it work, lets see some data", sort of things. Just found these last night. (Like I said, out of touch) Seems like it has a 25khz switching freq? What's the high end sound like? Must be nasty!? Now "AB"'s that are biased close to "B" are prone toward yuck also, but ..... And what about rf?
All the amps I have posted / discussed on this forum have been class A and AB.
I have built most types of amplifiers. Played with UCD and layed a few boards out for USD to try them out. You would expect switching frequencies in the region of 3-400kHz, rather than 25kHz.
My experience was:
- Very, very low power dissipation in the switching devices. Even at fairly reasonable power levels
- Moderate dissipation in the driver circuitry (probably because I was using pretty "average" FETS for the application) - i.e. a few watts.
- Not very pretty output waveforms - a couple of volts RF ripple on the output may not be a big deal in an audio sense, but really messed with my sense of aesthetics.
- Clipping waveforms did not have the clean look / recovery that I am familiar with in linear designs.
You could knock up your own Class D amps - at 100W you could get away with quite modest "output" devices. I should go back and play again with this, as class D is popular with a lot of people on this forum.
If I understand you, the XO is going to be as complex as the power amp... if you go analogue, architecture will be a big deal to keep noise under control. If you go digital then it will be easy, and noise not such a big deal, but your hip pocket will take a pounding!
I have built most types of amplifiers. Played with UCD and layed a few boards out for USD to try them out. You would expect switching frequencies in the region of 3-400kHz, rather than 25kHz.
My experience was:
- Very, very low power dissipation in the switching devices. Even at fairly reasonable power levels
- Moderate dissipation in the driver circuitry (probably because I was using pretty "average" FETS for the application) - i.e. a few watts.
- Not very pretty output waveforms - a couple of volts RF ripple on the output may not be a big deal in an audio sense, but really messed with my sense of aesthetics.
- Clipping waveforms did not have the clean look / recovery that I am familiar with in linear designs.
You could knock up your own Class D amps - at 100W you could get away with quite modest "output" devices. I should go back and play again with this, as class D is popular with a lot of people on this forum.
If I understand you, the XO is going to be as complex as the power amp... if you go analogue, architecture will be a big deal to keep noise under control. If you go digital then it will be easy, and noise not such a big deal, but your hip pocket will take a pounding!
Re xo, ya op amp heaven, heavy on the metal r's, watch the c's. I could go with commercial dsp products Beringher (sp), Rane, etc , but if I home brew it'll need to be analog, I just don't have the facilities to program a dsp, and no way am I going to try to solder something with a .05 pitch or less. The lugs on your power transformer are just about the right size for me. I can layout my own 4 layer board though! Just can't fab it.
RE the amp again, I really need to see some broad scope data before I commit. If I am right about that switching freq, I would really like to see the filtering! And even if they can filter, what does this do to damping, output z, phase response, etc? Dang, I hate compromise almost as much as I hate money.
Where can I get a bazillion computer processor sinks real cheap?
That's the other thing...Going with an "AB", I must use convection only cooling, no fans. Fans are noisy, break a lot, and suck up as much power as the amp does. On the other hand, try to modulate a dc fan with sound. Think about that one.
RE the amp again, I really need to see some broad scope data before I commit. If I am right about that switching freq, I would really like to see the filtering! And even if they can filter, what does this do to damping, output z, phase response, etc? Dang, I hate compromise almost as much as I hate money.
Where can I get a bazillion computer processor sinks real cheap?
That's the other thing...Going with an "AB", I must use convection only cooling, no fans. Fans are noisy, break a lot, and suck up as much power as the amp does. On the other hand, try to modulate a dc fan with sound. Think about that one.
Ah yes but one can get around the huge transformer and capacitor bank issue by using offline switch mode supplies for the amplifiers. If RF is an issue with switch mode amplifiers (shouldn't be in a good design) then consider what a huge transformer with a massive capacitor bank causing a very small angle of conduction through a diode bridge will do when those large peak currents are passed though the house wiring.
As for crossover 20 channels is not beyond a modern computer but the software would be a pain to setup and you would need 3 sound cards to get enough outputs. Cheap though.
As for crossover 20 channels is not beyond a modern computer but the software would be a pain to setup and you would need 3 sound cards to get enough outputs. Cheap though.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.