Building the ultimate NOS DAC using TDA1541A

Re: Out with the 'ol Copland ...

NjoyTHEMUSIC said:
shhh :shhh:

Well, after all that I've just gone and made my next upgrade... :birthday:

Gonna have to do some workin' out, ready to carry the new WADIA'850 over the threshold (!) :yes:


- Life's a bitch sometimes {...and then some} !!!

cheers guys,

:wave2:
-Andy-


And in doing so you've brought the thread back on topic.
 
rfbrw said:


This way lies madness:crackup:.
In this thread the TDA1541A is clearly King of the hill. In the dddac threads it is the TDA1543. With some it is the TDA1545A. In the 300+ page long Marantz thread, some are getting excellent results with delta sigma dac SM5872. I've always felt the PCM63 to be the dogs but the Isodac mob will tell you it is the PMD100/SAA7350 combo. The Naimites have the TDA1305. I'd say that doesn't count for much but then again the same dac does a pretty decent job in the Dacmagic. For the APL lot it is AKM or nothing. There are also those oddities who like the TDA1547 but they are only visiting earth so I wouldn't bother too much with what they say, assuming you can understand them. Waiting in the wings is the PCM1704 which seems to capable of excellent results with or without a filter. And then there is the PCM1738 and the 1792 and the AD1853 and the CS4398 and the..... Then there the business of digital filters both off-shelf and proprietary.
Don't even think about looking for a definitive answer, there isn't one.


Bingo...

do I hear implementation!

cheers

T
 
Hi all,

Seems I am posting at the wrong thread,

I skipped some days as I have been busy testing all kinds of tap settings, sony format converter and a sony format timing chain. I am still working on a system to reduce jitter during transport of the differential I2S signal over (relatively) long distances 2...25 meters.

Some diyaudio members have come over to listen to both, the octal D-I DAC and the sonic resonators this weekend.
 
listening

Hi everybody,

Today I went listening at EC Designs. It did cost me about 500 kilometer today (around 312 miles) on my motorbike. But it was worth the effort!

You cannot imagine how natural the octal dac sounds. Really wonderfull, even at a loud volume, there is no listening fatigue. I would like to build the device.

By the way, I think, you cannot give any comment without listening to it.

And, the sonic resonators of EC Designs are very impressive!

Greetings

Dutchman
:D
 
exploring the limits

Hi all,

Project update,

In order to explore the limits of the Direct Interpolation technique, I have setup a 16 DAC version using the sony format. Taps on 3BCK multiples...using the TDA1543. The TDA1543 is specified for 4 times oversampling. Now it's running at a virtual sample rate of 705.6KHz comparable to 16 times oversampling! (so the sample rate is virtual, otherwise the TDA1543 wouldn't work), and a 20 bit theoretical resolution. I am pretty sure it will also run on 1.4112 MHz virtual sample rate / 21 bit theoretical resolution (32 DAC's comparable to 32 times oversampling). What about the TDA1543 reference pin? seems it musn't be connected directly as high frequency interference occurs. So I connected a 1 KOhm series resistor between each reference pin and the summing point (reference for the I/V converter). This also caused a significant drop in power consumption. 16 DAC's running at 7.5V only need around 750mA. Power consumption is down to 5.6W (0.35W / chip). The I/V OP-amp resistor had to be lowered to 360 Ohms. The setup is differential, so the output is fully DC coupled / no output capacitor needed. Oh yes, I used cheap OPA2132 op-amps.

Somehow I get the feeling this is no oversampling since the TDA1543 can't run at 16 times oversampling. So it looks like both the octal D-I DAC and the 16 DAC version are NOS DAC's afterall.

They meet the NOS DAC criteria, NON oversampling 44.1 KHz sample rate and filterless. But it has such high resolution, that the problems associated with NOS DAC's are practically non existent with this setup. It also has true phase linearity, very important for natural sound reproduction. The low 705.6 KHz residu is no longer a problem for both power amplifier and tweeters. High frequencies are reproduced crystal clear now.

The output signal looks very smooth on the oscilloscope, even with high frequencies. Frequency sweeps from 20Hz to 20KHz sound ok. The sound quality, despite the (economy version) TDA1543 DAC chip's, is very smooth and relaxing, crystal clear and open, it's doesn't sound like a digital system anymore.

So now it appears a very cheap D-I DAC can be constructed, using 8 or 16 DAC chip's that could fit on a euro sized board, timing chain, I/V conversion, diff amp and power supply regulators included that is. Both a philips and a sony version could be made, so it can be directly connected to a CDPRO2 using I2S.
 
multiple DAC chip's and resolution

Hi dddac,

Thanks for your reply [post#589]

Yes, the TDA1543 sounds better than I had expected, especially when multiples are used. With the standard NOS DAC setup (multiple TDA1543's in parallell) you get about 441 sample points for a 100 Hz sinewave, with a 1.25 KHz sinewave about 35 sample points. Each sample point can have 65536 possible values theoretically.

Now with the 16 DAC D-I technique, a 100 Hz sinewave has 7056 sample points, a 1.25 KHz sinewave about 564 sample points. Now each sample point can have 1048576 possible values theoretically. DAC current about 750mA, power consumption about 5.6W.

The resolution with 32 DAC D-I technique, 100 Hz sinewave, 14112 sample points. 1.25 KHz sinewave 1128 sample points. Samples can have 2097152 possible values theoretically. DAC current about 1.5A, power consumption about 11.25W.

With 64 DAC chips, 100 Hz sinewave, 28224 sample points. 1.25 KHz sinewave 2257 sample points. Samples can have 4194304 possible values theoretically. DAC current about 3A, power consumption about 22.5W.

This is quite a bit different than just parallelling DAC's, the output is way smoother and needs no filtering. The phase response is still true linear, and the trebles are crystal clear without interference.

That is what I mean by the D-I technique produces a smoother output signal that needs no filtering. The high frequency rolloff and linear interpolation errors still exist, but in exchange you get very high resolution and true linear phase response. So it seems it's either true sinewave shapes at high frequencies at the cost of phase errors. Or the D-I technique "problems" mentioned above, when one has to work with a 44.1 KHz sampled signal.

I added a photograph of the messy prototype setup, it's clean unfiltered 1.25 KHz sinewave 564 sample point output, and a view of the power supply to give a indication of power consumption of the complete 16 DAC setup.
 

Attachments

  • di16.jpg
    di16.jpg
    90.6 KB · Views: 1,639
Very interesting this TDA1543 experiment. One big pro is the easy implementation of it. Matter of fact is that i bought a cheap 2nd hand cdp with this dac, and i didn't dislike the sound, its unmodified allready slightly more transparent compared to 1541. John, where did you get the dac chips from, Reichelt?

I am happy to hear also Dutchman liked the 1541 D-I dac.;)
 
I am happy to hear also Dutchman liked the 1541 D-I dac.

Yeah...a pitty that he's on a hurry and couldn't share more details :D

The TDA1543's can have both an active or passive I/V stage, and the setup can be both differential (DC coupled) or single ended (capacitor needed). When a single ended setup is desired, the tap's for the even numbered DAC's serial data don't need to be inverted.

Most excellent news!
Did Dutchman heard the 1543 version?

Gratefully yours,
M
 
Hi all

Last Friday I visited ec-designs, curious about his system. I liked it a lot, but as everything was new for me I can not say anything about the quality of the D-I Dac on its own. What I can say is that the guy knows what he is talking about - and he even succeded in explaining me the debat between the NOS and OS.

I also heard the 8x TDA1543 version, but this was on another system, another room. But sure, not bad either! I was surprised to learn that with the 1k resistor on the REF pin the power consumption of the TDA's decreased by a half, so no cooling is needed.

And overall - what EC designs accomplished in 2 to 3 years of work is astounishing.
 
agent.5 said:
Just another crazy idea. If you are using 64 TDA1543 chips, the required I/V resistor must be very low, how about use a microphone transformer and use the DCR within the transformer as IV?

I have no idea if this will work. But it may.



It will work fo DC only than .... For AC you will have the transformed impedance from the load being the R in I/V ...

plus the DC will saturate the coil

ergo, won't work...
 
I/V stage OP-amps

Hi stefanobilliani,

Thanks for your tip [post#592]

I used a active I/V stage (OP-amp) for precision, so I needed a average reference voltage of all DAC chip's for the NON-inverting input since the output had to be fully DC coupled with absolute minimum drift. I can't use ground as the DC component would be too high. I tested some cascode setups with both transistors and JFET's and compared them to the op-amp stage. Both noise and drift was higher, and it was less accurate. So I continued testing op-amps that were fully stable at the high frequencies produced by the combined DAC outputs. The AD823 seems to work very well in the I/V stage.

But I just noticed another interesting op-amp on this site, the LT1469. As I need a very fast op-amp, this one just might be it.