Amplifiers

[GamerAndds]

My homemade 2000 watt stereo amp project(1000W per side). 
Should I use balanced or unbalanced inputs?
pros/cons
spk fuses?
how to find true output?
sourcing big VU meters?
http://1000wattampproject.blogspot.ca/

[Smokey]

Nice!
That's a killer project.  Power electronics is where the fun is anyway.  Keeps you on your toes!

It sounds like you are paralleling a ton of output transistors.  Do you have a method of matching so they will share evenly?  Bipolars have that nasty habit of running away in a hurry if one takes too much current.  Once that first one fails, it's all over.

I recently finished Bob Cordell's book, http://www.cordellaudio.com/book/ and highly recommend it to anyone interested in power amplifier stuff.  It has some really good explanations of pretty much everything in a power amplifier if you are looking for some circuit ideas.  It's not really a circuit cookbook, but does a great job of explaining all the sub-circuits and their interactions.  The book goes over both Bipolars and Fets, but he's a big fan of mosfets (silly spell-check wanted to correct "mosfets" into "marmosets") for power amps and I mostly agree with him on that one.  There are a lot of benefits.

Another good book is Randy Slone's, which actually does have whole schematics for some amps.
http://www.amazon.com/High-Power-Audio-Amplifier-Construction-Manual/dp/0071341196/ref=sr_1_1?s=books&ie=UTF8&qid=1364627453&sr=1-1&keywords=0071341196

If you end up doing any SPICE simulations, watch out for crappy power device models.  Most of the spice models for power devices range from completely wrong to only slightly terrible, especially mosfet models that don't use one of the advanced models like EKV or BSIM.  The normal Fet models, including most of the LTspice defaults, don't do weak inversion and sub-threshhold accurately which you need to simulate crossover distortion and such.  Be really careful with spice here as there is a really good chance this is one of those times it will lie to you.

diyaudio.com is another good resource if you have the time to figure out who knows what they are talking about and wade through all the other posts.

Beware the people that think audio stuff is mystical.  It's just circuits.  Now RF on the other hand ...... That stuff is still black magic!

[Eonir]

If you end up doing any SPICE simulations, watch out for crappy power device models.  Most of the spice models for power devices range from completely wrong to only slightly terrible, especially mosfet models that don't use one of the advanced models like EKV or BSIM.  The normal Fet models, including most of the LTspice defaults, don't do weak inversion and sub-threshhold accurately which you need to simulate crossover distortion and such.  Be really careful with spice here as there is a really good chance this is one of those times it will lie to you.

This is indeed quite the problem! One time I had to model a certain MOSFET used in a high-power bridge SMPS with ZVS. In that model, I needed to model some critical parasitic capacitances of the MOSFETs. None of the default models in LTSpice accomplish that. I ended up building a new model entirely by myself, which took me many, many days. It worked fairly accurately, but I don't want to end up doing that a second time.

Wherever you can, find accurate models of MOSFETs and other such devices. Many manufacturers provide you with very good models!

[Smokey]

The manufacturer supplied models are almost always generated with switching applications in mind and therefore don't use the advanced linear models like ekv or bsim.  This means that even the manufacturer models won't work for linear applications like classAB power amplifiers.  They just fundamentally don't take into account the behavior of the devices at low currents in the subthreshold levels. 
So what are your options if the manufacturer doesn't have an ekv or bsim fet model available?  Well not many.  You can try to find an advanced linear model that someone put together through experimentation but those are rare.  You can try to figure out the model parameters from the data sheet, but as people have figured out the data sheets don't always have all the information you need and the actual model parameter equations are not well defined in the public domain.  Your best option is to send a sample device to an independent test lab and have them construct the spice model for you.  You just better have a load of cash handy.

