[Luckbad]

I made a list of current Multibit/R2R DACs, maybe someone will find it useful.

Great list! Only thing that comes to mind immediately that I don't see is Lampizator: http://www.lampizator.eu/Fikus/DACs.html

 

[kvik]

Not sure why I forgot Lampizator. The list was originally inspired by a NOS DAC list on Audiostream. I decided to look for further entries, and the list slowly grew, particularly the last year or two. As I haven't seen a similar current R2R upshot anywhere else, I thought it might find some use on HeadFi also, rather than just keeping it local (Hoved-fi). Anyway, I will of course add Lampizator.

 

[glina]

Not sure why I forgot Lampizator. The list was originally inspired by a NOS DAC list on Audiostream. I decided to look for further entries, and the list slowly grew, particularly the last year or two. As I haven't seen a similar current R2R upshot anywhere else, I thought it might find some use on HeadFi also, rather than just keeping it local (Hoved-fi). Anyway, I will of course add Lampizator.

For your list, the Audio Note 0.1x uses the TDA1543 while DAC1 up to DAC5 all use the AD1865. All have the CS8414 spdif receiver and of course, tube output.

 

[MojoAudio]

I guess each of us has to draw the line where we consider a "hybrid" DAC chip to be R-2R.

Personally I can hear notable differences between these hybrids and pure R-2R DAC chips.

As for the TDA1543, it was originally offered as the first "budget" DAC chip - this was what first allowed companies to market a CDP for under $500.

Among the things that made the TDA1543 so revolutionary and so low-cost to implement were the many integrated features.

It has a built in S/PDIF receiver, a built in demultiplexer (L and R signal separation), built in digital filters and output stage, and a single +5V power supply. Comparing this to the TDA1541 which required two additional ICs for S/PDIF receiving and demultiplexing, external digital filters and op amps, as well as a five voltage power supply, companies now had a way of building a CDP in a fraction the size and at a fraction the cost.

Fast forward to the revival of the R-2R by DIYers, and now you have a single DAC chip that can be used with multiple chips paralleled to improve linearity and lower noise that can be powered by a simple battery power supply or a single-voltage linear power supply. So if a TDA1543 is implemented in this modern paralleled fashion with a single +5V ultralow-noise power supply the result is a relatively high-performance DIY DAC for a relatively low cost.

This is part of why there are so many low-cost modern TDA1543 DACs with such relatively high performance.

Personally I prefer a more modern DAC chip, such as the AD1865, but since it requires a S/PDIF receiver and four dedicated ultralow-noise power supplies (+ digital, - digital, + analog, and - analog), it never gained the DIY popularity or budget R-2R NOS DAC popularity of the TDA1543. The 18-bit AD1865 not only has literally 4X the bit-depth resolution of a 16-bit TDA1543, it has a SNR over 120dB, better channel separation, better linearity, and a less "grainy" sound.

Now you know why companies like Audio Note have been selling NOS R-2R DACs with the AD1865 chip for literally decades.

For those of you that didn't get why an 18-bit DAC has 4X the bit-depth resolution of a 16-bit DAC, let me elaborate on basic digital math.

Each time you add one bit you literally double resolution:

16 bits has a numeric value of 65,536
20 bits has a numeric value of 1,048,576
24 bits has a numeric value of 16,777,216

Roughly this would mean that a DSD64 or a 24/96 PCM recording would have over 30X the resolution of a 16/44.1 Red Book recording.

So how come we don't perceive this insanely higher digital resolution in HD audio the same way we perceive the difference between an old-school 20" TV with and a modern 4K video played on a 70" HDTV?

Because unlike the fixed pixels used in video, the resolution of these audiophile formats is much higher than the resolution of most input sources, power supplies, and output stages of our DACs.

Take one of these factors: power supply noise.

Based on a 2.5V output on a DAC (higher than average), below are the voltages power supply noise must be below in order to hear the LSB:

• 16-bit LSB noise floor voltage = 76uV
• 18-bit LSB noise floor voltage = 19uV
• 20-bit LSB noise floor voltage = 4.75uV
• 24-bit LSB noise floor voltage = 0.3uV

For a reference, a common LM317 regulator, the quality used in most commercial electronics, has about 150uV peak-to-peak noise, and the world’s lowest noise power supplies (we’re talking NASA, not audiophile) have about 5uV of peak-to-peak noise. That means even with the most sophisticated linear power supplies or batteries available today, 20-bit is theoretically the highest playback resolution and dynamic range possible.

