DirectStream Volume


#1

I ordered the DirectStream about a week ago and had a few questions about the volume control. I plan on using the DS as a DAC only.

  1. Should I set the Volume to 100 to avoid any loss of resolution?
  2. Do I set the Gain to High or Low?
  3. Is the volume control analogue or digital? I've read that if the volume control is digital the signal must be converted to PCM.

Thanks


#2

The DS has more headroom than typical DACs (i.e., more bits that can be thrown away without audible effect) and you should not lose resolution if the volume control is not set to 100. As for the gain, that depends on your system. Set it so that the gain best suits your amp (if connected directly) or preamp (e.g., so the usable preamp volume range is somewhere in the middle). The volume control is digital but there is no PCM conversion.


#3
asiano said I ordered the DirectStream about a week ago and had a few questions about the volume control. I plan on using the DS as a DAC only.
    1. Should I set the Volume to 100 to avoid any loss of resolution?
      1. Do I set the Gain to High or Low?
        1. Is the volume control analogue or digital? I've read that if the volume control is digital the signal must be converted to PCM.

        Thanks


        Steve M answer is correct. It’s a no resolution loss volume control and is never converted to PCM.

        #4
        asiano said
        1. Is the volume control analogue or digital? I've read that if the volume control is digital the signal must be converted to PCM.
        One cannot change volume on a 1-bit data stream, be it PCM or DSD/PDM. It is the 1-bit format which is the issue, not PDM or PCM.

        The DirectStream converts incoming data to 28.22MHz at 30 bits (calling this 10x DSD is a bit confusing; the sampling rate is 10 times that of standard DSD, but referring to DSD implies a 1-bit architecture.)

        The 30 bit data stream allows for volume changes. After the volume change, the data is down converted and noise shaped to 2x DSD 1-bit and the FPGA output goes to the low pass filter for conversion to analog. The output is transformer coupled (like most tube amps).

        As a side note, the DirectStream oversamples 10x DSD rates as this allows integer conversions (no numbers to the right of the decimal point) for both sample rate families of 44.1 base rate and 48 base rate. Integer conversion provides for simpler math and no rounding errors. This is why there is no loss of resolution. It is a very cool design.


        #5

        Slightly expanding on Elk’s post:

        I’ve put a FAQ on the site about worrying about PCM vs DSD: http://www.psaudio.com/ps_faq/converting-dsd-to-pcm-bad/

        It’s a matter of definition as to whether the volume is implemented with PCM or DSD.

        If the volume control wasn’t in the DS the DS would still have the same architecture and do the same math.

        When we say PCM, noised shaped samples at a 2.8224MHz or a 28.224 MHz sampling rate isn’t what we usually think of.

        When we say DSD, a 55 bit noise shaped encoding isn’t what we usually think of.

        The volume control is done with 55 bits at a 2.8224MHz rate but in the multibit part of the sigma delta remodulator. As I alluded to, that remodulator is in integral part of going from single rate DSD to double rate DSD - a process that some might argue is not an intrinsically PCM process.


        #6

        Thanks, Ted, for clarifying and correcting.

        I almost got it right. 55 bits at a 2.8224MHz is very fun as a concept.

        My recollection is the FPGA is 30 bits. Am I misremembering? If not, where is the additional 25 bits introduced? I am probably a bit confused as to the data path.

        Thank you once again for putting up with us and so actively participating in the forum.


        #7
        Elk said I almost got it right. 55 bits at a 2.8224MHz is very fun as a concept.

        My recollection is the FPGA is 30 bits. Am I misremembering? If not, where is the additional 25 bits introduced? I am probably a bit confused as to the data path.

        You aren't misremembering - I use a 24 bit volume (with 2^20 being unity) so I need at least 54 bits there (the extra bit is a technical detail.)

        One of the advantages of an FPGA is that you can use whatever resources you want when you need them. The width of the samples changes all over the place as needed. For example in a FIR filter the most obvious implementation needs a precision of (width-of-input * width-of-coefficients + log2(number-of-coefficients)) but the result is no more accurate than the input.

        Similarly the upsampler uses more the 30 bits internally (at 28.224MHz), and the sigma delta modulator then needs more bits than just the 55 bits to avoid saturation…


        #8

        Got it.

        Thanks!