[NiHaoMike]

That power level should be attainable using 4 MOSFETs or IGBTs per channel. (MOSFETs can do the entire audio band, while most IGBTs are only fast enough for bass.) You should take a look at the homemade EV inverters since the basic concept is similar. (Indeed, audio amplifiers are in fact motor drives!) It's very easy to parallel MOSFETs, while only some IGBTs can be easily paralleled. (But at only 1kW, it should not be necessary to parallel IGBTs.)

You can save a lot of cost and size by designing it to run directly from mains. With an 8 ohm load, full power would be about 90V RMS, so the 120V input would yield a 170V rail that would give plenty of headroom for peaks. (90V is high voltage anyways and must be insulated as such, so not having isolation doesn't really impact safety.)

BTW, if you can relax the power requirements just a bit (or split them over more channels), TI makes amplifier chips that can do 600W in about the size of a quarter, though they work on a 48V rail and so you'll need a 2 ohm load.

[GamerAndds]

Thank you Smoky for those links, as for power matching(it just mean getting trannies with close tolerances to each others specs?)..... i'm young leave me alone  . Thank you for telling that to me though as my later projects include a preamp with 4 way x-over and 3 stereo (Small wattage 50-100W) class A amps to run some tweeters and high/low Mids(The beast has a home running 2 12 subs rated for at least 500 watts each). Ill have to keep my eye open for that book as it seems to intrigue me. There's only a few types of books I actually read and  that looks perfect... novels is not one of them unless they get really geeky.

To be honest about mosfets(I can see your problem with spell check ).... there cool, which requires a smaller heatsink (or the ability to add more ) they even have (sometimes) easier drive circuits., but I don't like them for all-range amps. There either on or off, nowhere in between. I honestly don't think they can truly capture the full potential of the highs.
BUT
for subwoofer amps... that's a whole different ball game.
When choosing this project it really baffled me what to do. Old school and tubes or new school solid state... but then which one bi's or fets.
Anyhow if this interests you keep up with my blogger account. I try and keep it updated. I even got some 8 AWG wire for my DC feed.


Amp Specs
10 trannies per side (Comp pair(5NPN,5PNP))
Dual channel (Left and right or if your crazy enough bridged)
1000W, Note I have brought up the voltage (+/- 10 V) so a new measurement will have to be taken.
Total (Calculated) Voltage- +/- 80V, thats a total of 160V!!!
Total (Calculated) current to spk: 12.8 AMPS
1.5 kVa Hammond pre-made transformer (182T60), (http://www.hammondmfg.com/182.htm)
3*22,000uF caps per side. (6 in total- 36DY series from Vishay)
Total of 20 output transistors. NPN-PNP
220/110VAC support. (I will be using 110)
Homemade Heatsink (with a Fan?).
Auto-On circuit with soft start pwr supply.
12VDC standby.

[NiHaoMike]

MOSFETs can easily do the entire audio band. (It's some IGBTs that are more limited.) Also don't bother making your own heatsinks since you can easily find some premade ones for CPUs. (Those out of a discarded server are especially good!)

At that power level, you might as well just direct rectify the mains and do the isolation at the signal level. (The output doesn't need to be isolated since it's high voltage anyways.) It would turn out much cheaper, more efficient, and smaller.

[GK]

Bipolar transistors are generally less problematic to parallel than MOSFETS. All they need is emitter degeneration. Vertical power MOSFETs designed for audio (such as the old Toshiba parts) are not readily available. Laterals are stupidly expensive (ie Exicon). If you want to use cheap verticals (ie the "complementary" IRFP240/IRFP9240 pair) matching devices for parallel connection is a good idea due to the large parameter spreads. Source degeneration is next to useless due to the low gm as the resistors have to be impractically large in value to equalize current sharing between poorly matched devices.

The right IGBT's are certainly "fast" enough for audio power amplifier service out to 20kHz. See:

http://www.google.com.au/url?sa=t&rct=j&q=alexander%20current-feedback%20audio%20power%20amplifier&source=web&cd=1&cad=rja&ved=0CC8QFjAA&url=http%3A%2F%2Fwww.analog.com%2Fstatic%2Fimported-files%2Fapplication_notes%2F58052492001115525484056221917334AN211.pdf&ei=leJXUYuWL4iiiAeb54DoCw&usg=AFQjCNFkDmt2BFXvrWzr8XDtg3mhGLA7cQ&bvm=bv.44442042,d.dGY

Also, the suggestion that the amplifier be run directly on rectified mains is a terribly bad one. Who wants their speakers and speaker wires at mains potential (not to mention any source plugged into the amplifiers inputs)?
 

[mzzj]

That power level should be attainable using 4 MOSFETs or IGBTs per channel. (MOSFETs can do the entire audio band, while most IGBTs are only fast enough for bass.) You should take a look at the homemade EV inverters since the basic concept is similar. (Indeed, audio amplifiers are in fact motor drives!) It's very easy to parallel MOSFETs, while only some IGBTs can be easily paralleled. (But at only 1kW, it should not be necessary to parallel IGBTs.)

You can save a lot of cost and size by designing it to run directly from mains. With an 8 ohm load, full power would be about 90V RMS, so the 120V input would yield a 170V rail that would give plenty of headroom for peaks. (90V is high voltage anyways and must be insulated as such, so not having isolation doesn't really impact safety.)

BTW, if you can relax the power requirements just a bit (or split them over more channels), TI makes amplifier chips that can do 600W in about the size of a quarter, though they work on a 48V rail and so you'll need a 2 ohm load.

Are you talking about switching class-D amplifiers and every one about class AB amplifiers?

 

[mzzj]

Also, the suggestion that the amplifier be run directly on rectified mains is a terribly bad one.

I Agree.

IMO only suitable use for non-isolated audio amplifier is in  integrated subwoofer "cube" with built-in amplifier and no speaker terminals available to outside. Even then you have to consider about isolation requirements and select the subwoofer driver to fullfill all the required safety standards. AFAIK Velodyne did something like this in their high power series.

[NiHaoMike]

Bipolar transistors are generally less problematic to parallel than MOSFETS. All they need is emitter degeneration. Vertical power MOSFETs designed for audio (such as the old Toshiba parts) are not readily available. Laterals are stupidly expensive (ie Exicon). If you want to use cheap verticals (ie the "complementary" IRFP240/IRFP9240 pair) matching devices for parallel connection is a good idea due to the large parameter spreads. Source degeneration is next to useless due to the low gm as the resistors have to be impractically large in value to equalize current sharing between poorly matched devices.

The right IGBT's are certainly "fast" enough for audio power amplifier service out to 20kHz. See:

http://www.google.com.au/url?sa=t&rct=j&q=alexander%20current-feedback%20audio%20power%20amplifier&source=web&cd=1&cad=rja&ved=0CC8QFjAA&url=http%3A%2F%2Fwww.analog.com%2Fstatic%2Fimported-files%2Fapplication_notes%2F58052492001115525484056221917334AN211.pdf&ei=leJXUYuWL4iiiAeb54DoCw&usg=AFQjCNFkDmt2BFXvrWzr8XDtg3mhGLA7cQ&bvm=bv.44442042,d.dGY

Also, the suggestion that the amplifier be run directly on rectified mains is a terribly bad one. Who wants their speakers and speaker wires at mains potential (not to mention any source plugged into the amplifiers inputs)?
 

I see MOSFETs paralleled all the time in various power supply circuits. (And for just 1kW per channel, it's possible to find MOSFETs that don't need paralleling to handle it.) Isolation is not particularly helpful since the output is high voltage anyways and must always be insulated as such. (It's a different story if you're running on less than a 50V rail.) For the input, you could design it to only accept optical input (which is inherently isolated) or you could use gate drive transformers or isolated gate drive chips between the driver and output.

If you still want isolation, try looking at some surplus telecom and 48V data center power supplies. It's pretty easy to find good deals on those and they're very well built.

Are you talking about switching class-D amplifiers and every one about class AB amplifiers?

