• These commercial threads are for private transactions. diyAudio.com provides these forums for the convenience of our members, but makes no warranty nor assumes any responsibility. We do not vet any members, use of this facility is at your own risk. Customers can post any issues in those threads as long as it is done in a civil manner. All diyAudio rules about conduct apply and will be enforced.

Reference DAC Module - Discrete R-2R Sign Magnitude 24 bit 384 KHz

I'm sure SOX is OK for what it does. But I don't agree with the analysis around it in the article.

I certenly hope that Sören can provide a basic textbook filter as I dont belive there is any hidden gems in special filters.

In Nyquist and the Sinc I trust.

:)

//
I know for a fact that one of the commercial r2r dacs on the market has a lot of development effort invested in its filters. I expect effort here to hugely influence the results with this DAC.
 
Here are some links from Digital HW designer John Swenson on digital filters in general and using SoX to generate the filters he used in the DAC he designed for Bottlehead:

Upsampling Impressions

Here's the rundown on the DAC

Taking orders :)

Taking orders :)

Taking orders :)

Taking orders :)

I both second SoX as an upsampling filter generator AND John's thoughts as a starting point for at least one direction for an optimized filter search.

Greg in Mississippi

Thx Greg !
Great reading :)
 
Here are some links from Digital HW designer John Swenson on digital filters in general and using SoX to generate the filters he used in the DAC he designed for Bottlehead:

Upsampling Impressions

Here's the rundown on the DAC

Taking orders :)

Taking orders :)

Taking orders :)

Taking orders :)

I both second SoX as an upsampling filter generator AND John's thoughts as a starting point for at least one direction for an optimized filter search.

Greg in Mississippi

Hi Greg:
Thanks for posting those links to John Swenson's posts. You beat me to it and saved me a bunch of time. Funny, I see that the 3rd "Taking Orders" link goes to my post on tuning the filter in my studio.

To Soren:
SoX is a great place to start and will be handy for creating filters for your FPGA. I have listened to and tweaked the parameters of a lot of DFs, and can assure you that such is the largest part of making a DAC sound good with Redbook material. Personally I am coming from an entirely NOS PCM1704K--and so I do all interpolation in software (to 352.8/384). I am a Mac guy, so the two best players for realtime SRC on the Mac are:

a) Audirvana+, which integrates a licensed version of the excellent iZotope SRC engine, including giving the user access and fine granularity control over Slope, Cut-off, Max. Filter Length, the mis-named Anti-aliasing (which is just final attenuation depth of the filter), and crucially Pre-Ring (which is balance between pure minimum phase all post-ring, and linear phase which of course is equal pre/post ring). Extensive tuning with the parameters--using especially instruments with sharp attack and piano--can yield impressive results. Thus, when it came to tuning for the PCM5142 used in the Bottlehead DAC Greg linked to, we were able to make what would have been an average DAC chip into a great sounding piece by turning off the built-in filter and doing SRC in the FPGA--using parameters tuned by ear with SoX.

b) HQ Player by Signalyst (Windows, OS X, Linux) contains about the most advanced filters and dither available. Its Poly-sinc-* family of filters and well-tune dither options blow away anything I have been able to create, and the s/w has even more advanced sigma-delta modulators for outputting high-rate one-bit streams to DSD DACs. It of course uses a LOT of computing power to perform some of its feats, and thus his filter engine is not something that can be licensed for FPGA use.
Spend some time over at ComputerAudiophile.com and you will find excellent discussions of both. The developer of HQ Player goes by Miska on the forums, but his real name is Jussi. He is very experienced and I think you may find him to quite helpful.

So back to SoX. It and iZotope are possibly the best two filter generators available. The standard filters offered by SoX are just okay, but when you get into its custom parameters then great filters can be built. Of course then you need to output a list of coefficients to load into your FPGA. My friend John Swenson wrote a s/w routine to do just that. I will ask him if he is open to sharing that with you for your project.

I'm very glad to see such enthusiasm for your DAC. I am sure it won't be long before I order one for myself!

Regards,
--Alex C.
 
Member
Joined 2003
Paid Member
Hi Greg:
Thanks for posting those links to John Swenson's posts. <SNIP>

Welcome Danny_66 & Alex.

And thank you Alex for some great info on filter tools and optimization.

I don't offer John's ideas as a filter to use, but more of a roadmap for experimentation. In one of the links, John talks about having tried NOS and liking some aspects of it, but not other aspects. My read of his thoughts on this is that he has worked to identify filter designs that provide the needed filtering while not producing the negative aspects of most of the 'on-chip' filters and preserving the 'aliveness' one hears from a NOS setup.

