@ Thorsten
BTW.
Look at some measurements from stereophile here.
This is Linn DS Klimax and Bryston DAC
Stereophile: Linn Klimax DS network D/A processor
Stereophile: Bryston BDA-1 D/A converter
Have a look at the waveforms.
The Bryston features a CS4398 or two, The Linn a Wolfson Micro.
To me the waveforms look exactly how they sound at FS also
BTW.
Look at some measurements from stereophile here.
This is Linn DS Klimax and Bryston DAC
Stereophile: Linn Klimax DS network D/A processor
Stereophile: Bryston BDA-1 D/A converter
Have a look at the waveforms.
The Bryston features a CS4398 or two, The Linn a Wolfson Micro.
To me the waveforms look exactly how they sound at FS also
But that is a completely different project.
This projects scope is to design a single board DAC, with every component available from one single vendor except the PCB. And it is still to be better or equal to ie. Benchmark DAC 1, and it has to be @ 200$ or less.
This claim is nonsense. That's because those things you cite are categories, and categories cannot be experienced. What you experience are instantiations of those categories - a particular opamp, a 2SK170 etc. So its possible to discard indiividual instances of FETs, opamps, tubes and the like based on experience, but throwing out all members of the categories you cite is indeed prejudice.
If you have a hypothesis as to what all the members of a category share and cite that as a reason for rejecting all instantiations then we might make progress with dispelling your prejudices😛 You might begin with FETs...
DAC design (open source) projects - scope and forum setup
We need a project scope document, which is a start- instead of reading many pages to try figure it out. Hurtig put some good energy into this at the beginning.
I have some issues with forum limitations and group projects I posted here:
http://www.diyaudio.com/forums/forum-problems/162368-open-source-projects-need-done-differently.html
It looks like future changes to the diyAudio site could make group projects easier. For now, the thread needs to renamed (open source single-board DAC proj.?) and broken up into the power supply, ASRC, DAC, audio output etc. Maybe we can put together a wiki when the design settles down?
I am in favour of starting a modular DAC design (open source) project, with threads split up for each module. I have limited time to figure out the best way to proceed, does anyone want to help?
We need a project scope document, which is a start- instead of reading many pages to try figure it out. Hurtig put some good energy into this at the beginning.
I have some issues with forum limitations and group projects I posted here:
http://www.diyaudio.com/forums/forum-problems/162368-open-source-projects-need-done-differently.html
It looks like future changes to the diyAudio site could make group projects easier. For now, the thread needs to renamed (open source single-board DAC proj.?) and broken up into the power supply, ASRC, DAC, audio output etc. Maybe we can put together a wiki when the design settles down?
I am in favour of starting a modular DAC design (open source) project, with threads split up for each module. I have limited time to figure out the best way to proceed, does anyone want to help?
The time used for any DAC project is always unreasonable, if it is not just to build on opinions.
One of the coming days we ae going to test the ESS DAC out, to see if it is suitable for the project. We´ve build a discrete analog stage for it.
But if it wn´t do as supposed to, we are back to the beginning, that´s pretty much how things are, when building electronics.
One of the coming days we ae going to test the ESS DAC out, to see if it is suitable for the project. We´ve build a discrete analog stage for it.
But if it wn´t do as supposed to, we are back to the beginning, that´s pretty much how things are, when building electronics.
Hi,
Yes, quite interesting.
Some remarks about the respective measurements.
First, the Linn uses a Wolfson Micro DAC, but uses the common "brickwall" filter. Linn's output stages can be seen to be comparably non-linear using HD testing, compared to Bryston's discrete op-amp's.
On the other hand, when looking at stuff that actually matters for sonics, the Bryston DAC (from a company I shall not mention and of a model I shall not mention to avoid being acused of "bashing" them) has much worse HF IMD, which, as it has a more linear analogue stage likely comes directly from the DAC and it gets BY FAR worse if the ASRC is engaged.
