CHANGSTAR: Audiophile Headphone Reviews and Early 90s Style BBS

  • December 31, 2015, 09:40:31 AM
  • Welcome, Guest
Please login or register.

Login with username, password and session length
Advanced search  
Pages: 1 2 3 [4] 5 6 7 8 9 ... 11

Author Topic: slimline dac with the AK4396, and an XMOS USB receiver (>= 384 PCM/64/128 DSD)  (Read 17059 times)

0 Members and 2 Guests are viewing this topic.

AstralStorm

  • Speculation and Speculums
  • Able Bodied Sailor
  • Pirate
  • ***
  • Brownie Points: +250/-164
  • Offline Offline
  • Posts: 559
  • Warning: causes nearby electronics to go haywire

Uhm, group delay is the corollary of phase distortion, specifically phase is the differential of group delay. What are you trying to say?

78k is almost unfiltered. Might as well skip it or reduce it to first order.

Filter ringing is not really a source of IMD per se, but it can make existing IMD louder and thus more active.

I recommend digital filtering instead.
Logged
For sale: Hifiman HE-500; Paradox; Brainwavz B2. PM me if you would like to buy them.

Peef

  • Powder Monkey
  • *
  • Brownie Points: +11/-0
  • Offline Offline
  • Posts: 17
  • frends

Sorry, bad wording on my part. I meant to say that it is not possible to design a low pass filter that will have both constant delay, and constant phase for all frequencies. Phase delay is phase divided by frequency, and group delay is that expression's first derivative. The only way for group delay to be constant, if there's already some delay in the system, is for the phase shift to be proportional to frequency.

I can think of tree ways of making sure group delay is constant within the passband: using all pass filters for phase correction, increasing the cutoff frequency to push delay far enough from the highest frequency of interest, and using constant delay filters. Because of its delay response and impedance requirements, I haven't been able to successfully optimize it using reasonable LC values. That's why I shifted the cutoff to 78k.

Even with the high cutoff frequency, there's plenty of filtering due to the high filter order. Here's a comparison of a 1st order with Fc of 30kHz and the 7th order Bessel:

http://i.imgur.com/wJyvTvg.png

In a NOS DAC, both filters are pretty useless (the 1st order having an extra 4dB or so of attenuation at 44.1k). At 8x oversampling, the Bessel is down 50dB while the 1st order is just down 22dB. Hardly unfiltered, but not quite a 16 bit system yet. :)

More attenuation wouldn't hurt. Jung designed an interesting 7th order linear phase reconstruction filter using a linear-phase FDNR filter. I forget why I didn't end up doing a passive implementation of that, but it has excellent characteristics, if not a bit of high frequency rolloff. It's somewhere in this 100 page pdf: http://www.analog.com/library/analogDialogue/archives/39-05/Web_Ch5_final_PtB_F.pdf

Very curious about the IMD stuff, any idea where I can read more about it?
Logged

Peef

  • Powder Monkey
  • *
  • Brownie Points: +11/-0
  • Offline Offline
  • Posts: 17
  • frends

And (sorry for the double post) here's the lastest iteration. This one's a linear phase, based on the Jung FDNR reconstruction filter.



The simulation results are excellent. 80dB of attenuation with <1us of delay at 20 kHz. I think this is a more interesting compromise than the -1dB attenuation presented with the FDNR at 30 kHz. I think this is equivalent to 14 bits of dynamic range or so? Not enough for DSD or high res, but getting pretty close to the pcm63k's noise floor. Much better than the third order FDNR found in most older DACs.

Huge sim result, vs previous Bessel: http://i.imgur.com/i120ad4.png
Logged

ultrabike

  • Burritous Supremus (and Mexican Ewok)
  • Master
  • Pirate
  • *****
  • Brownie Points: +4226/-2
  • Offline Offline
  • Posts: 2384
  • I consider myself "normal"

Thank you very much Peef. I'm not as knowledgeable in analog design as some others here, and I'm learning stuff from your posts :)p5.

Most of the Bessel's (and Bessel like) I've seen seem active (but involve mostly resistors and caps). Pretty cool you are doing it fully with passive components.
Logged

Chris1967

  • Able Bodied Sailor
  • Powder Monkey
  • ***
  • Brownie Points: +28/-0
  • Offline Offline
  • Posts: 89

Finished the Salas in the weekend...









sounds nicer now... more detail and "air"...
Logged

Hroðulf

  • Able Bodied Sailor
  • Pirate
  • ***
  • Brownie Points: +262/-1
  • Offline Offline
  • Posts: 538
  • MOT: Sonarworks

Looking good, I'd rather had tried to stuff everything in a more compact enclosure. So far how high would you rate the DAC?
Logged
If it keeps on ringin', levee's goin' to break..

