Signalyst DSC1

Jump to Latest
S

Schussor

Member
Joined 2005
This is more typical - use this then

TNT said:
Appalling distorsion figures for those trafos....

//
Click to expand...

Here is a passive DI box ART Pro Audio

Note the specs are an order of magnitude lower. The trannies inside are not better. This model has a switchable 3rd order Bessel low pass at 30khz. So if the losses at the trannie are not enough, then you can switch this filter on.

Street price about $40

This one is better spec'c and provides +14db gain
Ebtech LLS-2 2-Channel Line Level Shifter | Sweetwater.com
http://www.ebtechaudio.com/llsheinf.pdf

Cost $80
 
S

Signalyst

Member
Joined 2014
signalyst.com
Schussor said:
The SK circuits for the filters should not be difficult to duplicate with RC LP filters, you may end up having one more gain stage depending on the number of poles and the Q of the filters.
Click to expand...

It becomes heavy to drive such low-impedance network (to keep noise down).

A successful implementation of this passive RC + makeup gain is in Nelson Pass' First Watt B4 crossover. It can do 4th order filters with 2 gain stages and IIRC there might be a buffer at the output. I think the schematics were offered online at his DIY forum, might still be there.
Click to expand...

You mean this?
FIRST WATT B4

By quick glance the block diagram in the manual looks exactly like standard SK design.
 
S

Schussor

Member
Joined 2005
blackhole said:
At the output of Amanero I have only one 33 Ohm resistance in series, a capacitor in parallel and the transformer. I don't need a coupling capacitor because of transformer. I don't need a higher resistance (this also because of transformer).

The isolation from Usb ground is made by a Gefen USB extender. Power amplifier ground is isolated with the output transformer.

The S/N is not exceptional, should be around 85/90 dB, but this configuration has been made for a tube power amplifier whose real S/N could be around the same. So no record dynamic range. Moreover the output signal is lower than 2 V
The output transformers of tube power amplifier will work also as a final filter before the signal is sent to the speakers (Audiostatic eletctrosatic modified).

The sound signature is of a very transparent sound. The sound is the opposite of dark but is not bright, because high frequencies are smooth and "liquid". There is an impression (as in all DSD reproduction I have tried) of a bass a little less present.

Confronting to iFI DSD micro in DSD there is no story: this configuration is a lot better, immediately more transparent. iFi idsd micro (to my knowledge) should use the analogic internal filter of the DAC chip and an active buffer. In my case on the signal path there are not transistors, chips, no feedback, only one resistance, one capacitor in parallel, one transformer.
Click to expand...

I have been looking to see how this can be done easily

At this point I am looking at using a canned USB-to I2S DSD converter such as the Gustard U12 or U10 (TCXO 0.1ppm clocks and XMOS with sample rate/type indicator display) like bay 111470778001

And the DI boxes for transformers, either
The one with 14db gain,
Ebtech LLS-2 2-Channel Line Level Shifter | Sweetwater.com
Or the one with the Bessel filter but no gain
ART DUALZDirect | Sweetwater.com

Then my ony issue would be to make an HDMI to TRS cable and figure out what wires to use off the HDMI carrying the I2S DSD signal. Do you happen to have a URL for that?
 
G

gentlevoice

Member
Joined 2007
Paid Member
Hi all,

Very interesting thread, indeed ;-)

As it is I bought the components to make this DAC some months ago and now am gettiing closer to making a DSD DAC. This made me look closer into Jussi's schematic and I'm honestly not quite sure I understand what happens in the four shift registers ... :confused:

Any of you can just briefly explain what happens? It seems to me that there's some kind of averaging done (?) when then 1 bit signal passes through the registers ... Or may it be something else?

I am comfortable with the analog output stage - just would appreciate a brief explanation of the what happens in the shift registers.

Cheers ;)

Jesper
 
M

mcluxun

Member
Joined 2010
gentlevoice said:
Hi all,

Very interesting thread, indeed ;-)

As it is I bought the components to make this DAC some months ago and now am gettiing closer to making a DSD DAC. This made me look closer into Jussi's schematic and I'm honestly not quite sure I understand what happens in the four shift registers ... :confused:

Any of you can just briefly explain what happens? It seems to me that there's some kind of averaging done (?) when then 1 bit signal passes through the registers ... Or may it be something else?

I am comfortable with the analog output stage - just would appreciate a brief explanation of the what happens in the shift registers.

Cheers ;)

Jesper
Click to expand...
Your first Digital to Analog Converter build | Hackaday

not exactly the same thing, but this link could be a good visualization of the use of shift register.
 
G

gentlevoice

Member
Joined 2007
Paid Member
Hi mcluxun,

Thanks for your reply & link - but looking at the link & the video it seems as if it's more to do with an R2R DAC (which could be interesting as well) than with a shift register...? Or maybe I missed something?

I somehow wonder why Jussi didn't just use a flip-flop (e.g. potatosemi) and a suitable filter after the flip-flop ... :confused:

Bewildered ;)

Jesper
 
abraxalito

abraxalito

Member
Joined 2007
The shift register is an 'analog' FIR filter - otherwise called a transversal filter. Bruno uses one on the output of his latest and greatest Mola-Mola DAC. One of the benefits of using the transversal filter is reduced jitter sensitivity as DSD decoding is notably extremely jitter sensitive.

I suspect that Jussi's circuit will suffer in terms of dynamics - getting a clean enough power supply is going to be a challenge as a shift register DAC will on average have about 6dB PSRR.
 
G

gentlevoice

Member
Joined 2007
Paid Member
Hi abraxalito - thanks for your input ... Without knowing I had the impression that it performed some kind of filter function. Interesting also that Bruno uses this in his latest DAC - IMHO very often his approaches are indeed both unusual and inspiring (not least with Grimm audio).

BTW Jussi made some measurements of the DSC1 that he posted an computeraudiophile:

Preliminary measurement results for the DSC1 DAC - Blogs - Computer Audiophile

It's interesting to see that the DAC has some HF noise - wonder if this may be related to inter-output timing differences in the '595 shift registers? I've noticed that soekris R2R DAC (also?) has some HF spikes (although at lower frequencies) that is suggested could be such timing differences ... Or maybe it could be inductive/(capacitive) crosstalk between the tracks/components which look as if they are quite closely placed ...

I suspect that Jussi's circuit will suffer in terms of dynamics - getting a clean enough power supply is going to be a challenge as a shift register DAC will on average have about 6dB PSRR
Click to expand...

I am not familiar with this but Jussi's DSC doesn't load the shift registers very much (15k) so maybe it is of less influence here ..??

Cheers,

Jesper
 
abraxalito

abraxalito

Member
Joined 2007
Very difficult to estimate the audio band noise floor given only an FFT with unknown number of bins. Typically though an FFT has 64k bins so is subject to an FFT gain of at least 30dB. Since the 'grass' in those plots is above the -120dB mark this tends to indicate a lower than 90dB SNR - not at all impressive.
 
S

Signalyst

Member
Joined 2014
signalyst.com
abraxalito said:
Very difficult to estimate the audio band noise floor given only an FFT with unknown number of bins. Typically though an FFT has 64k bins so is subject to an FFT gain of at least 30dB. Since the 'grass' in those plots is above the -120dB mark this tends to indicate a lower than 90dB SNR - not at all impressive.
Click to expand...

My analyzer is not very good for estimating SNR due to limited 16-bit resolution of the high speed ADC...
 
S

Signalyst

Member
Joined 2014
signalyst.com
gentlevoice said:
It's interesting to see that the DAC has some HF noise - wonder if this may be related to inter-output timing differences in the '595 shift registers? I've noticed that soekris R2R DAC (also?) has some HF spikes (although at lower frequencies) that is suggested could be such timing differences ... Or maybe it could be inductive/(capacitive) crosstalk between the tracks/components which look as if they are quite closely placed ...
Click to expand...

Yes, it is some capacitive switching glitch leakage, but luckily I know how to improve on this front in future versions. You need to take into account that this is first version of the design... :)

This is problem with lot of commercial DACs too, for example if you look at iFi iDSD Micro:
iFi iDSD Micro measurements - Blogs - Computer Audiophile

One reason is that many tend to use just analyzers like AP that have limited bandwidth and thus don't see how things look like outside the 100 or 200 kHz analysis bandwidth, which is essential for designing a DAC.
 
S

Signalyst

Member
Joined 2014
signalyst.com
gentlevoice said:
I somehow wonder why Jussi didn't just use a flip-flop (e.g. potatosemi) and a suitable filter after the flip-flop ... :confused:
Click to expand...

The shift register I used has a separate output latch, so I saved bunch of components for not duplicating the latch. I tried to keep the circuit as simple as possible while at the same time having sensible demonstration on technical level how I think a DSD DAC should be designed.
 
S

Signalyst

Member
Joined 2014
signalyst.com
gentlevoice said:
Any of you can just briefly explain what happens? It seems to me that there's some kind of averaging done (?) when then 1 bit signal passes through the registers ... Or may it be something else?
Click to expand...

Yes...

Shift register forms a "scrambled thermometer code" that is then directly converted to analog through a latch. It averages 32 samples of data, so the output gets value between 0 and 32 and thus there are 16 possible negative output levels, a zero level and 16 possible positive output levels. (converted output value is log2(33)=5.044 ENOB for Nyquist sampling)

It is also an analog FIR filter with benefit of having optimal step (transient) response without any ringing and also reduced jitter sensitivity. This allows both cutting down the the noise shaping noise at higher frequencies and relaxes requirements for the following analog filter stages.

With careful design of the following analog filter stage, the transient response is possible to get very good:
Squarewaves from DACs - Blogs - Computer Audiophile
While noise leakage at higher DSD rates becomes very small.
 
You must log in or register to reply here.
Top Bottom