So, based on the HF IMD results I suspect the Bryston DAC will sound considerably more edgy and annoying than the Linn unit.
On the jitter front Linn's product also shows a much narrower central peak, suggesting it's representation of the sinewave is much less "smeared" by noise. And the jitter noisefloor levels of the Linn Unit are around 15dB better than the Bryston, again something I'd be tempted to place at the doors of the DAC in the Bryston unit.
So, based on the measurements presented in Stereophile I would probably choose the DAC deployed in the Linn DS, not the one in the Bryston DAC, the improvements in measured performance attributable to the DAC Chip seem overwhelming. That is before you account for the fact that the Chip in the Linn DS has several selectable filter responses for the Digital Filter, including the new audiophile Darling (according to HP, RH, JA and others) Appetising filter, or whatever they call it - sounds more tasty anyway... 😀
Ciao T
Yes, quite interesting.
Some remarks about the respective measurements.
First, the Linn uses a Wolfson Micro DAC, but uses the common "brickwall" filter. Linn's output stages can be seen to be comparably non-linear using HD testing, compared to Bryston's discrete op-amp's.
On the other hand, when looking at stuff that actually matters for sonics, the Bryston DAC (from a company I shall not mention and of a model I shall not mention to avoid being acused of "bashing" them) has much worse HF IMD, which, as it has a more linear analogue stage likely comes directly from the DAC and it gets BY FAR worse if the ASRC is engaged.
So, based on the HF IMD results I suspect the Bryston DAC will sound considerably more edgy and annoying than the Linn unit.
On the jitter front Linn's product also shows a much narrower central peak, suggesting it's representation of the sinewave is much less "smeared" by noise. And the jitter noisefloor levels of the Linn Unit are around 15dB better than the Bryston, again something I'd be tempted to place at the doors of the DAC in the Bryston unit.
So, based on the measurements presented in Stereophile I would probably choose the DAC deployed in the Linn DS, not the one in the Bryston DAC, the improvements in measured performance attributable to the DAC Chip seem overwhelming. That is before you account for the fact that the Chip in the Linn DS has several selectable filter responses for the Digital Filter, including the new audiophile Darling (according to HP, RH, JA and others) Appetising filter, or whatever they call it - sounds more tasty anyway... 😀
Ciao T
Hi Kurt,
One more item to test.
As said, the WM8805 can reject jitter much better than the more common chips, so trying it is very well worth doing.
As you still want an ASRC, based on a discussion elsewhere here someone recommend the SRC4194, as it has some interesting features build in that can be accessed in hardware mode, including the "direct downsampling" and the ability to select the length of the aliasing filter if used.
I think these features would be worth having, regardless what DAC is used after the SRC, compared to using the SRC4192 or the AD counterpart. Direct downsampling is available in the SRC4193 as well, but it needs a MCU to control. The latency control (length of filter) does not exist in the 4193.
As a final suggestion, using the kind of powersupplies ESS shows for the analog pin's of their DAC's on any modern DAC analog supplies as well as for the reference voltage (if accessible) is a significant improvement for sound, using discrete low noise shunts with current sources is even better, but harder to implement than just adding a bunch of 5532's with one per supply pin...
With a 5532 and suitable low turnover frequency RC filters a powersupply with 0.5uV RMS noise in the audio bandwidth is possible.
With discrete shunts (well, sort of) I can get 0.1uV BTW.
By comparison a LM317 has around 250uV noise.
Ciao T
One more item to test.
As said, the WM8805 can reject jitter much better than the more common chips, so trying it is very well worth doing.
As you still want an ASRC, based on a discussion elsewhere here someone recommend the SRC4194, as it has some interesting features build in that can be accessed in hardware mode, including the "direct downsampling" and the ability to select the length of the aliasing filter if used.
I think these features would be worth having, regardless what DAC is used after the SRC, compared to using the SRC4192 or the AD counterpart. Direct downsampling is available in the SRC4193 as well, but it needs a MCU to control. The latency control (length of filter) does not exist in the 4193.
