Signalyst DSC1

Imho they all do (cmos based). It manifests itself only at high dsd rates (preferably dsd512, 256 is a lot less problematic) and only at levels around -30 to -60dB.

Just came to my mind that maybe some implementations that suffer from noise are using 3.3V 595's? I chose NXP AHCT595 running from 5V supply for very specific reasons.

...I've done quite a bit of measurements on my unit... Specifically at DSD512 and DSD256 rates. But I'm sticking quit a lot to DSD256 because it gives noise floor < -160 dB up to ~100 kHz. And the switching rate doesn't get too high for settling times, and allows using EC modulators with current processors.

It's an Inter Symbol Interference problem, not a low pass filter or single ended vs differential problem and is gone when using AMSDM7 512, or low dsd rates.

Problem with my original low-pass filter is that if you look at the datasheets, channel separation between the channels of dual op-amp decreases as function of frequency (capacitive coupling). This of course creates a problem when used in a low-pass reconstruction filter. So I'm pretty much back to using AD797's for that.

I'd love more special versions of the AMSDM7 512 type modulators, for this exact reason. If you've got some time left or need something for the wish list, this is it;-)

It is quite strange beast indeed. And now I could create AMSDM7EC as well, maybe I'll look into that. Although other EC modulators cannot run at 512 rates, that one has more parallelization opportunities.

It could work well with both DSC DACs as well as some other discrete implementations and the very latest AKM chips.
 
AHCT works nicely in terms of switching speed, and also works running from 5V PSU while the data source is running at 3.3V (Amanero). So for HCT check for the low/high logic transition levels and timing parameters compared to NXP's AHCT595.


P.S. Amanero firmware versions had bunch of clocking issues along the line, so some of the noise problems may have been related to that. There were number of firmware versions that were outright noisy at higher DSD rates (lot of hiss) due to this.
 
AHCT works nicely in terms of switching speed, and also works running from 5V PSU while the data source is running at 3.3V (Amanero). So for HCT check for the low/high logic transition levels and timing parameters compared to NXP's AHCT595.


P.S. Amanero firmware versions had bunch of clocking issues along the line, so some of the noise problems may have been related to that. There were number of firmware versions that were outright noisy at higher DSD rates (lot of hiss) due to this.

i am using beaglebone with botic driver

from what i can get from datasheets (philips)
both HCT and AHCT can run from 2-5.5v
HCT max clock 57Mhz / prop dly 25ns -> 52ns
AHCT max clock 170Mhz / prop dly 3.8ns->4.6ns

not sure what to think....

from the "quality" point of view is better to use newer nexperia device in ACHT format ... or older philips in HCT flavour ?
from nexperia datasheets the newer ahct model from them is not far from the specs of the older hct.
 
Last edited:
Hi all
Has anyone compared AK4490/3/7 in direct DSD mode with the same transformer output with DSC2?
Yes, the AK4497 sounds good too.
 

Attachments

  • 16-11-2019 18-42-24.jpg
    16-11-2019 18-42-24.jpg
    129.3 KB · Views: 421
  • 16-11-2019 18-43-24.png
    16-11-2019 18-43-24.png
    187.9 KB · Views: 431
i am using beaglebone with botic driver

from what i can get from datasheets (philips)
both HCT and AHCT can run from 2-5.5v
HCT max clock 57Mhz / prop dly 25ns -> 52ns
AHCT max clock 170Mhz / prop dly 3.8ns->4.6ns

not sure what to think....

from the "quality" point of view is better to use newer nexperia device in ACHT format ... or older philips in HCT flavour ?
from nexperia datasheets the newer ahct model from them is not far from the specs of the older hct.

There is pretty distinct difference of driving the chip from 3.3V vs driving it from 5V. But it also matters for the conversion stage. So I've been running from 5V always. Some of the more recent AKM chips push this further by using 7V for the conversion stage. With 5V supply using AHCT and 6 dB gain in the final analog stage things are quite good.

Delays should be in 5 ns range.
 
Hi all!
Andrea and me have new design dac with input Ct7302,
Please see some . Thanks

STAR Pure DSD DAC

1 x Star Pure DSD DAC module
1 x Controller LCD module
1 x XLR and RCA output module
1 x Silver remote

This not include Amanero USB module (you can buy directly) and this not include 6.5V 20VA transformer.

INTRODUCTION

The high resolution PCM at 96-192KHz 24bit offer the possibility to have a good quality digital signal exceeding the big limits of the CD.
The DSD is very different from PCM, using a sampling frequency 64 times higher than that of a standard Compact Disc, offers a frequency response of up to 100 kHz and a dynamic range of up to 120 dB across the entire audible range.
The DSD on paper should be better than hi-resolution pcm like the 192KHz 24bit because the two conversions (from and to the analog) are more simple (see this article from Korg website) but the AES Convention Paper 6086 and AES Convention Paper 5395, seems to contradict this.
The DSD signal can be directly used to reconstruct the original analog signal with a simple RC filter but with this method we get many digital noise also using a good digital and analog filters so forget this method.
I have lost many time and money to test these solution, using different USB to DSD interfaces, I2S isolators and D-type flip-flop.
The little Lampizator and RT audiodesign companies has succeeded to reduce the noise and produce a DSD DAC with this method but with some limits.
This project here exposed has another approach, it is based on the DSC1 that is an Open Hardware, "discrete" delta-sigma D/A-converter developped by Signalyst, the DSC1 is able to convert the DSD signal to analog without noise and without a DAC chip.
The DSC converters does not work with a PCM signal so it is necesary convert the PCM to DSD on the fly during playing so it is necessary to use computers with adequate performance.
In this project we have used the CT7302PL single chip dual channel audio digital to digital bridge with a sample rate converter to convert PCM signals to DSD.
Schematic1.jpg


