full differential UcD modulator with extra integrator

Trying to create hi-End class D amplifier.

Fully differential structure allows to work the bridge , half-bridge, or a bridge with mono polarity power supply output stage (TAS5261).

Full bridge citcuit prefered, because it compensates for partly nonlinearities in power supply, modulation, inductor. Differential noise shaper further increases the loop gain of 6 dB.

modulator 4 layer 32x60mm
3 FDA THS4131 and comp AD8561
 

Attachments

  • MC model.GIF
    MC model.GIF
    28.4 KB · Views: 2,141
  • BTM.GIF
    BTM.GIF
    45 KB · Views: 2,110
Last edited:
Connector left - power. Right - feedback IN and 2 PWM outputs, SOT223 -LM317/337.
circuit attachment top post only for checking the stability and performance simulation in microcap11.
After some time, the pсb will be made in China, then I will carry out tests.
Тhe project is really very interesting, it is necessary to solve several problems. For example, it is not clear whether the use of hysteresis must be in high-speed comparator?
And some others.
 
Last edited:
Sous,

I know you have done some very good amp earlier, but this thread is ... well ... a little confusing.

- Maybe the most disturbing problem is spelling. I guess "brighe" and "bright" = bridge, but I don't know why do you use all of these forms? (FYI: Bright means shiny.) Or is it "Right" (=opposite of left)?

- What kind of parts are X4 and X7?

- The overlapping labels also make the schematic hard to follow.

- The limiter is quite overcomplicated.

- You said "bridge compensates for partly nonlinearities in power supply, modulation, inductor", but this is not true, it compensates only asymmetry. Nonlinearity in inductor is symmetrical, so it is not compensated by bridge topology, and you have only 1 modulator, the errors of the modulator are not compensated this way.

- You said "Differential noise shaper further increases the loop gain of 6 dB.", but I don't see it on your schematic. I don't know what do you think "differential noise shaper" is. Actually noise shaper relates to D/A or A/D converters, while this circuit is fully analog.

- The circuit is not fully differential. The modulator is single. This means all the theoretic limitations (loop gain, ripple) come from the basic topology still apply. Effectively you only multiplied every voltage by 2.

- The circuit is very big compared to its functionality.

But the selection of power stage is perfect I think. I also design with it, but a really fully differential design, with 3 level modulation, modulator+power stage on a little more area.

For example, it is not clear whether the use of hysteresis must be in high-speed comparator?

I don't understand what do you mean.

1: you think you must use hysteresis, but not sure if it must be in the comparator or elsewhere? or
2: you are not sure if you have to use histeresys which is already exist in that stage?

Well, UcD is basically free of histeresys, but a small amount is acceptable, and beneficial, if switching noise is not low enough.
 
:( Оh, sorry for my terrible English
As you can imagine, the first attachment is not a complete circuit, a simplified model to simulate (MC)

X4,X7 is a unity gain buffer

PCB demonstrates the possibility a compact tracing.

I do not know how to make a limiter is better than this.:D

At the expense of compensation, you are right. correct to say that the bridge reduces the impact of the volatility of supply voltage and pumping effect. Inductors on the combined core have a lower nonlinearity.

I honestly do not understand why you think that is not a differential modulator? It has differential inputs and outputs. Of course, it turns out synchronous, but it is rather a positive effect on results. Not synchronously, such a scheme can not work. Let it be correct - differential feedback modulator.

I meant "integrator"= noise shaper.

simplified scheme already exists - it's called n-core ;)
want to implement it on the IC аnd monopolar PS.
 
Last edited:
:( Оh, sorry for my terrible English
As you can imagine, the first attachment is not a complete circuit, a simplified model to simulate (MC)

I know. And I see you've set the loop gains and levels well (except that R22 not equals to R6), to keep the integrator correcting only errors, not driving modulator stage. This is impressive, and helps to evaluate the prototype, however not neccessary i think. But this is not important.

X4,X7 is a unity gain buffer

PCB demonstrates the possibility a compact tracing.

