Using a digital amp as a DAC
Posted 28th October 2016 at 03:19 AM by abraxalito
Updated 4th November 2016 at 03:03 AM by abraxalito
Updated 4th November 2016 at 03:03 AM by abraxalito
I've been mighty impressed with the STA333BW board (just 62rmb on Taobao, see an earlier post for the link) driving headphones via a transformer (ferrite cored step down 7:1, I've tried various sizes). So much so that I don't much feel the need to upgrade my headphones beyond my Superluxes (HD668B). The sound I'm getting is immensely satisfying - rich (which I got from my TDA1387-based DACs, but not from my phone with its ES9018K2M, that doesn't qualify as rich in tonality) and dynamic beyond anything I've so far heard. I noticed in a recent thread post by TNT (here - https://www.diyaudio.com/forums/vendors-bazaar/283672-new-audio-op-amp-opa1622-6.html#post4868193) he says he doesn't much like listening on headphones due to the 'in-head' quality of the sound. Yeah I used to get that, transformers pretty much get rid of it giving an out-of-the-head-all-around-holographic kind of experience, along with stupendous dynamics. I like listening to classical piano and that's very demanding on dynamics.
On connecting the headphone output directly to my amp though the sound wasn't at all like I was getting on my headphones - it was harsh, gritty. There's a good reason why it wasn't decent - the high level of ultrasonic output from the standard filter (a 2nd order LC) because after all this DAC is in effect a classD amplifier from the point of view of how the output's generated. Audio Precision has gone into the foibles of connecting a classD amp to its own measurement kit - in short it successfully knobbles anything containing opamps. Perhaps even discrete transistors, but the jury's out on that until I've done further testing with amps which don't contain opamps (or chipamps).
Here's a link to their technical justification of the AUX-0025 passive line filter : https://www.ap.com/analyzers-accesso...ement-filters/
I decided to build my own variant of this filter, I took as a basis the tables for a Chebyshev filter in Arthur Williams' book on filter design (here : https://www.amazon.com/Electronic-Filter-Design-Handbook-Williams/dp/0070704341/ref=sr_1_1?s=books&ie=UTF8&qid=1477624063&sr=1-1).
This book is highly recommended for anyone interested in filter design, I've been relying on it for decades. The used ones showing at $3 look to be a total steal.
I went for a 6th order though it might turn out that 4th order is sufficient, I'm quite paranoid about out of band signals getting into my amplification stages. I've attached the schematic - the first two components (L and C) replace the output filter that's already present on the board (22uH and 470nF). The output response plot shows we're 100dB down by 100kHz, the switching component being above this the layout and component parasitics are going to determine the rejection rather than anything Spice can show me. In modding the output filter you'll also want to tweak the various other components attached to the output (snubbers and EMC caps) as they're designed to work when the load's 8ohms. Here the load's considerably higher - this filter works into 1600ohms, to ground, equivalent to 3200ohms balanced. You'll need 4 of these filters as the output's balanced.
I'm going to get to work on a version of this filter using off-the-shelf inductors as I wound these rather odd values myself on 14mm gapped ferrite cores. Stay tuned for updates
Update - as promised here's the schematic for building the output filter using off the shelf inductors (MSS1210 from Coilcraft). I've also standardized the capacitor values at a small cost to the passband flatness. Best use NP0 types as they have the highest Q, X7R will work though if that's all you can get. If using X7R choose the highest voltage ones you can get, they'll have the lowest ESR.
Note that the working impedance has gone down to 750ohms, that's because finding inductors with enough Q isn't easy at values above 10mH. As a 7:1 output transformer gives a 49X decrease in the impedance, the output impedance of this as a DAC will be nicely low (sub 50ohms) - so much so that it'll also drive headphones unless you have planars. Planars can most certainly be driven but you'll need a larger output transformer with a reduced step-down ratio. Drop me a PM if you're a planar fan and want the details of an OPT to work with yours.
Update 2 - There are some high enough Q inductors from Toko which extend to higher inductances which have been brought to my attention. So I've tweaked the first filter to use these Tokos.
RS has a bargain price for these inductors at present, get them while you can - https://ie.rs-online.com/web/p/leaded...33383038333626
Mouser also has Fastrons which are similar and just as good - for example this is the 33mH one - https://www.mouser.com/ProductDetail/...o5Zi941i8Ro%3d
and here's the 27mH - https://www.mouser.com/ProductDetail/...wO6WAqKQ%3d%3d
On connecting the headphone output directly to my amp though the sound wasn't at all like I was getting on my headphones - it was harsh, gritty. There's a good reason why it wasn't decent - the high level of ultrasonic output from the standard filter (a 2nd order LC) because after all this DAC is in effect a classD amplifier from the point of view of how the output's generated. Audio Precision has gone into the foibles of connecting a classD amp to its own measurement kit - in short it successfully knobbles anything containing opamps. Perhaps even discrete transistors, but the jury's out on that until I've done further testing with amps which don't contain opamps (or chipamps).
Here's a link to their technical justification of the AUX-0025 passive line filter : https://www.ap.com/analyzers-accesso...ement-filters/
I decided to build my own variant of this filter, I took as a basis the tables for a Chebyshev filter in Arthur Williams' book on filter design (here : https://www.amazon.com/Electronic-Filter-Design-Handbook-Williams/dp/0070704341/ref=sr_1_1?s=books&ie=UTF8&qid=1477624063&sr=1-1).
