Class D and triangle wave linearity

One thing, the 555 exponential triangle wave (as shown) is taking the sweet spot out of the curve. While far from perfect, it is also the best triangle wave you can make using this technique.
.
I used an op amp intergrator to get the trinagle waveform and this is more linear than just using an RC. The RC is exponential and the integrator is linear.
 

tom66

Joined May 9, 2009
2,595
You miss the point of projects like this, no specialized parts, minimum parts count. Keep It Simple Stupid (KISS). Done correctly you will not need heatsinks, the whole point of of Class D is PWM. The transistors are on/off, and not running in analog. Shoot though would be what generates heat, if the transistors lasted long enough.
Shoot through doesn't generate much heat, if it lasts only for a short time. You might be talking about 1-1.5W in each FET at full power; a small heatsink will be more than enough.

You will need a heatsink especially if you are using logic ICs as gate drive as the turn on/off will take a long time during which the FET is operating in a linear region. Also if you are using high on-resistance FETs. No specialised parts really limits performance; even a cheap gate driver like the TC1412 (which is available in DIP and costs just over $1 per part) would significantly improve the performance of this project. And with those which integrate shoot through protection, then you would not need any heatsinking.
 

rogs

Joined Aug 28, 2009
279
No specialised parts really limits performance
Not by definition - although often true!

Some of the commercial products based on the designs from this guy: http://www.hypex.nl/docs/papers/AES118BP.pdf have some superb performance figures, and do, IMHO, sound absolutely great.
And those designs are based totally on common discrete components - no 'special' devices at all.

Not really going to win any prizes in the 'minimum component count' category though...:)
 

THE_RB

Joined Feb 11, 2008
5,438
Again, this is not really the point. This not meant to be a practical circuit per se, but a teaching circuit. If it works it could wind up as a learning project.
...
Ahh I get it now. You are making a 555 based class-D amp! :)

I saw one in a book many years ago. It might have been in one of those "101 circuits using a 555" type books. From memory it applied the analog sound input to the threshold pin and that varied the duty cycle.

...
If I were truly serious about this I would definitely use a linear ramp generator.
Do you have a draft schematic so far? If you were able to use the middle 1/5th of the RC range (like the middle 1v of a 5v device) the curve is plenty linear for a simple "educational" class-D amp.

Otherwise a 1-transistor CC source (to charge RC) plus a comparator IC (with 2 comparators) can make both the linear sawtooth ramp osc and gating comparator. That's the entire front end of your amp with just 1 tran and 1 8pin IC?
 

PaulEE

Joined Dec 23, 2011
423
You can use a 555 and a few pnp transistors for a pretty handsome ramp signal, also. Keep both transistors at the same temp. for frequency stabability.
IRF530/9540. Low Rds(on) at logical level input voltages.

My 2¢.
 

Ron H

Joined Apr 14, 2005
7,014
You can use a 555 and a few pnp transistors for a pretty handsome ramp signal, also. Keep both transistors at the same temp. for frequency stabability.
IRF530/9540. Low Rds(on) at logical level input voltages.

My 2¢.
Maybe you should make that your 1¢.:D
Neither of those MOSFETs is spec'ed for logic level drive. Both are spec'ed for Vgs=10V.
They are also 100V parts, which is overkill for a small class D driver. I believe that Rds and gate charge are both higher than what you can get in lower voltage parts.
 

tom66

Joined May 9, 2009
2,595
For a gate driver, try this.

Rise time is ~530ns, fall time is 250ns; the main limitation is the pull-down. A lower pull down works but dissipates a lot of power.

If you have a 12V supply use a 4.5V rated (logic level) MOSFET as the driver only puts 8V on the gate, which isn't always enough to turn on some MOSFETs.

Dissipation in a 1mohm, low gate charge MOSFET (~30nC), is <0.2W, with a 1 ohm load. I highly recommend a low gate charge MOSFET. Going up to 150nC, the dissipation is almost 1.5W, yet the device had the almost the same on-resistance.

A gate driver would manage 20ns/20ns at full 12V to 0V. That would probably be 0.01W dissipation, no need to even think about a heatsink.
 

Attachments

Last edited:

PaulEE

Joined Dec 23, 2011
423
Maybe you should make that your 1¢.:D
Neither of those MOSFETs is spec'ed for logic level drive. Both are spec'ed for Vgs=10V.
They are also 100V parts, which is overkill for a small class D driver. I believe that Rds and gate charge are both higher than what you can get in lower voltage parts.

Hey now! You can squeeze around 5 amps at Vgs = 5 volts!

They aren't overkill if they want a 1kW amp...hehe.

In retrospect, these might not be super-great at higher frequencies due to the gate capacitance and whatnot.

They worked great for a high power pulse-width modulation peltier cooler driver circuit, and class D is basically that and a filter.

-1¢ :D
 

Ron H

Joined Apr 14, 2005
7,014
For a gate driver, try this.

Rise time is ~530ns, fall time is 250ns; the main limitation is the pull-down. A lower pull down works but dissipates a lot of power.

If you have a 12V supply use a 4.5V rated (logic level) MOSFET as the driver only puts 8V on the gate, which isn't always enough to turn on some MOSFETs.

Dissipation in a 1mohm, low gate charge MOSFET (~30nC), is <0.2W, with a 1 ohm load. I highly recommend a low gate charge MOSFET. Going up to 150nC, the dissipation is almost 1.5W, yet the device had the almost the same on-resistance.

A gate driver would manage 20ns/20ns at full 12V to 0V. That would probably be 0.01W dissipation, no need to even think about a heatsink.
A push-pull emitter follower (2N2222/2N2907)driven by a CD40106 is a lot faster and simpler, IMHO.
 

Thread Starter

Wendy

Joined Mar 24, 2008
22,141
As far as current goes the speaker and the low pass filter dictates. Ideally the 100Khz (or whatever) is blocked, leaving the audio on the speaker.

The power supply is going to be 12VDC to 15VDC, within the range of the 40106. Also within the range of the MOSFETs. I never mentioned 5V, that came from others, I think I said the chips will dictate the power supply voltages.

I could be wrong, I frequently am, but generally a CMOS gate can drive complementary MOSFETs. A 100Khz may be a bit fast, but an oscope should show the tale. I'm not so sure of BJT types, the 1/10 base current for saturation is a killer. If we use BJT as emitter followers then we loose around 1.2V minimum from the power supply, and in either case limit the max current to 600ma. They could drive the MOSFETs of course, but I'm not convinced that the CMOS gates couldn't do the job.

This circuit would need a clipper circuit to prevent overdriving, but I think I would leave that off.

So lets assume a square wave, 15VDC, through this setup. I calculate a maximum current of 3.75A. This is the maximum, it will generally be much less. I don't know where the 5A figure is coming from. The RMS power at max will 1.9W, also assuming a square wave.
 
Last edited:

PaulEE

Joined Dec 23, 2011
423
Well, maybe with some parts, but I wouldn't count on all of them working at 5V. The max VGS(th) is 4V on both parts. That's at Ids=250uA.:rolleyes:
Good point!

I think the original circuit pulled the gates to +12v with 2222's.

The VNP35N07 is a neat logic level Nmos with protection circuitry, but it might not react fast enough to the audio...but not sure.

Class D amps are a very cool thing to build, though.
 

tom66

Joined May 9, 2009
2,595
A push-pull emitter follower (2N2222/2N2907)driven by a CD40106 is a lot faster and simpler, IMHO.
If you drive it with 12V, but then all the logic needs to operate off 12V.

I would think a 555 timer ought to be able to provide an decent gate drive.
 

Thread Starter

Wendy

Joined Mar 24, 2008
22,141
You mean the LM393 (dual comparator)? It is good for 36V. Two chips, 40106 and LM393, I didn't spell it out before but that is what I was thinking.

Any inverting Schmitt Trigger will produce a decent exponential triangle wave.

I actually was originally thinking a couple of 556's but I saw an opportunity to reduce parts count.

There is a really simple way to kill shoot through using the spare comparator (two extra resistors), but it would also kill the lower dynamic range of the audio. I'll hold it in reserve.
 
Last edited:

Audioguru

Joined Dec 20, 2007
11,249
Some people like me enjoy music and are audiofiles. We dislike distortion.
Some people NEVER listen to music and hear only the screaming or cheering at a game.

Do an AB test with "average" people and the result might be that 20% distortion is tolerable. BS!
ANY amount of distortion is intolerable today when nearly zero distortion is easily made.
 

Ron H

Joined Apr 14, 2005
7,014
Bill, if you run into a problem with driving MOSETs with a CMOS device, it will be worst at the duty cycle extremes. You need fast rise times to get the short pulses. You need high current to drive the Miller capacitances, and the CMOS outputs can only provide currents on the order of tens of milliamps. As I said before, you can probably find MOSFETs with low enough gate charge to work reasonably well with CMOS drive, as long as you are content with Rds(on) on the order of 100 milliohms, more or less (OK, I didn't say all that previously).
The 1.2V loss incurred with emitter followers is not a problem with a 12V supply. Both MOSFETs still get adequate turn on and turn off levels. ±600 mA is a vast improvement over the capabilities of a CD4000 series gate.
 

Ron H

Joined Apr 14, 2005
7,014
Some people like me enjoy music and are audiofiles. We dislike distortion.
Some people NEVER listen to music and hear only the screaming or cheering at a game.

Do an AB test with "average" people and the result might be that 20% distortion is tolerable. BS!
ANY amount of distortion is intolerable today when nearly zero distortion is easily made.
Yeah, string telephones sound good to me.:cool:
 

Audioguru

Joined Dec 20, 2007
11,249
Yeah, string telephones sound good to me.:cool:
Telephones sound awful, especially cell phones. ALL the high frequency sillibants of speech are gone so it is difficult to understand what is being said.

ee de'ingner' o' 'ele'one' 'ould be 'o'. (Just vowels.)
Translation to wideband audio: The designers of telephones should be shot.
 
Top