As a final suggestion, using the kind of powersupplies ESS shows for the analog pin's of their DAC's on any modern DAC analog supplies as well as for the reference voltage (if accessible) is a significant improvement for sound, using discrete low noise shunts with current sources is even better, but harder to implement than just adding a bunch of 5532's with one per supply pin...
With a 5532 and suitable low turnover frequency RC filters a powersupply with 0.5uV RMS noise in the audio bandwidth is possible.
With discrete shunts (well, sort of) I can get 0.1uV BTW.
By comparison a LM317 has around 250uV noise.
Ciao T
the whipping boy for K & H's Project...
Hi K & H,
Just to illustrate my point about the original project...
If anyone would like an upsampling DAC using CS8416 and CS4398, using 3-pin regulators and op-amps and fully assembled PCB with good quality parts, pretty good mains transformer included as well, the total before shipping is 120 USD.
I agree that this project now has moved away from this kind of thing and it is a good thing too, as paying 200 USD to assemble something myself that I can purchase for 120 USD already done would strike me as a waste of time.
BTW, I came across this DAC today as I followed up on a project that includes this DAC and a Tube Stage for DAC's I designed. But I thought it rather illustrated my contentions very well.
I'd probably suggest using WM8805, SRC4194, TE7022 USB Chip and whatever DAC with nice supplies for all the power-supply pins (could be as simple to implement as a bunch of 5532) and discrete analog stage with suitable shunt regulators as the outline spec that would make it worthwhile to build this manually, instead of just buying the Gigaworx DAC of E-Bay.
Ciao T
Hi K & H,
Just to illustrate my point about the original project...
If anyone would like an upsampling DAC using CS8416 and CS4398, using 3-pin regulators and op-amps and fully assembled PCB with good quality parts, pretty good mains transformer included as well, the total before shipping is 120 USD.
An externally hosted image should be here but it was not working when we last tested it.
I agree that this project now has moved away from this kind of thing and it is a good thing too, as paying 200 USD to assemble something myself that I can purchase for 120 USD already done would strike me as a waste of time.
BTW, I came across this DAC today as I followed up on a project that includes this DAC and a Tube Stage for DAC's I designed. But I thought it rather illustrated my contentions very well.
I'd probably suggest using WM8805, SRC4194, TE7022 USB Chip and whatever DAC with nice supplies for all the power-supply pins (could be as simple to implement as a bunch of 5532) and discrete analog stage with suitable shunt regulators as the outline spec that would make it worthwhile to build this manually, instead of just buying the Gigaworx DAC of E-Bay.
Ciao T
I also find it interesting
The DS is using op-amps in conjunction with output transformers - as the matter of fact "high end Lundahl output transformers" - as if high end transformers do exist at all.
You cannot reallycompare the jitter levels, as the Linn is a computer with internal DAC, which probably is connected to the computer output via I2S interface. The Bryston is an universal DAC with SPDIF reciever, which always will result in somewhat higher jitter.
But besides that, every other measurement done by Stereophile shows the Bryston pretty much superior to the Linn. Both waveform and spectrum at low and high levels are much better on the Bryston as well as distortion, and I do not doubt that linearity also is superior.
And then none of the jitter levels is above -125 dB, so what are we talking about?
And the I also do know the sound of the Crystal chip very well, and it is not edgy at all, in the contrary I think the Linn, like all other Linn products are very very polite. But that might be subject to personal preferences and the quality of the rest of audio chain.
Now I´ve heard various inventions from Linn incl. very expensive ones, and what can one say? If music should sound even more relaxed and even more PRAT´ing, the musicians have to be dead when recording, or your hearing aid muust be turned off.
My house is pretty full of musical instruments, and my record collection is pretty full of AB stereo recordings. To get live music to meet the canned thing on records, no "politeness" is needed, just plain neutrality.
Why not SRC4392 with the ability of sending back clock signal to the source , then just disable ASRC block, use it as a plain RX / TX . Not sure it'd work but worth a look.
Hi,
Actually, they do exist and they have a by far lower amount of distortion than any Op-Amp or discrete circuit and they can be made essentially flat in the 10Hz-100KHz bandwidth.
So, what is your concern then, if not linearity or frequency response? What makes transformers that are essentially transparent to bot measurements and listening not suitable for high end (I mean other than prejudice, real technical reasons please).
Well, I am actually referring to the noisefloor in these measurements:
Linn:
Bryston:
At issue is not a particular jitter component, but the level of the noisefloor (though there is also much more jitter anyway).
Another area of concern is the HF IMD:
Linn:
Bryston:
BTW, the jitter measurements and the IMD Measurements where done with ASRC on, but the ASRC failed to lower jitter levels.
Another issue is visible in these graphs:
Linn:
Bryston:
This effectively shows among other factors the "instanious distortion" of the DAC. It appears that the Digital Filter/noise-shaper in the DAC used by Bryston causes much greater distortion, ooops, sorry, out of band noise that what Linn used, even though I personally think either graph is not great.
In the end it is a question how measurements relate to what we hear of course. I personally feel that the measurements I singled out tell us a lot more about what things sound like, than the one you emphasise.
Ciao T
Actually, they do exist and they have a by far lower amount of distortion than any Op-Amp or discrete circuit and they can be made essentially flat in the 10Hz-100KHz bandwidth.
So, what is your concern then, if not linearity or frequency response? What makes transformers that are essentially transparent to bot measurements and listening not suitable for high end (I mean other than prejudice, real technical reasons please).
Well, I am actually referring to the noisefloor in these measurements:
Linn:
Bryston:
At issue is not a particular jitter component, but the level of the noisefloor (though there is also much more jitter anyway).
Another area of concern is the HF IMD:
Linn:
Bryston:
BTW, the jitter measurements and the IMD Measurements where done with ASRC on, but the ASRC failed to lower jitter levels.
Another issue is visible in these graphs:
Linn:
Bryston:
This effectively shows among other factors the "instanious distortion" of the DAC. It appears that the Digital Filter/noise-shaper in the DAC used by Bryston causes much greater distortion, ooops, sorry, out of band noise that what Linn used, even though I personally think either graph is not great.
In the end it is a question how measurements relate to what we hear of course. I personally feel that the measurements I singled out tell us a lot more about what things sound like, than the one you emphasise.
Ciao T
Hi,
I'm in great favour of "sending the clock back", but for most cases of DAC's this would mean modifying the transport and limiting the maximum available quality modes to the modified unit.
Non of this applies to PC based sources of course. I now use Asynchronous USB with up to 24/192 (in theory more, well past 32/384) sample rate capability. So all this jitter this and that is once and for all eliminated.
But of course such a DAC is compatible only with computer sources.
Ciao T
I'm in great favour of "sending the clock back", but for most cases of DAC's this would mean modifying the transport and limiting the maximum available quality modes to the modified unit.
Non of this applies to PC based sources of course. I now use Asynchronous USB with up to 24/192 (in theory more, well past 32/384) sample rate capability. So all this jitter this and that is once and for all eliminated.
But of course such a DAC is compatible only with computer sources.
Ciao T
USB cable length is limiting once someone has the PC in another room, with a hole in the wall. I dont see the need for 32bit 384, the advantage is not even apparent,
- SD DA cuts bitsd down and extends time resolution infinitely, also starts noiseshaping process long before the actual 192khz.
- Multibit DA wont do anytingh with 32bit , without HW OS you need software to feed it 354/384 signal.
- Multibit dithered DA, wont need nor the bits nor such frequency, the degree of intentionally applied noise at 16bit 176khz is effectively non existant. (-150 inband, -120 oob)
Most of those 32bits just toggles randomly anyway. A true 27bit accurate S-D ADC would need a builtin low bit DA block built out of superconductors. Good for high speed measurements.
