1. Things you need to know about the new ‘Conversations’ PM system:

    a) DO NOT REPLY TO THE NOTIFICATION EMAIL! I get them, not the intended recipient. I get a lot of them and I do not want them! It is just a notification, log into the site and reply from there.

    b) To delete old conversations use the ‘Leave conversation’ option. This is just delete by another name.
    Dismiss Notice

Some observations about the Benchmark DAC1

Discussion in 'audio' started by darrenyeats, Apr 24, 2013.

  1. darrenyeats

    darrenyeats pfm Member

    IMO the DAC1's true potential is realised when the extras (XLR output attenuators, physical volume pot) are switched off, and it is fed a digital signal that is attenuated, upsampled well (e.g. SoX with -v) to 110.6kHz then dithered at 24 bits. If used via XLR, also the XLR calibration trimmers (NOT to be confused with output attenuators) must be used.

    XLR is a balanced connection, this offers a high degree of common mode noise rejection, which in some systems has sonic benefits.

    (1) to (7) took me to "now the sound is not annoying".
    Finally (8) took me to "now it sounds good".

    1. Use XLR but disable XLR output attenuators by setting to 0db (factory is -20db).

    2. Use 'calibrated' mode which bypasses the volume pot.

    3. Use the internal 'calibration trimmers' to reduce the XLR output levels. It should be possible to reduce output by at least 15dB, I got more like 20dB. See notes *.

    4. Use digital attenuation in the transport. It should be dithered at 24 bits. The ideal way to do this is discussed in point 8!
    Why? "DSP headroom"; see pictures in post 11 later in thread; see Figs 1, 2 and 13 at https://benchmarkmedia.com/blogs/ap...chmark-dac2-vs-dac1-side-by-side-measurements.
    Why dithered? Dither makes sure the errors take the form of noise instead of distortion.
    Why dither at 24 bits? The DAC1 accepts input, and operates, at 24 bits.
    Make day-to-day use of the digital volume control in the transport and never go above a certain level of attenuation****

    If you own a Squeezebox Touch see http://forums.slimdevices.com/showthread.php?104629-Dithered-volume-control-for-Squeezebox and http://forums.slimdevices.com/showthread.php?105197-All-day-battery-power-for-Touch-%A324.

    5. Pay attention to cabling and power supply. See here: http://www.pinkfishmedia.net/forum/showpost.php?p=2456579&postcount=126.

    6. I found S/PDIF TOSLINK sounds better than coax with Squeezebox Touch into the DAC1 - maybe due to the combination of galvanic isolation of TOSLINK and jitter reduction of the DAC1. I would not use USB: http://forums.slimdevices.com/showt...nly)-exist-now&p=787260&viewfull=1#post787260

    7. Let it warm up to a stable temperature, half an hour should be sufficient. See http://www.soundonsound.com/sos/jul05/articles/benchmark.htm (sidebar 'A Warm Feeling'). There is some evidence that leaving a DAC switched on longer can measurably help clock stability, I think this will be inaudible but for reference:
    http://www.audioquest.com/wp-content/uploads/2014/04/Phase-Noise-Jitter-Report-0317-14.pdf.

    8. If you have a Squeezebox Touch with Linux server**, upsample to 110.6kHz using SoX prior to feeding files into the DAC1. The sox command should include attenuation (e.g. gain -6.02059991327962390427 which to SoX is a 1 bit shift) followed by 'rate -v -b 89 110600' (this will use the default linear phase filter, -v means very high quality, -b 89 gives an optimally smooth drop off between 19kHz and 22kHz assuming a 44.1kHz rate - for an informative thread see http://www.pinkfishmedia.net/forum/showthread.php?p=3040013#post3040013). Ensure the sox command ends with 'dither' directive. This can be done to music files beforehand. If you have a Squeezebox Touch: you will have to install EDO to support this rate; you could also in convert.conf set up the sox call to convert on the fly all streamed music; ideally combine with (4) so bit shift, upsampling, volume attenuation and dither are done in one SoX call for maximum precision - recommended! PM me for detailed instructions!

    The DAC1 can accept arbitrary rates (between 28kHz and 195kHz) because of its ASRC. The manual states the TOSLINK works up to 96kHz, but it seems to work at 110,120,130kHz - however not 140kHz***!