HARDWARE

The core of this design is the CT7302PL with the following specifications:

SPDIF support up to 768 KHz/32bits mode
I2S support up to 768 KHz/32bits mode
DSD support 1x/2x/4x/8xmode
DoP support 1x/2x/4x mode,throughI2S
DoP support 1x/2x/4xmode, through SPDIF
PCM SRC / ASRC fully support 32 KHz~ 768 KHz.
PCM / DSD / DoP to PCM / DSD / DoP converter with SRC / ASRC
Crystal reference clock / MCLK / DoP / Mute status output
Auto De-emphasis32KHz/44.1KHz/48KHz.
Auto detecting PCM / DoPformat.
Auto detecting I2S / DSD interface.
Volume control with fading in/out.(+18dB ~ -110dB step 1/32 dB)
SRC / ASRC 32 bits dynamicrange(192dB), THD+N (-174dB).



After the CT7302PL chip there is the DSC in the differential version designed by Pavel Pogodin with the output transformers to filter and to convert the output in un-balanced and balanced signal.
The 74AHCT595 are high-speed Si-gate CMOS devices used for the serial-to-parallel data conversion.
This module use a mute circuit to eliminate any click in the transition from one track to another using the TS5A3167 (SPDT) single-pole double-throw analogswitch .



This module need a AC power supply in the range 6.5V - 7V and it can power the Amanero USB module with a dedicated 3.3V.
On Amanero module to exclude the USB power supply and use an external 3.3V is necessary cut a wire or remove on-board regulator.
The USB interface used is the Amanero with the normal firmware to support Native DSD.
ph6.jpg


ph2.jpg


The Controller module included in the kit use the 16F877 chip, a 20MHz Enhanced 8K Flash Microcontrollers.
It drive a 4 x20 characters LCD and it is connected with the I2C bus to the CT7302PL.
A separeted power transformer is used to manage the standby function.
ph3.jpg


OUTPUT TRANSFORMERS

The output transformers are necessary to mix the differential signals coming from the 74AHCT595 in phase and out phase.
These are also used to give un-balanced and balanced output signal, the module integrate a relay to switch the ground connection.
The models has been selected by Pavel Pogodin after many test and also I have compared these with other models with much higher cost always preferring these.
These have a turn ratio of 2:1+1, made with a permalloy core with a thickness of 0.16mm and an inductance about 340H.
The declared frequency response is 10Hz - 195 KHz at -0.2dB for the yellow and 20Hz - 80 KHz at -0.2dB for the black type
Yellow type has more "air" and black are probably more balanced but it all depends very much on the record or even his mood.

Taobao
AliExpress
Alixpress type 2

REMOTE

Remote1.jpg


SOFTWARE

This hardware is compatible with any PCM and DSD signal so to play music you can use any audio player.

Here follows some player in order of sound performances.
HQPlayer by Signalyst available for Windows, Mac OS and Linux, there is a Server and a Desktop version at a cost of 145,18 euro
Daphile Audiophile Music Server for x86 hardware, it is free but it need a dedicated hardware
JRiver Media Center available for Windows and Mac OS at a cost of $49.98
Foobar2000 freeware
In same case it is necessary increase the output level so use in the setting menu the V+ and -V buttons to set +6dB. +12dB or +18dB, I suggest a max of +6dB because otherwise the distortion can increase.
There is also a function to set an amplification of +6dB for DSD tracks because some player give a lower level for these.
All the settings can be saved in the EEPROM so these will be set at the next power on (press mute button on Setting menu).


MEASUREMENTS
Output at 0dB near to 1.5Vrms.
DSC3_Int_12MHz.jpg


REVIEWS

STAR DSC took first place in the contest organized by the forum Vnav
2019_ketqua_DAC.png

Here is the result of a test in October 2019 with a friend of mine with a lot of experience in recording and playing audio.

Yesterday evening the first serious listening test about the new dsc module with CT7302 vs NAA (Network Audio Adapter) DSC module.

The quality of the last DSC with CT7302 is very high, the same details and the same sound stage of the NAA DSC module.

The NAA DSC module with the HQplayer filters is a little bit more softer but it is not easy define if it is more or less near to the real sound.

Using the CT7302 with jRiver the sound is very natural and analog.



PHOTOS
ph7.jpg


ph8.jpg


ph9.jpg
 
Last edited:
Very intersting...how does it sound with the Fifo ?


I don't have a comparison of the sound, with or without FiFo, I built it with FiFo from the beginning ... Dac is wonderful sounding, I don't have any desire to come back to any of the very expensive dacs I previously owned...



With Iancanada McDualXO I had the opportunity to "play" with clock crystals. The original units provided by McDualXO are just decents, Crystek crystals are definitely better ... more organic sound ... slighly "rounded", then I added a Pulsar clock ... and it's perfect, precise while remaining organic


In any case I've the suspect that DSC2.5.2 output stage, perfectly match the ingress stage of my amplifier, sound is so magic ...
 
ppy

You have noted the quality of resistor matters for resistor network but have you compared boards using different resistor tolerances for the network?

We know the tolerance does not effect THD but filter response between register outputs, you could say it is less important but not sure if it could be deemed unimportant.
You believe its worth using precise resistors here?