Chris1967

  • Able Bodied Sailor
  • Powder Monkey
  • ***
  • Brownie Points: +28/-0
  • Offline Offline
  • Posts: 89

Thanks Hroðulf...

About the enclosure... it was very cheap (13 euro only), very sturdy sides, double bottom, and very easy to work with front and back (soft thin aluminium) which makes it easy to drill holes etc...

It has happened in the past that i couldn't cram everything into smaller enclosures, and had to buy again, plus the fact that i would prefer to have electronic components further apart, for isolation purposes.

I keep the noisy ones to the one side and gradually move to the audio out side to be the furthest apart.

I believe also in mains filtering... this takes up some space also, and it is noisy.

The regulators tend to heat up, so i also needed space for the heat sinks, plus enough space for convection...

In the future i will make an aluminium face-plate without handles, and i will try to incorporate the lock and sampling frequency leds on it... i think this will look pretty cool and it will be useful also.

Well how to i rate this dac...

I haven't had a lot of commercial product available to me so i cannot convey comparisons, but i must say that it has lots of detail, space and stero definition, is very natural sounding...and it causes no listening fatigue...

For me these are high priorities.

I will compare it soon, to two very hi spec, over engineered Buffalo III dacs, and will report back...

To me diy is a solution to get high value, out of spending relatively little, and i think this dac board, with the usb controller, is just that.

It is no giant killer (and i don't think it will kill the Buffalo's..), just a very very honest product.

I have made me three (relatively cheap) diy dac's in the last 4 years... i have gained a lot of listening experience, and bought some time allowing the technology to become maturer...

In the future i will most definitely go for a commercial product... but meantime, i am enjoying my diy projects... :)p1
Logged

MisterRogers

  • It's a beautiful day in his neighborhood! He is a master DAC builder after all...
  • Able Bodied Sailor
  • Pirate
  • ***
  • Brownie Points: +82/-1
  • Online Online
  • Posts: 257

Very nice build Chris! By itself, the XMOS/AK stack with good value (as you have) is a very good performer for the price. You need to hear this with a JG Filter Buffer - it elevates this kit such that I prefer it to my more complex Sabre 9018 builds (PH regs, isolation/reclocking, NTD1 IV, etc.).

JG boards are hard to come by though. I mentioned awhile back that Joro/Lyuben (the designers of this kit) have been working on their version of a filter board. It's very flexible; allowing you to build it like a JG (2nd order harmonic filter), or apply filtering to other parts of the spectrum. I have a prototype board, parts on the way, and hope to build it this weekend.

When I have permission from Lyuben, I'll upload some photos/more information about this filter board.
Logged

Hroðulf

  • Able Bodied Sailor
  • Pirate
  • ***
  • Brownie Points: +262/-1
  • Offline Offline
  • Posts: 538
  • MOT: Sonarworks

Do keep us posted!

In the meantime take a look at these http://www.diyinhk.com/shop/audio-kits/46-17uv-ultralow-noise-dac-power-supply-regulator-33v55v-1ax4.html

Might be a good fit for the digital side of things. I wouldn't bother with the bipolar version since it uses two positive regs which isn't too kosher.
Logged
If it keeps on ringin', levee's goin' to break..

Chris1967

  • Able Bodied Sailor
  • Powder Monkey
  • ***
  • Brownie Points: +28/-0
  • Offline Offline
  • Posts: 89

Very nice build Chris! By itself, the XMOS/AK stack with good value (as you have) is a very good performer for the price. You need to hear this with a JG Filter Buffer - it elevates this kit such that I prefer it to my more complex Sabre 9018 builds (PH regs, isolation/reclocking, NTD1 IV, etc.).

JG boards are hard to come by though. I mentioned awhile back that Joro/Lyuben (the designers of this kit) have been working on their version of a filter board. It's very flexible; allowing you to build it like a JG (2nd order harmonic filter), or apply filtering to other parts of the spectrum. I have a prototype board, parts on the way, and hope to build it this weekend.

When I have permission from Lyuben, I'll upload some photos/more information about this filter board.

This is good news MrRogers!! after a few hours of listening i think i will go for your advice and install a JG buffer...

I am waiting for you to post about the one Lyuben and Joro are planning.

Any other information on similar boards and a place to order them would also be helpful.

Thank you!!
Logged
Pages: 1 2 3 [4] 5 6 7 8 9 ... 11