Building the ultimate NOS DAC using TDA1541A

[dddac]

Re: resistor tolerance

So instead of 470 Ohm you could use 10 X 4K7 in parallel, in this case it is advisable to use 0.1% metal film resistors to maintain 1% overall tolerance.

I am surprised you make this statement after all the discussion of paralelling DAC's

every paralelling of a random process will reduce tolerance with SQRT ( n ) improvement. where n is the number of parallel items.

Thus, 10 resistors will bring down 1% tolerance to aprox 0,3% ......


I have another question for you as well, to put things (at least for me) in perspective. The language you use for describing the differences between the DACs and other improvements you made are of those typically heard in magazines etc. I prefer to be a bit backholding on it, but that is a personal matter and I don´t mind too much. Any way, have you been able to comprate your DAC with high resolution media, like Audio DVD or SACD ?

Would you say that your DAC sound better or still you do not arrive at that level of resolution ?

doede

 

[ecdesigns]

Hi, MGH,

Thanks for your reply [post#180]

>The complete octal D-I dac consumes about 28W of power (OPA627 I/V converters and diff amp). By optimizing power supply circuitry i could reduce this to about 20W. TDA1541A temperature averages around 40 degrees centigrade at 20 degrees centigrade room temperature.

> I have build power amplifiers on epoxy board about 20 years ago, they still look fine and function perfectly. The temperatures in the D-I DAC are rather low, there are no components that get so hot they could damage the PCB's. But since the design is fully modular, analog sections (Quad DAC boards and diff amp) could be built on teflon boards if desired.

> USB interface (post #47)

>Battery power, Good point MGH, I have to take this into account in the D-I DAC design. Options are lead acid batteries or NiMh rechargable D cells.

 

[MGH]

Thanks ecdesigns,

I forgot you had plans for USB.

Yes, battery is a good plan - I owned an Ack! DAC which was battery powered and had the quietest digital playback I've ever heard. I'm not sure which type of battery sounds better, but I've heard lead acid is a more stable power source over time. What do you all think? Perhaps Doede could give his expertise on this.

If you can get Teflon PCB with thick copper tracing with gold plating, that would be a first for a DIY kit and look awsesome, plus a PCB that will last indefinately although I doubt it would have a sonic benefit. Unfortunately, I've PCB that has became brittle and is falling apart after 10 years of use - may be bad luck or the heat and humidity here.

Based on your octal DAC, extra cooling doesn't seem like a necessity. But after adding tube stage, the DAC may need some good ventillation.

I'm really looking foreward to seeing your final design.

 

[ecdesigns]

DAC module schematics

Hi, all

Here is the schematic diagram of a single DAC module I used in the D-I DAC. The quad-DAC board has 4 of these identical modules on one PCB. The design is compact and straight forward. A printed circuit board layout will follow next.

J1 and J2 are 10 way single row pin headers
U1...U3 voltage regulators +5V, -5V and -15V respectively
C1...C8 are tantalium cap's 4.7uF / 35V. Black marker on PCB indicates minus, long wire / marker on the tantalium cap's indicates plus
C9...C17 are 1206 size X7R SMD cap's for power supply decoupling
C18...C31 are 100nF film capacitors from BC components for TDA1541A decoupling
C32 is 470pF polystyrene cap (will be modified later for DEM clock synchronizing, probably 100pF capacitor with jumper and addittional pin)
R1...R3 are 22 Ohm resistors (MRS25)
U4 is 28 pin socket with turned pin's for TDA1541A

 

[ecdesigns]

DAC module PCB layout

Hi all,

Here is the double sided PCB layout of the DAC module, so you can have a look at it, or use it to make your own double sided PCB.

From left to right: silkscreen component side, silkscreen solder side, solder side tracks, component side tracks

Exact board size is 1.4" X 3.95" or 35.56mm X 100.33mm. The board holds 2 separate parts: the main board and a small voltage regulator insulating PCB (the part with the 3 rectangles). This part is placed between the DAC board and the 3 voltage regulators, rectangles facing towards the voltage regulators. You can also place 4 of these identical modules on one circuit board to make a quad DAC PCB like I did, but pin spacing between 2 modules must be exactly 1.45" or 36.83mm so it will fit on the main board later on.

Tracks are kept as short as possible, power supply decoupling cap's are placed directly at the semiconductor pin's. Each DAC module has it's own fully separated power supply (crosstalk), with very short connections to the DAC chip (this method already proved itself in the prototype). Digital and analog ground are fully separated. Ground planes and tracks are used for minimizing interference. Modular setup enables easy star grounding scheme. The main board holds addittional (power supply and I2S) filters to obtain both a very clean power supply and clean digital signals.

First 5 trough metalizations have to be placed (wires)
The 28 pin socket, resistors and voltage regulators have to be soldered at both sides of the PCB. All other components either on the component side (pinheaders) or solder side (cap's)
C9...17 are SMD 1206 cap's that are soldered directly on the solder side

DEM clock synchronizing: The DAC modules can use either the standard 470pF cap's or DEM clock synchronizing with BCK. With BCK synchronizing a 100pF capacitor is used instead of the 470pF, and this cap is routed to a 176.4KHz clock (divider placed on main board). I have to figure out how to do this in a way that causes minimal interference. So either an existing pin is assigned for this, or an addittional pin is placed very close to the 100pF cap.

 

[guido]

Hi,

You did not search.... In short:

Supply for current reference is between -15 and -5.
This should be clean.

'audio' current comes out of the output pins (obvious) AND the +5 supply pin (left and right together). So your 'power supply caps' really are a means to get the audio currents to ground (coming out of the +5), where some caps at the output stage close the circle. I would not like to try to figure out where all the currents go in your setup...

and there is more to find.

 

[tubee]

and there is more to find.

Yes, there's a lot you can (or have to) do with a TDA1541

Example: pin 4 is connected with pin 2, so standard wiring. With the "A" version pin 4 can be connected to Gnd.

 

[ecdesigns]

Hi, dddac,

thanks for your reply [post#181]

>The parallelling of resistors in the I/V stage is soly to reduce inductivity. I indicated the use 0.1% resistors to be on the safe side, that's all.

> Sound improvements were made step by step over many years, and many listening sessions. I used the oscillograms and FFT scans to give an impression of the D-I DAC performance. I can only say that I already had a few High-end enthousiasts come over to listen to the octal D-I DAC. Here some remarks they made:

- Details can be heared more clearly
- The sound is so realistic, as if you were "there"
- Sound is much more "open"

I use the sonic resonators [post#8] instead of standard loudspeakers. The latest highly optimized version reveals much more information about the sound. They re-create room acoustics in a very realistic way. Now when I use a DAC or amplifier for comparison, differences can be clearly heared.

> SACD/DVD to CD D-I DAC comparison.

Years ago I compared my twin DAC with 8th order filter to a SACD1000 player (SACD has 2 layers, the CD layer and DSD layer). The SACD owner said that the twin DAC sounded more open and more realistic....

I haven't compared the D-I DAC to SACD/DVD yet, but I am very curious at the result.

 

[ecdesigns]

Quad DAC improvements

Hi, guido / tubee

Thanks for your reply [post#186 / #187]

Improvements for the quad DAC module are very welcome. The quad DAC module has to be redesigned anyway (DEM clock synchronization). The power supply could perhaps be improved by using band gap reference diodes (LM336-5.0), the series resistors could be placed on the main board. Tubee was right, pin 4 (nc) of a TDA1541A can be connected to ground.

 

[ecdesigns]

Timing chain module schematics

Hi, all

Here is the schematic diagram of the timing chain module I used in the D-I DAC. It supports dual, quad and octal D-I DAC. The design is straight forward, but could be simplified (post #52 by dddac). A printed circuit board layout will follow next.

J1 ... J4 are single row pin headers
U1 voltage regulator +5V
C1, C2 are tantalium cap's 4.7uF / 35V. Black marker on PCB indicates minus, long wire / marker on the tantalium cap's indicates plus
C3...C19 are 1206 size X7R SMD cap's for power supply decoupling
U1...U14 are 74HC(T)166 synchronous serial shiftregisters
U15 is 74HC(T)00 quad NAND gates (indicated as LS in schematics)

Power supply:
+8VDC...+15VDC
Inputs I2S:
BCK (must be buffered), DATA, WORDSELECT

Outputs WORDSELECT:
WS, WS8, WS16, WS24, WS32, WS40, WS48, WS56

Outputs DATA:
D, D8, D16, D24, D32, ND32, D40, ND40, D48, ND48, D56, ND56

 

[ecdesigns]

Timing chain module PCB layout

Hi all,

Here is the double sided PCB layout of the timing chain module, so you can have a look at it, or use it to make your own double sided PCB. Perhaps this module is simplified later on, but this version works fine.

From top to bottom: silkscreen component side, solder side tracks, component side tracks, silkscreen solder side.

Exact board size is 5.75" X 2.15" or 146.05mm X 54.61mm.

Power supply decoupling cap's are placed directly at the semiconductor pin's. The timing chain board has it's own 5V power supply regulator.

First 3 trough metalizations have to be placed (wires)
U1...U15 (14 / 16 pin sockets), and voltage regulator have to be soldered at both sides of the PCB. I soldered the sockets at the solder side first, then heated the pin's at the solder side and applied soldering tin at the component side when the pin was heated up enough. This is necessary as the sockets are placed close together, it avoids damaging the socket's plastic frame. Check/measure for short circuits after each socket is soldered in place, because removing short circuits when all sockets are placed could become quite problematic.

All other components either on the component side (pinheaders) or solder side (cap's). Note that J1 pin's must be soldered on both sides.
C3...19 are SMD 1206 cap's that are soldered directly on the solder side

 

[tubee]

Hi Ecdesigns;

Off topic here but want to let you know:

i am just finisching my surround-back speakers. Dont want to make it very expensive, bought Monacor car power 25 Eu/pair CRB40 Neodymium dome tweeters for highs. Did not like the sound, but after removing the metal grid and laying them it horizonal with a radiating cone, it is really surround radiating now! I saw you made something like that, the Sonic resonators, here only the tweeter is radiating around, and i can tell now; very good sound, the dispersion cone works as a phase cone too, so hearable less distorsion/phase errors, and for a 12 Euri tweeter astonisching dynamic. Did experiments longer time ago, with a straight cone shape, now modified to an exponential shape, better fitting to the dome.(have a lathe here)

Post it all when i have photo's of it in the speakers departement.

 

[ecdesigns]

Sonic resonators

Hi tubee,

The sonic resonator is much more than some speakers and a few cones (see post#8). As you keep turning up the volume there is hardly any sound coloring or distortion until the amplifier clips. At very low levels, everything stays perfectly balanced (even bass reproduction). Because small speaker chassis can be used, this system is very fast. It took years to fine-tune as you have to correct many more parameters compared to conventional speakers. The sonic resonator acts as a wideband spherical radiator, "filling" the entire listening room with a very realistic sound, basically eliminating the need for rear speakers. The "sweet spot" as you all know it no longer exist, it extends to the entire room and beyond. The sound improvement you noticed is minute (during prototyping I tried something similar) compared to what these fine tuned resonators are capable off, so imagine how they sound like .

 

[tubee]

Hi Ecdesigns

Yes good sound demands years of fine tune, i know. What filters do you use? Series? Right now i am listening to the 360 degrees radiating tweeters: (on my stereo amp) very good sounding, indeed the stereo is more through whole room, even compared to the heavily modified T120K tweeters i use in the normal stereo pair, they radiate also very good, but only to the listener, not in the whole room. When i walk now around the speaker i hear highs all around. Agreed: this is an experiment for me which is demanding for more! : Full range radiating 360 degrees.

Update: the sound comes from above also, its if the tweeters communicate with each other.

 

[X.G.]

hi,Ecdesigns

your work is great.

I think you are on the right way.....I had experienced in the TDA1543 and TDA1541 NOS DAC.IMHO,the sound of NOS DAC is nature,but less datail,less focus,and the sound of OS DAC is more detail,but less natrue.I also tried TDA1541 NOS DAC+high order low-pass LC filter,its sound is more detail,more focus than without the filter,but less danamic,and the higher the order of the filter,the more less the danamic.Before you posted the thread,my friend told me that he got the natrue & detail sound in his new 'NOS DAC" which is TDA1541+CS8420 in upsample mode,much much better than the sound of his PCM1794 Evaluation board.


http://www.hifidiy.net/bbs/dispbbs.asp?BoardID=2&ID=24855&replyID=&skin=0

so I guess that the DAC without Digital filter easily get the natrue sound,the DAC with less imagine signal easily get the detail sound.

X.G.

 

[ezkcdude]

Dude, that doesn't make any sense to me. How do you have a NOS DAC with upsampling? Do you call that a UNOS (upsampling non-oversampling) DAC?

 

[MGH]

Hi X.G.

I get jist of what you're trying to say, but why do you provide a link that very few, if any, of us could understand? Most of us here can't speak or understand Chinese.

 

[ecdesigns]

Hi, X.G.

Thanks for your reply (post#195)

I had a look at the URL you posted (especially at the circuit board and schematic diagrams). You indicated about the same things I already noticed:

- a standard NOS-DAC has less detail but more natural "open" sound
- a NOS-DAC with filter has more detail, but it's sound is not so natural / "open" anymore
- a oversampling DAC has more detail, but the sound is not so natural "open"
- a "NOS-DAC" with a CS8420 sample rate converter (upsampling) has both, more detail, and the sound is more natural / "open"

When I designed the D-I DAC, I tried to get a smoother output signal (less harmonics in the audio range and no need for filtering). This is done by using "copies" of one sample and place them on regular distances within 64BCK clock cycles. This way you get more "samples" within the 64BCK cycle. Then the outputs of all DAC's are added together, so the copied samples also form a higher resolution (amplitude). See post#34 for a graphic representation using only 2 DAC's. With 4 or 8 DAC's basically the same happens. With an octal DAC you get a theoretical resolution of 19 bits and a virtual sample rate of 352.8KHz.

I am quite happy with the sound now, I tried some other "tricks" like non-linear interpolation, but the sound quality even seem to get worse. I know it's not perfect according to theoretical analysis, but it's a great improvement compared to the standard NOS-DAC.

I also build several tube output stages in the past days. One of them is a differential triode amplifier with a transistor stabilizer on the annode voltage. The differential triode amp is then fed into a cathode follower. I use a "hybrid" setup with OPA627 I/V converters and a tube output stage. The sound is very transparent, and you also get a differential output (well I needed 2 double triodes anyway), but I can't seem to eliminate the noise that is produced by the tubes. So now I am trying to create a very high amplitude signal (easy when using op-amp I/V converters) and attenuate the output, this seems to reduce the noise quite a bit.

I also added a photo of a 8000Hz sinewave and FFT scan of the unfiltered octal D-I DAC.

 

[dddac]

- a standard NOS-DAC has less detail but more natural "open" sound
- a NOS-DAC with filter has more detail, but it's sound is not so natural / "open" anymore
- a oversampling DAC has more detail, but the sound is not so natural "open"
- a "NOS-DAC" with a CS8420 sample rate converter (upsampling) has both, more detail, and the sound is more natural / "open"

I tried all 4 of these versions myself past years with PCM63 and TDA1543 and came to similar conclusions. Although # 4 was not such a big thing.

For me the way out on keeping open sound and increase detail and spatial sound reproduction, was trying to increase the low level details by paralleling DAC chips. I use the TDA1543 as you know, as this is relatively easy to do, relatively cheap and it will get you straight 2 Volt Output level with no Transformer, Tube, OPAMP stages and still low output impedance.

It is amazing, but the difference in detail and soundstage between 12, 24 or even 60 DAC-chips is every time a clear step forward. Going from 1 Chip to 12 is listening to a totally different system .... I have not tried going above 60 as power consumption gets disproportinal to the expected results though ....

Basically what you are doing with the 8 x TDA1541 is similar. You keep the open NOS sound, but improve the detail.

Now the addition of time shifted samples is a very interesting EXTRA addition to this and I am working out this idea somewhat differently (I do read your punch line ) and will produce a small board based on SMD chips, as I have a professinal company doing small series for me at a very reasonable price. This keeps things easy to integrate in existing systems.

Well done, and I hope to be able to do some listening tests my self soon with my own DAC solutions during the next months or so

Doede

 

[MGH]

Hi ecdesigns,

When you finalize your kit, will you include the 1541A chips in the kit or do we need to get our own supply? thanks.