Hi maxlorenz,
Rewrite data basically does the same as defrag. The re-arrangement of the data (that remains unchanged with same checksum) may help to reduce or change the interference spectrum generated by disc access.
This interference is always audible on USB DACs as it is impossible to block all digital audio source interference. The interference is passed on through USB jitter and / or USB signal AM modulation (ripple on the digital signal). This is why there are -always- audible differences between digital audio sources and USB interlinks. The data remains the same but the interference spectrum on this data in the form of jitter and ripple voltage varies and is unique for each digital audio source.
The problem with SD-cards is that these contain a wear levelling system to spead R/W cycles evenly over the memory cells in order to maximise SD-card service life. This also means that rewrite has no control over where the data will be stored on the SD-card as this is determined by the wear levelling algorithm. Similar would apply to SSD. It works on most HDDs. When the audio file is moved it starts to defrag again so the rewrite effect is gradually lost.
So in short, rewrite is similar to defrag and it cannot work on devices with wear levelling.
It is much easier to buffer a complete track in system RAM, then playing the track from system RAM instead of from HDD, SSD, SD, or USB stick.
TDA1543 sounds most musical and involving of all Philips DAC chips but it does not offer detail and refinement like the TDA1541A. This has to do with 10 times higher THD of the TDA1543. The TDA1541A does offer detail and refinement but sounds rather analytical and uninvolving compared to the TDA1543. Every DAC chip has its typical sound that relates to its unique switching noise and distortion spectrum and none of these DAC chips sound transparent. The spectrum can be tweaked by injecting extra distortion but this in turn reduces detail and refinement. So it’s basically a hopeless situation.
That’s why I abandoned DAC chips and designed my own D/A converter. I used novel circuits that simply eliminate the many flaws that exist in popular (discrete) D/A converters by design.
Rewrite data basically does the same as defrag. The re-arrangement of the data (that remains unchanged with same checksum) may help to reduce or change the interference spectrum generated by disc access.
This interference is always audible on USB DACs as it is impossible to block all digital audio source interference. The interference is passed on through USB jitter and / or USB signal AM modulation (ripple on the digital signal). This is why there are -always- audible differences between digital audio sources and USB interlinks. The data remains the same but the interference spectrum on this data in the form of jitter and ripple voltage varies and is unique for each digital audio source.
The problem with SD-cards is that these contain a wear levelling system to spead R/W cycles evenly over the memory cells in order to maximise SD-card service life. This also means that rewrite has no control over where the data will be stored on the SD-card as this is determined by the wear levelling algorithm. Similar would apply to SSD. It works on most HDDs. When the audio file is moved it starts to defrag again so the rewrite effect is gradually lost.
So in short, rewrite is similar to defrag and it cannot work on devices with wear levelling.
It is much easier to buffer a complete track in system RAM, then playing the track from system RAM instead of from HDD, SSD, SD, or USB stick.
TDA1543 sounds most musical and involving of all Philips DAC chips but it does not offer detail and refinement like the TDA1541A. This has to do with 10 times higher THD of the TDA1543. The TDA1541A does offer detail and refinement but sounds rather analytical and uninvolving compared to the TDA1543. Every DAC chip has its typical sound that relates to its unique switching noise and distortion spectrum and none of these DAC chips sound transparent. The spectrum can be tweaked by injecting extra distortion but this in turn reduces detail and refinement. So it’s basically a hopeless situation.
That’s why I abandoned DAC chips and designed my own D/A converter. I used novel circuits that simply eliminate the many flaws that exist in popular (discrete) D/A converters by design.
OK. I got it.
I got questions.
About Mosaic:
Do you find differences between well done 16/44 and "Hi Res" files on it?
It is not possible to make a 16/44 - 24/192 Mosaic SD player, without sound degradation?
About Dual-Mono:
What about introducing low value R between logic chips interconnections and also using ferrite beads on their power rails, to prevent interference?
The Dual-Mono DAC is indeed very involving and dynamic, as I said, and with silver wire jumpers for the digital section and OCC cables for interconnection and speakers (now trying DIY speaker cables, very slow to burn-in) I achieved very open sound, which is not what we usualy expect with the veteran.
This is the best poor-man's DAC that I tried so far.
The sounds jumps at you in the fff passages, but it is not always upfront like other DACs are as the soundstage has proper depth, plenty of it...it is very hard to go to work...
Now listening Mahler's fift; very enjoyable, though I'm not a fan of Mahler.
Best wishes,
M.
I got questions.
About Mosaic:
Do you find differences between well done 16/44 and "Hi Res" files on it?
It is not possible to make a 16/44 - 24/192 Mosaic SD player, without sound degradation?
About Dual-Mono:
What about introducing low value R between logic chips interconnections and also using ferrite beads on their power rails, to prevent interference?
The Dual-Mono DAC is indeed very involving and dynamic, as I said, and with silver wire jumpers for the digital section and OCC cables for interconnection and speakers (now trying DIY speaker cables, very slow to burn-in) I achieved very open sound, which is not what we usualy expect with the veteran.
This is the best poor-man's DAC that I tried so far.
The sounds jumps at you in the fff passages, but it is not always upfront like other DACs are as the soundstage has proper depth, plenty of it...it is very hard to go to work...
Now listening Mahler's fift; very enjoyable, though I'm not a fan of Mahler.
Best wishes,
M.
The audible differences are marginal, 96 Khz sample rate may be useful to move the reflected mirror images further away from the audio spectrum but that’s about it. I personally see no advantages in using higher bit depth. But we provided 24 bits (limited by the laws of physics) resolution on the Mosaic 24 USB DAC so people can experiment with this.
Higher sample rate increases jitter sensitivity that in turn limits resolution. The break even point is somewhere around 96 KHz.
With the 18 bit auditory system resolution and the resolution limitations of audio sets at practical volume setting it makes little if any sense to use more that 16 bits resolution for playback.
So 16/96 would be the optimal, small files, maximum practical performance, also very suitable for lossless streaming through the internet (bandlimit).
It is technically possible when using the 4-bit protocol (annual license fees) and FLAC, one could even go as high as 24/384. In practice 16/96 would be more than sufficient and this would be no problem when using the SPI interface.
Yesterday we ran tests with digital silence test tracks on both, the Mosaic 16 player (SD-card) and Mosaic 24 USB DAC. The digital silence test tracks make it very easy to hear digital audio source interference during reading.
We used a battery powered mackbook air for streaming.
We turned the volume all the way up (2 x 160W with 90dB sensitivity speakers). Interference of both, Mosaic 16 player and Mosaic 24 USB DAC remained inaudible when listening centimeters away from the speakers.
Connecting the mains powered charger to the mackbook air introduced slight hum that was only audible centimeters away from the speakers.
The Mosaic DAC is based on a completely different concept compared to existing DACs. It offers up to 90% digital audio source interference blocking. Now it becomes interesting to use USB for transparent sound reproduction.
So the problem of digital audio source interference was finally solved at the D/A converter by means of the Mosaic DAC concept.
Low value resistors between locic building blocks can help to reduce interference spectrum bandwidth (RC low pass filtering with logic building block input capacitance).
One has to be careful with using ferrite beads in power supplies as this effectively increases power supply output impedance.
Hi Jaffrie,
Nothing is degraded, the data is just moved to non-contiguous blocks during writing. Where it is moved depends on the wear leveling algorithm:
Wear leveling - Wikipedia, the free encyclopedia
So one basically cannot make sure that all data is stored in contiguous blocks during writing. There would be a bigger chance of obtaining contiguous blocks when writing files to a brand new SD-card for the first time
When reading the data, it is not a continuous “flow” annymore but there are “hickups” as the time between access of the next data blocks now varies. These data read “hickups” can reach the DAC electronics and cause audible degrading with conventional D/A converters.
Digital silence test track can be used to verify audible source interference.
If one hears humming or buzzing sounds when the digital silence track is played, it prooves that the audio source interference comes through.
Nothing is degraded, the data is just moved to non-contiguous blocks during writing. Where it is moved depends on the wear leveling algorithm:
Wear leveling - Wikipedia, the free encyclopedia
So one basically cannot make sure that all data is stored in contiguous blocks during writing. There would be a bigger chance of obtaining contiguous blocks when writing files to a brand new SD-card for the first time
When reading the data, it is not a continuous “flow” annymore but there are “hickups” as the time between access of the next data blocks now varies. These data read “hickups” can reach the DAC electronics and cause audible degrading with conventional D/A converters.
Digital silence test track can be used to verify audible source interference.
If one hears humming or buzzing sounds when the digital silence track is played, it prooves that the audio source interference comes through.
-ECdesigns- wrote:
Thanks again for your input.
YGM.
Best wishes,
M.
PS: indeed, I have noticed that the TDA1543 DACs I own accept (or seems to accept) well 24/96 and 24/192 tracks. I imagine they simply neglect the last bits and the datasheet says it accepts up to 192KHz...these tracks sound in general better than "common" 16/44, perhaps because of the better registered and better engineered material. Nyquist was right or what?
Thanks again for your input.
YGM.
Best wishes,
M.
PS: indeed, I have noticed that the TDA1543 DACs I own accept (or seems to accept) well 24/96 and 24/192 tracks. I imagine they simply neglect the last bits and the datasheet says it accepts up to 192KHz...these tracks sound in general better than "common" 16/44, perhaps because of the better registered and better engineered material.
Nyquist was right or what?why not finally a fifo buffer to avoid any jitter from the source ? It is more and more avaliable ?
Ram is a buffer and are made also from micro caps iirc, so even Ram memory needs to be near the dac chip or not ? Rams need to be time re synchronised (I mean the I2S buffered in a Ram memory)
John, did you made a R2R discrete dac chip ?
Any answer please about feeding the tda1541 3 voltages rails please ? Better to tie the caps gnd decoupling firstly referenced on the pin 5 (AGND) or close to the pin 14 (DGND) first (even if both are tied together somexhere to share their ground on the pcb : better to tie the pin 14 on the analog ground or make a huge dgnd continuous ground on the pcb and tie it on the pin 5 near the 4 first pins (1 to 4 TDA pins) ?
Ram is a buffer and are made also from micro caps iirc, so even Ram memory needs to be near the dac chip or not ? Rams need to be time re synchronised (I mean the I2S buffered in a Ram memory)
John, did you made a R2R discrete dac chip ?
Any answer please about feeding the tda1541 3 voltages rails please ? Better to tie the caps gnd decoupling firstly referenced on the pin 5 (AGND) or close to the pin 14 (DGND) first (even if both are tied together somexhere to share their ground on the pcb : better to tie the pin 14 on the analog ground or make a huge dgnd continuous ground on the pcb and tie it on the pin 5 near the 4 first pins (1 to 4 TDA pins) ?
Hi Eldam,
The source must be as clean as possible. By loading an entire track into a RAM buffer (at the digital audio source) and playing it from RAM, disc I/O switching noise spectrum can be avoided.
FIFO-based reclockers don’t work as the I2S switching noise simply passes such circuit through crosstalk (stray capacitance, EMI, power supply rails, ground plane). The digital audio source simply has to be as clean as possible when using conventional D/A converters.
Post #5426
“The Mosaic DAC is based on a completely different concept compared to existing DACs”
It’s a discrete D/A converter but it is -not- based on the conventional (segmented) R2R ladder DAC concept. The (segmented) R2R ladder DAC concept contains design flaws that prevent transparent sound reproduction.
In the most ideal case one could approximate PCM1704 sound quality with a discrete segmented R2R ladder.
TDA1541A uses same supply voltages to power both digital and analogue circuits.
Decouple +5V, -5V, and -15V to pin 5 (AGND) using 33uF electrolytic cap in parallel with 100pF (805 size) bypass cap.
Decouple +5V, -5V, and -15V to pin 14 (AGND) using 33uF electrolytic cap in parallel with 100pF (805 size) bypass cap.
33uF electrolytic cap in parallel with 100pF (805 size) between -5V and -15V.
Use pin 5 for all analogue signal references.
Use pin 14 for all digital signal references.
Use 1uF 1210 size SMD film active divider decoupling caps in parallel with 100pF 805 size NPO cap and route these directly to pin 5.
Make absolutely sure that no other GND return currents share these active divider GND return currents as this couples noise into the extremely sensitive active divider outputs. So a solid ground plane would not be optimal for this as other GND return currents can now be coupled into the active divider return currents. Cover all active divider decoupling circuits with suitable lacquer in order to keep wiring dry and prevent DC leakage currents due to moisture / humidity.
Synchronise the DEM oscillator with WS, one can use up to 4 times WS frequency as long as this frequency does not exceed 200 KHz.
Use shortest possible traces / wiring and route the PCB as if it were a RF circuit. Always use a solid ground plane that does not contain any embedded traces nor isolated via’s. It is only allowed to connect via’s directly to this solid ground plane. The holes of isolated via’s will compromise the ground plane impedance. Buried via's and a 4-layer PCB are required.
Use Nichicon UKA electrolytic caps throughout the DAC ciruit and power supplies.
Only use NPO / COG ceramic caps.
The source must be as clean as possible. By loading an entire track into a RAM buffer (at the digital audio source) and playing it from RAM, disc I/O switching noise spectrum can be avoided.
FIFO-based reclockers don’t work as the I2S switching noise simply passes such circuit through crosstalk (stray capacitance, EMI, power supply rails, ground plane). The digital audio source simply has to be as clean as possible when using conventional D/A converters.
Post #5426
“The Mosaic DAC is based on a completely different concept compared to existing DACs”
It’s a discrete D/A converter but it is -not- based on the conventional (segmented) R2R ladder DAC concept. The (segmented) R2R ladder DAC concept contains design flaws that prevent transparent sound reproduction.
In the most ideal case one could approximate PCM1704 sound quality with a discrete segmented R2R ladder.
TDA1541A uses same supply voltages to power both digital and analogue circuits.
Decouple +5V, -5V, and -15V to pin 5 (AGND) using 33uF electrolytic cap in parallel with 100pF (805 size) bypass cap.
Decouple +5V, -5V, and -15V to pin 14 (AGND) using 33uF electrolytic cap in parallel with 100pF (805 size) bypass cap.
33uF electrolytic cap in parallel with 100pF (805 size) between -5V and -15V.
Use pin 5 for all analogue signal references.
Use pin 14 for all digital signal references.
Use 1uF 1210 size SMD film active divider decoupling caps in parallel with 100pF 805 size NPO cap and route these directly to pin 5.
Make absolutely sure that no other GND return currents share these active divider GND return currents as this couples noise into the extremely sensitive active divider outputs. So a solid ground plane would not be optimal for this as other GND return currents can now be coupled into the active divider return currents. Cover all active divider decoupling circuits with suitable lacquer in order to keep wiring dry and prevent DC leakage currents due to moisture / humidity.
Synchronise the DEM oscillator with WS, one can use up to 4 times WS frequency as long as this frequency does not exceed 200 KHz.
Use shortest possible traces / wiring and route the PCB as if it were a RF circuit. Always use a solid ground plane that does not contain any embedded traces nor isolated via’s. It is only allowed to connect via’s directly to this solid ground plane. The holes of isolated via’s will compromise the ground plane impedance. Buried via's and a 4-layer PCB are required.
Use Nichicon UKA electrolytic caps throughout the DAC ciruit and power supplies.
Only use NPO / COG ceramic caps.
Hello.
To see if i understand this correct, is it two 33uf and two 100pF per rail(+5, -5 and -15v) there one pair ground to pin 5 and one pair ground to pin 14?
Is smd caps better then 2.5mm film caps for the 14 decoupling?
Best Regards //Daniel
wow, thanks John for the road map...
So Finally you came again in your last tda1541 proto with active dem sync again and gave up passive 100 pF DEM Sync cap and 0.33 uF x14 DEM caps ! Ok
I always use UKA for the anlog stage of a TDA1541 DAC, with good results, I will try it for the tda1541 itself (thanks for the tip) .
Ok, so first decoupling the -/+ 5V & -15V with direct traces to pin 5 AGND... and only connect the star ground AGND to pin 14 on one point only (so without mixing AGND and DGND in several points for the 3 voltage rails at different places of the continuous dgnd plane) :
but would you join pin14 (dgnd) on the pin 5 directly or trying to draw a standalone trace (à la star ground you made) to connect pin 14 to pin 5 ?
I'm trying to understand if a layout should isolate pin5 from the close pins 1 to 4 (dgnd signals) for example by making an island around pin 5 with a specific trace from pin 5 to pin 14 ?
Of course I talk using a diy two layers pcb only : my idea is to divide in two one of the layer : the left aera (pin 1 to 14) will be the DGND on the upper pcb face with the pin 5 upper via not connected to it . On the bottom face the AGND star ground with all the agnd traces to pin 5 but this bottom pin 5 via is isolated from the left side and connected directly to pin 14 by a specific trace on the bottom (star ground technic) . Of course pin 14 is connected to the uper left side (DGND continuous plane) by its via/hole.
So I will have a pin 1 to 4 with a continuous plane for DGND return path (upper side) while having pin 5 isolated because tied close to pin 14 only ! The idea is to make the jonction between pin 5 and pin 14 near pin 14 and not near pin 5 !
Thank you for your answers, I know this is a non usefull return in the past for you talking about the TDA1541 !
I still love the sound balance of this chip even if the 1543 is lively, the1545A according to me having a better uper treble !
AD1862 is still avialble in good quantity at Rochester and for many despite its 20 bits sounds far better than the PCM1704 (24 bits) ! I will soon finish a layout than Painkiller member drawed with the ad1862 and have a point to benchmark the two chips despite a different analog stage (but very close : Pedja Rogic design given on his blog).
I follow all you thread since the beginning but the analog stage you described is upon my personnal skill and understanding (which is low as non technician) !
cheers, and thank you again for your threrad : all those explanations makes the dacs layout understanding more involving !
Eldam
Last edited:
Sorry to answer, I'm not John.... but having trying many caps : of course you can : far better than the old OSCON SP.... and not talking about new Oscon SEPC you talk about which must be avoided here (in fact some other like the SEP are better here despite the worst esr blalblabla, but even better are the United Chemicon Chemical ASA serie (polymer) and some others also...
The local supply decoupling caps will depend on the main power supply caps... so I should try in your shoes the Panasonic FC for the main (not FM... you could try the FR) and the BG you have near the pin 15, 28 & 26 !
Some fools would sayt it's an old debate, but you can waste a good design with the bad caps around the TDA 1541 . It's sounding signature can have a huge gap from the worst to the best in relation to the mixed caps used everywhere around it !
Ah I will try also myself the tip given by John about the Nichicon UKA for the 3 voltages rails !
So no SEPC or polymer caps. I'll buy and use Nichicon KA
And what about by-pass caps for power supply local decoupling? I have RedBaron board for through hole assembly without SMD pads for these caps. Lead spacing is 5mm.
So no SEPC or polymer caps. I'll buy and use Nichicon KA
And what about by-pass caps for power supply local decoupling? I have RedBaron board for through hole assembly without SMD pads for these caps. Lead spacing is 5mm.
I don't know about the Red Baron (certainly a name which could be avoided now
but a strong second degree sense of humour!) but would trust Oliver's experience !Wlowes member also testimonied about it and Black Gates with good results !
With the few tests I made around the TDA (but each hifi system has its own balance so hard to say which caps are the best : you should try test & error system on good bases
) : all BG series (if you have NX or best N serie for the local 3 rails, I found it like everybody specially good here !)I don't say you should avoid polymer but the specific too dry SEPC ! in this particular case ! I find the standard BG even better than the old Oscon SP (itself better than the newer SEP/C/F series !) ! The -15 V is important : BG N or NX gives better result !
But recently with a shunt feeded by a Nichicon FC (that I liked here much than a FM, FR serie being in second best position of the three !) : no local decoupling cap on the -15 V while the +/- 5 V has a BG local decoupling (with the same main Nichicon FC feeding the shunt); Which is strange as I had a BG NX on local -15v decoupling position : so each system is unique for the final setup about caps !
But hey this is in my specific hifi : nothing universal here ! For PS the ESR will tell you nothing to find the good tonal color and balance according yor whole system : this is a compensation !
Polymers stay good for dac chips but for local voltage decoupling try the Jean-Paul technic : Vishay SAL rpm (Farnel) : 3.5 mm pitch but thick leads (while the size cap stays small) ! Avoid MKP but maybe the Wima MKS2 only in the 2.5 mm leads pitch ! Will try this last on the 14 dem caps if no smd layout !
Those experiments was made with an analog stage feeded by the Nichicon KA! i/v resistor is the Rhopoint the designers advises : John, Thorsten, Oliver,...
Last edited:
Hows about yee old '45 !?! ie TDA1545!
http://s3-us-west-1.amazonaws.com/w...n-WatsonLecture-NEWS-WEB_0.jpeg?itok=sV62je7m
Ahem, ...Some of the j u m p of the '43 but with most of the refinement/detail of the the '41.
I believe I said s o m e w h e r e that ... the '45 wins for bang/buck ! Utilizing 'calibration' a la 1541, but in a simple [1543ish] package, for yer DIY pleasure!
Dear EC, you are 'maxin' out! And that's awesome ..
But DIYers often follow the scroungy middle -- that leads to a reasonable MID- MAX, ...
'braxy's got a lot of interesting info about Vref [lotsa low capacitance] and Vcc [ditto]. ... add your own exotic IV stage ...
In a diff thread perhaps?
"Po man's" 1541 ... 154*5*
Last edited: