Digital Drive Upsamplers, DACs, jitter, shakes and analogue withdrawals, this is it.

---

[ Follow Ups ] [ Post Followup ] Thread: [ Display  All  Email ] [ Digital Drive ]

---

[ Alert Moderator ]

---

RE: The 4 ways to decode digital

Posted by Thorsten (M) on October 14, 2015 at 19:22:40

In Reply to: The 4 ways to decode digital posted by J. Phelan on October 13, 2015 at 14:11:32:

Hi,

> -Commercial DAC-chips. Resistors, representing 1 to 6 bits, except
> Burr-Brown 1704 (24). But the Burr has only 18 bits of resolution.
> The most common type of DAC -inc. Delta-sigmas and hybrid/multi-bit.

BB PCM1704 is true 24 Bit R2R sign magnitude (actually two 23 Bit DAC's!). It's signal/noise ratio is only equivalent to around 18 Bits worth, but that still compares well with many discrete implementations.

BB DSD1793 is a hybrid Chip using 6+1 Bit with the top 6 Bits being true multibit (if Thermometer code with dynamic element matching instead of R2R) and lower bits in Delta Sigma. It has lower noise and THD than the PCM1704 if correctly implemented.

In some ways this hybrid approach is a genius way of doing it (and BB/TI still hold active and valid patent on this approach), because classic Multibit DAC's are worst at zero crossing (Glitch - kind of like Class B crossover distortion) and classic Delta sigma DAC's are worst at high modulation and thus high levels.

So each technology is applied where it is advantageous and where it's drawbacks are minimised. It's a crying shame that TI has abandoned this in more recent DAC Chips for the same generic switched capacitor filter everyone else uses, thus making sure that "all DAC's sound the same"... For now the older parts remain in production (even the PCM1704) but prices are steadily increasing, as do lead times, so mene tekel.

Other "Hybrid" DAC's" (including TI's latest) apply DS modulation to the whole low bit number multi-bit array (doing it differently would get their pants sued off by TI). So there is not quite the same balance of virtues and vices, but the technology is much more cost efficient to produce where switched capacitor filters are used.

Of course there are still pure Delta Sigma DAC's (even though they are exceedingly rare now and relegated to "junk level" status. Personally I found some of the original pure DS DAC's (like the Nippon Precision SM5872 found in the Marantz CD-63 II/67/5000 etc.) excellent, as a rule they all came straight off the bit-switches with no integrated analogue stages. Designing a competent analogue stage was very challenging (which is why all these parts have been long discontinued) but if you got it right it was very good.

In many ways the recent trend to "Chipless DSD DAC" is a return to this old style of doing things.

> -In-house chips. Using (maybe 4) "latches" instead of bits, this approach
> relies on field-programmable gate arrays (FPGAs) for DSP. A system that
> emphasizes instructions (vs. operations). Possible to remove a few clocks
> and use a passive output stage. Playback Designs and PS Audio idea.

Not in house chips. They simply use programmable and discrete logic.

The whole subject of clocking and passive output stage or not are not specific to these designs but apply equally to most monolithic integrated "off the shelf" DAC Chips. It is just a question how you implement things. If I put the DSD1793 into Delta Sigma Mode I can loose the system clock, for example.

The originator of this style of DAC was actually Cirrus Logic with their CS4303 back in '93. Their Chip contained the modulator (the part that now goes into the FPGA etc.) and used external Latches to re-clock the datastream and perform the actual DAC function. They later integrated more and more functionality on chip and went to cheaper technologies, so making sure no designer could bollox up performance but also limiting the degrees of freedom in optimising designs and the subjective performance.

The modern pure Delta Sigma Systems with FPGA/CPLD/DSP modulators are nothing but a remix of the 1993 demo PCB for CS4303.

BTW, you forgot the Fishy DAC, which uses Async Upsampling in FPGA to 3.125MHz (not an integer multiple of either 44.1 or 48kHz but an interger divison result of 100MHz) and then up-samples further to 100MHz and then uses a 32-TAP FIR analogue filter to create an analogue output.

If you replace "100MHz" with "22.5792/24.576MHz" and "32 Tap" with "8 tap" you incidentally have a BB DSD1793 in DS(D) mode.

> -Discrete resistors (no chip). Doesn't need a current-to-voltage converter
> or digital filter. Totaldac is one.

Few R2R DAC's (be they monolithic IC's or discrete) NEED a current to voltage converter. However, if non are used commonly linearity suffers badly.

Specifically TDA1543 and TDA1545 and AD1865 DAC Chip's are all known to be usable without any form of I/U converter, others can be used likewise. More recent DAC's often have protection diodes on chip for the output limiting usable signal swing to around 0.5V (instead of the common 2V), but a low ratio stepup transformer can take care of this.

> -Reformatted then switched. Lyngdorf and NAD digital amps.

Switching Amp's are quite old. Sharp may have been the pioneer here. The main Chip maker that is widely used in these systems is TI (especially in the high end) who also pioneered the use of variable powersupply as volume control and continue to manufacture the chipsets.

Several other chip makers also have had direct digital input solutions since the late 1990's, including (but not limited to) SGS Thomson and Hitachi/Renesas, TI has the patent on the PSU Volume control though so for this you need to go to TI. A problem with this type of Class D Amp is that switching speeds are way too low and that usually no feedback is used, so they measure quite "funky". The TI power-supply volume control reduces this problem notably.

Those kind of Amp's are implemented both in lowly "Consumer grade" active desktop computer speakers and in high end systems. The chip-sets are the same.

So, bottom line, all these systems rely on "off the shelf" Chips and all have "off the shelf" Chip's.

Ignoring the different DAC Technologies and subclasses (Multibit, Delta Sigma and Hybrid) the key differences are that dedicated DAC Chip's usually offer excellent measured data but often (but not always) restrict the designers freedom to create unusual solutions, while discrete solution often (but not always) measure less well but allow a greater degree of freedom in design and component quality.

Sadly dedicated AUDIO DAC chips are becoming more and more the same and are optimised for measured performance while being low cost and low power consumption. Subjective sound quality is rarely if ever an issue being even considered.

A quality design will show it's quality regardless of what specific solution is adopted (presuming some of the basics have been taken care of in the case of dedicated DAC Chips - which is becoming more difficult) and a poor design will not be rescued by discrete DAC's and FPGA's.

Thor

At 20 bits, you are on the verge of dynamic range covering fly-farts-at-20-feet to intolerable pain. Really, what more could we need?

This post is made possible by the generous support of people like you and our sponsors:

The Cable Cooker

---

[ Top of Page ] [ Contact Us ] [ Support/Wish List ] [ Copyright © 2025, Audio Asylum, all rights reserved ]

---

[ General ] [ Speakers ] [ Tubes ] [ Vinyl ] [ Digital ] [ Hi-Rez ] [ Video Asylum ] [ Cables ] [ Tweaks/DIY ] [ Music ] [ Films ]

---