Here's a couple of experiments with a Scarlett 2i2, just seeing what you guys think or if i made any mistakes.
I recorded the Kubelik sine sweep, powered by a Didden superreg and i2s from the JLsounds board.
Using REW, sweep up to 40khz just to see the steepness of the anti imaging filter in action. Going into the line-in of the interface (10k input impedance probably?) Volume was about 90%, just to give it a bit of headroom. For each image i showed, i re-ran the test a couple of times, just to see how much it changes and what is within measurement error, so here is what i got:
First here is 16 bit 44khz
We see a rising FR (actually, a bit too much, i know there is NOS droop compensation but this is unusual how low it goes and how its overcompensated, so i am courious what you think about it. maybe something else is at play.) We see some actually high distortion spikes in higher order harmonics, THD being around 9% at 8.6khz
Second interesting thing is increasing bit depth from 16 to above, marginally but repeatably decreases the THD in the 200-3k range (purple and green = 24 and 32bit, orange is 16bit). That is, 24 and 32 are the same. Is it correct to interpret this as, that it has greater than 16bit range (and so benefits from a larger word length) in this frequency span?
Third thing is increasing sample rate seems to remove the higher order harmonics in high frequencies, but up to a point where it then starts raising the whole floor.
Yellow -44.1k
Red - 48k
Pink - 96k
Green - 192k
Azure - 352k
96khz is probably the optimal point, 192k is ok, and you can see 352k for yourself 🙂
This is not that unexpected, just doing oversampling on the computer instead of a little chip on the board. Not unexpected that it cant handle 352k. I suppose it is hard to say now that it is still NOS when playing 96k?
Maybe this type of measurement is more suited for speakers, but impulse response at 16bit-44.1k
And then same settings, but 96khz:
And for reference, using the same exact settings, but the loopback of the interface outputs into itself.
Even if the measurements are not "pefectly accurate" in absolute terms (it is no APx555, and even then..it is only a man made machine), at least they can be compared in relative terms which this reference recording shows (that its not just a glitch in the methodology, or the ADC, or whatnot), right? That being said i am only starting out at measurements so any thoughts and feedback is appreciated.
Maybe some of the things here are old news for some people, but its interesting to me 🙂
I recorded the Kubelik sine sweep, powered by a Didden superreg and i2s from the JLsounds board.
Using REW, sweep up to 40khz just to see the steepness of the anti imaging filter in action. Going into the line-in of the interface (10k input impedance probably?) Volume was about 90%, just to give it a bit of headroom. For each image i showed, i re-ran the test a couple of times, just to see how much it changes and what is within measurement error, so here is what i got:
First here is 16 bit 44khz
We see a rising FR (actually, a bit too much, i know there is NOS droop compensation but this is unusual how low it goes and how its overcompensated, so i am courious what you think about it. maybe something else is at play.) We see some actually high distortion spikes in higher order harmonics, THD being around 9% at 8.6khz
Second interesting thing is increasing bit depth from 16 to above, marginally but repeatably decreases the THD in the 200-3k range (purple and green = 24 and 32bit, orange is 16bit). That is, 24 and 32 are the same. Is it correct to interpret this as, that it has greater than 16bit range (and so benefits from a larger word length) in this frequency span?
Third thing is increasing sample rate seems to remove the higher order harmonics in high frequencies, but up to a point where it then starts raising the whole floor.
Yellow -44.1k
Red - 48k
Pink - 96k
Green - 192k
Azure - 352k
96khz is probably the optimal point, 192k is ok, and you can see 352k for yourself 🙂
This is not that unexpected, just doing oversampling on the computer instead of a little chip on the board. Not unexpected that it cant handle 352k. I suppose it is hard to say now that it is still NOS when playing 96k?
Maybe this type of measurement is more suited for speakers, but impulse response at 16bit-44.1k
And then same settings, but 96khz:
And for reference, using the same exact settings, but the loopback of the interface outputs into itself.
Even if the measurements are not "pefectly accurate" in absolute terms (it is no APx555, and even then..it is only a man made machine), at least they can be compared in relative terms which this reference recording shows (that its not just a glitch in the methodology, or the ADC, or whatnot), right? That being said i am only starting out at measurements so any thoughts and feedback is appreciated.
Maybe some of the things here are old news for some people, but its interesting to me 🙂
Something very strange going on with the frequency response, its showing about 30dB boost at the top end. Which would sound hugely bright. I'm guessing the THD calculation is being confused by the image frequencies.
The DAC chip can run at 384kHz sample rate but it would need more current to the 74HC86, I've only allowed enough to handle up to 96k comfortably to keep the thermals under control.
The DAC chip can run at 384kHz sample rate but it would need more current to the 74HC86, I've only allowed enough to handle up to 96k comfortably to keep the thermals under control.
Hi Zbunjen
In REW when you use one device for digital out (your USB to I2S) and another device for analog in (your soundcard analog in), how do you get your reference input that is required for to do a frequency sweep?
The very low dbFS levels at your first three screenshots should alarm you of the measuring problem.
I haven’t found a way to do a frequency sweep with such a configuration. I can only do a step level and step frequency test where external reference input is not required.
George
In REW when you use one device for digital out (your USB to I2S) and another device for analog in (your soundcard analog in), how do you get your reference input that is required for to do a frequency sweep?
The very low dbFS levels at your first three screenshots should alarm you of the measuring problem.
I haven’t found a way to do a frequency sweep with such a configuration. I can only do a step level and step frequency test where external reference input is not required.
George
Attachments
Hey y’all - I’m doing a balanced build and am curious if anyone has thoughts on an elegant way of connecting i2s to two boards.
Was thinking of doing something like what the puredsd folks do with the dsc2.5 when stacking
http://puredsd.ru/media/252_4.jpg
Basically run long pins from the top board to the bottom board and wire i2s from my optical receiver to the top board.
Is that sane as long as I keep the total distance to the bottom board to under 10cm?
Was thinking of doing something like what the puredsd folks do with the dsc2.5 when stacking
http://puredsd.ru/media/252_4.jpg
Basically run long pins from the top board to the bottom board and wire i2s from my optical receiver to the top board.
Is that sane as long as I keep the total distance to the bottom board to under 10cm?
It could be something like that, because in reality i definitely do not find it to be boosted like that. Further, it dissapears on higher sample rates suggesting it could be something to do with aliasing images..
The tda1541 board i made shows the same thing on 44khz.
I guess i guess i dont have a reference IN. Actually REW suggests how to do it already, it just wants the reference input on the other channel (L) im not measuring with. But i dont have an appropriate cable on hand, so that's something to think about..
This is actually a good point i wanted to mention, kubelik is capable of 2V rms yet this is at 90% volume on the I2StoUSB and ~80% gain on the interface (above that noise floor starts to rise uncomfortably). Yet as you mention, only about -55dbfs.
The TDA1541 board mentioned (somehow 5th and 7th harmonic higher than 2nd) measured with the same setup shows up around the same -50dbfs level and it is just a 100R I/V, no filtering, no droop compensation or anything (when running it at higher sample rate, droop becomes negligible, and images pushed even further so i ignore it). Since its 100R, it can only put out about 0.137V p-p. But also notice, nice a flat FR, with same setup as last post.
Also nice measurements in your post, you made those? The distortion profile having a rising 3rd above 1khz seems to agree.
Sounds fine to me - the 10cm recommendation can be relaxed I reckon when only running at 44.1kHz. I routinely use longer than 10cm together with a Y splitter and haven't had any issues so far. The inter-board spacing is about 2cm when you stack them.
Really interested, although I am not now in a phase to solder all these things. So interested for pre soldered. So here is the story.
I own many DACs and the ones that ASR (no science and no audio review) measures best (yes and the topping D90SE mentioned and the others more expensive). So when received I will able to perform listening tests with the other well known brands.
My idea is to connect RPi4 I2S directly to this, so i will have the shortest path (I don't know if this means anything to digital worlds, i am an analogue guy.)
I also need to leave the analogue part unsoldered so I will use only top parts and silver solder. I usually put WBT RCAs or Furutech so I will use them
My aim is to try to beat the other DACs.
One question. Using the direct I2S, will I be able to play DSD512 and PCM384? Or what are the limits? (I am not dying to have them, I just want to know, since I am a little disappointed with DSD, even 512. Well that's it. How much to Greece?
I own many DACs and the ones that ASR (no science and no audio review) measures best (yes and the topping D90SE mentioned and the others more expensive). So when received I will able to perform listening tests with the other well known brands.
My idea is to connect RPi4 I2S directly to this, so i will have the shortest path (I don't know if this means anything to digital worlds, i am an analogue guy.)
I also need to leave the analogue part unsoldered so I will use only top parts and silver solder. I usually put WBT RCAs or Furutech so I will use them
My aim is to try to beat the other DACs.
One question. Using the direct I2S, will I be able to play DSD512 and PCM384? Or what are the limits? (I am not dying to have them, I just want to know, since I am a little disappointed with DSD, even 512. Well that's it. How much to Greece?
Alex - if you want pre-soldered then Kubelik won't be it. If you want a DAC to beat other DACs then Celibidache would be the one, you can buy that pre-soldered : https://www.diyaudio.com/community/threads/celibidache-nos-dac.379964/
As regards the I2S sample rate, the DAC chip can handle it however the 74HC86 driving the DACs will need more current to deliver such a high frequency to the DACs. So will need its supply current resistor reducing (and increasing in size because of the extra power dissipation so best mounted off-board). DSD is another format entirely and cannot be played. Shipping to Greece depends on speed of transit - slower is around $14 and faster, you can double that figure.
