what is a 1 bit DAC?

danger boy

i've seen this mentioned for years.. but i don't understand what it means... of course there are other DAC chips out on the market. is this a low end old school DAC chip? is there better ones out there, or is this the standard in most components?

madmax

This is all from seeing the numbers as they came out. There could be a better explanation but what I remember were 8 to 16? bit dacs (parallel input designs) then a high speed 1 bit (serial? design) which was fast enough to do things in a string rather than several input bits (parallel). Anyone else?
madmax

billbillw

As I understand it, (and I'm by no means and expert) the bandwidth of digital audio is equal to the bit rate multiplied sampling frequency. CDs are saved in 16-bit format at 44.1 kHz or roughly 705 kilobits/second (if I'm doing my math right).
When you do the Digital to Analog conversion, you have a choice of doing multi-bit or single bit. Usually the design of the DAC has much greater bandwidth than the original CD. For example, a 24-bit/192kHz DAC has approx 4.6 Mbits/sec bandwidth. That's where you get your "oversampling". It allows for noise shaping that trys to match an analog curve better. The other choice is to do single bit conversion at much greater speeds. To equal the 24-bit DAC above, a single bit DAC would be operatining at greater than 4 Megahertz to get similar bandwidth. I believe Sony or Yamaha started the single bit DAC trend, and it has been called many things like bitstream, S-bit, High Density Linear converter, etc. Over the years though, it seems that most of the audiophile players have stuck with the multibit DACs and they are generally thought to sound better. The trend through the middle to late 90s was to have multiple DACs in a "differential configuration" to achieve a more realistic sound.

RuSsMaN

Check out the big brain on Bill!

danger boy

Bill thanks for the explanation... that seems to make sense to me now... I've just seen this mentioned while looking at CD players alot over the years.

heiney9

Here you go Danger a link that explains it all pretty well.

http://www.howstuffworks.com/question620.htm

H9

billbillw

heiney9 wrote:

Here you go Danger a link that explains it all pretty well.

http://www.howstuffworks.com/question620.htm

H9

That does a little better than my explanation. Thanks for the link.

LuSh

One of the best DAC's in the world uses Bitstream. And that's the Museatex Bidat, precursor to the EMM Labs DAC6, which also read's redbook in DSD format. And DSD if I'm not mistaken was entirely based on Bitstream technology. Ed Meitner used Bitstream code to create SACD.

heiney9

LuSh wrote:

One of the best DAC's in the world uses Bitstream. And that's the Museatex Bidat, precursor to the EMM Labs DAC6, which also read's redbook in DSD format. And DSD if I'm not mistaken was entirely based on Bitstream technology. Ed Meitner used Bitstream code to create SACD.

That is correct that DSD uses Bitstream and it's part of the creation of SACD. PDM (Pulse density modulation) rather than PCM (Pulse code modulation)

Notes by Colin Miller: You don't need analog filters with DVD-A either at 24/96, 24/192, or 16/44.1, if you don't mind a bunch of ultrasonic noise going downstream. Single bit D/A conversion has the same issue, except that the quantization noise inherent to the front end is FAR higher in amplitude, but also at a higher frequency. Both PCM and DSD use dither on the front end, and any time the bit depth changes, i.e., 24 to 22, 4 to 1, etc., you randomize the quantization error, and turn it into noise. The spectrum of the noise added will depend on the nature of the dither applied. The dither, must, though, be about 1/2 the LSB. With 24 bit A/D converters, the amount of dither can be very small. With 1 bit, or even 4 bit conversion, the dither must be just 6 dB lower than the LSB. With a single-bit system, whether that data stream is generated through conversion from PCM, or what not, the associated noise will be -6 dB from full scale.

What DSD allows you to do is have a higher sampling frequency at the same data rate, so that you can focus that noise above the range you actually want, precisely so that you can filter out that noise. Note, though, that if you want to have any kind of dynamic range, the bandwidth limits will be far below the Nyquist limit. One can yield the same benefit of moving the quantization noise above the interested range with PCM through "over-sampling." DACs that take information at 44.1, and have common 8x oversampling (built into the DAC) multiply the sampling rate to 352.8 kHz. With 48 kHz, the rate would be at 384 kHz. You could say that after applying digital filters, the effective data rate would be equivalent to DSD at 9.2 MHz, but that would be somewhat misleading. Both formats SHOULD be filtered by an analog circuit, but neither HAVE to be.

So how does this relate to any superiority between the formats?

In terms of linearity, PCM's greatest limiting factor is the tolerances of the resistors that determine the amplitude of the DAC output, as it is important that each bit be a certain proportion to the rest of the bits, so that each bit's extra output is exactly half that of the next more significant bit, and twice that of the next less significant bit. This becomes very difficult, and requires laser-trimmed resistors, but manufacturing technology is improving the cost-effectiveness of high-quality multi-bit DACs. However, many DACs avoid this problem by converting the data stream to that of single bits, for reasons below.

Single-bit technology does not have any linearity problems in terms of matching bit amplitudes, because there is only one bit, and it matches itself perfectly. However, since the amplitude is more a function of time, the linearity is more heavily dependent upon the clock accuracy.

Both formats can be highly linear, though, if done well. In terms of resolution, it's not an easy comparison, because the resolution is inversely related to the noise floor, and the noise floor is a product of the spectrum of dither applied, and the filtering applied afterwards. As interesting as it is to look at the technology involved, if you really want to compare apples to apples, compare the analog outputs of two otherwise identical players with similar quality of manufacturing and design. Finding and identifying those units in itself would be a task.

However, that might be pointless, since both formats have more dynamic range than the analog electronics can yield, and greater bandwidth than any humans can make use of. I would argue that the greater benefits are measured in terms of implementation. PCM is far easier to apply DSP to. DSD is easier to build converters for, as the data stream can be fed directly through a filter, or if you don't mind a whole heap of noise, directly to whatever input you've got.

1-bit isn't "Delta-Sigma" (which is an encoding strategy) - 1-bit in this case is Pulse Density Modulation (PDM). The samples are represented with a frequency modulated square wave, with pulses of equal duration. This is as compared to Pulse Width Modulation, where the duration of the pulses is modulated rather than the frequency of the pulses. One could say that PDM is a frequency domain approach, while PWM is a time domain approach.

Analog sources coming into the SX100 are sampled at 2.8224 MHz as well, and there are a few sources (made by Sharp) that output 2.8224 MHz digital bitstreams, such as from their SACD player. In fact, the Sharp components are the rare set of products right now that have digital outputs from SACD players. More will follow I am sure.

H9