Transforming TPA3255 into ABD modulation.

[Hayk]

The TPA3255 is AD modulation and without option for BD modulation to transform it into ABD.
The DS shows a table of relative phase of the outputs, see below, where in SE mode intended for 4 independent channels, the A and C channels are in phase, as well as B and D, this can make BD modulation but not perfect bridge.
I modified today my board , M1 and M2 for 1 and swept the differential outputs from the opamp to be A,C and B,D instead of A,B and B,D. It is working.
Tomorrow, I will replace one bridged channel filter for ABD type.
The modification picture is not the last version.

 

[JayGilb]

Good luck. I personally would not use enameled wire for the modifications.

 

[Hayk]

What is wrong with such wires.

 

[Roland68]

Good luck. I personally would not use enameled wire for the modifications.

It's actually ideal for such delicate modifications. I've been using enameled wires for this kind of work for about 40 years, not just in audio and IT, and especially at higher temperatures. You should use enamels specifically designed for winding motors/armatures; then you won't have any problems.

 

[JayGilb]

What is wrong with such wires.

Nothing. Just a personal preference as stated, especially where wires cross, I prefer a plastic insulated wire.

 

[Audionaut]

The TPA3255 is AD modulation and without option for BD modulation to transform it into ABD.

What would be the Advantages of ABD?

 

[Hayk]

All classD amps have a discontinuity in the output transfer function because of the necessary dead time, see picture below.
With ABD modulation, it becomes continuous with slight nonlinearty similar to mosfet outputs.

 

[Hayk]

I have very good news.
I left one channel with on board filters and the second channel with my ABD filter with a switch to select the output while listening.
As the carrier is now in phased bridge, the carrier residual which was very high with AD mode, now it has cleaned off.
The sound with the on board filter is by far cleaner than before. Switching back and forth at low level, about 1w, I didn't hear significant change, as if the switch is stuck. Unfortunately, yesterday, I catched cold and I am not fit to listen anything loud today. I will try to record some.
The scope image is that of the on-board filter residual which is just the EMI noise of my neighbor's refrigerator.

 

[Hayk]

If with music, I (sick) cannot hear any difference between the 2 filters, with pink noise, there is very clear difference. The ABD has extra Haa high frequencies which can be due to higher resolution. The sound file is windows m4a

These are the spectrum of noise with volume zero. The ABD is far more cleaner.
Of course the frequency response is much high that should reach 100khz, it is for tomorrow.

 

[Hayk]

I recalled that, before, I was dealing with tiny sounding amp prior of modification to 4xSE. A powerful radio sound, I should say. The brightness is gone, the sound has large soundstage now.
The chip must have some kind of flaw when it switches to BTL, it must do something mutual to the channels, or else it would not be necessary to have BTL and 4xSE modes.
I have a commercial PA 2x400w. Tomorrow I will compare with it to see if it still bright.

 

[Hayk]

The 3255, hides a secret about its dead time management. The Pure path is explained as adaptative to driver delay skew to keep the dead time minimum and avoid reject at production. This technique allowed high performance and also lower cost.
I wanted to adjust the bias ac current necessary, it is similar to biasing class B by listening at low level. Using the speaker as headphone with mw power, to my astonishment, it needed only 10ma with 2.2k resistor. Of course by my generosity, I doubled it to be 1k heating 0.5w where tpa3110-16 required 100ma. This means, TI has adopted an adabtative dead time vs switched current invented by STmicro. At low currents, where little charge needs to be cleared the dead time for STA516 for example, is 10ns and reaches maximum 50ns.
As I switched to BD modulation, at quiescent, the current now is that of the bias,10ma instead of 870ma with AD modulation, making it run with very low dead time.
I noticed the output switched waves are not clean rectangular, see picture below. I will try snubers as used by NXP if I can make it better shape. It needs high current 2.5Ap current capable 220pF.

 

[Hayk]

I installed this monster and adjusted the gain to be identical. I listened a Cuban, wide spectrum music at moderate level, I can't say the 2 amps have identical sound but I will exegrate if the difference is clearly audible by switching back and forth. The 3255 in ABD mod has slightly more pronounced high frequencies but both are identical in all else.
I will try to find a way to put in evidence the difference by recordings.
Something for 100% sure, it is no more loud radio sounding amplifier as you hear in the below video

 

[Hayk]

I recorded the 20s of the beginning the Shaft theme. The level is low, but the best I could do.

 

[Hayk]

The ground wire was wrong, it should run along the output wires. See pictures.
No need for snubbers.
Today, I will record longer sessions to post in MP3. I am impatient to get out of my flu, to hear really, how high is the resolution of this amp to compare with TPA3110. This modest chip, has 80db PSRR vs 65db for 3255. This means, it has 4 times more error corrector, but 3255 needs less error correction and the winner is........

 

[Hayk]

Here is the ABD filter without the CM choke, replaced by 1mH.
It is optimized for 8 ohm. If 4 ohm used as is -3db falls at 20khz. For perfect 4 ohms the last 2 10uH inductors need to be shorted optional.
This is simulation to start the PCB, values can change on real circuit.
The 7A inductors made of TDK pq2012 with flat wires cost only $0.8 from Ali. The capacitors, I will let the user chose Wima, or Epcos from 10mm or15mm pitch.

 

[Hayk]

I listened again with better ground wire the pink noise. I couldn't hear now any difference. I recorded on Audacity, the first half is TADA amp, the second 3255 with ABD filter. The spectrum analyzer doesn't see any difference neither.
Today, because of the flu, I have a forest of cicadas in my ears that listening loud music is out of question.

 

[Hayk]

I have designed TPA3255 based amp in PBTL. Now I know it can not sound good, I will not built it although I have the PCB and all the components. In this amp, to drive in differential signals with opamps, instead, I used another kind of interface that I developed for TPA3116. It is a differential VI converter, as it drives the TPA3255 with high impedance in audio range, it provides extra 20db NFB. Below is the circuit. I will try to replace the opamps with this circuit, using the part of the PCB and see if it gives any further enhancement as there are some reviewers who pretend to hear better sound from $70 discreet opamps.

 

[Hayk]

I prepared last evening the 3110 vs 3255 ABD, to compare. This morning, while the cicadas are still sleeping, for 15mn I could clearly hear the difference. Not any frequency depending or texture, it involves the presence, the 3110 is more forward. The 3110 is direct at its input, what I can guess, the replacement of the opamps of 3255 will effect this character only.
Some people, who appreciate harmony rather than melody, don't like front staged sound, the back staged gives more 3d effect and better for background music.
Using tube triodes at the input, can bring more forward.

 

[Hayk]

Once, I have designed for LM3886 a tube input stage, with Russian low cost dual triode 6n23, an selectable front-rear staging circuit. I will design a similar type with 3 position, front, neutral and back stage selectable. This triode can function in VI mode down to 30v, 48 will be fine. It will require to add either a 6v >200mw transformer or step down 48 to 6.3v DC converter. Such circuit increases considerably the value of the amp.

 

[Hayk]

Filter PCB. 80x80mm. From JLCPCB costs $4 for 5 pieces.