At that power level, class D is really the only way to go. Nowadays, the cost breakeven point is on the order of a few tens of watts. The efficiency breakeven point is much lower than that, down to a few watts.

[mzzj]

Bipolar transistors are generally less problematic to parallel than MOSFETS. All they need is emitter degeneration. Vertical power MOSFETs designed for audio (such as the old Toshiba parts) are not readily available. Laterals are stupidly expensive (ie Exicon). If you want to use cheap verticals (ie the "complementary" IRFP240/IRFP9240 pair) matching devices for parallel connection is a good idea due to the large parameter spreads. Source degeneration is next to useless due to the low gm as the resistors have to be impractically large in value to equalize current sharing between poorly matched devices.

The right IGBT's are certainly "fast" enough for audio power amplifier service out to 20kHz. See:

http://www.google.com.au/url?sa=t&rct=j&q=alexander%20current-feedback%20audio%20power%20amplifier&source=web&cd=1&cad=rja&ved=0CC8QFjAA&url=http%3A%2F%2Fwww.analog.com%2Fstatic%2Fimported-files%2Fapplication_notes%2F58052492001115525484056221917334AN211.pdf&ei=leJXUYuWL4iiiAeb54DoCw&usg=AFQjCNFkDmt2BFXvrWzr8XDtg3mhGLA7cQ&bvm=bv.44442042,d.dGY

Also, the suggestion that the amplifier be run directly on rectified mains is a terribly bad one. Who wants their speakers and speaker wires at mains potential (not to mention any source plugged into the amplifiers inputs)?
 

I see MOSFETs paralleled all the time in various power supply circuits. (And for just 1kW per channel, it's possible to find MOSFETs that don't need paralleling to handle it.) Isolation is not particularly helpful since the output is high voltage anyways and must always be insulated as such. (It's a different story if you're running on less than a 50V rail.) For the input, you could design it to only accept optical input (which is inherently isolated) or you could use gate drive transformers or isolated gate drive chips between the driver and output.

If you still want isolation, try looking at some surplus telecom and 48V data center power supplies. It's pretty easy to find good deals on those and they're very well built.

Are you talking about switching class-D amplifiers and every one about class AB amplifiers?

At that power level, class D is really the only way to go. Nowadays, the cost breakeven point is on the order of a few tens of watts. The efficiency breakeven point is much lower than that, down to a few watts.

GK is talking about mosfets and igbt's in linear class-AB use and you are talking about switching class-D...

Class-D is more efficient and cheaper if you manufacture 100000 units per year but its also hell lot more difficult to design and implement. 1kW full-range Class-D audio amplifier is definitely not your first  hobby project...

There is still big difference in safety between high power isolated power amp with no input signal and power amp with loudspeaker terminals at +170Vdc relative to ground waiting for a change to kill you!
And I Am pretty sure that safety standards wont allow directly mains powered audio amplifiers ...

[NiHaoMike]

You cannot trust the output of even an isolated amplifier to be zero since a fault can cause quite a lot of voltage to appear unexpectedly. If the output is done using all mains rated connectors and cable, it will pass the safety requirements. You can also design an interlock loop on the output so disconnecting it opens the input and output contactors.

A 1kW linear amplifier isn't going to be a good beginner project either. If anything, switching stuff is easier at that power level.

[GamerAndds]

Actually my transformer is pretty much isolation for 110V except its center tapped, I pretty sure if you run my amp (when finished) at three quarters it would trip any 15 amp circuit...., esp (Mr. Rod Elliot) actually (on their 1.5kw amp) recommends (STRONGLY) that you use a 220vAc line.

Lemme clarify my views for mosfets.....
I DONT LIKE THEM FOR AB Amps, class D`s are ok. I was also very close to choose mosfets for mine instead of bi`s.

[GK]

I see MOSFETs paralleled all the time in various power supply circuits.

A powersupply is a different beast to a linear power amplifier. In a power amplifier you want the current sharing to remain accurate down to low drain currents to minimize crossover distortion.