As regards the I2S sample rate, the DAC chip can handle it however the 74HC86 driving the DACs will need more current to deliver such a high frequency to the DACs. So will need its supply current resistor reducing (and increasing in size because of the extra power dissipation so best mounted off-board). DSD is another format entirely and cannot be played. Shipping to Greece depends on speed of transit - slower is around $14 and faster, you can double that figure.
Hi alexopth1512,
It shall be very interesting to read your findings regarding one of the Abraxalito DAC vs D90SE!!!
I wish one of these DACs can be able to beat say a D70S, a middle class DAC I found quite capable (very little limitations in absolute terms regardless budget, perhaps a tad more impact and depth in the lower bass and better nailed instruments in a wider soundstage). D70S is possibly a tad under D90SE...
As already mentioned though, you should look into Celibidache or, if "on budget just to give it a try", the latest edition of the Deca DAC costing less than half as much and that is believed to be very close to Celibidache. Perhaps an easy untweaked version first, to get a taste of it and a baseline, then of course you could play...
Kubelik costs only 20$: obviously there is a limit to everything... if it beats a D50S with a proper PS it would be already a wonder, but who knows...
Interesting, keep us posted!
Claude
It shall be very interesting to read your findings regarding one of the Abraxalito DAC vs D90SE!!!
I wish one of these DACs can be able to beat say a D70S, a middle class DAC I found quite capable (very little limitations in absolute terms regardless budget, perhaps a tad more impact and depth in the lower bass and better nailed instruments in a wider soundstage). D70S is possibly a tad under D90SE...
As already mentioned though, you should look into Celibidache or, if "on budget just to give it a try", the latest edition of the Deca DAC costing less than half as much and that is believed to be very close to Celibidache. Perhaps an easy untweaked version first, to get a taste of it and a baseline, then of course you could play...
Kubelik costs only 20$: obviously there is a limit to everything... if it beats a D50S with a proper PS it would be already a wonder, but who knows...
Interesting, keep us posted!
Claude
Just a tiny point here - Deca DAC is slightly more than half the price of Celibidache. Unless you want the 3rd order filter - but then it won't be so close to Celibidache.
Yep, sorry, was talking too broadly... thanks for correcting.
So slightly more instead of slightly less and indeed with the more complex advanced filter.
Perhaps not as absolute as Celibidache, but a good portion and flavour of it at a budget price to get quite an idea in which ballpark these DACs play?
Enjoy music and please share your findings
Claude
So slightly more instead of slightly less and indeed with the more complex advanced filter.
Perhaps not as absolute as Celibidache, but a good portion and flavour of it at a budget price to get quite an idea in which ballpark these DACs play?
Enjoy music and please share your findings
Claude
How can I order one kit? Since the new version of Diyaudio I have no idea how to PM. Does "conversation" mean actually email?? - Sorry for my ignorance.
Hi, i will proceed with Celibi, but it will take some time before the test, because first my D90SE just broke down and I am sending to Audiophonics with RMA and second I have 3 other projects waiting but I expect to have it ready in a couple of months.
So Abraxalito, PM me price in euros but only if you can include VAT and PM me, I want to talk the details. I ll try to use wise, although I haven't yet used it before
And one other thing. I don't like very much D90SE.....
What else than D90SE do you have?
I haven't listen to it. What is it you don't like and did you have a D70S perhaps?
Sounds like a good plan and loioking forward to reading you, whenever you are ready of course (no problem)
Claude
I haven't listen to it. What is it you don't like and did you have a D70S perhaps?
Sounds like a good plan and loioking forward to reading you, whenever you are ready of course (no problem)
Claude
No I haven't D70S. From other DACs, my old one is McIntosh, I also have SMSL from d300, d500 to R2R new one, musician R2R and some others.
What I don't like to D90SE is that seems not "musical" at all. Really quiet DAC, highs are very "subtle", a lot of detail but lifeless... I usually come back and try it again, which I will do once more. If you wonder why I don't give it is because I never sell my gear, I still have the amp I bought and soldered as a kit 40 yeras ago.
I have just installed my new power amps 150W full Class A monoblocks with SITs and I wanted to try it again but unfortunately I broke it 🙁
From DACs I have the story was to try to "catch" my turntable sound, which I have not managed so far, although I haven't gone to really high prices.
It is true that in some aspects the sound was really really interesting and like my amps and preamps I like different things in each one....
Excellent, let us know your findings with this DAC when the time is right
Thanks for the detailled explanation
Claude ( I built my own DAC ages ago also in order to catch up with my turntable)
Thanks for the detailled explanation
Claude ( I built my own DAC ages ago also in order to catch up with my turntable)
I sold my turntable and all my vinyl when I emigrated from UK. I had the intuition that I could eventually design a DAC that wouldn't make me miss them.
