About tube DACs?

[cotdt]

kad, I can't seem to understand what point you are trying to make.

 

[kad]

kad, I can't seem to understand what point you are trying to make.

Please forgive my convoluted logic?

What drove me to reply was the OP's statement:

No, I´m trying to get best possible sound for "reasonable" price. And pretty much all the best DAC's that I have heard have been using tubes.

So basically, the answer to your question is: I'm trying to minize transistor based coloration & distortion from the signal chain, since it seems to be worst kind of coloration for me.

This peaked my interest and triggered my reply. While I don't pretend to know which is 'better' than the other, as I believe there are specific tools best suited to each particular task; my point was that coloration is a property of vacuum tubes technology. Whether that adds or detracts from the enjoyment of music is a completely subjective and personal matter for individuals. Further amplification of the subjective topic here.

Back to topic: Speaking of buffering after the DAC resistor switch array, this damn fast buffer (now probably defunct since NS was bought by Ti?) would probably be a candidate worthy of consideration?

damn fast op amp datasheet and application note, data sheet, circuit, pdf, cross reference | Datasheet Archive
Abstract: National Semiconductor Semiconductors Linear I.C.'s- Operational Amplifiers LH0033C LH0033C, LH0063C LH0063C Fast and Damn Fast Buffer Amplifiers GENERAL DESCRIPTION The LH0033C LH0033C and LH0063C LH0063C are high speed, FET , amplifiers, op amp isolation buffer for driving reactive loads and high impedance input buffers for high , , see AN-48 AN-48. * FEATURES Damn fast (LH0063 LH0063) 6000V/uS Wide range single or dual supply operation , degrees Fast rise times 2 ns High current gain 120dB High input resistance 101°0 ABSOLUTE MAXIMUM ...

Now 6000 volts per microsecond slew rate, that ought to be able to provide decent square waves?

Just attempting to contribute to this discussion and these diyAudio community forum knowledge base, best as I can.

Cheers to all!

~~~~~~~~~~
To love is to be happy with.
~ Joe Vitale

I was just pursuing what I enjoyed doing. I mean, I was pursuing my passion.
~ Pierre Omidyar

Love is life. All, everything that I understand, I understand only because I love.
~ Leo Tolstoy

Clarity of mind means clarity of passion, too; this is why a great and clear mind loves ardently and sees distinctly what it loves.
~ Blaise Pascal, 1623-1662

 

[cotdt]

That's true... tubes will have their own coloration. The OP needs to be more specific.

How would one know that the reproduction will not have colorations? There are just too many factors involved.

 

[AlliumPorrum]

BTW I think the OP meant a tube output circuitry, not entirely a DAC made with tubes?

Well I didn't even know that there are so many different kinds of tube DACs...

But, if I have understood correctly, when we talk about "normal" commercial tube DACs, we are talking about tube output circuitry, aren't we?

And just to be sure; actually I really don't care about different technologies or implementation methods or if there are tubes or not. But, as I said, the best commercial DACs that I have heard have always been using tubes (on output?). And that's the only reason why I'm interested on them. I have heard lots of 3-5k€ solid state DACs, and none of them comes even close to DACute, for example. And I prefer the balanced MHDT over any < 2k€ DAC that I have heard.

 

[marce]

As the signal is digital up to the output of the DAC, the tubes are only going to have any affect on the analogue output. If you managed to build a DAC section from tubes, as long as it didn’t affect the signal integrity of the digital waveform, it would not make any difference to the output. But the question is could you build a DAC out of discrete tubes that would be as good as a silicon based solution, nope, and not with the rise times of todays digital signals, the stray parasitics between the tubes (caused by the long traces required to get the signals from a to b) would probably destroy any signal integrity, plus size and power requirements would be ridiculous. So for the actual conversion a DAC chip is going to be the best solution, with a tube output stage if you wanted. Not only that but building a discrete tube DAC stage is very likely to be noisy in EMC terms, so youd require a hugh great metal case to try and keep the extra EMC noise from polluting everything around you.
DF96 was probably referring to tubes getting less upset over stray RF than some op-amps, ie the op-amps likely to be used for the output stage of a DAC.

 

[kad]

Apropos, here is another high-end tube output DAC thread I stumbled upon. This may be of use from the perspective there is yet another "Maverick D1 Pre-amp" which unfortunately is a bit obscure to me, since diyAudio doesn't return much more info when searching for this term?

I was trying to disable the "direct" switch which bypasses the volume knob. The switch has led to trouble in the past, it will get flipped and blast my speakers with more wattage than they were ever designed for...

Anyway, I cut the cable directly next to the plug thinking that it would disable it... It didn't.. In the pictures it looks like a plug but it actually is not, which is why i had to cut it. There is plenty of room for me to re solder it.

But is there anyway for me to disable it?

The "direct" Switch is all the way to the left.

1. I would suggest carefully reconnecting whatever you cut.
2. It is possible to disable a bypass switch but you must find the appropriate portion of the switch used. Only then can you disconnect the wire going to that terminal.
3. It might be helpful if you removed the volume / bypass switch assembly? For inspection of the PCB traces? If the potentiometer and switch go to the PCB, it is possible no wire can be disabled. Thus you must look at the traces on the PCB and find the appropriate cut.

Please focus onto the potentiometer LEVEL (volume control) R47 as an example of the way your volume control works.

The bypass switch simply shorts the portion of the resistor on the wiper arm gliding over the surface. Thus 'direct' I suppose?

And then again, I could be wrong? Good luck!

~~~~~~~~~~
The less you know, the more you believe.
~ Bono, Irish Musician/Social Activist/Lead Singer of U2

There's no time to stop for gas, we're already late.
~ Karin Donker

No one but a fool would measure their satisfaction by what the world thinks of it.
~ Oliver Goldsmith

 

[AlliumPorrum]

As the signal is digital up to the output of the DAC, the tubes are only going to have any affect on the analogue output. If you managed to build a DAC section from tubes, as long as it didn’t affect the signal integrity of the digital waveform, it would not make any difference to the output. But the question is could you build a DAC out of discrete tubes that would be as good as a silicon based solution, nope, and not with the rise times of todays digital signals, the stray parasitics between the tubes (caused by the long traces required to get the signals from a to b) would probably destroy any signal integrity, plus size and power requirements would be ridiculous. So for the actual conversion a DAC chip is going to be the best solution, with a tube output stage if you wanted. Not only that but building a discrete tube DAC stage is very likely to be noisy in EMC terms, so youd require a hugh great metal case to try and keep the extra EMC noise from polluting everything around you.DAC.

Ok, now please let's forget these "discrete tube DACs" or such. As I said in the first post, "What I'm looking for is a tube DAC kit that would sound quality wise be a clear step up from MHDT, but wouldn't cost quite as much as those high end commercial tube dacs such as DACute".

If I have understood correctly, both MHDT and DACute use "normal" solid state DAC ships, and tubes on the analog output section.

 

[kad]

Well I didn't even know that there are so many different kinds of tube DACs...

1. But, if I have understood correctly, when we talk about "normal" commercial tube DACs, we are talking about tube output circuitry, aren't we?

2. And just to be sure; actually I really don't care about different technologies or implementation methods or if there are tubes or not. But, as I said, the best commercial DACs that I have heard have always been using tubes (on output?). And that's the only reason why I'm interested on them. I have heard lots of 3-5k€ solid state DACs, and none of them comes even close to DACute, for example. And I prefer the balanced MHDT over any < 2k€ DAC that I have heard.

1. You are correct, the output section (analog) could be done with one tube as exemplified in the above picture. So you would achieve your 'tube sound neutrality or warmth' resulting in a satisfying sound you prefer.

2. I do not believe such preamplifiers with tube output would / should cost that much more. At least in part component cost, it's probably only another $20 or $30 more? However, with the marketing appeal of 'tubes' this could go hundreds more? But certainly not thousands? And if they are, then it is a ripoff.

The most expensive part in any preamp (which has to use DACs nowadays anyway) will probably be the Digital to Analog Converter chip itself. Depending on if it is 16-bit, 24-bit or 32-bit deep.

It is simple to visualize mathematically:

2^16 = 65,536
2^24 = 16,777,216
2^32 = 4,294,967,296

From Wikipedia regarding Signal to Noise Ratio

So for a 16-bit DAC the Least Significant Bit (LSB) noise floor is: 96.3 dB

Similarly, for 24-bit DAC, the level of quantization can resolve: 144.5 dB

Finally, for 32-bit DAC, the level of quantization can resolve: 192.7 dB

But practically, and in reality, no analog circuitry to my knowledge can have a SNR as low as that, provided by the digital domain. Which means even a 32-bit DAC will be submerged into the analog noise floor. The best I've seen so far in audio implementation sits at around -105 dB noise floor.

So the info might be there digitally, except it is completely obliterated (buried within) by analog circuitry noise floor.

Anyway, I personally believe 32-bit DAC sounds better. But the difference in purchasing price isn't linear. 32-Bit DAC are much more expensive, due to the precision integrated circuit (solid-state) DAC. Not the tube output.

Also DAC require extremely tight precision 'resistor ladder' temperature controlled etc... It is a natural side effect of solid-state microelectronics to have all 'transistors' and intermediate passive components such as resistors matched by the simple fact they are built at the same moment using the same substrate on the chip die. It is an oversimplification, but it makes sense? Whereas in discrete macroelectronics, each resistor, transistor, tube, or whatever must be manually 'matched' with corresponding parts of effective similar value.

Another reason why I think solid-state DAC on a chip are superior to discrete component implementations.

~~~~~~~~~~
Amplifiers always oscillate; Oscillators always amplify.
~ Murphy’s Law

Experience is the comb that Nature gives us after we are bald.
{Belgian Proverb}

Everything should be made as simple as possible, but not simpler.
~Albert Einstein

 

[kad]

Ok, now please let's forget these "discrete tube DACs" or such. As I said in the first post, "What I'm looking for is a tube DAC kit that would sound quality wise be a clear step up from MHDT, but wouldn't cost quite as much as those high end commercial tube dacs such as DACute".

If I have understood correctly, both MHDT and DACute use "normal" solid state DAC ships, and tubes on the analog output section.

Yes you understood correctly.

I still would like to see actual implementations using tube output and a solid-state DAC chip be posted into this thread. I am particularly interested in the schematic diagrams? To see how it's done?

Just out of curiosity?

~~~~~~~~~~
Be less curious about people and more curious about ideas.
~ Marie Curie, 1867-1934, Polish-French Physicist and Chemist

It is not who is right, but what is right, that is of importance.
~ Thomas Huxley, 1825-1895, English Biologist

Only the curious will learn and only the resolute overcome the obstacles to learning. The quest quotient has always excited me more than the intelligence quotient.
~ Eugene S. Wilson, 1905-1981

 

[kad]

As a side note, the reason I believe 32-bit DAC signal processing sounds better than 24-bit or 16-bit is the 'staircase' effect above the SNR floor.

You can visualize the 'staircase' quantization between different types of DAC by the ratios (numbers of steps) above.

It follows that even a 16-bit Analog to Digital Converter (ADC) used to quantify any analog waveform, such as used in the recording industry to produce mere "CD Quality" will have gross amplitude quantization steps (65,536) of them. These are known as .wav files. A CD .wav file can be compressed without any bit-loss via the lossless FLAC codec. Usually the standard specifies 44.1KHz sampling rate.

Then comes "DVD-Audio" quality, mostly 16-bit quantization @ 48KHz sampling rate. But also allows 24-bit quantization @ 192KHz sampling rate.

Then comes "SACD" quality which is a single bit quantization (the prior bit defines the next level) but is sampled at 64x44.1KHz. Currently this is the ultimate in terms of Audio Quality.

The problem with analog circuitry is the noise is cumulative. Which means during the recording & mastering process, each track adds its own noise floor, substantially lowering the available dynamic range.

With digital recording, the noise isn't added, so ultimate recordings do not suffer from distortions, deviations, due to mastering recursive manipulation.

Here is a real-life example of why I believe 32-bit A/D and D/A sounds better. I bought a new receiver this month a Pioneer Elite SC-65.

It utilizes 32-bit DSP (Digital Signal Processing).

All my digital audio media accumulated over the years, now truly sounds better! Sweeter. More definition. Things I never heard before, such as the pluck of a string, or the slightest brush of a snare, now become clearly audible. It is like rediscovering a whole new library of music!

Really!

Pioneer Hi-bit 32

This YouTube will describe the process and is worth a watch:
Pioneer AV receivers - Hi-bit 32 - YouTube

This YouTube will describe the high-level technical details concerning the SC-65
Pioneer Elite SC-65 - YouTube

The nice thing about that receiver is that it has 9.2 analog outputs. So this is the DAC I'm satisfied with, for my entire sound system.

Hey, I just go for what really works, within my price budget.


~~~~~~~~~~
Anger and intolerance are the enemies of correct understanding.
~ Mohandas Gandhi

The true barbarian is he who thinks everything barbarous but his own tastes and prejudices.
~ William Hazlitt

An idealist believes the short run doesn't count. A cynic believes the long run doesn't matter. A realist believes that what is done or left undone in the short run determines the long run.
~ Sydney J. Harris (American Journalist and Author, 1917-1986)

 

[AlliumPorrum]

Hey, now I found one interesting option: Audio Note Kits - L3.1 DAC Kit

Does anyone here have experience on these? Not a cheap option anyway since balanced 3.1 costs way over 2k$, but how about the sound quality...?

 

[Julf]

As a side note, the reason I believe 32-bit DAC signal processing sounds better than 24-bit or 16-bit is the 'staircase' effect above the SNR floor.

There is no "staircase effect" as the DAC will have a reconstitution filter (sinc function).

It follows that even a 16-bit Analog to Digital Converter (ADC) used to quantify any analog waveform, such as used in the recording industry to produce mere "CD Quality" will have gross amplitude quantization steps (65,536) of them. These are known as .wav files. A CD .wav file can be compressed without any bit-loss via the lossless FLAC codec. Usually the standard specifies 44.1KHz sampling rate.

Then comes "DVD-Audio" quality, mostly 16-bit quantization @ 48KHz sampling rate. But also allows 24-bit quantization @ 192KHz sampling rate.

Then comes "SACD" quality which is a single bit quantization (the prior bit defines the next level) but is sampled at 64x44.1KHz. Currently this is the ultimate in terms of Audio Quality.

Eh?

FLAC allows 24@192, and that is definitely higher resolution than the 2.8224 MHz x 1 bit (64 x 44.1 kHz, but only 1 bit) DSD encoding used on SACD.

It utilizes 32-bit DSP (Digital Signal Processing).

Only 32 bit? For DSP you do need to use a much higher bit depth to avoid rounding and accumulation errors (precisely the cumulative errors you pointed out with the analog signal). That's why most DSP's are 48, 64 or even more bits - but the output is still only 24 or 16.

All my digital audio media accumulated over the years, now truly sounds better! Sweeter. More definition. Things I never heard before, such as the pluck of a string, or the slightest brush of a snare, now become clearly audible. It is like rediscovering a whole new library of music!

Really!

Enjoy the cool-aid!

 

[DF96]

Dither has a role to play too, but let us not introduce too many facts into the discussion.

I'm not quite sure what this thread is now about. 'Valve after DAC chip' DACs are quite common, aren't they? When properly designed they may sound as good as a conventional circuit; valves don't have to insert noticeable distortion, although some people seem to prefer it when they do. So what does the OP actually want that is not provided by existing designs?

 

[AlliumPorrum]

I'm not quite sure what this thread is now about. 'Valve after DAC chip' DACs are quite common, aren't they? When properly designed they may sound as good as a conventional circuit; valves don't have to insert noticeable distortion, although some people seem to prefer it when they do. So what does the OP actually want that is not provided by existing designs?

Nothing.

Once again, this thread is NOT about:
- some strange "discreet" DAC designs, or any DAC *designs* at all
- tube vs. solid state comparisons
- technical issues, measurements or SNR´s

What I would like this thread to be is:
- experiences and recommendations on high quality DIY tube kits
- comments on tube DIY kits against high end commercial ones

Thanks!

 

[cotdt]

You might be interested in taking a look at the Lampizator page. He is both DIY and commerical, and his tube output stage is well regarded.

 

[DF96]

Lampizator's designs and views are certainly not universally "well regarded". It is a while since I last looked at his website, but "best avoided" would be my verdict.

 

[kad]

Nothing.

Once again, this thread is NOT about:
- some strange "discreet" DAC designs, or any DAC *designs* at all
- tube vs. solid state comparisons
- technical issues, measurements or SNR´s

What I would like this thread to be is:
- experiences and recommendations on high quality DIY tube kits
- comments on tube DIY kits against high end commercial ones

Thanks!

Since you are the OP, I shall respect your wishes. Please forgive the next few replies (I feel compelled to correct some disinformation); after which I'll only focus onto your wishes which also happens to correspond to mine.

Cheers!

PS: Beer And Food and Music is good.

~~~~~~~~~~~
Think much, speak little, write less.
{French Proverb}

This world is comedy to those that think, a tragedy to those that feel.
~ Horace Walpole

I would propose that life is very much like a game and that a comparison between the two is very enlightening. In any game, you need to be fit, possess certain skills and have a good knowledge of the rules in order to participate. In life, we need to have a good knowledge of the rules that apply and to use this knowledge continuously in order to assess and make decisions in every situation that we face. The more intensely we play the game of life, the better our knowledge and application of the rules, the better the results we will achieve, much like a football player who is skilled, trains hard and knows the rules.

During the course of our lives we progress from being dependent on our parents for food and shelter and emotional support, to being independent as we become young adults. We learn to make decisions for ourselves, and as we see the results of our decisions we get feedback and the resultant experience adds wisdom to subsequent decisions.

The next stage is to develop relationships with others, or interdependence, and it is through these relationships that significant personal growth can occur. Many of us however have a low sense of self-esteem which holds us back in relating to others.

How then do we make significant progress in our lives, as evidenced by personal growth or an increase in wisdom? We look at the results we have and are achieving in our lives. If our results are unsatisfactory, we reassess our beliefs and make adjustments to our actions. Additional important feedback is our emotions (fear, anger, unhappiness, disappointment etc.) which acts as warning signs and which leads us to evaluate either our rules or beliefs, or our actions. Our target is to lead a joy-filled, intense yet balanced life, leading to good life results and continually increasing wisdom.

Author's Bio:
Tony Orbin is a chemical engineer with a passion for the field of self-improvement. He is in the process of starting to give courses in this field. Tony can be contacted on aorbin@iafrica.com.

 

[stephensank]

Hey, now I found one interesting option: Audio Note Kits - L3.1 DAC Kit

Does anyone here have experience on these? Not a cheap option anyway since balanced 3.1 costs way over 2k$, but how about the sound quality...?

I have a good deal of experience on a very, very similar, earlier AudioNote dac, with the same AD1865 dac chip & i/v transformers, but a different tube selection. I was not terribly impressed, even after upgrading many of the unit's shortcomings. I think they have taken a step down from that earlier unit on this one, in choosing the oddball 5687 tube for output. There is no good reason to use this tube versus a conventionally-pinned-out 12a*7 or 6DJ8 type, and I have always found it extremely close to impossible to find 5687's that both sound good & are not badly microphonic. All of the good, vintage ones seem universally microphonic. I have also found AN's circuit boards to be of fairly poor quality, with very thin aluminum cladding & poor quality plate-through holes. The transformers they use for i/v, despite what they tout about them, seem to be the limiting factor on sound quality, at least with their AD1865 based units. They don't seem to be nearly as good as they should be, or have ill-chosen primary impedance for the dac chip and/or ill-chosen pri:sec ratio.

My suggestion would be a Monarchy 24 dac, with a fair amount of upgrading/alteration. Very reasonable average used cost, and not too hard to make into a seriously superb sounding dac. Main thing they did wrong is to not factor in that the SRPP circuit used after the passive i/v resistor only sounds it's best at one specific load impedance, with the distortion rising sharply more than one percent above or below that value. So, the key is to either feed it to a preamp with a known input impedance that calculates, with the unit's internal 220k output load resistor, to the right(42k by my running measurements) ideal load impedance for the circuit, or add a very high quality buffer stage, with it's own input Z calculated to total 42k with the 220k unit's load. I prefer the latter approach. Sounds really spectacular. The failure to account for the SRPP's specific load demands is no doubt why the 24 has gotten mixed reviews, from glowing to tepid, as the stock unit will sound very, very different according to each preamp's input Z.

 

[kad]

Dear All:

Please forgive my multi-posts here, but I'm hitting the limit on the number of quotes within a single post reply. Thus I shall organize them by paragraphs.

As a side note, the reason I believe 32-bit DAC signal processing sounds better than 24-bit or 16-bit is the 'staircase' effect above the SNR floor.

There is no "staircase effect" as the DAC will have a reconstitution filter (sinc function).

Oh yes there is a staircase distortion. The DAC output low pass filter is designed to have a time constant following the Nyquist-Shannon sampling theorem.

Which basically states if one samples at 44.1KHz (44,100 samples per second) then the highest sinusoidal waveform that can be faithfully quantified and then later reproduced is 22.05KHz.

This only relates to the sampling interval. Then comes in the amplitude of the sampled signal, in the case of a 16-bit ADC/DAC means 2^16 or 65,536 levels of amplitude that can be captured.

What you call the reconstitution filter (analog low pass) must then be able to pass from 0Hz-dc to 22.05KHz ideally without altering the signal in any way, but must reject any distortion component above 22.05HKz by at least 95 dB, which is practically impossible to achieve using analog methods.

You can get a simple low or high pass filter using R/C constant (via a capacitor and resistor) of the first order @ 6dB/octave type. Second order filter @ 12dB/octave. Third order filter @ 18dB/octave. The problem is phase shift. The more filters sequences, the more phase distortion of the signal. What 'tube' audiophiles would call 'pace' or 'coherance' or 'sound stage'... While some others, more eclectics, use the proper terminology 'phase'.

Here is an excerpt on the subject:

Further details can be revealed by actually reading the PDF.

Much more detailed info here:
Fundamentals of Sampled Data Systems - Analog to Digital Conversion
(Of use only if one reads it, alas)

Oh do you remember the first high-power amplifier brand? Phase Linear (of Bob Carver's fame)?

And there are less anti-aliasing glitches with a smoother waveform rather than a gross one. So oversampling the 16-bit quantization at 32-bits depth produces substantial and tangible results. Towards perfection.

Have you even actually auditioned what I am talking about? I suspect not.

Bottom line: your 'reconstitution filter' is what affects the sound quality, or 'coloration' for some? And the exact same principles would occur whether using solid-state or vacuum tube technology (as far as the low pass filter is concerned).

PS: It would be nice towards the diyAudio members community, if you would post some independent verifiable substantiation to support your claims and beliefs. Just say'n.

~~~~~~~~~~
If you shut your door to all errors, truth will be shut out.
~ Rabindranath Tagore, 1861-1941, Bengali Poet and Writer

Education is the key to unlock the golden door of freedom.
~ George Washington Carver, 1864-1943, American Botanist

Excellence in any department can be attained only by the labor of a lifetime; it is not to be purchased at a lesser price.
~ Samuel Johnson, 1709-1784, English Author and Critic

 

[kad]

Continued from above.

(Apologies, I forgot to include a link to the PDF, so here it is.)

It follows that even a 16-bit Analog to Digital Converter (ADC) used to quantify any analog waveform, such as used in the recording industry to produce mere "CD Quality" will have gross amplitude quantization steps (65,536) of them. These are known as .wav files. A CD .wav file can be compressed without any bit-loss via the lossless FLAC codec. Usually the standard specifies 44.1KHz sampling rate.

Then comes "DVD-Audio" quality, mostly 16-bit quantization @ 48KHz sampling rate. But also allows 24-bit quantization @ 192KHz sampling rate.

Then comes "SACD" quality which is a single bit quantization (the prior bit defines the next level) but is sampled at 64x44.1KHz. Currently this is the ultimate in terms of Audio Quality.

Eh?

FLAC allows 24 @ 192, and that is definitely higher resolution than the 2.8224 MHz x 1 bit (64 x 44.1 kHz, but only 1 bit) DSD encoding used on SACD.

How so? Since the 'master' recording usually is in DSD format? And require conversion from DSD to PCM for most?
SA-CD.net - FAQ

Are you saying that a FLAC encoded @ 24 bits / 192Khz is definitely superior to SACD? Oh really?

Then please tell us where to find a DSD compatible DAC, high-end or low-end, tube or otherwise for under $2K? With a true 32-bit DAC and 32-bit Signal Processing?

While I might propose a potential solution?
Pioneer DSD SC-67

~~~~~~~~~~
Americans drink to get drunk, whereas in France, getting drunk is just a consequence of sampling too much wine you really like.
~ Ted Breaux, New Orleans Absinthe distiller

The mark of an immature man is that he wants to die nobly for a cause, while the mark of the mature man is that he wants to live humbly for one.
~ W. Stekel

There seems to be some perverse human characteristic that likes to make easy things difficult.
~ Warren Buffet