TDA7293, exempt of crossover distortion.

[Hayk]

This chip can function 2 units in parallel, called modular. Bellow is the amp and the crossover with single stand alone.

In stead of directly, if they are linked by a 10uf capacitor the result is this.

For more engineering explanation refer to:

Discovered a way to "erase" the crossover distortion in class B amps. - Page 1

Discovered a way to "erase" the crossover distortion in class B amps. - Page 1

www.eevblog.com

 

[audio_tony]

This looks like Class S (Sandman) - Technics (Class AA) used a similar topology.

 

[Hayk]

I know very well this circuit, It has a low power class A controlling a high power class B. Here, there is no any class A amp just a pair of class Bs.
See the explanation on EEVBLOG.COM.

 

[KL....]

This chip can function 2 units in parallel, called modular. Bellow is the amp and the crossover with single stand alone.
View attachment 480667

In stead of directly, if they are linked by a 10uf capacitor the result is this.
View attachment 480668

For more engineering explanation refer to:

http://[URL]https://www.eevblog.com...ase-the-crossover-distortion-in-class-b-amps/

[/URL]

View attachment 480666

Thank you @Hayk, nicely done, is this a modular TDA7293 implementation of what you are suggesting or is it a variation? To ask....

• what level of efficiency are you expecting, such as 80%/90%/95%/etc?
• what THD+N/SINAD are you expecting, such as >=95Db/etc?
• what output wattage are you expecting into 2R/4R/8R/2uF loads?
• the TDA7293 is a MosFET implementation, isn't it? What would you expect with GaNFET implementation/s? Can this be integrated into ClassD implementation?
• and anything else that you would like to mention :=)

From eevblog, also for discussion, hope that you do not mind.... Discovered a way to "erase" the crossover distortion in class B amps

• The class B amps, as are the majority of push pull solid state amps, get the output impedance increase at the crossover region. To cancel the nonlinear transition I parallel with another output stage of small power linking with a capacitor. The capacitor will advance the phase of the current in the auxiliary output so that when the primary is at the crossover, the auxiliary has already passed that region and its output impedance is lower to overtake the loading. Bellow is a simulation in class C to observe better the behavior although non biased outputs do not erase in perfection as in class B. In real world, I used 2X TDA7293, where the second with unity gain output paralleled with a 10uF switched. The scope pictures shows stand alone crossover distortion and the second with paralleled, sorry it's noisy. See the pictures here

• This is a simulation of the circuit I described above. It has 0v bias and running at 50hz in open loop. You can notice it is exempt of crossover distortion/region. The bias voltage, not current, is necessary here only for charging the capacitors of the transistors in less time for higher frequencies. 1v is sufficient for lowest distortion at 20khz with 32ma current bias.

 

[Hayk]

Thank you KL for your interest with this project.
This amp is intended to be Accurate Compact Audio Amp intended for desktop home studio monitor. It has 2 modules to plug in and determine its characteristics.
It is designed to output 2x30w@8ohm and 2x35w@4ohms.
One module is error corrector of my creation that reduces the THD+N by 200 times. A second module can be configured in 2 ways, either it have high even harmonics for leisure listening
https://www.diyaudio.com/community/threads/chip-amp-mosfet-composite.403832/
or high forward sounding for attentive investigating.
This chip amp is exempt of internal termal feedback which result fabulous bass.
It is equipped with speaker protection.
Bellow you can see, the PCBs are prepared and the enclosure has the power supply in lower seperate but bolted to the upper amplifier part.
From JLCPCB, you will get 5pcs for $4 for each board.
The total cost will be less than $150.
I also have prepared a low cost type with LM3886 and error corrector only smashing THD+N by 400 times, no modules, no crossover distortion eraser but costs less than $50 with 2x48v 2A PSUs for 2x40W@4ohm.
Be patient as I am dealing with 7 project in parallel, I'll post on YouTube explaining how to built and give comparative listening.

 

[KL....]

Thank you KL for your interest with this project.
This amp is intended to be Accurate Compact Audio Amp intended for desktop home studio monitor. It has 2 modules to plug in and determine its characteristics.
It is designed to output 2x30w@8ohm and 2x35w@4ohms.
One module is error corrector of my creation that reduces the THD+N by 200 times. A second module can be configured in 2 ways, either it have high even harmonics for leisure listening
https://www.diyaudio.com/community/threads/chip-amp-mosfet-composite.403832/
or high forward sounding for attentive investigating.
This chip amp is exempt of internal thermal feedback which result fabulous bass.
It is equipped with speaker protection.
Bellow you can see, the PCBs are prepared and the enclosure has the power supply in lower separate but bolted to the upper amplifier part.
From JLCPCB, you will get 5pcs for $4 for each board.
The total cost will be less than $150.
I also have prepared a low cost type with LM3886 and error corrector only smashing THD+N by 400 times, no modules, no crossover distortion eraser but costs less than $50 with 2x48v 2A PSUs for 2x40W@4ohm.
Be patient as I am dealing with 7 project in parallel, I'll post on YouTube explaining how to built and give comparative listening.

Thank you @Hayk, there will be many questions, won't there be?
You have indicated the following....

• The TDA7293 has 0.005%THD 1khz 5W which, in this configuration, gets reduced to 0.0001% plus THD of 2N7000 of 0.0005%, result should be less than 0.0006% in total.... is that THD+N/SINAD or only THD?
• It is designed to output 2x30w@8ohm and 2x35w@4ohms....
  • this suggests that the amp is Load Dependent/Variant (Open Loop) and Current Limited, doesn't it? Can you clarify/expand on this and further explain?
  • the TDA7293 is capable of 70~100W/8R (TDA7293/100W into 8Ω @ THD =10%, with VS = ±40V) but your implementation is 2x30w@8R, isn't it? Can you clarify/expand on this and further explain?
• Heatsink requirement/s....
  • what do you recommend?
• PS requirement/s....
  • what do you recommend?
• the TDA7293 is a MosFET implementation, isn't it? What would you expect with GaNFET implementation/s? Can this be integrated into ClassD implementation?

https://www.diyaudio.com/community/threads/chip-amp-mosfet-composite.403832/.... are these comments still relevant for TDA7293 implementation?

• I love the design of this amp with the 3886 + fet input. My only apprehension is the floating PS ground being neg speaker output, which essentially makes it cap coupled. There has to be a way to DC couple it and still retain the midrange magic you describe. It would be better to get the PS caps out of the speaker signal path, since the caps can be an avoidable source of HD at lower frequencies. The slight bump in the LF area (warmer mids you describe) is probably caused by increased resistance and H2 generated from caps in series instead of solid ground. Have you tried bridged output with second chip or even 4x 3886 and only single supply rail? You could even do fully balanced input this way, probably better common mode rejection too.
• The purpose of the MOSFET is to simulate the transfer function of single ended Triode. It is not a VAS as the load, the power opamp negative input is very low impedance. The MOSFET acts as V/I converter and the power opamp as I/V converter. The total NFB @20khz is 70db instead of 24db for the LM3886 alone with 26db gain. In post 1 you can see the noise is about 46db lower than stand alone.
  

TDA7293 Specs/Features

• Multipower BCD technology
• Very high operating voltage range (±50 V), DMOS power stage
• High output power (100W into 8Ω @ THD =10%, with VS = ±40 V)
• Muting and stand-by functions, No switch on/off noise
• Very low distortion, Very low noise
• Short-circuit protected (with no input signal applied), Thermal shutdown, Clip detector
• Modularity (several devices can easily be connected in parallel to drive very low impedance/s)

 

[Hayk]

Do not confuse this new amp with the old DIYAUDIO one.
It is possible to make the TDA7293 roar 60w in single. This will require big transformer, big chassis, big heatsink, why then chip and not discreet?
About the output power with 4 ohms.
If you look the specs of Japanese 70's amps, you see similar power 8 ohm vs 4 ohm. The technique is to use low power transformer, 100w Toroide, here and small reservoir capacitors, 3000uF/rail in total so that the power supply sags and avoid the transformer to burn with 2x60w output.
To obtain very low THD+N, by using the Error Corrector module. You get 240 times lower THD+N. The mosfet 2n7000 version does reduce the distortion of the chip but it is adjusted to generate on purpose its own harmonics.

How to make chip amps sound as beautiful as SE triodes do? - Page 1

How to make chip amps sound as beautiful as SE triodes do? - Page 1

www.eevblog.com

I'll posted on EEVBLOG.COM how my error corrector, that can be implemented on existing amps, to design, to add 2 zeros in the THD+N number. It is very simple.
GaN transistors can't be used for linear applications. For classD amps, I have 2 creations, both are lower grade amps but my class ABD modulation although 2x25w has extreme high resolution that I qualify it, 21 century sound. Unfortunately, unlike this amp, I could not design at will, the kind of sound I would like with classD chip amps as the feedbacks are inherent.
If you want to design higher power amp using 3x7293 or more, first built mine when the final schematic is published, then you design the PSU and the heatsink size for you required output power.
To get genuine TDA7293 from Ali, I find it on this store.

I just found this on AliExpress:
US $2.16 | New Original TDA7293 TDA7293V ZIP-15 120V 100W High power amplifier audio amplifier chip

https://a.aliexpress.com/_c38EoriT

Everything I bought from this vendor are genuine.

 

[KL....]

Do not confuse this new amp with the old DIYAUDIO one.

• It is possible to make the TDA7293 roar 60w in single. This will require big transformer, big chassis, big heatsink, why then chip and not discreet? About the output power with 4 ohms. If you look the specs of Japanese 70's amps, you see similar power 8 ohm vs 4 ohm. The technique is to use low power transformer, 100w Toroide, here and small reservoir capacitors, 3000uF/rail in total so that the power supply sags and avoid the transformer to burn with 2x60w output.
• To obtain very low THD+N, by using the Error Corrector module. You get 240 times lower THD+N. The mosfet 2n7000 version does reduce the distortion of the chip but it is adjusted to generate on purpose its own harmonics.

• How to make chip amps sound as beautiful as SE triodes do? - Page 1

  How to make chip amps sound as beautiful as SE triodes do? - Page 1

  www.eevblog.com
• I'll posted on EEVBLOG.COM how my error corrector, that can be implemented on existing amps, to design, to add 2 zeros in the THD+N number. It is very simple.
• GaN transistors can't be used for linear applications. For classD amps, I have 2 creations, both are lower grade amps but my class ABD modulation although 2x25w has extreme high resolution that I qualify it, 21 century sound. Unfortunately, unlike this amp, I could not design at will, the kind of sound I would like with classD chip amps as the feedbacks are inherent.
• If you want to design higher power amp using 3x7293 or more, first built mine when the final schematic is published, then you design the PSU and the heatsink size for you required output power.

Thank you, @Hayk, re Do not confuse this new amp with the old DIYAUDIO one.... good to know but you did post the link, didn't you?

• re PS.... do you like SMPS? If yes, which one would you use, for your TDA7293 configuration?
• re I'll posted on EEVBLOG.COM how my error corrector, that can be implemented on existing amps, to design, to add 2 zeros in the THD+N number. It is very simple....100 times lower THD+N/SINAD, is that correct? Can you provide a SINAD example of your TDA7293 configuration?
• re GaN transistors can't be used for linear applications. For classD amps, I have 2 creations, both are lower grade amps but my class ABD modulation although 2x25w has extreme high resolution that I qualify it, 21 century sound.... perhaps you will show and explain these, when ready?
• re If you want to design higher power amp using 3x7293 or more, first built mine when the final schematic is published, then you design the PSU and the heatsink size for you required output power.... how efficient is your TDA7293 configuration? What level of efficientcy are you expecting, such as 80%/90%/95%/etc efficient?