This book is highly recommended for anyone interested in filter design, I've been relying on it for decades. The used ones showing at $3 look to be a total steal.
I went for a 6th order though it might turn out that 4th order is sufficient, I'm quite paranoid about out of band signals getting into my amplification stages. I've attached the schematic - the first two components (L and C) replace the output filter that's already present on the board (22uH and 470nF). The output response plot shows we're 100dB down by 100kHz, the switching component being above this the layout and component parasitics are going to determine the rejection rather than anything Spice can show me. In modding the output filter you'll also want to tweak the various other components attached to the output (snubbers and EMC caps) as they're designed to work when the load's 8ohms. Here the load's considerably higher - this filter works into 1600ohms, to ground, equivalent to 3200ohms balanced. You'll need 4 of these filters as the output's balanced.
I'm going to get to work on a version of this filter using off-the-shelf inductors as I wound these rather odd values myself on 14mm gapped ferrite cores. Stay tuned for updates
Update - as promised here's the schematic for building the output filter using off the shelf inductors (MSS1210 from Coilcraft). I've also standardized the capacitor values at a small cost to the passband flatness. Best use NP0 types as they have the highest Q, X7R will work though if that's all you can get. If using X7R choose the highest voltage ones you can get, they'll have the lowest ESR.
Note that the working impedance has gone down to 750ohms, that's because finding inductors with enough Q isn't easy at values above 10mH. As a 7:1 output transformer gives a 49X decrease in the impedance, the output impedance of this as a DAC will be nicely low (sub 50ohms) - so much so that it'll also drive headphones unless you have planars. Planars can most certainly be driven but you'll need a larger output transformer with a reduced step-down ratio. Drop me a PM if you're a planar fan and want the details of an OPT to work with yours.
Update 2 - There are some high enough Q inductors from Toko which extend to higher inductances which have been brought to my attention. So I've tweaked the first filter to use these Tokos.
RS has a bargain price for these inductors at present, get them while you can - https://ie.rs-online.com/web/p/leaded...33383038333626
Mouser also has Fastrons which are similar and just as good - for example this is the 33mH one - https://www.mouser.com/ProductDetail/...o5Zi941i8Ro%3d
and here's the 27mH - https://www.mouser.com/ProductDetail/...wO6WAqKQ%3d%3d
Save
Total Comments 13
Comments
-
"transformers pretty much get rid of it giving an out-of-the-head-all-around-holographic kind of experience"...
How are transformers used here?
//Posted 28th October 2016 at 02:23 PM by TNT
-
To step down the output voltage (running at 20V its 14VRMS peak, balanced) to something compatible with a 60ohm impedance (2VRMS max, unbalanced).Posted 28th October 2016 at 10:35 PM by abraxalito
-
is it mean when using doesn't need any I/V stage ?Posted 31st October 2016 at 09:06 AM by oxoxbyx
-
No I/V stage for this DAC as its voltage out (though no opamps). I rather suspect the lack of opamps is one of its key sonic advantages...Posted 31st October 2016 at 09:13 AM by abraxalito
-
the datasheet doesn't provide complete details about this dac, this is delta sigma ? or multi bit like TDA1543 etc ? so for this dac implementation we should something like step down tranformer instead of direct out to speaker ?Posted 31st October 2016 at 10:23 AM by oxoxbyx
-
Internally its delta-sigma yes but sounds at least as good in timbre as an R2R. Yes a step down transformer is needed at the very least because its balanced out.Posted 31st October 2016 at 10:47 AM by abraxalito
-
is STA350 have same dac chip inside and another STA chip ?Posted 1st November 2016 at 02:49 AM by oxoxbyx
-
No DAC chip inside any of these STA parts, they're PWM chips which switch the power supply as their Vref.Posted 1st November 2016 at 09:41 AM by abraxalito
-
hows sound when directout without transformer coupled to speaker ?Posted 6th November 2016 at 10:28 AM by oxoxbyx
-
I didn't like it much at all but then I didn't give it a really low impedance supply. I have the D2 amp playing into speakers now and its getting there but I have 2:1 trafos and 40,000uF supply capacitance. I need to go looking for more......Posted 6th November 2016 at 12:45 PM by abraxalito
-
Do you think that a DIBox will be a good idea to interface these Digital Amplifiers DACs with headphones ? I have two Behringer DI100 and it seems they can support 3000W at their inputs, attenuate the output signal, still leaving it in balanced mode...
Of course, I know that there is a trafo (the OT-1) in this device… And I remember what I wrote elsewhere in the thread, but anyway do you think this idea deserves a try ?Posted 29th November 2016 at 04:34 PM by ondesx
-
In my understanding DI boxes are for signal level isolation so might not perform well at all into low impedance headphones. If they contain transformers they'll be very small sized ones. But there can't be any harm in trying the experiment, transformers are hard to break.Posted 30th November 2016 at 02:22 AM by abraxalito
-
Thank you for your input, I'll give it a try asap…Posted 30th November 2016 at 07:38 AM by ondesx
