You can directly convert some set of upper bits or you can in effect add noise to the lower bits, by randomizing the voltage you modulate, which as I understand the Ring DAC it’s mostly doing the latter, and cross sampling to average out any variance in the output resistors.
Signal processing just gets complicated and unintuitive when you start talking about noise, the problem with strict 1 bit DS, and why no one does it anymore is that all the errors stack up on top of each other, and then fold down through sub multiples of the sampling frequency, adding noise doesn’t reduce the overall error, but it spreads it out, so it isn’t one big peak.
Cross sampling which you see in high end R2R DAC’s as well, just averages out errors, which does reduce the error.
What I was trying to say was even knowing all the details of how it works, doesn’t tell you how it will sound, all it can really tell you is if you think the solution is “clever” I for example think the DCS ring DAC is “clever” engineering, doesn’t mean I’d own one.
IMO It’s not that R2R DAC’s have a sound because they are R2R DAC’s, but rather designers choosing R2R as a solution tend to be targeting a sound, in the same way people expect Tube amps to sound a specific way, but tube amps can be designed to incorporate lots of feedback and have very clean clinical sounds, they just generally aren’t, you use tubes in a design because in low/zero feedback settings they are more linear amplifiers, if your going to design an amp with a ton of feedback, you might as well use solid state.
I don’t think it’d be fair to call anything that converts to 1-bit PDM/DSD or anything running PWM an FPGA DAC, as technically then you’d encompass all ESS DACs for instance which are PWM
Agreed… I’d only suggest the term be used for an FPGA specifically programmed to convert to 1-bit PDM/PWM. This makes the FPGA logically a replacement for a delta/sigma DAC chip in a component diagram of the audio device.
But then what about stuff like dCS, Chord, and many others that use FPGAs and in fact are often the brands people think of when they hear the term “FPGA DAC”, but are not 1-bit?
It makes far more sense to just call stuff what it actually is.
An eversolo DAC-Z8 is an ESS DAC
An RME ADI-2 is an AKM DAC (or newer ones ESS I guess)
A Chord DAC is a discrete multibit ‘pulse array’ DAC
A Meitner MA3 is a discrete 1-bit DAC
A dCS DAC is a discrete 5-bit DAC (brand name ‘Ring DAC’)
And then there’s of course plenty of examples of R2R DACs with discrete 16-24 bit R2R Ladders
All of these have FPGAs, only one is 1-bit
But the point is it never is.
The FPGA is a logic device. It isn’t the actual DAC. All the DACs people think of as “FPGA DACs” are using the FPGAs for DSP and controlling/outputting to the ACTUAL converter circuitry.
The FPGA alone is not a replacement for the DAC chip.
Thanks for your responses; we really don’t need to belabor this further. My real interest is validating my understanding of these various DAC implementations and developing an understanding of the taxonomy of DACs. A taxonomy, anyway, probably should be a prerequisite to a discussion on naming.
Key to my understanding of non-R2R implementations is whether or not it is true that a PDM or PWM digital signal can be directly converted to analog simply by passing it through an appropriate analog low-pass filter (would this be called a reconstruction filter?). If true, then is it fair to say that a delta-sigma chip DAC is basically doing this? Is “Delta-Sigma” just the name of an algorithm to go from PCM to PDM/PWM. Do such DAC chips contain this analog low-pass filter or is it external to the chip?
Thanks in advance for any expertise you can impart. Your back and forth with Polygonhell has been informative but perhaps a bit over my head in areas .
I like the simplicity of this as a starting point for understanding various DACs. Do you think it would be fair to say that, in a way abstracted from the implementation (chip, ladders, FPGs, etc.) that all DACs fall on a continuum between switching resistors and pulsing voltages? That is, can you imagine a theoretical DAC implementation parameterized by some value to determines the amount of each of these two strategies used? Would GoldenSound agree with you (I’m not sure after reading your back-and-forth with him)?
The ‘binary’ yes/no aspect is “Is it doing delta-sigma modulation”. ie: Is it a FULL bit depth R2R ladder that can just convert >=16 bit samples exactly as-is, with no digital processing. Or is it doing DSP to modulate to a higher sample rate at a lower bit depth.
The continuous spectrum aspect is then ‘what bit depth is the DAC operating at’. Stuff like AKM/ESS chips are 5-7 bit
dCS is 5-bit
Chord I think is model dependent as the number of elements in the pulse array differs as you go up their product stack (though this could also be a speed & randomization change rather than a bit-depth change, I don’t believe they’ve said publicly)
Meitner and Playback designs are 1-bit
Though whilst R2R is generally a case of “R2R is R2R”, delta-sigma DACs have a wide variety of ways of doing the actual conversion even at the same bit depth.
For example an AKM DAC converts to what is effectively really really high speed 7-bit PCM. Whereas many of the burr-brown chips convert the top 6-bits exactly as-is (akin to an R2R DAC), and then use delta sigma modulation for ONLY the lowest bit.
Some DACs might convert to an output that is only ever fully ‘on’ or fully ‘off’, similar to 1-bit PDM/DSD, but use higher bit depths to change the width of the pulses so it’s teeeechnically not ‘1-bit’ (although would arguably be analogous to much higher rate 1-bit)
There are also many 1-bit DACs that whilst the actual data itself is 1-bit, the converter mechanism works in a way where each new bit changes the state of the output up or down one ‘step’, acting as a hardware level low-pass filter and then behaving more like what you’d expect from a straight up 5-bit design for instance.
I talk a bit about this in my Meitner MA3 review at 2:45
Hi, I´m wondering about your statement in one of your latests videos in which you mentioned something like this: “If I must sell off some of my equipment, I wanted to keep the Zähl EQ 1 for sure.” I got the impression it´s a very important tool for you. Now I don´t EQ at all, want to start but software EQ is not an option, hence I´m not equing… I looked at the EQ1 a bit (was curious about it when you first mentioned it) and consider buying one.
In a perfect world you would just go ahead and make a video… NOW
since that´s not going to happen how would I go about using that (I have just a HP system no 2ch)
Looks pretty straight foreward but all videos are from pro audio folks, producers and such and I wonder if anbody elese uses an EQ1 in his system.
It’s all the same controls as a digital parametric EQ, so you’d configure things the same as you would there, it’s just being done in the analog domain instead of digitally.
The drawback is you only have four bands (or 6 if you’re using it with an HM1), so if you’re wanting to do complex corrections it’s not the most ideal tool, but personally I don’t usually do complex EQ and mostly make more targeted or broader adjustments so 6 bands is plenty for me.
The benefit is that I don’t need to rely on any software, so can use any playback hardware/software I like, and all the adjustments respond in realtime whereas most digital options you have to make an adjustment and then it’ll ‘jump’ or briefly pause and apply them. I find that the constant/immediate response helps me to dial things in by ear much faster and easier
That’s why I like the Schiit Lokius. Well thought out points on the spectrum and Q values. I find it’s much more likely that I need to adjust something fast and easily on a particular track or album than I need to make my headphones sound different.
If I wanted the headphones to sound different, I would have bought different headphones. My most used 3 just don’t require sophisticated EQ at all. (Rosson RAD-0, ZMF Auteur Classic, NectarSound HIVE estat)
Hey @GoldenSound! I came across this image about where jitter can occur and was curious if you have any comments on it. It seems correct to me but I’m no expert:
.