 

[Hayk]

You need to use the DS of both 7293 and 7294 for Pd.
Although it is bootstraped it has high drop out. In my configuration, the main, the master is not loaded so it doesn't contribute to output power. Where have we seen masters work hard as slaves should. You can add as many slaves you want without decreasing the slew rate because each slave will provide the master VAS it's dc bias current increasing the VAS bias hence increase the transconductance of the VAS mosfet.

 

[KL....]

You need to use the DS of both 7293 and 7294 for Pd.
Although it is bootstraped it has high drop out. In my configuration, the main, the master is not loaded so it doesn't contribute to output power. Where have we seen masters work hard as slaves should. You can add as many slaves you want without decreasing the slew rate because each slave will provide the master VAS it's dc bias current increasing the VAS bias hence increase the transconductance of the VAS mosfet.

Ah @Hayk, thank you, yes, the master is a VAS, the slave is the output stage, and the slave will provide the master VAS it's dc bias current which increases the VAS bias (GFB), hence increase the transconductance of the VAS mosfet. The output Impedance/V/I can only be delivered by the salve/s, isn't it? Quite Novel, thank you....

 

[Hayk]

Exactly. The master's output is linked to the slave(s) output by a 10uF capacitor only. This why the master who is controlling the output remains cool hence exempt of termal feedback that compresses the low frequencies and increase the low frequency distortion. Here, the slave(s) only are sweating. This chip can be powered by +/-50v. You can use 3 slaves on heatsink and the master on its own small one and get 2x200w @4 ohm with +/-45v.
Only when you decide the output power then you can know what power supply you need.

 

[KL....]

The master's output is linked to the slave(s) output by a 10uF capacitor only.

Thank you @Hayk, yes, the choice of 10uF capacitor will be quite critical, won't it? Which 10uF capacitor do you utilise and recommend?

To ask, why can GaNFETs not be utilised for this implementation?

 

[Hayk]

For SMPS used in Audio are these 3 types. The bottom one is 250W 2x36v +2x15v +12v all isolated. It requires small modification to add 36v zener so that both polarities are regulated.
The right one is the holly grail of SMPS, it has PFC input stage to feed the isolated step down with a clean 400v and the step down is LLC noiseless 250w 48v.
The top one is a new type and lower cost 500w 48v, it has out of feedback loop a huge ferrite inductor and 6800uF capacitor that cleans off all ripples including 100hz.
You need to replace the 6800uf 50v with 63v.

 

[Hayk]

Across the 10uF capacitor, there is near 0v, it works only at crossover and high frequencies in current mode only. On the model I used Nichicon high current 16v capacitor but on the PCB, it is 1206 smd 50v X7R MLCC ceramic capacitor used also for psu decoupling. Ceramic capacitors are not good with voltage variations but excellent with high impulse currents as is the case here.
Here we are dealing with chips, GaN is discreet component that cannot be installed on heatsink and has very low power dissipation. It is designed to operate at high speed switching with very low Rds.

 

[KL....]

1206 smd 50v X7R MLCC ceramic capacitor used also for psu decoupling. Ceramic capacitors are not good with voltage variations but excellent with high impulse currents as is the case here.

Thank you, fascinateing, a resonant/G/LFB (HF?) mechanism, very/very good/low THD+N. You mentioned Class C and/or S, didn't you? You mentioned TDA7293 & LM3886, can this be done with other chips, if yes, which ones? What is the Ripple and PSRR of your implementation? Would you expect any issues with 2R or 2uF loads, if yes, what would they be?

I was hopeing that this/our discussion would entice others to join the discussion but.... we shall see, won't we? Quite Novel, isn't it?

 

[Hayk]

Only classB can be implemented.
With LM3886, not easy to implement crossover eraser. The LM3886 has a secret feature that can considerably reduce the termal feedback. By using an error corrector reducing the THD+N by 400 times compared to the DS application, it gives a very good result but inferior quality to 2X TDA7293. It is intended for low budget DIYs who cannot afford $150 so that with less than $50 can get tiny high quality amp.
Another candidate for crossover eraser is the Chinese replica of TDA2050-LM1875. 10pcs cost less than $2 Inc. ship. These power opamps are unity gain stable and can be paralleled without resistors. It can be mounted in 2x5 bridge to output 200w@4 ohms where one on each branch is secrified to drive the other 4 and erase the crossover.
The crossover distortion eraser does effect the PSSR and the 2R 2uf loads.

 

[KL....]

termal feedback

Thank you, what is termal feedback?

inferior quality to 2X TDA7293

Thank you, how inferior, in what way/s?

Another candidate for crossover eraser is the Chinese replica of TDA2050-LM1875. 10pcs cost less than $2 Inc. ship. These power opamps are unity gain stable and can be paralleled without resistors. It can be mounted in 2x5 bridge to output 200w@4 ohms where one on each branch is secrified to drive the other 4 and erase the crossover.
The crossover distortion eraser does effect the PSRR and the 2R 2uf loads.

Thank you, cheap chips :=). With the TDA7293, what PSRR and the 2R 2uf loads do you expect. With the Chinese replica of TDA2050-LM1875, what PSSR and the 2R 2uf loads do you expect.

What level of Efficientcy do you expect from the TDA7293 implementation. In what way/s do you expect a PS to affect this implementation?
What level of Efficientcy do you expect from the Chinese replica of TDA2050-LM1875 implementation. In what way/s do you expect a PS to affect this implementation?

Only classB can be implemented

Thank you, ah, ok

You have mentioned that MosFET chips are not suitable. What are the reasons for their unsuitablety?

 

[Hayk]

Thermal feedback is a big problem in designing chip power amps. The transistors are temperature dependent, when the outputs go hot and cold, the low level circuit transform the temperature variations into a feedback signal. The designer will make it to be negative or else it generates distortion. But now, when a sudden beat is to amplify, the sudden heat will make the low level to decrease the output power and compress the bass notes. This is why, when the master is not loaded, it doesn't compress.
Exempt of crossover distortion and without thermal feedback, the amp is as good as discreet classA amp.
I didn't mention that mosfet chips are not suitable.
The Chinese TDA2050-LM1875 have temperature protection but no current limiters.
The slew rate is limited 2.5v/us.

 

[KL....]

Thermal feedback is a big problem in designing chip power amps. The transistors are temperature dependent, when the outputs go hot and cold, the low level circuit transform the temperature variations into a feedback signal. The designer will make it to be negative or else it generates distortion. But now, when a sudden beat is to amplify, the sudden heat will make the low level to decrease the output power and compress the bass notes. This is why, when the master is not loaded, it doesn't compress.

Thank you, Ah yes, nice that the Master does not suffer from Thermal drift/Thermal feedback although the Slave/s will though, won't they? Why are they not affected?

Exempt of crossover distortion and without thermal feedback, the amp is as good as discreet classA amp.

Yes, without ClassA heat & inefficiency? What level of Efficientcy do you expect from the TDA7293 implementation.

I didn't mention that mosfet chips are not suitable.

Apologys, my mistake, perhaps that should be 'not as suitable', is that correct? Can you explain this further, in more detail and if you have tryed Mosfet chips?

The Chinese TDA2050-LM1875 have temperature protection but no current limiters. The slew rate is limited 2.5v/us.

It would appear that the TDA2050 LM1875 interchangeable with each other? 2.5V/us is is a moderate-to-high slew rate, isn't it?

The TDA7293 slew rate is 15V/μs, LM3886 slew rate is 10V/μs.... for this implementation, what would you consider as a minimum acceptable slew rate and what would you like the slew rate to be?

 

[Hayk]

The slaves are only outputs controlled by the constant temperature master.
On #9, I posted necessary curves to determine the efficiency and the required heatsink size. See again below.
For example. 8 ohms loaded, +/-35v supply, while pouring 32W output, it will require 32w also to dissipate, hence efficiency 50%. With 1W output it needs another 12w dissipation, efficiency 7.7%.
2.5v/us is good for 20w, but when bridged it will double, then good for 80w 8 ohms or 160w 4ohms.
TDA7293 is a mosfet chip.