    You can also try a similar filter but flat to 20kHz (for 16/44) using "-b 92.5" instead.

    --
    Steps 1-3 is how you can get an XLR output level of around 2VRMS and output impedance of 60 ohms *at the same time*. You can't achieve this any other way than steps 1-3 with a DAC1, and you can't achieve it at all with a DAC2/3 (because there is no step 3 possible with a DAC2/3). Reasonable output level and good output impedance are relevant if you're using your DAC as a DAC/pre. Output impedance is less important if using a separate pre or integrated amp.

    Other options that give an XLR output level of <=2VRMS and output impedance of <=60ohms *at the same time*:
    Various Weiss DACs
    Various dCS DACs

    ...!

    * (Point 3) Benchmark confirmed to me by email that the 'calibration trimmers' don't raise output impedance. (By contrast the 'output attenuators' do raise output impedance, for me a sign they are in the "wrong" place i.e. after the preamp output stage.) To achieve point 3 you will need test tones, a DMM and a precision screwdriver set - and possibly an amp - to calibrate left and right channels. If you need help then PM me.

    (Tip: considering output levels of modern sources, it's good FIRST to use any input sensitivity settings on your power amp or actives to minimise loudness. Then, set the DAC1's output levels. That's the order of priority. But you could well find it appropriate to use quietest settings at both stages, like I did. Also see note "****" below!)

    ** (Point 8) not tried Windows server; also all this might be possible with Signalyst HQ Player, but I'm not sure if it supports non-standard rates or if PC audio cards generally support non-standard rates.

    *** (Point 8) this is evidence to me that the Squeezebox Touch is really outputting non-standard rates. It may be that Squeezebox Touch and DAC1 are a "peaches and cream" combo because this works!

    **** Actually, digital -25dB seems to result in better sound than -5dB despite the technical loss of SNR - at least this is the state of affairs after doing steps 1-8 in my case. The sound is effortless, I don't know why. One should be flexible for quieter recordings. Personally, I most often range between -30dB and -20dB, sometimes higher or lower - but I never go higher than -17dB. If you think this is foo, then at least stay below about -5dB since that at least gives necessary DSP headroom!

    --

    Benchmark say the volume pot is transparent, the output attenuators are transparent, the digital interfaces all sound the same etc ... but the DAC1 measurements on Stereophile are taken in its stripped back state with volume pot DISABLED and output attenuators DISABLED. And also, it seems, whilst using its TOSLINK input. This might explain why it measures so well and yet audiophiles complain about the sound (when listening with attenuators in circuit, volume pot in circuit and using USB, for example).

    See section "Addendum to Benchmark DAC1 Review" here https://positive-feedback.com/Issue26/benchmark_dac1.htm and section "The Jumpers" here http://www.positive-feedback.com/Issue48/benchmark_usb.htm I dismissed it as audiophile rubbish for one or two years but, since I experimented, I believe it really isn't.

    It is true that each digital interface is resistant to input jitter, but it's not true that each interface exhibits exactly the same jitter (Ken Rockwell has measured the TOSLINK as slightly better than the coax and far better than USB).



    Original post:

    Issue 1. Channel balance.

    Ken Rockwell reviewed the HDR some time ago and observed that channel imbalance using the potentiometer peaks at 1.25db around the low extremity of volume. The HDR has a good custom-made ALPS potentiometer: I suspect these comments apply to many other devices (although some stepped attenuators might be better for channel balance).
    http://kenrockwell.com/audio/benchmark/dac1-hdr.htm#meas - Channel Tracking

    Issue 2. DSP Headroom

    According to Ken Rockwell, many DACs (NOT just the DAC1) can be tripped up by extreme waveforms.
    http://kenrockwell.com/audio/benchmark/dac1-hdr.htm#meas - Square Wave Spectra (overshoot handling)

    Note this has been remedied in the DAC2 HGC
    http://www.benchmarkmedia.com/dac/dac2-hgc - High Headroom DSP - with 3.5 dB "Excess" Digital Headroom

    Note the 3.5db figure is not related to anything specific about the DAC1, it is a generic figure based on the amount of headroom needed for almost any extreme (e.g. hyper-compressed) recording. Natural music will most likely not be affected. This figure is applicable to many other upsampling DACs which have the same DSP headroom situation.
     
