DirectStream Impulse response


Just curious what the impulse response looks like from the DirectStream using a DSD source. I’ve seen proponents of DSD show that the impulse response looks very analog-like with almost no ringing, unlike PCM even at high resolution formats.

The early Stereophile test on the Directstream does show ringing, but that was with 44kHz PCM source:

So, does it show less ringing with DSD data? Thanks.


In the DS the PCM upsamplers from 22.05k, 32k, 44.1k, 48k, 88.2k and 96k are all similar to the plot by Stereophile, tho I’d quote a different description of that plot: “With the tone, the second, third, and fourth harmonics, respectively at 38.2, 57.3, and 76.4kHz, can be seen, albeit at respectably low levels. But there is no trace of the aliasing product at 25kHz, this being completely suppressed by the reconstruction filter.”

Here’s the DS’s output of a 960Hz square wave sampled at 192k:


You can see the expected ringing of the IIR upsampling filter - it would be the same if it were an analog filter - band limited inputs have ringing and the PCM inputs are all bandlimited.

For reference here’s the theoretically perfect bandlimited 1k square wave (the squares are the sampling points):Square.png

I don’t have a convenient DSD test tone right now, but in general it will be sharper and the ringing will die down sooner. That’s the nature of the beast.

Here’s a plot that you can see at various places around the net:



@tedsmith If I may, can you further explain square wave measurements and what they tell to us to non-engineer folks? I ask because I came across this website which, if I’m potentially reading it correctly, suggests DS has a treble boost character?

-Many thanks.


No, that discussion is not on point.

If you measure the DS you get a flat frequency response so any claims to the contrary which aren’t measurements don’t matter :slight_smile: The site is talking about analog test signals not sampled signals.

An A/D (and PCM in general) requires a bandlimiter with a very sharp cutoff, a sampled bandlimited square wave has what’s often called preringing, overshoot, post ringing, etc., depending on the exact bandlimiting filter. So any signal shape preserving filter (filter that renders phase accurately) has about 9% overshoot and some preringing and settling that looks like ringing (the Gibbs phenomenon.)

This isn’t to be worried about since we don’t listen to test tones :slight_smile: In the practical world the sample rate basically restricts the maximum slew rate (how fast a signal can go up or down) that faithfully representable. Put another way, you need higher frequencies to get steeper slopes and the steepest slope you can get is constrained by the sample rate.

Anyway given a signal that’s bandlimited by the A/D, a perfect DAC (for that A/D) will render it faithfully with out (additional) preringing, frequency response changes, etc. Even if a DAC gets the frequency response correct (many/most) do, it still may not get the phase response correct to get back the actual wave shape. For example some DACs consider non-linear phase response to be less important than, say, avoiding preringing. Since there’s no perfect filter different DACs are optimized for different things.


Thank you, Ted! Very much appreciated.


My feeling is the DS besides being king of imaging is also king of timing and therefore probably king of phase response