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

[Kokoriantz]

How to make a chip amp sound as beautiful as SE triodes do?
This is engineering forum and not a place to discuss aesthetics. The Japanese SET lovers analyzed the Marantz 8b in 60's to understand why it sounds as beautiful as a SET to discover the distortion spectrum to be similar . Since Jean Hiraga promoted this idea in his publications, it became for many designers a recipe to follow and that I will apply here too.
The surviving class AB chip amps are now very cheap and good ones, like cheap Italian table wine in cardboard container, makes many happy ,but it is not to compete with class A open loop triodes. These chip amplifiers have only 50db open loop gain at 20khz, once 26-30db feedback applied, it leaves only 24db NFB where 60db is needed for HQ sound.
I added a transistor to the chip to provide the nonlinear Ic/Vbe square law function and add extra 36db NFB to the closed loop. To show its efficiency, the chip stand alone with application circuit has 70uV noise in audio band spectrum, with transistor added it falls to 1uV, all other anomalies are corrected    with the same ratio of 70 fold.
I did realize this amp , with +/-25V and large heatsink the sound was perfect but nothing special, it passed the most difficult music, Mozart's Requiem in perfection, the high frequencies are exceptionally clean but the music lacks enthusiast . I replaced the heatsink with a tiny one and supplied  with +/-20V where the chip dissipates a near constant power of 12W eliminating internal thermal feedback. I truly assure you this tiny amp is adorable. It takes me about 2 weeks to tune an amp, I'll keep updating.
 I have already done similar circuit with LM3886 and 6N6P tube with 250V floating supply, this amp is as good and much simpler.

[oz2cpu]

>sound as beautiful  ??

if the goal is to ONLY amplify, and not add or remove anything,
tubes fail, transistors and opamps win

if the goal is to add and remove, affect the sound in a way that some people find nice,
then tubes will most likely win, since they add 2nd harmonics

[Kokoriantz]

Look at the spectrum, the main distortion is second harmonic as tubes do.

[janoc]

Make DSP filter replicating the distortion response of the tube amp?

However, ultimately you are going to be going against a brick wall because the issue there isn't the "beautiful sound" but that it isn't "tubes" and you thus won't be able to convince anyone who thinks that tubes sound/work better that your "chips" work as good or better as the tube amps.

It would be an exercise in futility, IMO as this isn't about measurements and science but beliefs and personal preferences.  And that's long before you go as far as outright audiophoolery...

[oz2cpu]

or just get the real thing and be happy,
the orange glow at night is so much also adding to the total love feeling.
https://fairman.dk/kt88amp.htm

[TheUnnamedNewbie]

Make DSP filter replicating the distortion response of the tube amp?

However, ultimately you are going to be going against a brick wall because the issue there isn't the "beautiful sound" but that it isn't "tubes" and you thus won't be able to convince anyone who thinks that tubes sound/work better that your "chips" work as good or better as the tube amps.

It would be an exercise in futility, IMO as this isn't about measurements and science but beliefs and personal preferences.  And that's long before you go as far as outright audiophoolery...

In the guitar/bass guitar amplifier world there has been a lot of shift in mindset I can see where people are embracing/accepting DSP-based emulators over tube amps. Esp when you consider the advantages, such as a DSP based amp being able to switch to a totally different model at the literal push of a button (on stage) without you having to drag around multiple 30kg amplifiers.

[Kokoriantz]

You are contradicting yourself by using dsp to simulate tube sound and convince anyone to be tube sound without tube. The transistor with its square law Ic/Vbe is similar to tube transconductance function, The spectrum of the distortion similar to SE triode, is it audiophoole or science? For you it is science if DSP simulated but audiophoole if done by transistor.
I have several SET amplifiers from 300B to 6L6GC to compare with. If these amplifiers cost several hundred of dollars for 8W, this amp cost the price of a pack of peanuts for 14W with very similar sound, The LM1875 (fake), I got 10 pieces for $1.2 including postage, I would feel guilty l not to post such a circuit that can satisfy most exigent ears.       

[Kokoriantz]

I explained, the chip amp is by far not a perfect amp. Before you add spices with DSP, the chip must be perfected with additional 36db NFB. A single transistor does both functions, why trouble with ADC,DSP,DAC to achieve such a simple square law function.

[Haenk]

Generally I think an amp should be just that - it should amplify as linear as possible. Your audio source should be as linear as possible. You can then adjust the listening experience by using loudspeaker that you like, or (more likely, as very much cost saving) modify your audio signal before the amp. As a good amp and good loudspeakers are the expensive parts of the reproduction chain, one usually want them to be as universal as possible.

Modifying an amp to sound like you like it best - why not. *You* are the listener. Howevery this might not work for anybody else.

[Alex Eisenhut]

Well if you can read French I recommend Francis Ibre's books if you can find them...

Might not have what you are looking for but full of interesting and useful information.

https://www.eyrolles.com/Sciences/Livre/bien-entendu-itineraire-d-un-audiophile-9782866611613/

[janoc]

You are contradicting yourself by using dsp to simulate tube sound and convince anyone to be tube sound without tube. The transistor with its square law Ic/Vbe is similar to tube transconductance function, The spectrum of the distortion similar to SE triode, is it audiophoole or science? For you it is science if DSP simulated but audiophoole if done by transistor.
I have several SET amplifiers from 300B to 6L6GC to compare with. If these amplifiers cost several hundred of dollars for 8W, this amp cost the price of a pack of peanuts for 14W with very similar sound, The LM1875 (fake), I got 10 pieces for $1.2 including postage, I would feel guilty l not to post such a circuit that can satisfy most exigent ears.     

But you know, it wasn't me who asked "How to make chip amps sound as beautiful as SE triodes do?" So who is fooling himself here? 

For you it is science if DSP simulated but audiophoole if done by transistor.

I have never said that. My point was that trying to coerce a transistor or "chip" amplifier to sound like a tube amp is ultimately a futile effort because you won't convince anyone who cares about the "tube sound" with it. Not because it doesn't or couldn't be made to work but because they believe their "golden ear" and no measurements will convince them unless they see a glowing tube in the box.

Whether it is a DSP doing it or you have some sort of magic discrete circuit managing to pull that off doesn't matter squat.

[Geoff-AU]

How to make a chip amp sound as beautiful as SE triodes do?

First you have to persuade yourself that the audio signal wiggling its way through many paths and branches inside an IC, is similar to how water is filtered when it passes through a bed of sand.  Then, and only then, can it compete with the immense power of electrons jumping across an airless gap between metals.

[james_s]

I just think tubes are neat. I don't know or care if they actually sound "better", but they look cool, smell nice, and there is something almost magical about them.

[Kokoriantz]

I agree with you 100%. I, "Son of TV repair" when child, I started by repairing dial thread of tube radios, I had great love for AD1 tube sound. The simpler it is, the fresher, "presence", ii sounds. Even direct heated gives fresher sound. The chip has 3 stages, each, degrading the transients but when feedback applied of very high NFB, the 3 layers, sound as a single stage. Here, the amp works as 2 stage, driver transistor and output chip. I have tried numerous composite types using high quality op amp as driver, I  got very pure sound but degraded presence, this why I used tubes beforehand and now transistor. I will try out jfets and Mosfets to see what it sound like.
To measure the presence, I use piano synthesizer software and measure the ratio of the first peak to that of the fifth and compare with the input. Comparing the input, output spectrum of such signals gives also good information.       

[Terry Bites]

You mean get the kind of distortion you prefer.

[james_s]

You mean get the kind of distortion you prefer.

That's about what it comes down to. It's worth remembering that there is no such thing as no distortion though. A solid state amplifier can be very clean but still everything in the chain from the room acoustics where the music was performed, the microphones and recording gear, the processing done while mastering it, the master recording media, and then everything in the playback chain. The speakers no matter how good are probably the largest source of distortion, I can believe that certain forms of distortion in the amplifier could result in a subjectively better or cleaner sound.

[TimFox]

To show how much misunderstanding exists about even vs. odd distortion, I had posted in an audio forum about the general opinion that odd (especially 3rd) order distortion sounds worse than even order distortion, with an analogy to 3rd order IM in communications (which generates spurious junk near the two original tones that cannot be filtered away like 2nd order junk near the 2nd harmonic of the two tones).
Someone agreed, because he calculated the third harmonic of A-440 as 1320 Hz (so far, so good) and then looked up the musical scale and found that 1320 Hz did not exist in the tempered musical scale (falling between E-1318.5 and F-1396.9).  Of course, the scale frequencies he found were for "equal temperament", which is an engineering approximation useful for tuning pianos that have only 12 notes per octave.  Harmonic distortion occurs at harmonic frequencies (integer ratios).

[Kokoriantz]

I don't know what miracle happened, I bridged two identical amps feeding the second by the emitter from the output of the first, the distortion remained exactly the same. Note how the second follows inversely the first high up to 200khz.
I don't know how to apply current feedback in bridge to eliminate the Thiel LR.   
Edit. The power measured is not 1W but 4W for 0.05% THD+N

[Circlotron]

I just think tubes are neat. I don't know or care if they actually sound "better", but they look cool, smell nice, and there is something almost magical about them.

I agree.
And to my way of thinking they have a number of similarities to an open fireplace, both functionally and aesthetically.

[precaud]

I just think tubes are neat. I don't know or care if they actually sound "better", but they look cool, smell nice, and there is something almost magical about them.

So the logical thing to do is just use them as a visual prop, then.

[Kokoriantz]

The LM1875 is not suitable to be bridged as the TDA2050 is. The power dissipation with +/-18v and 4 ohms gives near constant dissipation. The DS preconizes small heatsink to run at high temperature and output 30W, that is 60W bridged. I transformed the circuit to single supply 36V and now the the THD+N fall to 0.01%@1W, the resistor R25 adjust the subtraction of even harmonics to reach bellow 0.001% with 500 ohms. Decreasing the bias of the lower transistor, the THD hikes to 0.1%, this gives a wide adjustment means to tune the amp Tube-SS type sound.   

[james_s]

I just think tubes are neat. I don't know or care if they actually sound "better", but they look cool, smell nice, and there is something almost magical about them.

So the logical thing to do is just use them as a visual prop, then.

But that's just stupid, I absolutely loathe fake stuff. It's only cool if it actually works.

[Haenk]

I just think tubes are neat. I don't know or care if they actually sound "better", but they look cool, smell nice, and there is something almost magical about them.

So the logical thing to do is just use them as a visual prop, then.

Isn't there a great chinese tube-preamp available, doing just that? OPAMP doing the preamping stuff and the tube is just for showing off, powered up, but not connected.

[Warpspeed]

Oh no !
It needs to be a proper tube amplifier, built with oxygen free wire, gold electrical connections, shock mounted, facing East, and blessed with holy water.

[Gyro]

Oh no !
It needs to be a proper tube amplifier, built with oxygen free wire, gold electrical connections, shock mounted, facing East, and blessed with holy water.

No, a properly designed tube amplifier will do just fine. No need for the other crap.

[niconiconi]

You are contradicting yourself by using dsp to simulate tube sound and convince anyone to be tube sound without tube. The transistor with its square law Ic/Vbe is similar to tube transconductance function, The spectrum of the distortion similar to SE triode, is it audiophoole or science? For you it is science if DSP simulated but audiophoole if done by transistor.

I have to say I strongly agree with the OP. This project has a clear and objective goal: Design a solid-state audio amplifier in such a way that it produces a distortion somewhat similar to a vacuum triode amplifier. The OP did not claim that tube amplifiers have a better "audio quality" due to their mysterious properties, instead, just that their unique distortion produces a sound that some people find to be subjectively "beautiful". OP also used spectrum analysis to quantify the distortion. So I think it's unfair to call it "audiophoole". Both an analog circuit and DSP are valid solutions to produce this effect. You may find it to be a pointless exercise, but you can think of it as an exercise of the art of analog signal processing. It's pointless in the same sense that painting, fishing, or software demoscene are pointless.

Finally, my general thoughts on audiophoole: Historically, Hi-Fi audio have two types of things. The first type attempts to create a transparent signal path with minimum distortion, the modern trend is pushing for increasingly extreme THD, IMD, and other specs you can measure on an Audio Precision analyzer. On the other hand, the latter deliberately use distortion to create a subjectively "beautiful" sound or an artistic effect. However, both approaches got mixed up and both are labeled Hi-Fi, due to intentional or unintentional factors. For example, it's intentional for electric guitar amplifiers to create sound effects, it's unintentionally for loudspeakers since even faithful amplifier designs from the tube era still have significant residue distortions (which can give an amplifier an unique "personality" - which really is just distortion). Later, this becomes responsible for various questionable "audiophoole" practices, which are mainly attempt to create unintentional distortion by mixing and matching audio gears in uncontrolled experiments, and claiming a mysterious change of sound represents a better "sound quality" - which can't be farther from the truth...

However, in my opinion, if you openly admit that you just like amplifiers with distortion as a personal preference, and acknowledge that they're not really "High-Fidelity" by the very definition of the word, I think there's nothing wrong with that.

P.S: Interestingly, the exact same phenomenon exists in photography. Many fans of vintage lens claim that these analog-era designs are just better and each has an unique personality, while modern designs are lifeless. The reality is that, even many good lens back then had significant optical imperfections. Meanwhile, modern lens are as transparent as technically possible, as measured by their Modulated Transfer Function (MTF) - the optical equivalent of |S21|. But I think nothing wrong if you prefer distortion as an artistic effect, as long as you don't claim the "image quality" is mysteriously better.

[Kokoriantz]

In recent years, vinyl records made their comeback, not because they reproduce with higher fidelity then CDs, because they sound more agreeable. The professional quality pick up cartridge, Shure V, measured by Audio magazine shows 2% 2nd and 0.5% 3rd harmonic distortion with 3% IM distortion. Before 1947, the ideal amplifier was the one that had 5hz-100khz bandwidth and 0.1% distortion which was achieved by Williamson amplifier. Amplifier specs mentioned "distortion-inaudible" if it was less. To promote commercially better amplifiers, magazines convinced the audiophools that lower is the distortion, better it sounds, despite the distortion of the loudspeaker, which hardly goes bellow 1% at high and mid frequencies and reach above 10% at 50hz.           

[james_s]

I got into collecting vinyl several years ago after somebody gave me a box of records and I decided to fix up a half decent turntable I had been given earlier by someone else. Growing up in the early 80s vinyl records were the standard way people listened to music and being a kid both the records and equipment I had were worn out junk so I remembered records sounding scratchy and not very good. Now when I played a lightly used record on a decent turntable I was amazed at how good it sounds. Does it sound better than a CD? Probably not, but it sounds quite good, and there's something almost magical about such high fidelity sound coming out of something so simple. I mean if I woke up in the year 1000 BC with a record I could probably cobble together something to play it with, by the 1400's reasonable fidelity would be possible, especially if it was a record intended for early mechanical phonographs. No way that is achievable with a CD. Useful exercise? No, but it's fun to think about.

[niconiconi]

To promote commercially better amplifiers, magazines convinced the audiophools that lower is the distortion, better it sounds, despite the distortion of the loudspeaker, which hardly goes bellow 1% at high and mid frequencies and reach above 10% at 50hz.           

At this point, I don't think low-distortion loudspeaker playback (in the audio engineering sense) is even practical for the average home user, any attempt to reduce distortion at the amplifier further only has diminishing return. Ultimately, the room adds a lot more distortions, including resonances that are detrimental to the audio quality. For the purists, headphones are a much better option to tinker with. If you can find some good chips and closely replicate the vendor's reference designs, it's perhaps even possible to achieve their world's class datasheet performance in your own circuit board without a rack of test gears.

Years ago, I've watched a hilarious presentation by an audio engineer (not in English, otherwise I'd share it). In which he pulled an article about a Hi-Fi studio room build from a Hi-Fi magazine, and pointed out and debunked all the mistakes and misconceptions one by one. It was a pretty funny talk. The audiophools spent many thousand dollars to purchase top-of-the-line loudspeakers, DACs and amplifiers, including questionable accessories like boutique cables and an air ionizer (which was supposed to mysteriously improve the transparency of the air) for their purpose-built studio. But they were unable to identify serious acoustic problems of their studio, which could be fixed rather inexpensively for $1000 or so. They did try to treat the room but it was done in a totally incorrect manner. Because these practical problems (1) require background knowledge and experience to diagnose, and (2) the solutions are not expensive, (3) but they're also not easy to apply, since it requires room renovation, and (4) definitely cannot be fixed by selling you some audio gears. You rarely hear these discussions from audiophool vendors.

[Kokoriantz]

The only modification applied till now is rebiasing the transistor back to 0.2ma with +/-20v supply. It needs 5mn to heat up to get the tonality correct or else it is too bright.
I want to remind that I am using fake LM1875 costing $0.15 and all looks perfect. I didn't try short circuit or temperature protections yet.

[Kokoriantz]

Little by little, enhancing this amplifier by listening, I ended up to the configuration bellow. I paid great attention to the high frequency character as I am persuaded no good low frequencies can come out of chip amps, specially with split supply. The loop gain now is 70db at 20khz on simulator with LM3886 where it is 26db only stand alone. The very high frequency is so exceptional that I decided to make it tweeter dedicated amplifier above 1khz with adjustable damping from 0.5ohm to 5 ohms. I was about transforming it into single supply that I tried the MOSFET instead of the BJT. I had in drawer 2N7000. The sound metamorphosed into very high quality full tonal sound. The tweeter range is less pronounced but very smooth all give to be tube sound, but what astonished me is the very deep bass got resurrected. I want to remind that I am using Chinese TDA2050 replica costing $1.3 per 10 pieces. 
As the footprint of BC550 and 2N7000 are the same order, I will put on socket to be chosen for type of music.

[Kokoriantz]

I will try a different chip amp, the TDA1514A. This race is extinct since 2 decades, it was used in Cambridge Audio A1 first version and Marshal guitar amp. On web, there is lot of complaints since a decade to have replacement parts explode, even sources from reliable retailers advised by Marshal. These were original Philips components but defective factory outlet, I think, as I bought TDA7293 for 1.5$, it is original but slave master select pin is shorted to Vee to function only in slave mode. 
From nearby vendor, I found a Chinese replica with laser graved along the Philips logo, 1.5$ probably a decade old. whereas on Ali, I found also laser graved also 1.5$ but honest producer without any logo. I also bought a pair for 2.6$ with original type blue print no logo in the picture, but I received with logo and written Taiwan.
By measuring with multimeter, The two from Ali are exactly identical whereas the old replica has very slight difference.
This IC has 12db more open loop gain than LM3886, and has complimentary outputs with common emitters. On Youtube there is example of the sound recorded,impressive.       

[jonpaul]

If you really seek great sound, I suggest opamp made for audio like NE5534, use US made parts not  Chinese knockoffs.



Jon

[Kokoriantz]

Thank you Jean Paul for advice.
Using opamp to compensate opamp, is like leading a blind by another blind. The purpose of this composite is not to get low distortion numbers, but highest NFB possible. The problem as I explained in another thread, the compression that takes place because of 90°in open loop that only the NFB ratio can bring it back correctly to 0°. Just a single transistor is sufficient to bring the NFB at 70db@20khz.
Patents of components is only 20 years, after that every manufacturer can produce. When you by a BD139, you don't insist to be Telefunken, any other generic is fine.
The Chinese are producing so cheap TDA2050 equivalent where there is no any other source is available, so why not use it if it is good. Soon or late, the linear original chips are going obsolete, only the Chinese can provide us such replicas. 
Unfortunately the TDA1514 I got are all original manufactured in,96,99,and 2006. No new replica yet.
I prepared today a the same circuit but with LM3886, I will try comparatively and report it tomorrow. I think with such high NFB, the sound of the chip gets nutralized to give place to the transistor to express its character. Several years ago I tried 6N1P in transconductance mode with LM3886 with gain of 10db, I didn't get such extraordinary clean high frequencies.   
Unfortunately 

[Kokoriantz]

Here is an honest vendor on "Ali" selling both original and Chinese replica of TDA7294 for 1$ less.

https://www.aliexpress.com/item/1005003385598335.html?spm=a2g0o.productlist.0.0.2718603amDXkY1&algo_pvid=f2def68a-6334-45e9-ac2a-3cc4f34301e1&algo_exp_id=f2def68a-6334-45e9-ac2a-3cc4f34301e1-51&pdp_ext_f=%7B%22sku_id%22%3A%2212000029896853556%22%7D&pdp_npi=2%40dis%21USD%212.58%212.58%21%21%21%21%21%40210318c916706569554074855e8d23%2112000029896853556%21sea&curPageLogUid=UZLqRX2g2iOZ

I will order both types and compare.

