Rats, I had no problem making sense of the schematic or the detail discussion. I’m pleased and worried.
3 Likes
My head hurts just trying to pay attention in class.
Thanks Ted!
4 Likes
Thank god its just not me. I thought I was the only one being so confused. But at the same time looking at my bank account to see if I can afford one.
2 Likes
How about Ted’s discussion of finite impulse response filters and upsampling? Not quite as simple eh Al?
2 Likes
This is why I left electrical engineering after college and decided to go into computer science. No filters, op amps, spice plots and discussions about frequency.
I only need to be able to add 1 and 0.
1 Like
Perhaps some here would like to know something about the software architecture of the FPGA code.
Part 1 is simply the comparison of some of the numbers for the DS vs the DS Mk II
DS DS Mk II
input bits 24 bits 24 bits
elastic buffer size 1024 stereo samples 4096 stereo samples for quadrate
...
256 stereo samples for 44.1kHz and 48kHz
PCM upsampler
output frequency 352.8kHz 705.6kHz
output sample width 28 bits 28 bits (two more on the top and two more on the bottom)
clock rate 169.344MHz 225.792MHz
filter size 7959 taps 16,225 taps
coefficient size 30 bits 38 bits
multiplies per sec 1.528 billion 3.116 billion
multiply size 24 x 30 24 * 35
accumulator size 48 bits 61 bits
hardware DSP blocks 24 (out of 32) 16 (out of 80 per FPGA)
UI volume resolution 1/2 dB 1/8 dB
UI volume steps 100 963
FPGA volume ctrl width 20 bits 20 bits
SDM register size 56 bits 72 bits
SDM coefficient size 40 bits 47 bits
13 Likes
If I am reading this correctly, that’s a remarkable increase in volume control.
When you are ready and willing, it would be interesting to see the TSS added to this table.
3 Likes
For the time being I’ll be doing the same software parameters for the DS Mk II as the TSS until the Mk II is out of resources, similar to the DS Jr and the Sr.
8 Likes
Just trying to put this and your earlier post together.
44.1 → 705.6 → SDM to DSD256 for output?
48 → 384 → 56448 → SDM to DSD256 for output?
44.1 → 705.6 → 11,2896 → SDM to one bit for output.
48 → 384 → 56,448 → 11,2896 → SDM to one bit for output.
5 Likes
Fascinating stuff! I think I will get a DSII just to hear what these numbers will sound like.
2 Likes
24/44.1 → 28/705.6 → 28/11,2896 → 48/11,2896 (apply volume) → 1/11,2896
24/48 → 28/384 → 28/56,448 → 28/11,2896 → 48/11,2896 (apply volume) → 1/11,2896
5 Likes
Tho this does gloss over the fact that I really take 48, 96 and 192 to 768 (so they can share the 44.1, 88.2 and 176.4 to 705.6 upsampling filters). Then I go back to 384, then up. Going from 768 to 384 is lossless so the extra step doesn’t hurt anything and it shares code and coefficient memory with the 44.1 series of sample rates.
7 Likes
I will get an MKII just because it is your design.
You know what Stuart, I like you!
You’re not like the other kids, here in the trailer park.
8 Likes
Then (if I don’t have one already) you send it your your Buddy Mikey…to try.
Hopes for 2022!
To get the full benefits of galvanic isolation you need all inputs and outputs isolated, every groundloop in your system, related to the DAC or not, is a source of noise and can hurt the detail retrieval. The amount of degradation or improvement possible is very system dependent.
If USB is your only input plugged in (except possibly for TOSLink) then a USB isolator might be useful. But even a good USB cleaner can sometimes be helpful, e.g. the Matrix X-SPDIF 2. Some USB isolators hurt the sound quality so don’t assume any USB treatment is always good.
“new opamp circuit (with correct resistor/cap values)”? I presume you are talking about the Vocm mod? It provides a lower impedance for the video opamp’s output common mode, i.e. the firmness of the output’s ground reference. It helps bass a lot and perhaps a blacker background a little.
“the linear PSU for the digital section” I’m not sure what you are referring to here. All power supplies in the DS are linear. In the mod’s thread most of the PSU talk is about external analog board 12V supplies. Replacing the stock PS Audio power supplies will help with a black background, but AFAIK no one has explicitly done that. Replacing the analog board’s power supply (with something reasonable) definitely lowers the background noise.
Almost all of the software changes in the DS had the goal of reducing (and whitening) noise. Less noise gives a blacker background for the music. The sense of space grows with a blacker background and detail retrieval increases; performances seem a little more real. Most of the changes in the DS Mk II are intended to lower noise and give a blacker background. I’ll give more details in a future post.
Your diagram seems to be accurate; I wouldn’t say its complete, but it’s not missing a lot. This thread is my attempt to flush out some of the details.
7 Likes
Is this now another „Ted recommended and mentally approved“ mod? Not sure if I directly find what’s done in the thread, but I remember the topic.
Yes, I explained why it worked and what parameters were the best to play with as soon as Turbo mentioned he was working in that area. Picking better resistors there and adding some cap filtering on the Vocm signal is a reasonable thing to do. Using monster thruhole components for that purpose isn’t as effective as picking good surface mount caps and soldering them across the resistors.
5 Likes
Everything I touched when working on the TSS and the DS Mk II was done with reducing noise in mind. I thought about it with architecture choices, design choices, parts choices, layout choices, routing choices and similarly with all of the software work.
I started enumerating all of the noise reduction items, but instead I’ll just say that’s what this thread is about.
Which will have the biggest effect?
That’s like picking your favorite child 
Some of the changes like isolation will made huge differences in some systems and smaller ones in others. Power supplies, more and better regulators will also have a noticeable effect. But all of the component choices, layout and routing decisions, etc. make a non-trivial difference. Keeping more S/N everywhere in the software is also noticable.
10 Likes