DirectStream DAC - wide band sweep

Hi @tedsmith

Can you share a wide band 0 - 22.05 kHz sweep, from 44.1/32 source file, at 5 MHz bandwidth.

Something like this:

Why ?


5 MH bandwidth? If you are really talking about a sweep, the 5Mz doesn’t make any sense at all, the configuration for a “sweep” is the filter, e.g. 1/3 octave and the frequency range. If you are talking about FFTs, you need to specify the sampling rate, the number of bins, the number of FFTs averaged… Without these specs an FFT “sweep” isn’t useful. FWIW Stereophile has some measurements of older versions of the DS software.

My own FFTs have noise from my system that I’ve “calibrated” out, but you won’t know what’s from my USB cable vs what’s from the DAC.

In the abstract when measured like this over the audio band the DS has a flat noise floor whose level depends on the above FFT parameters (except for some small peaks at 120Hz, 180Hz… Depending on the software release there’s an ultrasonic rise to a level not much more that 30 to 40dB above that noise floor which starts at 70-100kHz, after an octave or so it falls to the flat noise floor of your measuring system (modified by the above FFT parameters.) approx at -140dFS with the parameters I typically use. I don’t typically store these, I’m usually looking at the audio band and/or the shape of the bump. As rajugsw asks: “why?” what are you really looking for?


A stick, perhaps?

1 Like

Do Any images (correlated with input signal) show up?

Or “humps” like shown in the OP.

That is a measurement of RME ADI-2 DAC when it is fed DSD256 (the AKM chip is in “DSD Direct” mode). No images but you see DSD noise humps. Which I believe are harmless as they are uncorrelated with input signal.

Uhhh…I dunno…Ted’s DAC just sounds so friggin’ awesome .I listen to it and don’t worry about the tech behind it (that’s his problem - not mine) :smiley:


I described the one ultrasonic hump. I placed that dips you see at about 80kHz and 110kHz instead of the 100k and 500kHz in your plot to extend the the audio band flatness. (Why waste them? :slight_smile: )

Images? are you asking about “aliasing” (reconstruction images from improper low pass filtering)? To look at those you want to zoom in to the audio band (and a little higher.) You need to play a test tone and look for those. I use a very slow frequency sweep over the extended audio band to check for such bugs in the DS.


Hi Ted. Yes

Yes, see below, for the same DAC. Fed PCM44.1 kHz input signal vs DSD256 input signal. Not my measurements btw.

So I was interested in seeing similar for DirectStream DAC. I would imagine no images, due to the very high upsampling rate and analogue filtering?

The DS uses a proper anti-imaging filter - there aren’t images like this in the ultrasonic. The PCM filters all reach below -140 by the Nyquist rate. They are very steep so there’s no tradeoff in audio bandwidth or imaging like some “soft rolloff” filters