Building the ultimate NOS DAC using TDA1541A

I need help from people more gifted than I am in digital electonics.

I built the I2S to simultaneous converter proposed by John from EC Designs in post #6315 on page 632, and later develpoed into a PCB by member Ryanj. I used my own PCB design.

I have built a TDA 1541A DAC with 50hz DEM clock built inside a Marantz CD67 player - so the CD67 is effectively a transport feeding an I2S signal to the separate DAC borad inside the player. I decided to convert from I2S to simultaneous operation based on comments from this thread. Problem was the circuit proposed by John did not give any audio from the left channel (DOL). Right channel was fine. Connecting DOR to left DAC input gave audio, so amp and DAC were working fine....just no output from DOL.

Looking on the scope I could see that the DOL abnd DOR outputs were not synchronised (relative to LE ouput) using 16 BIT I2S feed from inside the CD player. What I eventually did was to remove one of the 74LV164 stages (U4) from the left delay section. So the DOL delay is now 23 shifts versus the original 31 shifts with 4 x 74LV164 (output is now taken from pin 13 on the third 74lV164 U3 to input of 741G79 U7). This was the only feed that gave synchronised data.

Now I get perfect left and right channel output. So my question is: Was the original circuit proposed by John designed for 24 bit I2S data not 16 bit I2S data that you find in a standard CD player, and is this why the circuit did not work in my CD player? Does what I have done make sense?

To my ears the DAC now plays perfectly, but just want to make sure I am not screwing up somewhere.

Any comments appreciated.
 
Last edited:
Hi john jones,

Marantz CD67, from the manual:

Marantz CD-57 Service Manual (Page 7 of 20) | ManualsLib


Clock equals 16.9344 MHz (384fs).

44.1 KHz, 8 times oversampling and 48 bits / frame results in:

44100 * 8 * 48 = 16.9344 MHz.

So the Marantz 67 works with 48 bits / frame.


The I2S to simultaneous converter was designed for I2S, 64 bits / frame. That's why you needed to adapt it to make it work correctly with 48 bits / frame.
 
Hi John,


Sorry for interrupting, if I may get your attention:


- You specified 25V el-caps for 50Hz decoupling. All manuals however quote 7.3V max on the pins + el-caps unlike tantalums seem to have the leakeage currents stable at 0.01CV until the max volatge. Will 10V caps work? I've put them caouse where the smallest of the type I'm interested in (Elna Cerafine). A bit scared powering up not to blow them.


- How critical is the layout at 50Hz? My longest PCB trace to the cap is 32mm, comparable to other PCB's I've seen here. Thick copper, 75u + metalization so about 100u total. GND plane on the other side of the board.


Thanks!
 
Hi john jones,

Marantz CD67, from the manual:

Marantz CD-57 Service Manual (Page 7 of 20) | ManualsLib


Clock equals 16.9344 MHz (384fs).

44.1 KHz, 8 times oversampling and 48 bits / frame results in:

44100 * 8 * 48 = 16.9344 MHz.

So the Marantz 67 works with 48 bits / frame.


The I2S to simultaneous converter was designed for I2S, 64 bits / frame. That's why you needed to adapt it to make it work correctly with 48 bits / frame.

Thanks for the feedback John. Seems so simple when you explain it. Why didn't I think of that! Anyway DAC is now playing wonderful music.
 