This makes it quite easy to understand how a properly done 16-bit DAC with an ultralow-noise power supply can easily outperform many so-called 24-bit DACs

If you want more info on digital resolution, you may want to read my blog "The 24-Bit Delusion," which basically points to the fact that lower resolution R-2R is more than sufficient to hear all the resolution possible from modern HD recordings.

http://mojo-audio.com/blog/the-24bit-delusion/

 

[glina]

Among the things that made the TDA1543 so revolutionary and so low-cost to implement were the many integrated features.

It has a built in S/PDIF receiver, a built in demultiplexer (L and R signal separation), built in digital filters and output stage, and a single +5V power supply. Comparing this to the TDA1541 which required two additional ICs for S/PDIF receiving and demultiplexing, external digital filters and op amps, as well as a five voltage power supply, companies now had a way of building a CDP in a fraction the size and at a fraction the cost.

I'm sorry, but there's plenty of factual errors here. Unless you are mistaking the TDA1543 for something else, it seems you don't have much experience with that chip.

1. The TDA1543 does not have an SPDIF receiver, nor would it require one in a typical cd player. It has i2s input.
2. It doesn't have a digital filter. Typical application recommends the SAA7220P/B for up to 4x oversampling
3. It has no output stage. It is a typical current output dac chip and the recommended I/V conversions is op-amp based.
4. It may not be operated with 5V if you intend to use it with passive I/V (resistor based) conversion, as it will generate severe clipping. There are certain conditions to be met if you want passive I/V. You can read about them here: http://myweb.tiscali.co.uk/g8hqp/audio/TDA1543IV.html
In any case, this mode of operation violates the specifications and actually degrades the already poor specified THD and Noise rations. It doesn't matter though, as it sounds better this way.

I must also touch on the topic of power supply noise vs bit resolution. What you wrote would be valid only if the system had a power supply rejection ratio of 0db. This is generally not the case.

 

[MojoAudio]

I don't want to get into a debate with anyone, but I'm literally looking at two data sheets for the TDA1543 DAC chip distributed by Philips.

One is the "economy version" with I2S input, has built in 4X oversampling, requires no external parts, and only has voltage output.

The other TDA1543 data sheet is quite similar, but has an S/PDIF input as opposed to an I2S input. Apparently there was more than one version of the TDA1543 DAC chip.

I'm certainly no expert on every version of every vintage R-2R DAC chip, but based on the two distinctive TDA1543 data sheets published by Philips I have in my files, it leads me to believe that most if not all of my original statements were correct.

This would account for the dozens of TDA1543 DAC designs that use no receiver chip, no external op amps, and no digital filters...not much more than an S/PDIF input, a +5V power supply, and a pair of coupling capacitors at the output.

As for power supply noise vs. bit resolution, granted my math and explanations are purposely over-simplified to make it easier for someone without an EE degree to understand them.

But the simple fact is that if the peak-to-peak ripple (noise) at any stage in a DAC is greater than the voltage of any of your LSB, that the resolution from those bits can not be heard. Period.

And Power Supply Rejection Ratio (PSRR) has nothing to do with actual voltage of peak-to-peak noise in the final DC used to power ICs (DAC chips).

So my statement about the LM317 regulator is overly simplified, but the rest of the math regarding the maximum peak-to-peak noise voltage compared to the LSB voltages are spot on.

Now that some of you are bringing up other factors, how about background noise and listening levels?

A quiet room has about 30dB of background noise. This means we can not hear any sound much below 30dB in volume.

And most of us listen at no more than 100dB.

Subtract the 30dB background noise and we are left with only 70dB of audible dynamic range.

So when Sony and Philips developed the 96dB dynamic range of the Red Book format they knew what they were doing.

Many renowned recording engineers have written white papers stating that it is not the 16-bits but rather the 44.1KHz that is the major limiting factor in the Red Book format.

And that doesn't even get into the fact that most recordings are compressed down to less than 60dB dynamic range so that they can be enjoyed at modest volumes and played back on modest priced electronics.

Do most of you realize that the maximum dynamic range of a stereo LP record is less than 65dB?

That sure puts the 96dB dynamic range of a 16-bit recording into perspective

This accounts for why many of you have found that a significant percentage of original 16-bit Red Book CDs sound better than newer remastered 24-bit so-called "HD" releases.

And also accounts for why many of you that prefer the sound of 16-bit to 20-bit R-2R DACs over modern 24-bit and 32-bit DACs, which is the whole point of this thread, and was my whole point in writing all this techno-babble

My goal was to put technical facts into accessible terms a layman could relate to...I apologize to any of you uber techs for some of my over simplifications.

 

[glina]

I don't want to get into a debate with anyone, but I'm literally looking at two data sheets for the TDA1543 DAC chip distributed by Philips.

Would you mind posting the datasheet you are referring to?

There are indeed two versions. One is the TDA1543 we are discussing, another the TDA1543A which is the same chip but with "Japanese input format" instead of I2S. Otherwise, neither has SPDIF input, integrated digital filter, 4x oversampling nor a voltage output. In the original TDA1543 datasheet there is a typo though. They define AOR as right channel output, but AOL as left channel voltage output. I take this is why you thought this chip has a voltage output, but otherwise the datasheet is clear about the current output. There's also a mention of "Possible 4x oversampling", but if you read deeper you will find "This flexible input data format (I2S) allows easy interfacing with signal processing chips such as interpolation filters, error correction circuits and audio signal processor circuits (ASP). The high maximum input bit-rate and fast settling current facilitates application in 4 × oversampling systems. An adjustable current is added to the output currents to bias output operational amplifiers (OP1; OP2) for maximum dynamic range (see Fig.1)."

And I'm sorry to say, but implying that <5uV PSU noise is a necessity for 20-bit output at RMS level is nonsense. If this was the case, a basic workaround would be to work with inverted signals which would cancel out the common mode psu noise and lead to an infinite improvement. You are the constructor advertising his product, so I surely hope you know for a fact that this is not the case.

 

[erin]

And I'm sorry to say, but implying that <5uV PSU noise is a necessity for 20-bit output at RMS level is nonsense.

Can you please explain why you think its nonsense when using a single ended DAC?

If this was the case, a basic workaround would be to work with inverted signals which would cancel out the common mode psu noise and lead to an infinite improvement.

Its do-able, and has been done but its not really "basic" to make a differential DAC using R2R chips. It involves some "glue" logic, meaning more chips, longer track lengths, latency between channel data etc.
If it were "basic" wouldn't every company be doing it? I'm fairly sure it wont be an infinite improvement, I think it would theoretically result in 3dB greater signal to noise ratio, assuing the design and implementation is perfect, which in the real world nothing is.

 

[glina]

Can you please explain why you think its nonsense when using a single ended DAC?

Because it makes the assumption that PSU noise goes unattenuated into the signal. This is a false statement, based on the understanding that DAC signal is a time modulated function of the voltage input. This is not true. Thousands of commercial products have power supplies with multiple milivolts of noise, which should result in 60db SNR. You'll struggle to find measurements as bad.

 

[erin]

This would account for the dozens of TDA1543 DAC designs that use no receiver chip, no external op amps, and no digital filters...not much more than an S/PDIF input, a +5V power supply, and a pair of coupling capacitors at the output.

I think you are mistaken. All TDA1543 chips use a digital reciver chip or a USB to i2s reciver chip. They do not have onboard SPDIF decoding.

Now that some of you are bringing up other factors, how about background noise and listening levels?
A quiet room has about 30dB of background noise. This means we can not hear any sound much below 30dB in volume.

And most of us listen at no more than 100dB.

Subtract the 30dB background noise and we are left with only 70dB of audible dynamic range.

So when Sony and Philips developed the 96dB dynamic range of the Red Book format they knew what they were doing.

This is an interesting topic. You have to take into consideration the frequencies of the background noise. If background noise is pink or white noise, then all frequencies in the noise floor would be masked, but if background noise is mostly in the range of (having a guess) 100hz to 3000Hz then other frequencies above 3K and below 100Hz will be theoretically audible.

Many renowned recording engineers have written white papers stating that it is not the 16-bits but rather the 44.1KHz that is the major limiting factor in the Red Book format.

And there are other recording engineers and electronics engineers who say frequencies above 22Khz are not audible, and that 44.1khz 16bit is all that is needed.
I conceed there is debate on this topic. I tend to think 44.1khz 16bit is good enough.

Do most of you realize that the maximum dynamic range of a stereo LP record is less than 65dB?

But, like you said, most recordings dont have more than 60dB dynamic range, so and LP can, for most pop/rock music, convey the full dynamic range on the recording, and it is possible to hear music below the noise floor of an analoge recording!
If you check out the dynamic range database you will see that most recordings only have a dynamic range of about 16dB for "good" recordings with and average of about 12dB and "brickwalled" recordings have about 6-8dB dynamic range. Classical music is an exception - it has more dyamic range.

 

[erin]

Because it makes the assumption that PSU noise goes unattenuated into the signal. This is a false statement, based on the understanding that DAC signal is a time modulated function of the voltage input. This is not true. Thousands of commercial products have power supplies with multiple milivolts of noise, which should result in 60db SNR. You'll struggle to find measurements as bad.

I take your point regarding signal to noise, but we are talking about bit depth, and I have seen measurments of commercial 16 bit DACs which have shown them to only have a bit depth of 14-15 bits of resolution. The least 1-2 significant bits can get swamped with noise, and I have found this to be audible, particularly with the TDA1543 chip, which though pleasant to listen to, is not renowned for its "detail" levels.

 

[glina]

I take your point regarding signal to noise, but we are talking about bit depth, and I have seen measurments of commercial 16 bit DACs which have shown them to only have a bit depth of 14-15 bits of resolution. The least 1-2 significant bits can get swamped with noise, and I have found this to be audible, particularly with the TDA1543 chip, which though pleasant to listen to, is not renowned for its "detail" levels.

This is perfectly correct, but the reason for this is the noisy integrated current source or internal logic of the chip. I think you'll admit that even supplying the TDA1543 with 0.1uV power supply will not get you the lost bits back.

It is perfectly true to say noise at the output will limit the audible bit-depth, but it is very false to say that power supply noise = line level noise.

 

[erin]

This is perfectly correct, but the reason for this is the noisy integrated current source or internal logic of the chip. I think you'll admit that even supplying the TDA1543 with 0.1uV power supply will not get you the lost bits back.

It is perfectly true to say noise at the output will limit the audible bit-depth, but it is very false to say that power supply noise = line level noise.

[/QUOTE]

I do agree with you completely.

 

[MojoAudio]

Glina, my purpose in posting about the TDA1543 chip was not to get into a debate on the various versions of the TDA1543 chip, but rather to confirm the reason there are so many wonderful sounding low-cost DACs using the TDA1543 chip is that it requires fewer power power supplies and fewer external parts. The TDA1543 has built in L and R channel demultiplexing, a built in voltage output stage, a single voltage power supply, and it requires no external parts following its voltage output. Compared to most vintage 16-bit R-2R DAC chips the TDA1543 requires about 25% the overall circuitry to build a DAC.

Any argument there?

But since Glina brought it up, his points on "inverted signals" and "common mode PSU noise" are conditional and over simplifications.

That would imply that the only PSU noise that exists is "common mode" and would imply that the only electronics that could have low noise would have balanced circuitry.

Note that most of the DACs from that wonderful R-2R DAC list are single-ended and don't use common-mode noise cancelling.

Glina is also confusing the low-pass digital filter that is required at the output of most vintage R-2R DAC chips with the over-sampling input ICs which are not used in modern non-oversampling R-2R DACs.

But let me get back on topic...

To minimize any confusion, here is a summary of my major points:

1. Since background noise is about 30dB and most people listen at lower than a 100dB volume the 96dB dynamic range of a vintage 16-bit R-2R DAC chip can potentially yield indistinguishable resolution from much higher bit-rate DAC chips.
2. Since most recordings are compressed to less than 60dB of dynamic range that the 96dB dynamic range of a vintage 16-bit R-2R DAC chip can potentially yield indistinguishable resolution from much higher bit-rate DAC chips.
3. Since the 90dB SNR of most tube output stages is less than the 96dB dynamic range of a vintage 16-bit R-2R DAC chip it can potentially yield indistinguishable resolution from much higher bit-rate tube DACs.
4. Since a TDA1543 requires far fewer external parts and far fewer power supplies it can potentially be implemented into a smaller and lower cost circuit that will rival many larger and more expensive higher bit-rate DACs.
5. High bit rates (24 and 32 bits) are vital in the editing, mixing, and mastering of low-noise digital recordings, but are so far from significant when it comes to the resolution our ears can hear during playback it is ridiculous.

But that only confirms what all of you vintage R-2R DAC lover's ears have been telling you all along

That's all I wanted to share.

 

[glina]

Glina, my purpose in posting about the TDA1543 chip was not to get into a debate on the various versions of the TDA1543 chip, but rather to confirm the reason there are so many wonderful sounding low-cost DACs using the TDA1543 chip is that it requires fewer power power supplies and fewer external parts. The TDA1543 has built in L and R channel demultiplexing, a built in voltage output stage, a single voltage power supply, and it requires no external parts following its voltage output. Compared to most vintage 16-bit R-2R DAC chips the TDA1543 requires about 25% the overall circuitry to build a DAC.

Any argument there?

Yes! Seeing how you are not addressing my replies to you, the least I expect from you as a DAC designer is to read the datasheet. TDA1543 has no voltage output. It is a current output DAC, and operating it without an active output stage is only possible with a non standard (higher than 5V) supply.
You imply that I'm confusing a digital filter with a low pass filter, but I don't know why how you got this idea. It is irrelevant though, as the TDA1543 has neither of the two.

I generally agree with the rest of your points, but I can hear and appreciate the additional audible resolution of Hi-Res Audio (24b/96kHz or more) on a more modern DAC. Still, I always return to my plain TDA1543 for fatigue free listening pleasure.