Transforming TPA3255 into ABD modulation.

[Hayk]

I tested it, two issues. It is stable but very noisy. At not low levels, it cuts continously.
The sound is very refined.

 

[Hayk]

The noise is high impedance wires,switch capturing neighborhoods refregerator noise. The cut was due to capacitor 100nf //R6 (which is now 62k instead of 330k),is too high.
I installed the stability capacitors 68pf on the amplifier board, afterwards, I add to get best square wave transient.
What immediately new sound I heard, is the velvet voice of Peggy Lee. The fine beat of brass instruments indicates still higher resolution than the 3110 stand alone. To remind that the PSRR indicates the NFB. With 3110 24v it is 80db but 3255 it says >65db at the beginning of the DS. Now with extra 20db it has become 85db. The purifi is 90db.

 

[Hayk]

The sound of this board makes the precedent a gadget. I remade the PCB only high impedance version so that you can order from the PCB manufacturer, all mounted circuit. I think it costs $8 more as many don't like to solder smd components and you don't need to buy 100 pieces of each type of components.

 

[Hayk]

The noise still there but 20 dB less with very short wires. The resolution is exceptional but where has gone the fabulous bass. I feel it will take a week if not two to tune this circuit.

 

[voodooless]

Please, please, please use a controlled listening test...

 

[Hayk]

I have designed this circuit about 8 years ago to drive directly without coupling capacitors, TPA3116. The outputs had 9k DC resistance from the internal chip. Here, I don't have dc resistors, if an unbalanced current occurs the simulated inductors do not absorb exactly the source current and the output saturates to either rail. This why the capacitor is destroying the operation. Without capacitor, the low frequency transient becomes weak.
Tomorrow, I will try 2 independent simulated inductors.
The mater of noise, resolved.

 

[Hayk]

Please, please, please use a controlled listening test...

I wired this circuit without enthusiasm and made many mistakes due to carelessness. But once I heard the velvet voice of Peggy Lee with Goodman's orchestra behind, I decided to give my best for it, whether you may like it or not.
It has many issues, stability, tonal balance, installation, wiring....
and I am ready for it.

 

[JayGilb]

I wired this circuit without enthusiasm and made many mistakes due to carelessness. But once I heard the velvet voice of Peggy Lee with Goodman's orchestra behind, I decided to give my best for it, whether you may like it or not.
It has many issues, stability, tonal balance, installation, wiring....
and I am ready for it.

I understand you are in the development stage, maybe after the design/enclosure mature, you would consider sending it to Amir for testing ?

 

[Hayk]

Total failure. It gave exactly the same result , cutting.
Tomorrow, I will make it more meticulously with matched paired resistors, as I use 5% and I will add a 100k adjust to perfectly balance the gains.

 

[Hayk]

I could pinpoint the problem. In simulator, the BC550C has hfe of 800 whereas I have 450. By this not all the current poured by the CCS passes through the simulated inductors, some pass through the input bases decreasing considerable the input impedance hence less bass. I will used darlington tomorrow.

 

[Hayk]

Perfect.

 

[Hayk]

The very high resolution, bringing on surface sounds I never heard before in music I know since decades. The problem now, the voice is metallic, after 10mn, I had enough. It will not be easy to shape this diamond.

 

[Hayk]

The metallic voice was due to cfp darlington. I folded back and changed the ground configuration.
The high impedance captures too high level of noise inaudible, and heats the heatsink to 50°C in silence. Probably the noise is destroying the sound quality as, it is not giving the sense of luxury. Although the bass is much better but my tonal balance test music, B Manilow's I can't smile, is disagreeable to listen, too much highs not enough lows.
Sine wave frequency response is flat tested from 1Hz to 20khz. No oscillations on square wave.if

 

[Hayk]

I had to reduce the output impedance from 560k to 68k to get a good tonality. Now the extra feedback is 10db instead of 20db but still it is high resolution and the fabulous bass is back and Manilow's song is agreeable to listen.

 

[Hayk]

I am trying this and that to see what better I can make. I paralleled the input 1uf capacitors with Nichicon 10uF gold electrolytic. Did I get better lows, not at all, the high frequencies got smoother!!!!!
These are not cheap but it worth what it does.
I am listening with great leisure, very proud of myself.

 

[Hayk]

My 3255 board uses ordinary ceramic smd capacitors. I pulled the input wires before them and applied the Nichicon ones external. Today I got my overdose of listening, I'll see tomorrow what better it gives, maybe not much with this high impedance but surely with the low one.
The extraordinary feature of these Nichicons, they can function linearly with -2,5 instead of -0.6 ordinary electrolytic do.

 

[Hayk]

I discovered something very interesting in my simple low impedance circuit.
At the junction of the ccs and the input emitters, I can apply a current feedback from the 3255 outputs. I simulated for 10db extra NFB, the circuit does it. I will put a CM choke to get rid of 450khz, filter the 900khz and apply a differential current.
I will try it tomorrow.

 

[Hayk]

This is the feedback filter. The CM choke and 4.7nF capacitors eliminates the 450khz and the 390 ohms along the capacitors reconstruct the audio. The extra feedback ratio is 6.8k/2.2k.
Before I connect the feedback, I listened with Nichicons instead of ceramics, significant difference in sound quality. Already it is high end sound, will the feedback increase the resolution?

 

[Hayk]

Sometimes, you need to circumnavigate the planet, to find what you were looking for, was behind your heel.
By feeding with high impedance, the ceramic capacitor became transparent. This why I got better resolution. Now with Nichicons instead I, have as high resolution with adjustable bass pronounced.
I will not try out the feedback, as it will decrease the valuable 2nd harmonics.
I will developed the PCB only for this circuit. For those who need very low distortion numbers, I advise you use from TI tas5630 evm user guide.

https://www.ti.com/lit/ug/slau287a/slau287a.pdf?ts=1766794624824

 

[Hayk]

As many appreciate amplifiers by reading numbers, I will design one for such purpose. As the 3255 is variable dead time, in BD modulation it showed very little distortion at 10-20 mw and much lower noise than AD mod. As such power amp, no one will be using 105db efficiency speakers and distortion numbers are important at minimum 1w if not 5w, so forgetting the ABD mod and use it in BD. It eliminates a pair of inductors per channel and allows PFFB which is only 5db but the title has become highly commercial.
Is it post filter feedback?
Oh yes, it is.
I will use very low distortion opamps to convert single to differential and differential input, whatever it may sound as numbers are what celebrated here.
The 48v to 12v converter, I will use a ready made step down converter, to get rid of the noise from on board one. I will use JLCPCB listed components, so that one can order all built ready to be used by adding a fanned heatsink, 48v to 12v converter and 48v smps.