I do not know how to make a limiter is better than this.:D

What about the same quality but much simpler? I didn't say it's not good. It's good, but overcomplicated.

The tracing is compact, but considering it's on 4 layers it's very big. Or is the power stage on the other side? I guess not. Similar functionaly could be done on 2 layers with 1 layer is continuous plane, on the same area, manufacturable at home, without plated vias. :) (OK, only if one wouldn't use 2 high voltage and 2 low voltage rails...)

At the expense of compensation, you are right. correct to say that the bridge reduces the impact of the volatility of supply voltage and pumping effect.

Yes. As much as in case of 2 independent Single Ended amplifiers, driven at opposite phases.

Inductors on the combined core have a lower nonlinearity.

Maybe, it is not duscussed yet. I don't know if this meant as a general statement or you think something particular. But if true then it can be realised in an SE circuit just as well. It's not related to bridge circuits. Exactly the same current flows on each halfs, so you can connect them simply serial in a SE amp with exactly the same performance.

I honestly do not understand why you think that is not a differential modulator?

You have 1 comparator with 1 output signal that is duplicated later. You can call it differential, but it doesn't have any benefit over the basic circuit. Contrary to your design it is possible to use 2 different modulators that carries different informations and can be used to suppress unwanted modulation artifacts. That would be fully differential in my terminology.

It has differential inputs and outputs. Of course, it turns out synchronous, but it is rather a positive effect on results.

Compared to what and how?

Not synchronously, such a scheme can not work.

I thought so also, but some month ago I realised it can be done. And it is beneficial.

Let it be correct - differential feedback modulator.

Differential, but it's only 1 modulator, while everything else are doubled.

I meant "integrator"= noise shaper.

And why do you think it "further increases the loop gain of 6 dB"? Of course you can always set a higher gain, but at the expense of modifying oscillation frequency. Without modifying original oscillation characteristics you can have exactly the same limit of loop gain. In your design all voltages have its exact negative pairs, and at every point in your circuit the operation is determined by the difference of the node voltages, which means you have 2 times the voltage of a single ended design with the same rail voltages. Every voltages are doubled (except of course the single ended input), but the ratios (=gains) stay the same.

want to implement it on the IC аnd monopolar PS.

I appreciate it, and I also think single rail is good (for sake of simplicity, and in case of differential design for sake of symmetry also), but I extend this to the whole circuit, therfore I'm not gonna use many rail pairs, only as much as unavoidable. You're using here 4 supply voltages only in the modulator, and you will need a gate driver supply voltage. These with the main power supply counts 6 supply voltages (in case the comparator positive supply and logic input supply for the power stage will be common). Too many for a power amplifier I think. (My design consists 4, which is also a little too many for my taste, but this is I could do.)

I don't say your design is not good, just it's not especially better than the previous ones. Bridged, with a very good, integrated power stage, thats all.
 
Nice work, bro~
Thanks for sharing.
you are welcome!

I made the layout of the amplifier with a half-bridge inverter
Unfortunately, there not arrived the radiator.
Power s.+/- 34V (100W transformer 50Hz +/- 24V AC output)
cutting at the output signal of 85 W.

the upper line is the AMP, the bottom - through channel sound card


graph distortion of power max -78W
 

Attachments

  • rrrrrr.jpg
    rrrrrr.jpg
    189.5 KB · Views: 774
  • 15V RMS 3_9 Ohm.gif
    15V RMS 3_9 Ohm.gif
    305.6 KB · Views: 730
  • thd vs power.png
    thd vs power.png
    6.1 KB · Views: 421
Last edited:
you are welcome!

I made the layout of the amplifier with a half-bridge inverter
Unfortunately, there not arrived the radiator.
Power s.+/- 34V (100W transformer 50Hz +/- 24V AC output)
cutting at the output signal of 85 W.

the upper line is the AMP, the bottom - through channel sound card


graph distortion of power max -78W

pretty impressive. Left channel seems cleaner than the right