- SD DA cuts bitsd down and extends time resolution infinitely, also starts noiseshaping process long before the actual 192khz.
- Multibit DA wont do anytingh with 32bit , without HW OS you need software to feed it 354/384 signal.
- Multibit dithered DA, wont need nor the bits nor such frequency, the degree of intentionally applied noise at 16bit 176khz is effectively non existant. (-150 inband, -120 oob)
Most of those 32bits just toggles randomly anyway. A true 27bit accurate S-D ADC would need a builtin low bit DA block built out of superconductors. Good for high speed measurements.
Last edited:
@Thorsten
Transformers are prone to hysteresis, and that is just about the worst of any thing in magnetics at all.
A transformer is equivalent to a few km. of cable, just a bit worse than a capacitor.
Both are breaking the signal path.
Transformers are unwanted, which everybody in the business knows. Which is also the reason for Transformerless versions of i.e. DPA´s state of the art issues of their best microphones.
Transformers are used for isolation purposes, nor for sound quality enhancement purposes.
DPA, Neumann, Shoeps and others know this, which is the reason for the TL versions.
In analog tape recorders hysteresis was a known issue. High frequency bias was the solution to that. In transformers there is no bias, thus there is no low level information either. This is a law of nature, but if one is not interested in low level information and thus dynamic range, transformers can be an easy way out of isolation problems and HF problems.
Transformers are prone to hysteresis, and that is just about the worst of any thing in magnetics at all.
A transformer is equivalent to a few km. of cable, just a bit worse than a capacitor.
Both are breaking the signal path.
Transformers are unwanted, which everybody in the business knows. Which is also the reason for Transformerless versions of i.e. DPA´s state of the art issues of their best microphones.
Transformers are used for isolation purposes, nor for sound quality enhancement purposes.
DPA, Neumann, Shoeps and others know this, which is the reason for the TL versions.
In analog tape recorders hysteresis was a known issue. High frequency bias was the solution to that. In transformers there is no bias, thus there is no low level information either. This is a law of nature, but if one is not interested in low level information and thus dynamic range, transformers can be an easy way out of isolation problems and HF problems.
Transformers are able to break ground loops, handle very low signal levels as in microphones or MC pick-ups and "amplify" by up to 20x or more.
Where do you see the problem using them in a DAC at much higher signal levels ?
There is a tendency in this forum to create nonexistant problems and present useless solutions.
A transformer is a very linear element, the same can not be said about what you prefer: Semiconductors 😛
To give you an idea about my own DAC:
A hundred meters of wire connected between Iout of R2R DACs ( fed by unmanipulated data from a slaved transport ) and a power amp, provides:
I/V stage,
60dB/oct analog filter,
sinx/x compensation,
common mode noise rejection,
balanced to single ended conversion,
volume control,
+6dB amplification,
very low impedance output stage,
DC isolation between DAC and power amp,
*** **** ** ***** **** *** *** **** ***** **.
No digital data manipulation filter, transistors or opamps or tubes, resistive voltage dividers, coupling caps.
Just a hundred meters of wire and a bunch of passive VHQ components to ground.
Last edited:
Yes, if you have no experience how to correlate measurements to sound, then this is very true
The point is, that even in analog audio, active components are prefered.
I´ve had my transformers for MC, and they are really the worst thing I ever can imagine back into my audio chain, and the reason for that is excactly the same why MC cartridges are prefered over MM´s.
Huge amounts of induction and hysteresis.
A transformer is not linear at all, and everybody using them knows that, thus the best mics are transformerless, and the best step-ups are active.
Lately I´ve even heard a passive preamplifier based on transformers, and I must say, that I´ve never heard anything that resonant and high Q earlier, besides MC step - up transformers.
But this is simply a law of nature, I don´t realy know why this matter turns up in this thread.
To make it simple, no signal path transformer will ever find its way into anything originating from our heads, as it will never find its way into my system.
- Status
- Not open for further replies.