    Last edited: May 5, 2018
  2. AndyU

    AndyU pfm Member

    Just to say, having had a DAC1 HDR, I now have a DAC2 HGC and it sounds really wonderful. More than a good step up from the DAC1. It does something clever with analogue and digital volume control and passive attenuators and spare headroom. Just noticed they've brought out a slightly cheaper digital input only variant. Very highly recommended indeed.

    (but, to follow on from the o/p, if you use your DAC from it's USB input and have doubts about it's own volume control, you could use JRiver Media Center as your player. JRMC has a 64 bit volume control, properly dithered too, so is probably as good as it gets)
     
  3. darrenyeats

    darrenyeats pfm Member

    Thanks Andy! I use the Squeezebox Touch over optical S/PDIF into the HDR. The Touch has a 24 bit dithered volume.

    The way DAC2 HGC handles peak headroom can be mimicked by just keeping the digital volume at least 3.5db lower than max at all times. For this to work, digital volume control needs to be done upstream from the DAC1, in the transport, hence my solution in the OP.

    Glad you're enjoying your DAC2 ... it's clearly a SOTA device!
    Darren
     
  4. Werner

    Werner pfm Member

    One could argue that this does not really matter.

    If the source has the occasional excursion to 0dBFS and a potentially resulting freak waveform post-reconstruction then this will generally go unnoticed.

    If the source lives long-term near 0dBFS then it is innately ****ed up (sorry: mastered) and the additional distortion won't be of that much harm.
     
  5. Paul R

    Paul R pfm Member

    You're probably right. But it is an avoidable imperfection.

    Here's an example from a jazz CD (Art Blakey/Moanin'),

    [​IMG]

    Load a track into your favourite editor, upsample it, count the clips. Decide whether that is acceptable. Or use a DAC with a digital attenuator prior to the digital filter.

    Paul
     
  6. darrenyeats

    darrenyeats pfm Member

    Hi Werner,
    Broadly I agree with you, which is why I argue based on performance rather than audibility.

    This costs me nothing; and a lot of my favourite music is, unfortunately, badly mastered for loudness!
    Darren

    PS: I do hear an improvement but that isn't the basis of my argument, if you see what I mean, because the cause might easily be psychological with small differences like this. And a blind A/B would be tricky.

    But because I know intellectually it is better, it doesn't matter if the sonic difference is proven or not, I prefer this set up.
     
  7. AndyU

    AndyU pfm Member

    Another plug for JRiver - it will analyse any or all of your tracks for peak level; were you so inclined to follow darrens strategy you could pick the ones that hit or got close to 100% and resample them appropriately, or use replay gain, which would keep the digital attenuation only for tracks for which you believe it to be necessary.
     
  8. darrenyeats

    darrenyeats pfm Member

    The loud track "I Can Talk" by Two Door Cinema Club is already infamous from another thread. Here's the problem ... I love that album! Ha ha!

    I reduced gain by 3.5db and resampled to 192kHz using Sox.

    Audacity is reporting the new peak as -2.1db which means we've gained 1.4db of headroom. Give that it was DR4 to begin with, this is something! Normally as Werner said this sort of thing would be a few samples. But this track was originally a solid brick and now it has sprouted teeny fuzz, visually, throughout the brick ... for me it's plausible you could hear that.
    Darren
     
  9. Werner

    Werner pfm Member

    You could always listen to the un(strikethat)less compressed Jools Holland version.

    As for the nice and fuzzy waveforms: again, you won't hear that. The ear really has no single-cycle peak detector.
     
  10. darrenyeats

    darrenyeats pfm Member

    Hi Werner,
    I won't put up much of a fight about audibility ... I have no evidence to the contrary.

    Of course, once using digital volume, simply not maxing it out is no hardship in exchange for peace of mind; and I have one less reason to open my wallet for a DAC2!