[cncjerry]

You need those rocks you set on your speakers too.

[Kokoriantz]

If you want original ones from Mouser, they are not available and cost about 11$. Just 5 years ago Mouser was selling the 7293 for 2.5$. How much it will be in 10 years.

[jonpaul]

Rebonjour, the distortion modes of opamp, negative feedback and solid state amps was studied in 1970s, 1980s.

See papers in JAES, preprints, by Todd, Jung, and others.

The warmth of the classic valve amplifiers like 1960s Marantz is due to soft limiting, and even HD rather than odd HD.

Suggest you find some vintage audiophiles with Marantz or similar tube amps and perform listening and laboratory tests.

Finally some solid state amps with good performance are in diy or production eg from Nelson Pass.

Just the ramblings of an old retired EE

Jon

PS...I tribute to the enduring products and   fine engineering by my friends and mentors,

Sol Marantz, Sid Smith, Dick Sequerra..(RIP)

[Kokoriantz]

 Salue Jean Paul. I lived in Paris about a decade and met several audiophiles as we were regular customers of La Maison D'Audiophile shop including once Jean Hiraga who was selling there his Lectron amp. I built then my first 300B SET. To be more clear, The first SET I had I (son of TV repair) was 12 years old which was a chassis of German radio from late 30's with AD1/NF2.
The Marantz 8b is described in Jean Hiraga's book That you see the spectrum compared to 300B. The designer Sydne smith, used purposely a dirty IF/detector tube 6BH6, biased at very nonlinear region to provoke square law function. My opinion is that the frequency domain is not the reason but in time domain. when a pinched cord, drum beat, consonants, all have a high positive pulse first a followed by less negative to decay gradually, If the amp amplifies more the positive then the negative ones, the sound becomes better pronounced with more presence. It has nothing to do with saturation as it acts at low levels.
In my circuit I use the transistor with CCS to get the same square law as S. Smith did once upon a time, nothing new here under the sun.
The question is, why with BJT, I have less bass than MOSFET? I think the input impedance of BJT depends upon the amount of feedback it gets on the emitter. If the chip, because of internal thermal feedback for example pours less bass, than the input impedance will be very low and the input capacitor will act as high pass filter. With MOSFET, the impedance is always high.
Why the BJT has more pronounced high frequencies? I have no idea.
Here in Asia including Russians, fine female voices are the main reason of HIFI, they need to hear this very smooth high pitch as Mireille Mathieu or Mathe Altery kind of. A few weeks ago, two motorbike taxi drivers were listening Jazz music from an amplified speaker, Never I have seen Thai ordinary people listen to Jazz, They were simply listening to the very smooth clean sound of the saxophone that the speaker/amp designer succeeded to reproduce. 


   

[Kokoriantz]

Great, great deception. I made the LM3886TF work. First, the extraordinary high frequencies disappeared but the sound is very clean. The fabulous bass is halved but what makes me say no thanks, its very small sound, smashed dynamics. I listened about 15 minutes, I got bored.
I was all wrong saying the very high NFB neutralizes the chip, I was lucky to have a good working chip at start, If I have it started with this chip, probably I would have been discouraged and abandoned  the idea.
I will try next with TDA7293.

[Circlotron]

The Marantz 8b is described in Jean Hiraga's book That you see the spectrum compared to 300B. The designer Sydne smith, used purposely a dirty IF/detector tube 6BH6, biased at very nonlinear region to provoke square law function. My opinion is that the frequency domain is not the reason but in time domain. when a pinched cord, drum beat, consonants, all have a high positive pulse first a followed by less negative to decay gradually, If the amp amplifies more the positive then the negative ones, the sound becomes better pronounced with more presence. It has nothing to do with saturation as it acts at low levels.
In my circuit I use the transistor with CCS to get the same square law as S. Smith did once upon a time, nothing new here under the sun.

You must feel horrified when sitting glistening to a live performance where there  is no opportunity to put an amplifier that is doing it's stuff between you and the music! 

[Kokoriantz]

Because your loudspeakers are perfection, mine not.