I haven't heard this yet in my system, I am close (I hope) to being able to replicate John's filtering setup, and I think his parameters will be a good starting point for this DAC. I doubt the filter that John found best for the PCM5142 will be the right one for Soren's DAC... but starting there and trying variations, both in filter parameters and oversampling rates, will likely produce at least one filter variation that will work well with this setup.

There's another topic John discussed some recently, also related to his development of the BH DAC... that is of a set of mechanisms where upstream issues make their way through an isolation and reclocking scheme such as what Soren used on this DAC (and similar to what John used on the BH DAC)

Some of those posts are here:

Mac Mini version of a CAPS music server - Step by Step - Page 9

Mac Mini version of a CAPS music server - Step by Step - Page 9

Mac Mini version of a CAPS music server - Step by Step - Page 9

Mac Mini version of a CAPS music server - Step by Step - Page 10

Mac Mini version of a CAPS music server - Step by Step - Page 10

Mac Mini version of a CAPS music server - Step by Step - Page 11

DDR3 Flipping bits in software without accessing them "Memory Known Failure Mechanism called 'Row Hammer'" - Page 5

The minimum DSD DAC - Page 9

DDR3 Flipping bits in software without accessing them "Memory Known Failure Mechanism called 'Row Hammer'" - Page 7

He discusses it in terms of how to make the DAC more immune to upstream issues (of course he would, he designs DACs!). But I've looked at it as how do I get the best out of the setups I have today... and I keep coming back to having the best power from both a low-noise and wide-band low output impedance as being a critical thing, along with minimization of noise fed back into the AC lines (hint, hint, battery-powered servers? See here: "New" LiFePO4 Powered Server).

I think Normundss's plots showing noise at the output of this DAC even when it is powered off, but the USB hub he is using is powered on, points to the importance of this. John's BH DAC also uses isolation/reclocking (and a few other tricks, which he discusses at length on the BH forum) to reduce and/or eliminate the impact of upstream sources on the sound of that DAC. And he admits that some differences still get through, although he has continued to improve it if you read between the lines in the posts above.

I've read of other examples of DACs with sophisticated isolation & reclocking plus other tricks that still exhibit sensitivity to upstream problems... such as the PS Audio Perfectwave and Directstream DACs sounding best for most people using a good I2S source and being sensitive to feeding USB cables.

These seem to be the best tools available today... and still don't make the downstream DACs immune to upstream issues. And Normundss' plot shows that very clearly.

Flame suit on!

Greg in Mississippi
 
Last edited:
Member
Joined 2010
Paid Member
Enrico,

I like your project a lot. I would make just one suggestion in regards to the digital signal transformer. It would be great to make multipurpose and surface mount pads for it, that would allow usage of different transformers. Best in industry for this purpose are Scientific Conversions transformers used in all leading broadcast brands.

Scientific Conversion, Inc. - Transformers and Inductors

I will use SC947, see specs published. If interested I could help this group in obtaining the group buy. Audio Note transformer seems to be easy to mount in any position as long as there are holes, so that will leave everyone a choice.

Truth to be told, not that it matters much since there is FIFO on the board, but still why not do it the best we can.

Hi AR2,

I am glad that you like it and thanks for the link.
I agreed with you and I like too to have more options in the same board. In the website I can't found the technical datasheet around.
I need the exact pads position and dimension but the only information that I found is "9.5 mm W x 5 mm H SMD 8 pin package". This doesn't help me too much. You know where I can get the datasheet?

Thanks,
Enrico
 
To Soren:
SoX is a great place to start and will be handy for creating filters for your FPGA. I have listened to and tweaked the parameters of a lot of DFs, and can assure you that such is the largest part of making a DAC sound good with Redbook material. Personally I am coming from an entirely NOS PCM1704K--and so I do all interpolation in software (to 352.8/384). I am a Mac guy, so the two best players for realtime SRC on the Mac are:

a) Audirvana+, which integrates a licensed version of the excellent iZotope SRC engine, including giving the user access and fine granularity control over Slope, Cut-off, Max. Filter Length, the mis-named Anti-aliasing (which is just final attenuation depth of the filter), and crucially Pre-Ring (which is balance between pure minimum phase all post-ring, and linear phase which of course is equal pre/post ring). Extensive tuning with the parameters--using especially instruments with sharp attack and piano--can yield impressive results. Thus, when it came to tuning for the PCM5142 used in the Bottlehead DAC Greg linked to, we were able to make what would have been an average DAC chip into a great sounding piece by turning off the built-in filter and doing SRC in the FPGA--using parameters tuned by ear with SoX.

b) HQ Player by Signalyst (Windows, OS X, Linux) contains about the most advanced filters and dither available. Its Poly-sinc-* family of filters and well-tune dither options blow away anything I have been able to create, and the s/w has even more advanced sigma-delta modulators for outputting high-rate one-bit streams to DSD DACs. It of course uses a LOT of computing power to perform some of its feats, and thus his filter engine is not something that can be licensed for FPGA use.
Spend some time over at ComputerAudiophile.com and you will find excellent discussions of both. The developer of HQ Player goes by Miska on the forums, but his real name is Jussi. He is very experienced and I think you may find him to quite helpful.

So back to SoX. It and iZotope are possibly the best two filter generators available. The standard filters offered by SoX are just okay, but when you get into its custom parameters then great filters can be built. Of course then you need to output a list of coefficients to load into your FPGA. My friend John Swenson wrote a s/w routine to do just that. I will ask him if he is open to sharing that with you for your project.

I'm very glad to see such enthusiasm for your DAC. I am sure it won't be long before I order one for myself!

Regards,
--Alex C.

Yes, SoX is first step as source code is available, it would be easy to add a routine to output the FIR coefficients to a file where my utility can convert them to a format suitable for downloading onto the dam1021. Would be easier if John Swenson would be nice and help a little, but I can do it myself. My DAC filters sounds like they look much like John Swenson's, I also oversample to 352K/384K in one step, although mine is limited to 1016 coefficients.

If people want to start then I now have the tools on our website:

http://www.soekris.dk/1021filt.txt -- ascii file with all filters
http://www.soekris.dk/1021filt.skr -- binary file for downloading onto a dam1021
http://www.soekris.dk/MKROM.EXE -- 32 bit dos utility to convert txt file to skr file, usage "mkrom 1021filt"

Please note that messing with IIR filters without knowing what you're doing can result in oscillations which will blow your speakers and/or ears.
 
Member
Joined 2003
Paid Member
<SNIP>
... and I keep coming back to having the best power from both a low-noise and wide-band low output impedance as being a critical thing, along with minimization of noise fed back into the AC lines (hint, hint, battery-powered servers? See here: "New" LiFePO4 Powered Server).
<SNIP>

I need to clarify that the above power supply comments refer to the power of the upstream devices, whether a CD/DVD player or music server of some sort, whether using direct I2S, USB, S/PDIF, or TOSLINK outputs. Also the associated connected equipment, such as Ethernet hubs.

Greg in Mississippi
 
Member
Joined 2010
Paid Member
I think it would be more flexible if the u.fl's were on the upper side of the board, and the other side of the 26-pin connector. if it's hard to get room for them, then move the 3,3V and FSEL connectors a bit lower down, and rotate them 90 degree, then place the MCKL u.fl above them.

this is at least what i think would be best - but we all think different :)

Well, I will add the U.FL in the top of the board as you suggest and I will keep anyway the ones in the bottom.

Like that (almost) everybody should be happy ;)
 
Hi AR2,

I am glad that you like it and thanks for the link.
I agreed with you and I like too to have more options in the same board. In the website I can't found the technical datasheet around.
I need the exact pads position and dimension but the only information that I found is "9.5 mm W x 5 mm H SMD 8 pin package". This doesn't help me too much. You know where I can get the datasheet?

Thanks,
Enrico

Those Audio Note and SC trafos are costly and those use it are minority.

IMO, design trafo pads based on those popular and proven quality trafos will provide greater versatility, for example:

1) Murata DA101C, Newava S22083 (TH Ver, similar pads)
2) Murata DA101MC (SMD Ver)

http://www.murata-ps.com/data/magnetics/kmp_da100.pdf
http://www.newava.com/pdf/S22083.pdf
 
Last edited:
Member
Joined 2010
Paid Member
Those Audio Note and SC trafos are costly and those use it are minority.

IMO, design trafo pads based on those popular and proven quality trafos will provide greater versatility, for example:

1) Murata DA101C, Newava S22083 (TH Ver, similar pads)
2) Murata DA101MC (SMD Ver)

http://www.murata-ps.com/data/magnetics/kmp_da100.pdf
http://media.digikey.com/pdf/Data Sheets/Pulse PDFs/PE-65x12_Rev2006.pdf

Thanks for the link... I will give a look and try to accomodate
 
What is the "strategy" if the commercial has already an output traffo with its spidf ? We keep just the 330 ohms shunt and the 1.2 V PS layout ? (believe I already asked that but had a silence !)
I wanted to ask the same.:)

You can then connect it GND to GND, Signal to where the trafo + connects, but with a series capacitors, any value around 0.1 uF.

The important when using the LVDS receiver is that it need to be biased around 1.25V as the common mode input range is 0-2.5V.
Could someone do a drawing of this? It is not absolutely clear to me. Thanks!
 
Is it like this?
 

Attachments

  • Spdif.jpg
    Spdif.jpg
    13.8 KB · Views: 840
I need to clarify that the above power supply comments refer to the power of the upstream devices, whether a CD/DVD player or music server of some sort, whether using direct I2S, USB, S/PDIF, or TOSLINK outputs. Also the associated connected equipment, such as Ethernet hubs.