    What is certainly significant is that digital volume performs better. Up to around 1.25db of channel imbalance is a material issue which digital avoids. And according to Ken Rockwell's analyses, in THD terms digital volume control doesn't seem to do any worse than physical (even though I thought it would and that was one reason I bought the HDR).
    Darren
     
  11. darrenyeats

    darrenyeats pfm Member

    This is the difference I was referring to with that track earlier.

    [​IMG]

    Both tracks upsampled to 192kHz. Top one -3.5db applied beforehand (i.e. DSP headroom) bottom one no attenuation beforehand (no DSP headroom).

    I adjusted the levels for both afterward, to make them comparable.
    Darren
     
  12. AndyU

    AndyU pfm Member

    darren - How do you know your HDR has that amount of channel imbalance? Didn't Rockwells measurements show the 1.5db discrepancy in his unit was only at low listening levels - a pretty quiet 9 o'clock on the volume control? When you listen to a mono signal on your unit is it central or displaced laterally? Can you hear it move as you reduce the volume?
     
  13. Paul R

    Paul R pfm Member

    The problem with that track is that it has a lot of clips regardless, the processing means that they don't quite touch 0dBFs, but the flat top to the waveform gives it away. Most of what you get back with the attenuation is a 'ring' on the corner of the clipped wave.

    I've been trying to find an example of something that is dynamic and 'well mastered', just touched 0dbFS and has implicit digital overs.

    Paul
     
  14. darrenyeats

    darrenyeats pfm Member

    Agreed, that's why I said "up to" around 1.25db. Absolutely, you should set up gain structure so that typical loud listening occurs around 12 o'clock (where the imbalance is very small) but then again I don't always want to listen loud: that's the point of having a volume control.

    Having said that, HDR pot imbalance even at 10 o'clock is 0.5db.
     
  15. darrenyeats

    darrenyeats pfm Member

    The screen-shot demonstrates the kinds of re-sampled digital waveform from which the DAC will attempt to generate an analogue signal. All I'd say is, there is a difference between having DSP headroom and not having it and it looks like that. Maybe the Benchmark website has more information about the benefits of DSP headroom.
    Darren
     
  16. AndyU

    AndyU pfm Member

    How do you know your unit is imbalanced? Can you actually hear it? What happens when you play a mono signal? Does it pan as you move the volume control? Are you worrying about things you can hear and have measured on your unit, or are you worrying about things you can't hear and haven't measured and which may not actually exist? If you're really worried, why not get someone competent to measure your DAC - it'll only take a few minutes.
     
  17. darrenyeats

    darrenyeats pfm Member

    I am going by the imbalance in Ken Rockwell's unit (see OP for link) and the unit here:

    http://www.benchmarkmedia.com/discuss/forum/dac1-pre-dac1-hdr/dac-1-hdr-channel-imbalance

    which Benchmark states is "within spec".

    It does exist: all analogue attenuators suffer from some kind of channel balance issue, usually the stepped attenuators are better ... alas the HDR cannot accommodate a stepped attenuator.

    The key point is Ken Rockwell's analysis of the THD behaviour of the HDR pot versus a digital volume control, finding they are very similar. So the real question then changes to "why use a physical pot?"
    Darren
     
  18. AndyU

    AndyU pfm Member

    But you haven't got Ken Rockwells DAC1. Have you heard or measured any imbalance in your own DAC1? If you have, have you followed the recalibration procedure that Benchmark advise in the link you posted? Using, in their words, "an accurate calibration recording and a meter"? Or your ears. If you haven't heard or measured any imbalance in your own unit, then what's the issue?
     
  19. darrenyeats

    darrenyeats pfm Member

    Andy,
    You calibrate a particular position (sensibly they recommend 12 o'clock). The other positions will always be imbalanced even if the calibrated position is perfect (and both examples were excellent at 12 o'clock) ... that is the nature of the physical pot.
    Darren
     
  20. AndyU

    AndyU pfm Member

    How do you know how imbalanced yours is Darren? Have you measured it? Can you hear anything? Why assume that the difference someone else has measured on a different example exists in yours? Surely if you have any misgivings you should get your DAC checked out by someone with the test-gear to measure it and the competence to fix it if it is sub-standard.
     

Share This Page

  1. This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
    By continuing to use this site, you are consenting to our use of cookies.
    Dismiss Notice