Dsd without D/A stage....

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hallo, recently I've read a Ken Ishiwata interwiew (suono 3/05) above the DSD format and his own system.
In wich the signal coming from the SACD drive going directly trought a simple low-pass filter, without it go trought a D/A converter. I'm a little bit puzzled from this opportunity...isn't a DSD signal a digital bit sequence that need to be converted in a analogue stream? Some one can help me and suggest me some litterature on this format? Or explain how should work a system like the Ishiwata one?
Many Thanks
Domenico
 
DSDstream.gif


With a simple R/C network the Cap will becharged with every '1' and discharged with a '0'
As you can see from the picture above al lot of '1's wil give you a higher Voltage on the Cap and vice versa.

You would need a decoupling Cap to remove the DC offset (½Vdd)
Don't ask me about the values ;)

Greetz Simon
 
Administrator
Joined 2004
Paid Member
Its just a simple low pass filter that integrates the dsd pulse train. I am actually planning to do this to my sacd player in my next round of modifications. (See my thread for progress so far.)

A simple first order filter designed to roll off above 50kHz will do it, but the amount of RF present on the output would be substantial. I believe a third order filter would provide enough attenuation for most applications.

I am thinking of buffering the digital outputs of my dsd decoder ic and feeding that signal to a passive RLC filter designed to do the job - possibly if I can keep the impedances low enough I will not even need active buffering at the output of the filter. Using level translation would actually allow me to have voltage gain in the digital domain where in theory there is no added distortion.

The 0 baseline is at half rail and therefore the output must be ac coupled or the baseline offset nulled out. (easily done, but requires a good op-amp or two.)

Kevin
 
kyrochan said:
How could it be possible?
I think that a DAC should be a bit puzzled to have a so complex element (like a speaker) at its output, and should not be able to bring to its output something like 4-5 ampére at 23V
Please explain your tought
"In wich the signal coming from the SACD drive going directly trought a simple low-pass filter, without it go trought a D/A converter. "
I suppose by doing the same with high power push-pull output driven by DSD and may be without low-pass filter.
 
Administrator
Joined 2004
Paid Member
In effect you are talking about a power dac, and yes it can be done by varying the raw dc voltage to the power amplifier - note that you still need a post filter to reconstruct the signal after the digital amplifier. I think in this case an air core inductor of a couple of mH and a capacitor such that your cut off frequency is around 50kHz would work well.. The digital amplifier would need to consist of a couple of high speed complementary mosfets and some level translation to provide gate drive. The power supply would be the most complex part of the equation and would need very good regulation and stability. If you desire to avoid using a large output capacitor either a bridge configuration could be used or tightly tracking bipolar supplies.

In my system I am planning to do something similar at line level, except that the supply will be a low noise fixed 5V.

Designing a good power dac as you describe is not trivial, and one of the biggest issues you will probably face is linearity in the output inductors and caps - you do not want to rely on the speaker inductance to integrate the pwm at the voice coil, not to mention the idea of 2.8224MHz RF running around speaker cables at high amplitudes may cause emi problems with other components in your hifi.
 
stolbovoy said:
Didn’t anybody think about making high current DSD "DAC" to feed loudspeaker directly, controlling volume by altering rail voltage. Pretty interesting and simple at first glance…

I've mentioned this idea on the forum before, I think.

It is similar to what Tact used in their Millenium amplifier, which used a class D output stage, where the volume control set the rail voltage.

My suggestion is that this approach should be used with electrostatic panels. Essentially, this would let the air perform most of the integration, which is about as ideal as it gets, IMHO.

Try it out with a set of "cheap" Stax 202 headphones. 100Vrms delivers 100dB, so 10Vrms should give you 80dB. Pretty okay for a prototype, and simpler than trying to deliver 1KVrms (120dB) peak at 2.8MHz into a 120pF load..
 
I'm very interested in matter of driving conventional loudspeakers by digital signal, currently i'm finishing my PhD work - digital loudpeaker - and 1-bit sigmadelta (DSD) stream seems to be an alternative to driving high count of small transducers with PCM decomposed signal. So, if anyone finish any suitable amplifier for DSD, let me know :) thanks
 
kevinkr said:
In effect you are talking about a power dac, and yes it can be done by varying the raw dc voltage to the power amplifier - note that you still need a post filter to reconstruct the signal after the digital amplifier. I think in this case an air core inductor of a couple of mH and a capacitor such that your cut off frequency is around 50kHz would work well..
What will happen if you let loudspeaker to do the "reconstruction" part of the job?
 
kevinkr said:
The power supply would be the most complex part of the equation and would need very good regulation and stability. If you desire to avoid using a large output capacitor either a bridge configuration could be used or tightly tracking bipolar supplies.
It sounds as power supply for this solution is much more critical than for conventional design. Why?
 
Administrator
Joined 2004
Paid Member
The reactance of the speaker voice coil is not well defined at rf, and the interwinding capacitance, not to mention strays result in unexpected resonances which can result in significant self heating. In addition depending on the amount of inductance present the impedance of the driver can be quite high resulting in extreme voltage drive requirements or high currents if there is a lot of capacitance in the voice coil.

The driver's impedance is not defined at these frequencies so it would be hard to know what amount of power would be delivered to it, and hence its efficiency.

A woofer driven through a second order LC based cross-over could work fairly well depending on how lossy the inductor is at higher frequencies and where the self resonance is.

A tweeter might not survive depending on the impedance across its voice coil at the switching frequency..

Also levels of rf could be quite high and some rf would be radiated off of your speaker cables which could create local rfi problems for your hifi and telephone.

Designing a well regulated and stable supply at the currents required is not a huge problem, making sure that the impedance of the supply is low at the switching frequency so as to minimize losses and rfi is an issue. If you contemplate using bipolar supplies they will need to track closely in order to eliminate variable dc offsets in the output.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.