Greg in Mississippi

are u saying if the source had a clean enough power supply
that this unease in the sound of the dam dac will be gone?

not that I disagree upstream has a profound influence on sound quality mostly due to power supplies
just not sure it would change my opinion of one dac vs another
 
IWithout headphone, do some members have the same hole in the middle between the speakers ? Like if the speakers were toe towards the exterior (not in: towards the listener) ?
Sorry to re ask ? I had no answer about that, I called this "phase" but have not technical knowledge about this (it's certainly not the good word) but I have not this strange stage presentation with my others dacs and cd players : e.g. voice is not centrer but divided in each speakers comming from the laterral walls !!!! (not nice)
No, with speakers the sound stage is very natural, things in the center are just "there". No noticable change in position with differnt DAC.
 
Member
Joined 2010
Paid Member
Attached is the last layout of the board where in addition to the Audio Note I added the transformers Murata DA101C and Pulse PE-65612.

@AR2, I can't found the datasheet for the SC947-02LF. If you have it please send post here or send me by PM.

The U.FL connectors have been moved back on the top side.

The board is 39 x 42 mm and doesn't need any standoff, will be solid just plugged in the 26 poles connector.

I will produce anyway a small quantity to have it when I will receive my DAC. The board is so simple that I don't think that I will have any debug to do.

If you have any interest I can run a GB... increasing the number of board will decrease the unit price.
In case, in order to don't make confusion in all the technical discussions on going here, I will open a thread in the GB section.

Regards,
Enrico
 

Attachments

  • R2R_INT_Rev3_BOTTOM.png
    R2R_INT_Rev3_BOTTOM.png
    65 KB · Views: 803
  • R2R_INT_Rev3_TOP.png
    R2R_INT_Rev3_TOP.png
    109.7 KB · Views: 797
Is it like this?

You must add a 0.1 uF minimum cap (film or ceramic) between rca and spidf in+ (but if you are absolutly sure than your streamer has already one before or after its pulse transformer in serie on the signal : if any doubt put the 0.1 uF cap or more in serie - DC pulse signal must be avoided and a pulse transformer is not enough here-)

My understanding is the 330 ohms has to be removed ! (as Soren didn't talk of it in the answer made to me !)
 
Attached is the last layout of the board where in addition to the Audio Note I added the transformers Murata DA101C and Pulse PE-65612.

@AR2, I can't found the datasheet for the SC947-02LF. If you have it please send post here or send me by PM.

The U.FL connectors have been moved back on the top side.

The board is 39 x 42 mm and doesn't need any standoff, will be solid just plugged in the 26 poles connector.

I will produce anyway a small quantity to have it when I will receive my DAC. The board is so simple that I don't think that I will have any debug to do.

If you have any interest I can run a GB... increasing the number of board will decrease the unit price.
In case, in order to don't make confusion in all the technical discussions on going here, I will open a thread in the GB section.

Regards,
Enrico

You might want to send the schematics to me, just to check....
 
You must add a 0.1 uF minimum cap (film or ceramic) between rca and spidf in+ (but if you are absolutly sure than your streamer has already one before or after its pulse transformer in serie on the signal : if any doubt put the 0.1 uF cap or more in serie - DC pulse signal must be avoided and a pulse transformer is not enough here-)

My understanding is the 330 ohms has to be removed ! (as Soren didn't talk of it in the answer made to me !)

I wouldn't remove the 330R, it's still needed as the LVDS receiver is 100R input impedance and SPDIF is 75R....
 
Attached is the last layout of the board where in addition to the Audio Note I added the transformers Murata DA101C and Pulse PE-65612.

@AR2, I can't found the datasheet for the SC947-02LF. If you have it please send post here or send me by PM.

The U.FL connectors have been moved back on the top side.

The board is 39 x 42 mm and doesn't need any standoff, will be solid just plugged in the 26 poles connector.

I will produce anyway a small quantity to have it when I will receive my DAC. The board is so simple that I don't think that I will have any debug to do.

If you have any interest I can run a GB... increasing the number of board will decrease the unit price.
In case, in order to don't make confusion in all the technical discussions on going here, I will open a thread in the GB section.

Regards,
Enrico

DA101C and PE-65612 pads are quite similar, I think PE-65612 pads can be removed.

There is a datasheet for SC trafo on PC's site (not the exact model but solder pads should be the same):
digital_scicon

BTW, is your board equiped with a female connector so that it can be plugged into Soekris DAC's male input connector (see the pics below) ? May be 1 or 2 standoffs are needed at the Coax/Toslink connector side to provide stability. Thanks.
 

Attachments

  • Input Board connector - Soekris dac rs.jpg
    Input Board connector - Soekris dac rs.jpg
    133.5 KB · Views: 791
Last edited: