Building the ultimate NOS DAC using TDA1541A

DEM clock experiments

Hi omalnlk

Thanks for your reply,


The same happened to me with the 4xTDA1541 DAC. At the beginnig there were a thermal problem producing sometimes notable distortion. With a better cooling of the DACs the glitches were only on the CDs

I solved the thermal problem by using a large heatsink construction. The noise and glitches I mentioned are also picked-up by other DACs, like the twin TDA1541A DAC with 8th order Butterworth filter. Other recordings produce hardly any noise at these same high volume settings. The noise and glitches I mentioned were actually on the recording, and not caused by (DAC) HF noise.


I didn't find info about this in earlier posts. Perhaps the nearly 150 pages were to much for a one day lecture. Coul'd You post the circuit of your modules again? It's troublesome to search through the old posts.

Basically it's a synchronous divider (74HC161) that derives a 176.4 KHz DEM clock signal by dividing BCK. I use 2.8224 MHz BCK and divided it by 16.

This DEM clock signal is attenuated by two resistors placed in series (you need approx. 1Vpp). The center tap is connected to TDA1651A pin 16 trough a 470pF...1nF* polypropylene or polystyrene cap. Very important, pin 17 is decoupled to AGND using a 100nF film capacitor (polyester is OK).

* I used 470pF @ 352.8 KHz and 1nF @ 176.4 KHz

You might need to experiment with the attenuator resistor values. You could start with 680 Ohm in series with 180 Ohm (180 Ohm connects to AGND).

Today I continued experimenting with the DEM clock circuit, it seems that a 352.8 KHz DEM clock produces an even more focussed sound. Perhaps doubling the TDA1541A decoupling capacitor values (from 100nF to 220nF) would have a similar effect at 176.4 KHz, haven't tested this yet.
 
Hi tubee,

Thanks for your reply [post #1472]


Is this your newest dem-reclock explanation to be found here? Have been looking for a while in this huge thread.

http://www.diyaudio.com/forums/show...907#post1204907


Yes, but meanwhile I changed the component values, and experimented with other DEM clock frequencies.

176.4 KHz: Rdem = 680 Ohm, Rgnd = 180 Ohm, Cosc = 1nF > pin 16, Cdec = 100nF > pin17

352.8 KHz: Rdem = 680 Ohm, Rgnd = 180 Ohm, Cosc = 470pF > pin 16 , Cdec = 100nF > pin 17
 
Hi John,

With 4x oversampling (SAA7220) you need a DEM clock of 16Fs (705.6KHz), correct ? Can the chip operate with this, in a previous post you stated the DEM clock is specified between 100 and 250KHz ? I can't see any reference to the spec of the internal oscillator in the datasheet so perhaps you can help out. Based on your measurements Cosc would be around 220pF.

Also, what is the effect on increasing the decoupling capacitors. Is there a formula for calculating the values ?

Cheers,

Jon
 
DEM clock frequency / TDA1541A decoupling caps

Hi JonHarrison,


With 4x oversampling (SAA7220) you need a DEM clock of 16Fs (705.6KHz), correct ? Can the chip operate with this, in a previous post you stated the DEM clock is specified between 100 and 250KHz ? I can't see any reference to the spec of the internal oscillator in the datasheet so perhaps you can help out. Based on your measurements Cosc would be around 220pF.

I would advise to use either 176.4 or 382.8 KHz, not higher!

The DEM clock merely drives the active dividers trough a shiftregister, the binary weighted output currents of the active dividers are decoupled by a big capacitor, so the frequency of the DEM clock not necessarily relates to the oversampling factor used. It just needs to be high enough for the 100nF decoupling caps to average correctly.

However it's very important to have a DEM clock that has low jitter and runs synchronous with BCK (to avoid intermodulation with other frequencies used in the TDA1541A).

The TDA1541A typical DEM clock frequency is specified between approx. 100 and 250 KHz. When using an external DEM clock source, it's possible to use higher DEM clock rates (limited by active divider settling time). Based on both measurements at the active divider outputs (decoupling capacitor removed) and listening tests, I advise not to use DEM clock frequencies higher than 352.8 KHz.

The maximum DEM clock frequency is approx. 7 MHz (when oscillating on stray capacitance).

If you want High-End performance, I would advise not to use the standard external DEM clock capacitor as noted in the Philips datasheet, it produces massive jitter / intermodulation. This results in clearly audible sound smearing.


Also, what is the effect on increasing the decoupling capacitors. Is there a formula for calculating the values ?

The decoupling capacitors are used to filter output current ripple, caused by mismatch of the passive divider. There are high frequency components involved, so use small decoupling caps (low inductance) with low leakage current. Large audiophile capacitors that have relatively high inductance won't be able to optimally suppress the HF ripple currents.

When using 100...250 KHz DEM clock (standard DEM clock circuit) you best use 220nF decoupling caps (100nF is a bit too low). I also prefer to use higher decoupling values for MSB, I used 1uF in the DI 8. The reason for this is that MSB errors have a big impact on bit accuracy. Philips actually used different decoupling capacitor values in the TDA1540 application note, 470nF for MSB followed by 220nF, 100nF, 47nF, 22nF and a couple of 10nF capacitors. Based on this information, I would advise to use the following decoupling capacitor values:

pin 13 & 18: 1 uF
pin 12 & 19: 470 nF
other decoupling caps: 220nF

The effect of increasing decoupling capacitor value is lower ripple current on the active divider outputs, it results in lower bit error fluctuations and a cleaner more focussed perceived sound quality.


By the way, I experimented with the SAA7220P/A - TDA1541A combination. The SAA7220 is a real winner in producing high clock jitter, even when using an external low jitter clock. Did you synchronously reclocked BCK?


Cheers,

John
 
Re: DEM clock frequency / TDA1541A decoupling caps

-ecdesigns- said:
I would advise to use the following decoupling capacitor values:

pin 13 & 18: 1 uF
pin 12 & 19: 470 nF
other decoupling caps: 220nF

The effect of increasing decoupling capacitor value is lower ripple current on the active divider outputs, it results in lower bit error fluctuations and a cleaner more focussed perceived sound quality.


By the way, I experimented with the SAA7220P/A - TDA1541A combination. The SAA7220 is a real winner in producing high clock jitter, even when using an external low jitter clock. Did you synchronously reclocked BCK?

John

Unfortunately it is not allways possible to fit those bigger caps on the original pcb with modifications.

I tried a 1541 without 7220, didn't like the sound after longer hearing periods, fatigueing sound. When a SAA7220 is handled properly, with a bead and oscons/tantaliums CLC filtering in PS, and 11Mhz separate clocking from master clock to 7210/7220 and divided clock to 1541. This divider gives a cleaner clock to the dac, the 7220's clock i suppose to be jittery indeed.
It can sound reasonable good for digital OSsed 1541dac. But the digital sharp edge on treble allways persist compared to non-os, i think due to the impulse response with pre-ringing.
Btw added a small ferrite coil right after the output on 1541, this reduced sharpness some, but soundstage suffes from it too a little. Matter of right value choosing.
 
Dem Clock frequency

Hi John EC,

After our nice phonecalls I changed 2 days ago in
one configuration the dem clock frequency to 352,8 khz
using a bck of 2.822,4 khz (from your USB-I2S strip)
(74HC161 Q2 divided by 8)

I use two 1541 parallel 470 direct from Q2 to pin 16 (pin 17 0,1 to agnd)

In my other system (cd104) modified with 2 tda 1541a with a bck
of 2116,8 khz I use 264,6 KHZ derived direct from the (external)
crystal oscilator (running on 4233,6 KHZ (74HC161 Q3 divided by 16) feeding direct from Q3 to pin 16 via 470 pf (each TDA) and pin 16 via 0,1 mf to agnd. (no resistors !)
( tried 529.2 khz deviding by 8 with no good result)

Testing this configurations during two days with a lot of friends with good ears over I think this is the best one.

The to the 1541a "related papers" from Philips telling this about the Dynamic Element Matching :

Quote:

The interchanging network consists of Darlington
differential pairs which are optimized for base current
losses and interchanging frequency. The clock frequency of
the interchanging shift register is not related to the sample
frequency and is generated by a free-running emitter-ccn-
pled oscillator which operates at about 250 kHz without
affecting the accuracy.

Unquote

I think this is not true !!
So independent from me you changed also from 176.4 to 352.8
with the improved results.

(BTW: your USB-I2S strip is unbeatable !)
(BTW 2 I tested with a scope If you want photo's : tell me !)


Regards

Onno
 
Hi all,
Some pics of my DI16 nº1 ;)

my.php


You see, even my boxes are DIY :(


my.php


This is the beast...it looks like a bad guy :devilr: , with all those red R and black caps...:D
Maybe my heatsinks for the 7806 are overkill? :clown:

Cheers,
M

PS: the other picks I took are even worse...
 
Hi maxlorenz,


Here I got lost, sorry. It is not 50/50?
Playing with 64 DAC chips will not be possible (with this given buffer) then, unless we limit the max current to each chip, or add buffers to the buffers
I'm only guessing...

With 8 * 8 configuration, buffer 1 drives 8 x TDA1543, buffer 2 drives 56 x TDA1543
With 16 * 4 configuration, buffer 1 drives 4 x TDA1543, buffer 2 drives 60 x TDA1543

You could experiment using a second buffer parallel with buffer 2 or split-up BCK inputs in 2 groups, and adding an extra buffer.


Last night I listened again to DI16: it is becoming my favorite DAC. Wonderful tone and analog textures. You are right at saying that it sounds as it extracts all there is on the source. No way to test if this is true.
I think I will make an audio dedicated PC with passive cooling.

When USB interface jitter is lowered, and the 8 x 3 TDA1543 configuration is used, things can even get much better. The DI DACs are designed to retrieve every last bit of detail from CD recordings.



Please EC, tell me if I made a mistake...:angel:
(some components are on the bottom side)


It's quite an achievement that you have this DI 16 up and running. The PCB assembly looks fine too.

I have some suggestions for improvement:

It looks like you didn't insulate the RCA sockets from the chassis. Usually the RCA connectors are insulated from chassis. I used a small plastic plate that holds the RCA connectors, I made the holes in the housing just big enough to ensure reliable insulation.

There is a loose un-insulated green wire near the primary of the transformer at the bottom of the picture. I am not sure where it's connected to, but I would insulate it, just in case.

Consider adding a mains filter.

Did you place insulation washers underneath the two nuts at the bottom of the DI16C PCB? they seem to be very close to the traces.

I would mount the LM7806 heatsinks directly to the chassis (angled aluminum heatsinks), this way they use the entire chassis for cooling, and the heatsinks could be much smaller. There are TO220 insulation sets to isolate the LM7806 voltage regulators from chassis. The LM7806 heatsink is connected to GND.

The LM4562 chips probably need some small heatsinks too.


The USB receiver you used is not reclocked, I just looked up the schematics to make sure. The Tent clock is an optional 12 MHz clock (IC2) that only drives the PCM2707, it's used for USB timing, it won't have a big effect on the jitter performance of the PCM2707 PLL that generates BCK. BCK jitter would be somewhere around 500...1000ps.

So you are now only listening to a fraction of the DI 16 performance. If the jitter frequency components that degrade sound quality are lowered to 50...75 ps, it will make a huge difference.

The USB module I2S output signals are not buffered / single ended, so noise from the PC switched mode power supply, and ground loops could disrupt these I2S signals as well, adding jitter to the BCK timing signal.
 
Hi omalnlk

Thanks for your reply [post #1478],


The description of this speaker sounds really impressive. Coul'd You explain more about the acoustic-magnetic corrective element?

The correction element separates two resonator chambers, each resonator chamber has a different tuning frequency.

The interaction between both resonator chambers and the corrective action of the acoustic magnetic correction element gives the sonic resonator it's unique properties.

The acoustic-magnetic corrective element has multiple functions, it performs corrections, acts as an acoustically driven subwoofer, and transfers acoustic energy between both resonator chambers.

The acoustic-magnetic corrective element has a local feedback loop system.


Thank You for the info. Sooner or later I'll have to try one of Your DACs. Preferably on the CDPRO2

Since I started using the USBDI2S interface, I no longer see the necessity of using a CD transport. When a computer based music player is able to match or exceed direct I2S sound quality from a High-End transport, I rather use the very comfortable computer based music player. Using salling clicker remote control software on a cellphone enables one to control it with every suitable cellphone trough Bluetooth.

I added a picture of a cellphone remote control application, the CD covers are automatically transferred to the cellphone graphics screen.


It should be easy to transform the idea to 48 BCK/WS I²S Bus.

It needs a modified timing chain (different tap settings), since it still must be able to process 64 BCK/WS (USB), the timing chain needs to be dual format. Since it's not handy to keep switching between both formats manually, an auto detection circuit would be nice.

I already set up a schematic diagram for this, the auto detection circuit detects 48 BCK/WS and 64 BCK/WS without problems. I use multiplexers to toggle the tap settings.

But since I have no CDPRO2 at hand, I can't test if extra glue logic is necessary in addittion to the modified tap settings.
 

Attachments

  • cellphone1.jpg
    cellphone1.jpg
    61.8 KB · Views: 1,198
-ecdesigns- said:
Hi maxlorenz,

With 8 * 8 configuration, buffer 1 drives 8 x TDA1543, buffer 2 drives 56 x TDA1543
With 16 * 4 configuration, buffer 1 drives 4 x TDA1543, buffer 2 drives 60 x TDA1543

You could experiment using a second buffer parallel with buffer 2 or split-up BCK inputs in 2 groups, and adding an extra buffer.

When USB interface jitter is lowered, and the 8 x 3 TDA1543 configuration is used, things can even get much better. The DI DACs are designed to retrieve every last bit of detail from CD recordings.


The USB receiver you used is not reclocked, I just looked up the schematics to make sure. The Tent clock is an optional 12 MHz clock (IC2) that only drives the PCM2707, it's used for USB timing, it won't have a big effect on the jitter performance of the PCM2707 PLL that generates BCK. BCK jitter would be somewhere around 500...1000ps.

So you are now only listening to a fraction of the DI 16 performance. If the jitter frequency components that degrade sound quality are lowered to 50...75 ps, it will make a huge difference.

The USB module I2S output signals are not buffered / single ended, so noise from the PC switched mode power supply, and ground loops could disrupt these I2S signals as well, adding jitter to the BCK timing signal.

John,

I hope your comments on "huge" differences are based on comparing the two designs life and not based on trying to elevate your kits compared to my humble design based on paper study ?

Interestingly enough, the PCM2707 does respond audio wiselike on using the Tent Clock, rather than using the Crystal solution. This has been confirmed many times by people who built this design. If it sounds better, the goal is achieved by at least changing that part.

I would like to call for some more prudence on cracking my designs on expected audible outcome. You would be adding yourself to a line of people who judge the dddac on paper and drawing conclusions on expected measurements they never did, not to mention, never listened to it. Which is strange, as I very much like your approach of combining pure technic with intense listening sessions.

The ones who actually used my design are mostly very happy with the outcome. It can always be better, that is why I try to motivate people to tweak the design and get some more out of it, which lives up the hobby of course.

May be your comment on "huge" is just the way you express yourself.... I remember you talking about your earlier design sounded like a "transistor radio", which was hard to believe too ;) But not everyone would understand it in this way.

If the USB module would be that bad and jittery as you suggest and as you conclude therefore must be effecting negatively the sound, it would not be so much better compared to the best SPDIF like implementations, I have heard my self (and many others)

A solution to the buffer problem could be to use buffering at the input of the towers, as I do to enable larger quantities of towers to be used in parallel. In that case a 8 x 12 1543 implementation would be no problem for Max :angel:

keep up the good work, you do great stuff and I read your posts with more than average interest, but don't try to act as if your designs are worthy and others are absoluely not. That would do no justice and I consider this as unfair.

Doede

PS: this is the last one I got from someone who "upgraded" the PCM2707 with a Tent Clock:

WoW I have just install ,the sound improve a lot!!!! thanks for good gear......

If it is only for USB timing, how can this happen? I really don't know, but it seems to work :)
 
ECdesigns,

Thanks for your reply and nice words. Esthetically though, there is no difference if I take two days or two months to build it! :D


With 8 * 8 configuration, buffer 1 drives 8 x TDA1543, buffer 2 drives 56 x TDA1543.
You could experiment using a second buffer parallel with buffer 2 or split-up BCK inputs in 2 groups, and adding an extra buffer.

It is easy for you to say so. You know I would need help for this.
Lets begin with 8*4 chips first. :)

It looks like you didn't insulate the RCA sockets from the chassis.

They are insulated. The original photos have better resolution but are bigger and could not be uploaded.

There is a loose un-insulated green wire near the primary of the transformer at the bottom of the picture.

Good sight. This is R-Core Tx's shield wire, wich let me to the following question, about grounding. Your project has "floating" ground from chassis. Will it be an improvement in EMi immunity from attaching this shield wire to a ground chassis and using a 3 wire cable?

Consider adding a mains filter.

I allways use a "technical balanced power" filter as commercial filters degrade sound, IME.

I would mount the LM7806 heatsinks directly to the chassis (angled aluminum heatsinks), this way they use the entire chassis for cooling, and the heatsinks could be much smaller. There are TO220 insulation sets to isolate the LM7806 voltage regulators from chassis. The LM7806 heatsink is connected to GND.
The LM4562 chips probably need some small heatsinks too.

Thanks for the advices. I will see how I can manage to solve these aspects.

Dear DDDAC,
I can confirm that your USB kit sound better than SPDIF, wich is very dependent on transport quality. I will have to experiment with USB cables and PC's PSuplies to see if USB format is also influenced by these parameters.

Regards,
M
 
maxlorenz said:
Now I'm planing my audio dedicated PC. It should be relatively inexpensive.

M

M: A suitable pc for that would be a laptop. I have an old Compaq Armada E500, a free give away, ideal for music server. Repaired the USB connector, made it dual boot with win2k and ubuntu. (i am sick of XP, those yellow balloons with annoying messages...) Want more ram for it, but it uses special ram like compaq often did :(
The advantage of linux Ubuntu is you don't have to buy a special machine for it (apple), it can simply be 386-based, and runs very reliable.
Have to build a USB-I2S converter in time.

Problably you know this one allready: http://www.diyaudio.com/forums/showthread.php?s=&threadid=93315&perpage=10&pagenumber=1
 
Hi Tubee,
Thanks for your interest.
Know that I have thought about these problems quite a few times. Probably I will end with both, a laptop and a fanless PC but for now PC is cheapest as I don't trust second hand units around here.

Yes I know that thread and though I'm only a "user-level" guy I feel the beauty of Linux. My new informatic engeneer friend should help me here. :)

Regards,
M
 
maxlorenz said:
Hi Tubee,
Thanks for your interest.
Know that I have thought about these problems quite a few times. Probably I will end with both, a laptop and a fanless PC but for now PC is cheapest as I don't trust second hand units around here.

Yes I know that thread and though I'm only a "user-level" guy I feel the beauty of Linux. My new informatic engeneer friend should help me here. :)

Regards,
M

It was about 1.5 year ago when i switched over from M$ stuff to Linux Ubuntu. Just wanted to give it a try, because my newly installed win system, with all bells and wistlests: (Word,zip,adobe,pdf,scanner,network,printer,antivirus, adware,firewall)lasted only two weeks, problems simply started all over again. A new install cost me another day, why not spend this time into a new OS? Well it was the best decision i made for a long time: switch to Linux. Now i have still that old overclocked pentium 3, it flies over the net with epiphany browser. It costs some time to get into Linux, but the community is willing to help you with problems in forums.

Btw, now i know how powerful a 616Mhz PC can be: type in in terminal apt-get install... and it is working for you in a hurry, i haven't seen that ever with win. And that's the secret: the pc works for you again with Linux. (don't worry, further it all works graphical)

When you install a silent PS in the desktop, and add a big passive cooling on the processor you have a cheap and reliable music server, easy diy. Read in on the internet about different Linux distributions. When you want a little win-lookalike, Kubuntu could be the choice, but it needs a rather fast system imo.

Good luck
 
Dear Tubee,

my newly installed win system, with all bells and wistlests: (Word,zip,adobe,pdf,scanner,network,printer,antivirus, adware,firewall)lasted only two weeks, problems simply started all over again.

Believe me, i do know what you mean! :mad:

I have Ubuntu already but have only played around with it. I can't access internet throught it yet. When I 'll do I'll make it my main OS, no problem.
 
Hello ECdesigns,

yesterday I compared for a long time my tda1541 design to the tda1543 version, so I can make a decision which of your boards DI-8 / DI-16 is the better solution for me to try direct interpolation. I have 8 x TDA1541a left and so 60 pieces of the tda1543. The 1541 was always better.

With a power consumption of 50-60 mA the small IC of the tda1543 gets pretty warm in spite of massive cooling. So I put the DACs under oil. With the better cooling the TDA1543 version was under 30°C and achieved the same quality level as the tda1541. On most tracks the difference is inaudible. On some tracks it is a matter of taste but not a quality difference. tda1541 is more transparent and clear, tda1543 seems to have more powerful dynamics.

So I currently tend to the DI-16 version.

-ecdesigns- said:

The correction element separates two resonator chambers, each resonator chamber has a different tuning frequency.

I would love to hear this speaker. My active design needs four channel amplification. In tube technology this is quite place consuming and produces a lot of heat - 700W. During the summer I have to switch to other amps and don't like the sound of them (Class D). So I want to try a passive design or experiment with horn loading. So one tube amp would be sufficient.


-ecdesigns- said:

Since I started using the USBDI2S interface, I no longer see the necessity of using a CD transport. When a computer based music player is able to match or exceed direct I2S sound quality from a High-End transport, I rather use the very comfortable computer based music player. Using salling clicker remote control software on a cellphone enables one to control it with every suitable cellphone trough Bluetooth.

I already have a silent computer with no fan. The blue tooth keyboard from Logitech - search for dinovo edge - is a nice remote option too. I like the big PC screen for navigating through the music library. It will be a hard time to rip more than 700 CDs.

-ecdesigns- said:

It needs a modified timing chain (different tap settings), since it still must be able to process 64 BCK/WS (USB), the timing chain needs to be dual format. Since it's not handy to keep switching between both formats manually, an auto detection circuit would be nice.

I already set up a schematic diagram for this, the auto detection circuit detects 48 BCK/WS and 64 BCK/WS without problems. I use multiplexers to toggle the tap settings.

But since I have no CDPRO2 at hand, I can't test if extra glue logic is necessary in addittion to the modified tap settings.

The CD-PRO II should receive the best technology too. If that's not the case I can't sleep well ;-) Its not difficult to modify the timing chain. thank YOu for all the tips.

In Your design the I/V conversion is always done with a opamp. The tube stage is connected after that first opa. I want to try this first I/V conversion stage with a tube transconductance amplifier.

Is it difficult to obtain a low jitter crystal for the USB receiver board? Do You have a source? Are there other difficult to get parts?

Sad that all people interested in designing and building audio equipment are spread around the world. It were nice to know someone near vienna to exchange ideas about music, speaker, amps, dacs and turntables :cool:

Ciao
Peter