[Sherlock Holmes]

How to make a chip amp sound as beautiful as SE triodes do?
This is engineering forum and not a place to discuss aesthetics. The Japanese SET lovers analyzed the Marantz 8b in 60's to understand why it sounds as beautiful as a SET to discover the distortion spectrum to be similar . Since Jean Hiraga promoted this idea in his publications, it became for many designers a recipe to follow and that I will apply here too.
The surviving class AB chip amps are now very cheap and good ones, like cheap Italian table wine in cardboard container, makes many happy ,but it is not to compete with class A open loop triodes. These chip amplifiers have only 50db open loop gain at 20khz, once 26-30db feedback applied, it leaves only 24db NFB where 60db is needed for HQ sound.
I added a transistor to the chip to provide the nonlinear Ic/Vbe square law function and add extra 36db NFB to the closed loop. To show its efficiency, the chip stand alone with application circuit has 70uV noise in audio band spectrum, with transistor added it falls to 1uV, all other anomalies are corrected    with the same ratio of 70 fold.
I did realize this amp , with +/-25V and large heatsink the sound was perfect but nothing special, it passed the most difficult music, Mozart's Requiem in perfection, the high frequencies are exceptionally clean but the music lacks enthusiast . I replaced the heatsink with a tiny one and supplied  with +/-20V where the chip dissipates a near constant power of 12W eliminating internal thermal feedback. I truly assure you this tiny amp is adorable. It takes me about 2 weeks to tune an amp, I'll keep updating.
 I have already done similar circuit with LM3886 and 6N6P tube with 250V floating supply, this amp is as good and much simpler.

Get a very linear, low distortion, small signal AF amplifier and incorporate a DSP and software controller. Feed the output into a equally linear power amplifier.

Now, use software to emulate any number of historic valve (tube) based designs from yesteryear.

Get people to pay for "emulation packs" so they can have their favorite sound at the click of a switch!

[Coordonnée_chromatique]

Get people to pay for "emulation packs" so they can have their favorite sound at the click of a switch!

The space-time distorsions of the soundscape generated by the historic valve (tube) (and their associated loudspeakers) is impossible to reproduce with a software emulation IMHO.

[Kokoriantz]

On a planet, there are two straight lines to go from point A to point B. The short one, just a single transistor and a long one, all around the planet.
(janoc) on post 3 did proposed also the same. The question is, at what cost? Chip amps are intended to be low cost, compact solution. With the competition of class D, cheaper and better in time as the TPA3255, they are subject to extinction. This project is a survival temptation.     

[Sherlock Holmes]

Get people to pay for "emulation packs" so they can have their favorite sound at the click of a switch!

The space-time distorsions of the soundscape generated by the historic valve (tube) (and their associated loudspeakers) is impossible to reproduce with a software emulation IMHO.

I agree, but if we introduce an interrossiter, we can get part of the way there.

[langwadt]

Look at the spectrum, the main distortion is second harmonic as tubes do.

as an asymmetric output stage does ...

[Kokoriantz]

I think the Chinese TDA2050/LM1875 is not a replica of 30 years old technology. It can probably be a new design, with superior performance. I don't believe there is so much difference between LM3886 and LM1875 sound. If the 3886 is so bad, the 1875 can't be much better.
I have 3 different chips, I used only one, I will try the two other ones if get identical sound.
This circuit can be best suited for self powered full range, midrange or tweeters. I will proceed with single supply capacitive coupled version, maybe a pair of chips in parallel. What limits the power then is the power supply voltage. Low cost SMPS have max 48v can be adjusted up to 56v. If this Chinese chip can withstand it as LM1875 should, then I can forget looking other type of chips. 
There is a type of single supply that is best suited for low frequencies, is the floating ground, shown bellow, I call it floating supply as each channel requires its own supply. This type pulls full rectified signal current from the supply vs half rectified with the standard type which requires large reservoir capacitor to everage it.

[ethernet]

What a hero! Trial and error against a baseline. I have only just discovered valves (and germanium for radio restorations) and wish somebody used JFETS or the like to fake a valve. Someone did, with a P channel as well - but there were limitations. I speculated a 70 cent CPU could be added for active faking of a proper valve with some preset lookup tables. (over a resistor/capacitor averaging value). You could load up harmonics to get that sound. Presently hoards of RF radios are being robbed of valves and trashed so someone can have a nice sounding Beatles distortion, or whah whah pedal. I have seen only one person discussing adding mosfets to valves to make them better. Yes the valve class A designers knew this - but hey, valves were super expensive back when. There are plenty of youtube vids showing replacing a .cn preamp chip, with a better one. Did Marantz go over the top - to create a signature sound? Adding the right amount of harmonics is a just project. I also know a lot of work on matching paired transistors for AB to reduce that kink that was heard.  You will note Sony just recently managed to knock Bose Noise cancelling headphones off its #1 perch. I speculate these newer programmable FFT NC chips have the capability to really mess/simulate  the output of a near perfect preamp. Thinking some more, hearing aid wonderchips, will have to add active harmonics where there is total deafness. If you see a Marantz anything in the dumpster grab it. If you look on Ebay, you will see they are worth a fortune. I also note transformer weight , chokes and smoothing cap values also impact the final 'sound'.   

[ethernet]

Found a reference
http://www.geofex.com/Article_Folders/mosfet_folly/mosfetfolly.htm
"Where to NOT put MOSFETs in a tube amp. In general, not anywhere that you count on for gain or distortion" . See we agree!.

[vad]

How to make a chip amp sound as beautiful as SE triodes do?

Have DSP to add 2nd harmonic. That’s all that single-ended tube power amps really do. Audiophiles find excessive 2nd harmonic distortion beautiful and satisfying.

Added: SE transistor amp with MOSFET operating in quadratic region is a poor man’s way of adding 2nd harmonic without either the DSP, or tubes and heavy output transformers.

[Kokoriantz]

 Chip amps are intended for the poor people.

[Kokoriantz]

Thanks for the jfet idea, I have k117,k174, bf245 bf247. I'll try.

[bson]

Look at the spectrum, the main distortion is second harmonic as tubes do.

You also need to look at the phase... spectrum alone does not describe a response or transfer function.

[Kokoriantz]

You are evoking a very important issue. The Phase of the second and third order harmonics have paramount importance. As I already presented, that to my opinion, the time domain is more clear view than the frequency domain. If the nonlinear function amplifies more on the negative going signal, it does generate the same pattern harmonics but the sound will be much different. This happens with 3 stage SET amplifiers where the second harmonic phase is inverted. More dramatic case is the third harmonic phase. In phase, it will tend the sine wave towards square wave and reduce the dynamic as in time domain, the gain will decrease with level, the contrary is when the phase is opposite yielding dynamic expansion.     

[Kokoriantz]

I tried out the two other Chinese LM1875 and TDA2050. They are still much better the the first which was several years old version. The high frequencies are further step above in smoothness and resolution. Both are identical although I payed the TDA the double price. I listened 4 hours nonstop with great leisure without fatigue.
I will try now destructive tests, I have 20 pieces. If the same die is replacing the high current of TDA and high voltage of LM, then it should certainly be higher power than TDA. I will test the Short and thermal protections also.   

[vad]

Odd harmonics are unpleasant and should be avoided. Even harmonics, and especially the 2nd, are perceived by some people as beautiful and pleasant.

The simplest way of adding pure 2nd harmonic in DSP is to apply square function and then remove DC component: cos(w*t) squared equals (1 + cos(2w*t))/2. This power reducing trigonometric identity is also at works in SE triode amplifiers and SE FET amplifiers.

Similar Taylor transforms with 3rd, 4th and higher terms can be applied to add 3rd, 4th and higher harmonics, if needed.

Adding distortion in DSP is fun, can be done in controlled manner, can be adjusted to audiofool’s taste in software or firmware rather than in hardware as in case of tube and FET audiophile amps, and most importantly it can be turned off with a single line of code.

[Kokoriantz]

Odd order harmonics when of opposite phase are sound enhancers. First they decrease those of the loudspeaker which are in phase and results less angry voice. Increasing more, it acts as expander and reveals sounds you don't hear normally. Of course this can be done very easily with DSP.   

[Terry Bites]

Other crap is vital to wanktek industry.

[Kokoriantz]

I tested the Chinese TDA2050. It has drop out of 3V on each rail similar to LM1875. With 42V single supply I could get 19W before clipping , much lower than TDA2050. With output shorted, it got destroyed although only 2A power supply and 2500uF total supply capacitors.
High quality sound but fragile. I order other TDA2050, with the feet and backplate similar to the original, but without any logo, $1 for 3 pieces.
   

[Kokoriantz]

A nice fellow has posted on Youtube a comparative listening of original and Chinese LM1875. You can hear how the Chinese has better low frequencies pronounced (less thermal feedback) and how the high frequencies are less aggressive.

https://www.youtube.com/watch?v=sV_bvP6Hi5U&lc=Ugxv4ai9FnV22T0qVt14AaABAg.9jg_gmLxMEJ9jh35BJlKFY

[Kokoriantz]

This is single supply with floating ground circuit. It has further deeper bass compared to the split supply as the output current gets full rectified from the supply instead of half wave double current in standard single supply. I thought I invented this but I saw a Russian one from late 80's. I added a 5A regulator LM338, I didn't notice any difference, but with measurements it will result better S/N ratio.
I will develop two PCBs from this circuit, A full range audiophile version with 4×2200uF audio quality supply capacitors and a high frequency version using the low cost CCTV power supply, to be used as mid range tweeter amplifier. If the C1/C3 ratio is kept 3, it forms a Butterworth  high pass filter can be adjusted to 800Hz, above the value of the capacitors become in pF range and I don't know yet how to reach tweeter cut frequencies up to 4khz.
The floating ground needs to be earthed with a SMPS to eliminate the buzz, the problem with the CCTV PSU, it doesn't have any, I must figure out how to eliminate the buzz.

[Kokoriantz]

I could make the amp function buzz free without using the earth. It was my fault, I have grounded the stability compensating capacitor C4, it should use the + input reference. With 97db speaker my ear on the speaker, total silence without LM338 regulator. I will  add a common mode choke and reservoir capacitor instead of regulator for the tweeter amplifier.   

[Kokoriantz]

As I have now two models, split supply with TDA2050 and single floating supply with LM1875 both Chinese, I listened comparatively. The high frequencies of TDA are one step better than the LM, the better pronounced low bass of LM can be due to power supply. As I doubted the two are not identical, I proceeded to destructive test of LM1875.
Split supply by two 24v adjusted +/-19 to +/-28.8 with the +input ground and the -input pulled by 100k ohm to either rail for 5 minutes nothing happened with full 57.6v.
I linked the 100k to the output and ran without heat sink with 57.6v. First it is stable with unity gain, it started about 50ma and the thermometer reached 85°C to shut down and remained oscillating between 77°C to 82°C. Cooled and restart with the same quiescent current. It is perfectly temperature protected.
Tomorrow I will try short circuit test.

[Thunderer]

I tested the Chinese TDA2050. ... With output shorted, it got destroyed although only 2A power supply and 2500uF total supply capacitors.

One should be aware of the energy in the caps. The PS (rated 2A) will not limit the output power of the system.

For your case, 2500uF and 42V, you get 2.2J of energy in 1ms equivalent of 2.2kW to dissipate in the chip+wires (enough to fry lots of more powerful output stages - speaking here even of some MJ150xx power transistors), let's say we consider 10ms, and we still get 220W - too much even for a genuine chip.

Rule of thumb: when testing, one should disconnect the caps (or keep some 100uF - still 80W in the first ms  ) to prevent frying an output stage of an amplifier.

[Thunderer]

I could make the amp function buzz free without using the earth. 

Why not go for the classic ground lifter to eliminate ground loops. Care to insulate the 0V from the enclosure/chassis.

[Kokoriantz]

I proceeded short circuit test, no, it didn't survive. I opened the LM1875 $0.15, it is identical to TDA2050. Compare to the original ones picture bellow found on the web.

[SiliconWizard]

Many people have spent years in R&D, some others have done their PhD thesis around that. Have at it and have fun!

[Kokoriantz]

Thank you for your interest on the subject. I see two different types of ground circuit breaks, from Rod Elliott site and Hifisonic. One earths the chassis and links the whole ground with 10 ohm +100nF 630v + a 10A paralleled bridge diode. Other, the power stage ground is earthed along the chassis and the input signal ground is separated with 10 ohm + a pair of diodes or 100nF capacitor.
I deal my experiments with a non earthed DAC with optic link, to protect the DAC, I use 1k ohm with the earth wire, but the input and the output grounds don't have 10 ohms circuit breaking. If the DAC was earthed, of course it would be necessary to break.     

[Kokoriantz]

If you are referring  the intellectual property, these chips are in public domain since long time. If get a glimpse on R&D personnel in US chip design quarters, you will find majority of them to be Chinese. Why one of them wouldn't design with uptodate technology a power opamp and sold it to a Chinese factory?
 

[Kokoriantz]

I received the TDA7294 Sanxin version. The inscriptions on the chip has been erased but still visible to guess it is the right chip and had ST logo. On measurement it behaved very different to the original TDA7293. The output to Vee the original measures small leakage with analog ohmmeter but the Chinese one measures high impedance until I touch other pins and conducts about 5k until I short all the pins by a foil. This means, it is MOSFET outputs and a different design. The Tab which is -V signal  to Vee has 2.2M ohm vs 20M ohm for the original.   

[Kokoriantz]

I made the Sanxin TDA7294 work with split supply +/-24v. The sound is very high quality throwing the LM3886 into shame. The high frequencies are of superb but that outstanding character of the Chinese TDA2050/LM1875 is not reproduced. The low frequencies are more firm but less pronounced, it can be due to split supply. I will try it with dual +/-32v. 
For multiway amplified speaker, no doubt these Chinese can play magestral duo.

[Kokoriantz]

Using dual +/-33v supply didn't change anything. I don't know why the low frequencies are not as good as the TDA2050/LM1875. With signal generator it goes flat down to 10Hz. On the other hand it has the same spectrum of sound as my SET300B/2A3C with 5hz-100khz OPT. After listening two hours without any fatigue, the outstanding character of this chip that jumps undoubtedly  to the ears is the voice range. It sounds very similar to 45rpm records with very clear presence, but as full range amp it doesn't give luxurious sound as the tiny ones do. 

[Kokoriantz]

With my simple brain, I was thinking that a pair of SMPS for split supply, renders weak low frequency performance because each supply must start and stop as the current is half wave so with single SMPS with dual supply will have the current full wave and produce more powerful bass. It is not the case, because as I added a pair of 10,000uF low ESR reservoir capacitors, the bass got resuscitated. The high frequencies got better as I changed the grounding of the ceramic 100nF low level supply decoupling. 
This shows the superiority of single floating supply which doesn't need such expensive capacitors but on the other hand requires independent PSU for each amplifier.

[Doctorandus_P]

How to make a chip amp sound as beautiful as SE triodes do?

Put muffles in your ears and step on the tail of a nearby cat.

[Circlotron]

Using dual +/-33v supply didn't change anything. I don't know why the low frequencies are not as good as the TDA2050/LM1875. With signal generator it goes flat down to 10Hz.

I would say there are no  big differences in the performances of the two amplifiers. What maybe is different is you like one amplifier more than the other. How you feel about them affects how you perceive the sound from each one. To like something you built is fine. To claim it sounds better than something else without measurable proof is where it all falls apart. It’s like saying orange is better than blue. Not everyone will agree.

[Kokoriantz]

It is possible to record pink noise by microphone played by each amp and show the spectrum as the one bellow found on:

Such tests I do when tuning the same amp for best as I need to compare the same amp before and after modification after several days.