DA96ETF
Power supply: external, 5V/100mA (4.8 ... 5V5), connection USB-B (large). Matching low (ground-loop) noise, power supply is included.
Supply current: 25mA@44.1KHz (125 milliwatts), 35mA@96 KHz (175 milliwatts).
Digital audio interface: ElectroTos (Standard Toslink optical socket for use with / ElectroTos plug).
Supported protocol: ElectroTos low jitter protocol.
Note: the D/A96ETF does not support standard Toslink protocol, it will only produce noise when attempting to drive it with standard Toslink sources.
The DA96ETF pairs with U192ETL or UPL96ETL. These support the ElectroTos low jitter protocol.
-> ElectroTos low jitter protocol eliminates -unwanted- jitter in digital audio source, digital audio interlink, and all DAC circuits even when significant crosstalk is present. It has to be generated directly by software.
Sample rate support: 44.1, 48, 88.2, and 96 KHz.
Bit depth: automatic (ElectroTos protocol). U192ETL and UPL95ETL support 24 bit.
Internal digital audio interface system for the SIPO: (serial to parallel converter), 4 wire ElectroTos, ELE, ESC, ECK, and EDA.
Internal digital audio interface D/A converter: parallel interface.
Standard digital audio protocols are no longer used / generated anywhere in the DA96ETF.
D/A converter type: Fractal multi-pattern converter, 32 bit core / channel (this type of converter is -completely- different from all existing audio D/A converters).
Fractal multi-pattern converters offer extreme long term accuracy (with given component tolerances), low output impedance (375 OHms), and extremely low digital interface switching noise because of the parallel data interface (highest switching frequency equals the sample rate / ELE latch signal).
System: unfiltered NOS (fully passive output circuit). By using 88.2 or 96 KHz, 2 x oversampling can be obtained.
Output voltage: 3V6pp, DC-coupled.
Output connector: RCA, gold-plated, shielded connector soldered directly into the PCB for lowest possible noise / losses.
Output impedance: 375 OHms, can directly drive 250 OHms Beyerdynamic DT-990 studio or comparable headphones.
Recommended lossless volume control: Shunt volume control (shunt only), 375 OHm output impedance is used as fixed series resistor for the shunt volume control. This shunt regulator can be placed in parallel with Beyerdynamic DT-990 headphones for lossless volume control. Typical couput impedance 0.38 OHms (-60dB) ... 375 OHms (0dB). This way headphones can be driven without using any active linear circuit in the audio signal path, just a digital latch and a resistor network.
The DA96ETF consists of a logic board that contains all connectors, supply voltage filters, voltage regulators, ElectroTos receiver circuit, SIPO and indication LED (lock LED). The two Fractal multi-pattern D/A converters are located on a piggyback board that is plugged on top of the logic board, there is no wiring (very compact module, shortest possible connections).
I attached a sketch of the DA96ETF housing.

Hi ecdesigns,

may I ask why the DA96ETF does support sample rates up to 96kHz only?
Thanks

Matt
 
Hi ecdesigns,

on AS forum there was an interesting post last week:

I have heard back from ECD. They have done further improvements to the USB interface (U192ETL) to bring it up to a level where it "becomes hard to differentiate" from the UPL96ETL (which itself has been improved also).

If that is the case there will very probably be no further improvements to be obtained by optimizing the USB source as the U192ETL will be effectively "immunized" from upstream noise - solving issues of USB audio... I'm not going to quote them on the details of the way this is done, better leave this to them. It is quite a breakthrough IMO.

I understand the second batch of products will include all these changes and they will contact existing customers individually to manage returns/upgrades.

I am looking forward to listening tests with these upgraded units.

Can you please shed some light on these further improvements?
Thanks

Matt
 
Last edited:
Hi diggy,

what is the sampling rate supported by this device please

U192ETL USB to ElectroTos translator: 44.1/16 ... 192/24.



is it usb to usb, usb to I2S, usb to toslink ?

U192ETL is USB -> ElectroTos -> Toslink (S/PDIF protocol)

or

USB -> ElectroTos -> Toslink (ElectroTos low jitter protocol) for driving our Fractal DAC only.


I2S is not generated / used in any of our devices.
S/PDIF is only generated / used for compatibility with existing DACs.


ElectroTos interlink has a RCA plug on one side, interlink is coaxial, output is optical and compatible with Toslink. It operates with 3V3 pulsed DC.

It offers superb galvanic insulation, low jitter, and band limited (20MHz) output for minimum noise injection into the DAC.
 
Hi matthiasw,

may I ask why the DA96ETF does support sample rates up to 96kHz only?
Thanks

The very best performing S/PDIF receiver we found (we thoroughly tested & tweaked all popular S/PDIF receivers over and over again) was the DIR9001 with a tweaked loop filter and in combination with the ElectroTos low jitter protocol. The DIR9001 supports 96 KHz maximum.

Higher sample rates require higher clock and data rates in source, interlink and DAC circuits. Circuits are never perfect, the major issues with digital audio are stray capacitances and stray inductances in source, interlink and DAC. The combination of high frequency clock and data rates with these parasitics will cause major issues with digital audio that are still not fixed. One of the results is source dependency (every bit-perfect source still sounds different). When clearly audible source dependency occurs, digital audio is still far from optimal.

Every imaginable interference that is added to the always present digital audio related signals will make matters worse, think of (RF) ground loop noise and power supply (mains) noise. All data and frequency signals will crosstalk and create one big unpredictable interference spectrum that can reach into the GHz range (harmonics). Even Bluetooth, GSM and WIFI add to this mix. This complex and constantly varying large bandwidth spectrum will be de-modulated by P-N junctions for example (think of radio receivers). Then demodulation products can be formed from two or more inaudible frequencies that will now fall within the audio spectrum (same happens with radio, television and wireless phones, but here this is desired to get the information across).

This interference cannot be blocked by 20KHz brickwall filters because these frequencies fall within the audio spectrum itself. The result is audible degrading / change of sound even when using world's best synchronous re-clockers in a DAC.

This is the major (not the only) reason that bit-perfect sources still sound different with DACs that should technically be close to perfection. Source dependency is noticed by many audiophiles and the most hilarious fixes are used like audiophile LAN switches and filtering equipment that is more expensive than the digital audio set.


But it gets even much worse. Analogue audio was never perfect either, every amplifier (driven by an analogue audio source) sounds different. No analogue amplifier sounds or will ever sound the same and or perfect. There are too many variables to keep under control and there are fundamental problems with analogue signal processing that are not even discovered. I think of analogue audio circuits as non-linear circuits.

Now mix imperfect digital audio and imperfect analogue audio and one created one huge problem that is never going to be fixed. Think of digital and analogue as water and fire, these will never mix well.

Chances of success are a bit higher when going all analogue (tape / vinyl + analogue (pre) amplifiers) or all digital (no active analogue circuits allowed, only digital on/off switches).

So I personally arrived at the conclusion that the conventional digital audio setup with a digital audio source, DAC and (pre) amplifier will always be doomed to fail. Recently I gave up on (pre) amps combined with digital audio equipment and developed an alternative that might trigger a revolution in top performance digital audio. More on this later.


Can you please shed some light on these further improvements?
Thanks

We offer the UPL96ETL and U192ETL. After some time it became very clear that the UPL96ETL USB-memory-stick based player simply performs a lot better than the U192ETL, understandable, that's why we developed it in the first place, because of the never ending issues with USB audio. We still offer the U192ETL because most audiophiles prefer USB audio streaming and take related degrading for granted.

I was asked if I could give it another try and slightly improve the U192ETL so it comes closer to our UPL96ETL reference player. The U192ETL is an XMOS-based USB receiver with asynchronous feedback and master clocks that have better phase noise specs compared to the popular Crystek CCHD-759 clocks. We use a 8-core XMOS (XU208). Such configuration can be found in many audiophile DACs and this is perhaps the best performing solution for USB audio streaming, but it still fails to match UPL96ETL, SD-card players or even audiophile CD transports.

The irony is that both, UPL96ETL and U192ETL use exactly the same master clocks, same USB bus power supply, same USB interface and same synchronous re-clocker circuits. Find the 10 differences .......

Logically speaking it -should- be possible to get very similar performance from very similar circuits. So I was determined to pinpoint this decades old issue with USB audio and fix it once and for all.

After systematically testing -all- possible causes over and over again, one issue remained, this simply had to be it. I even mentioned this problem on this thread multiple times ... I should have known. The fix was a different matter, it basically boils down to blocking noise, but how?

I tried most popular digital isolators, opto-couplers constructed with Toslink transmitters and receivers. Nothing worked, only made matters worse. The theoretical galvanically insulated approach with master clocks on the other side of this barrier (feeding clocks back to the XMOS using a digital coupler) also failed miserably. For a while it seemed that this issue cannot be fixed and we simply have to live with the clearly degraded sound (compared to the reference) of USB audio.

After many sleepless nights I came up with one other option and this one finally worked, it fixed the USB issue. Not only that, when applied to the UPL96ETL reference, even that reference improved significantly. Now both, UPL96ETL and U192ETL perform on a virtually similar level. Sure there will still be small audible differences, nothing is perfect, but when these are very difficult to hear or cannot be heard at all by direct comparison with a reference, it is more than good enough.


Our listening test setups:

Macbook -> USB (UC250 interlink) -> Fractal PowerDAC -> 90dB Open baffle speakers (no analogue signal path, all digital).

Macbook -> USB (UC250 interlink) -> Modified DA96ETF (187.5 Ohm output impedance) -> SVC24 shunt volume control -> Beyerdynamic DT990-pro studio headphones (250 Ohm).

Macbook -> USB (UC250 interlink) -> DA96ETF -> SVC24 -> servo power amps -> 90dB Open baffle speakers.

We used unfiltered mains power supply, WIFI, Bluetooth all switched-off. No smartphones anywhere near the test setup.

We used Swinsian application

Swinsian — The Advanced Mac Music Player

We tested 44.1, 48, 88.2 and 96 KHz files.

We played all different kind of genres and different recording quality. Above systems make it very easy to hear big differences in recording quality and the highly destructive impact of DSP on already mastered audio files.


After getting annoyed with the flimsy USB interlinks with poor shielding and super thin wiring I decided to construct a better option. I use individually shielded data and power wiring pairs and thick wire with low DC resistance so I get a decent power supply at the USB receiver (reduced voltage drop / reduced load induced ripple voltage) and more reliable data transfer over larger distances. Note that with USB audio streaming, data transmission errors cannot be corrected. The individually shielded wire pairs minimise unwanted crosstalk between both, USB bus power wire pairs and USB data wire pairs.

I attached a picture of cable construction
 

Attachments

  • usb.jpg
    usb.jpg
    79.9 KB · Views: 618
Disabled Account
Joined 2019
Hello John,


Why the UPL96ETL supports "only" 99 Gb memory stick if I understood corectly ? Are some low factor size SDD memory or more bulky 2.5' SDD will not work either ? Where is the limitation please ? hardware drivers, programed chips in factory when not FGPA but chips as DIR901 for illustration?

There are micro SD card players used a lot with sucess in hifi, I assume more or less same tech are also limited but here some micro sdcard are now 1To but those players seems also limited to a max Gb they can read & not only by the not usefull acess to the datas - no library management-

By the way I have to say to you - congrats to the brother- that the soft you provid to manage the library on memory stick is a brillant idea many huge company didn't make yet despite an audiophile niche expensive market where performances matter first.

So now we have long life batteries with LiPoFe4 or Ni-Mh cells that can prevent the source to be grid walled. But from where comes all the problems you listed above - but the IME you explained & personnaly liked the hilarous word about wifi, bluetooth, etc ground-loops breakers- . We can shield a little though copper is expensive, we can not make any things but your already good choice to test extensivly the chips when having IE knowledge, same with passive components certainly the usual esr, esl not suffice to explain all ?
What is the trick using light memory banks : low voltage and low current that reduce the noises to a more acceptance factor ? I don't understand why cells are not used in the UPL96ETL despite now wall can be very good as Ifi, you and some others company can provide today, btw are they not smps technology but with low noise management ? Ok I understand even one feed a RaspBerry Pi with cells, the design is flawed when audio is concerned and nothing can fix the layout and the "glitchs" of the pcb layout & design for our audio purposes...

I think you have it all about the price management of your EC Designs hifi devices - congrats- :)... wanted to finish on that note cause it's rare !
 
Our listening test setups:

Macbook -> USB (UC250 interlink) -> Fractal PowerDAC -> 90dB Open baffle speakers (no analogue signal path, all digital).

Macbook -> USB (UC250 interlink) -> DA96ETF -> SVC24 -> servo power amps -> 90dB Open baffle speakers

Hi ecdesigns,

thank you very much for your answer, much appreciated :)

Which of the two listening setups above does sound superior?

What about the Fractal PowerDAC?

Thanks again

Matt
 
Last edited:
Disabled Account
Joined 2019
Ah, I'm a simple enthusiast and I got from HK & German company some A123 LiPoFe4 very fastly since IanCanada talked about it. I think while not usual - that option should exist in the device with a switch : free to the device owner to command its own LiPo.



Ni-Mh cells are also less good but can be find in a supermarkett . You can understand it's feasible if it's a factor of choice for audio quality - which I'm not sure it is though, hence the questions-
 
Disabled Account
Joined 2019
No,

First you talk if you were John, are you?
. Second I never talked about DIY but if you think than buying a cell to a website or a supermarkett is diy... what can I answer ???
Third I didn't ask for it, see it more as a technical question about batteries & sound which was not adressed to you btw. I just illustrated you it is possible but if a cell has no sense in relation to a sound improvment : please re read my questions to John.
Forth it's simple to have a switch toggle or push button position between AC for grid -wall and DC from an internal cell zone on a device : think about a remote for illustration : where is the diy option when putting two cells inside :rolleyes: ?

At a conclusion I believe you readed my question to John as a critic of him, his great job and great inputs according your post above:confused: it was not not, you just wrong everywhere imho.:D
 
Last edited: