My NON-discrete SODFA class-D amp

SSassen · 2005-06-04 1:49 am

Gentlemen,

After about a week of getting reacquainted with analog signal analysis and running through a whole set of batteries on my HP calculator I've come up with a NON-discrete class-D amp that should work well for my application. That application being a top-of-the-line subwoofer.

I'm about to start construction on the subwoofer, which will feature one of the new Adire Tumult 15" woofers (http://www.adireaudio.com/Home/TumultSeries.htm) in a closed box of about 35-liters. Of course with the Vas volume of the woofer, 266-liters, the F3 is around 100Hz and thus the drop-off is -12db/oct.

To compensate for that I use a simple filter that is the reverse of that curve, +12db/oct and hence will flatten the curve again. This principle is known as ELF or Bagend and is used by manufacturers such as Velodyne.

Of course the amplifier power needed is substantial, due to the +12dB/oct boost. For 110dB at 100Hz amp power is 50-watt, then at 50Hz > 200-watt and at 25Hz > 800-watt and so on. Obviously for efficiency and power handling reasons I want a class-D amp to power the sub.

I've been reading up on class-D amp design in the past week(s) and designed the following schematic by borrowing ideas from around the web, here at diyaudio.com, in application notes and at the German Hifi-forum.de website. What do you think, am I on the right track here? With a power supply of +/-80-volts and a 2-ohm impedance I should be hitting my mark of about 800-watt RMS right? And please let me know if there's any glaring errors in the schematic.

Class-D sub amp schematic
http://hardwareanalysis.com/images/articles/large/11520.gif

Looking forward to the (many) replies.

Best regards,

Sander Sassen
http://www.hardwareanalysis.com

classd4sure · 2005-06-04 4:56 am

Hi Sander,

Hysteretic self oscillator with pre filter feedback... yeah I think it fits the application and is a cool circuit/project.

I wonder if the integrating cap could drift or saturate, maybe a 1Meg in parallel with it or so, and the precision of it could possibly make a difference, as well as low input offset.

600 watts could be tough to get working reliably, guess you'll be looking at protection circuitry?

Regards,
Chris

JWFokker · 2005-06-04 8:40 am

Bagend as in "Bag End", home of Bilbo, and later Frodo Baggins?

ssanmor · 2005-06-04 8:55 am

The way you get the invert version of the PWM and at the same time level shift with T2 is very elegant!
First I thought it was a mistake, but now that I have analyzed it, I understand how it works.

Best regards,
Sergio

IVX · 2005-06-04 9:03 am

A) input buffer error.
B) hard to matching HIN-LIN falling edges. (maybe baker?)

SSassen · 2005-06-04 10:30 am

Chris, Ivan et al,

Thanks for the feedback, made a few changes to the schematic to incorporate some of the suggesties as well as a fix for some errors, the schematic in the fp is the latest rev.:

- a few polarized caps were placed in the wrong orientation
- input buffer had +/- inputs switched
- two zener diodes instead of one as BZX maxes out at 47-volt
- two zener diodes also better distribute the load

However by my calculations it will be a challenge to get to 800-watts RMS simply because 800-watts at 2-ohms needs 40V_eff and 20A_eff.

- 40V_eff * sqr(2) = 57V_s.
- max V_eff output is about 60% of VCC/VDD : VCC/VDD = +/-94V
- including losses across the mosfet's this will be +/-100V

Also I'm a little worried about power supply pumping due to the half bridge, maybe I'm better off doing a full bridge? I'd be happy to hear your suggestions about that as well.

Ivan, you mention it will be hard to control the falling edges of Hin-Lin, are you referring to the pre-driver circuitry or after the IR2110? And what do you suggest instead?

Best regards,

Sander Sassen
http://www.hardwareanalysis.com

SSassen · 2005-06-04 11:19 am

Alright, I took Ivan's comments to hart and went looking for a comparator with both a normal and inverted logic/ttl compatible output so I can better interface with the mosfet driver. The LT1016 seems made for the job and I have updated the schematic to include it. For every revision I will now increase the file number of the schematic so the highest file number wil automatically be the latest schematic.

class_d_amp_001
http://hardwareanalysis.com/images/articles/large/11521.gif

class_d_amp_002
http://hardwareanalysis.com/images/articles/large/11522.gif

class_d_amp_003
http://hardwareanalysis.com/images/articles/large/11523.gif

Thanks for all of your input, one more question though, can I interface the LT1016 directly with the IR2110, omitting the two transistors and four resistors? Or would it then go out with a bang?

Best regards,

Sander Sassen
http://www.hardwareanalysis.com

classd4sure · 2005-06-04 5:38 pm

Try it?

IR2110 has mos inputs do you know? Should consume little current unless switching.... maybe a small resistor would do

SSassen · 2005-06-04 6:08 pm

Chris,

Yeah, I already ordered parts, but I'd like for the circuit to be as simple as possible. I just don't feel comfortable to omit T1, T2 and the four transistors yet, as both parts of the circuit operate on a different voltage level. The GND of the IR2110 is tied to Vdd which will be at -50...80V whereas the LT1016 is at +/- 5-volts derived from Vcc and Vdd.

If I calculate the Vdiff between IR2110 GND and -5-volt at the LT1016 the difference in voltage is -50- -5 = 45 volts and that could very well cook the LT1016 hence I put the transistors there as a precaution. But maybe I'm just being too cautious, what do you think?

Oh, and you mentioned protection, you're indeed correct, it'll need both DC protection and a softstart and possibly overcurrent protection. That's why I picked the IR2110, as it features a enable pin (Sd) so you can disable the MOSFETS when needed which allows for a simple softstart scheme with just a few parts. However any suggestions there are also appreciated.

Best regards,

Sander Sassen
http://www.hardwareanalysis.com

classd4sure · 2005-06-04 7:42 pm

Hi,

1. Try increasing C19 drastically, 10 uF or more if you can.

Don't think you can omitt the transistors as you say there needs buffer after the op amp.

So a baker clamp like Ivan was sugesting on the buffers might be worth a shot, and certainly transistor selection could make a difference but those seem like a decent type to try it with.

Can you simulate this? Did you have any plans to?

Regards

SSassen · 2005-06-04 8:16 pm

Alright, another update, got some feedback from a few helpful subscribers to the hifi-forum.de website. The latest schematic is listed below. You'll see the transistors are now connected differently, this should allow for better interfacing with the mosfet driver. I've also switched to the IR2011 mosfet driver as it has better parameters overall and is better suited for high-frequency (>200KHz) operation.

class_d_amp_004
http://hardwareanalysis.com/images/articles/large/11524.gif

Best regards,

Sander Sassen
http://www.hardwareanalysis.com

SSassen · 2005-06-04 8:45 pm

Chris,

I'd love to simulate it, but to be honest I wouldn't know how, many of the parts I used aren't in the Spice library and to be honest I haven't used a simulator for a long time. If you'd like to help out and sim it for me you're most welcome of course. You can just omit the driver stage and the mosfets and connect the output of the comparator right up to the output coil in the sim if I'm not mistaken?

Best regards,

Sander Sassen
http://www.hardwareanalysis.com

SSassen · 2005-06-05 7:01 am

And another update, we're up to version 005 already with the latest changes being a simpeler input buffer and optimized values for the resistors around T1 and T2 to guarantee proper switching.

class_d_amp_005
http://hardwareanalysis.com/images/articles/large/11527.gif

Best regards,

Sander Sassen
http://www.hardwareanalysis.com

subwo1 · 2005-06-06 8:19 am

You may like to remember to adjust the values of R15 and R16 at some point in time.

Pierre · 2005-06-06 9:41 am

Sander, looking at your schematics, I have seen that the only way you have accounted for dead-time addition is the diode in antiparallel with the gate resistor.
I don't know if you have built this circuit and hence if you have found shoot-through currents, but that's something you may need to have in mind if you find that the mosfet reliability is not good when you test it.

Perhaps other people that have built similar designs can say that it is not necessary to add more dead-time before the driver. What are your opinions on this issue?

Best regards,
Pierre

SSassen · 2005-06-06 9:59 am

Subwo1, Pierre,

Yes, I might be looking at changing some resistor values to account for more deadtime, but the same mosfet driver (IR2011) and output stage is used by IRF in their reference design and their deadtime, and hence THD figures, are very respectable.

I've already ordered parts (at Farnell and RS) and should be getting them in shortly. I'll then construct a prototype on some proto-pcb and she if she'll run. If she does I'll hook her up to a load and see if I can get some good measurements on the scope and modify the design accordingly.

Best regards,

Sander Sassen
http://www.hardwareanalysis.com

phase_accurate · 2005-06-06 10:10 am

Hi Sander

I had a look at it and I'd say that I like it. I would also recommend a large parallel resistor to C7.

The working principle is from one of the guys on the German forum, isn't it ? What is his experience with DC drift ?

Regards

Charles

SSassen · 2005-06-06 10:24 am

Phase_accurate,

It is a bit of everything really, but indeed the feedback topology is from the German guys, as you can see the feedback loop feeds both the integrator and the comparator input and is slightly different from the classic self-oscillating class-D implementation.

As you're the 2nd person suggesting I add a large resistor across C7 to avoid drift, Chris also mentioned it, I added R2 - 4.7-MOhm, which should be sufficient I guess, see latest version of the schematic below. Fortunately I have all the tools here to evaluate the proto, such as a dual channel 100MHz storage scope, a DDS function generator and I just ordered a dozen 50-watt resistors to build a dummy load, and not to be forgotten new batts for my DMMs.

class_d_amp_006
http://hardwareanalysis.com/images/articles/large/11529.gif

Best regards,

Sander Sassen
http://www.hardwareanalysis.com

Pierre · 2005-06-06 1:20 pm

Sander, why do you like the last transistor arrangement more than the version-3?
I mean, what's the difference between having the base connected to the comparators and the upper resistor to Vcc (as it was before) and connecting the base to GND and the upper resistor to each comparator's output?

About level shifting, it is surely needed, unless you operate all your circuitry with the same supply as the IR2110, but then input should be capacitively coupled, as well as feedback signal. I don't know how that would perform during startup, however.

Best regards,
Pierre

phase_accurate · 2005-06-06 1:25 pm

I mean, what's the difference between having the base connected to the comparators and the upper resistor to Vcc (as it was before) and connecting the base to GND and the upper resistor to each comparator's output?

Apart form the fact that one of these arrangements is inverting and the other one isn't - there is a difference in speed. Driving the emitters should be faster since it eliminates the miller effect.

regards

Charles

My NON-discrete SODFA class-D amp

Member

Disabled Account

Member

Member

Member

Member

Member

Disabled Account

Member

Disabled Account

Member

Member

Member

Member

Banned

Member

Member

Member

Banned

Member