| rbroer |
| Some pictures and schematics first: |
|
|
| rbroer |
Sorry for rotation,
There is a 800 pixel width limitation and when I resize it becomes unreadable. |
|
|
| rbroer |
Basically the same I/V stage as for the TDA1541 posted earlier, just some adjusted values,
Like 2 LED's for setting the output on the TDA1543 output to aprox. 2.4V
This time it's a one board solution containing:
TDA1543 with REG101-5 smd LDO regulator,
Tent XO with REG101-5 smd LDO regulator,
74VHC374 for reclocking all I2S lines; note the double reclocking which even allows one to hook up an external asynchronous clock ;)
3 legged pre-regulator (one choosen for lowest noise, thanks Guido)
Of course lots of RC filtering, smd bypass caps (not shown on schematic, many 220u/10V Oscon's.
The TDA1543 has a combo of 1000u/10V Rubycon ZL (best bass), 220u/10V Oscon (HF filtering, creates "peace" in sound).
The I/V part uses DALE/RIKEN for Riv, and Philips KP for Riv.
The caps for filtering the TL431 references are 10u/10V Oscon this time.
The board only needs unregulated power supplies and I2S as inputs and outputs a clock signal for my cd player and the DC coupled analog outputs.
Sound:
Well, the same improvements as with the TDA1541 folded cascode I/V; there is more depth, width, microdetail in the sound.
This sounds incredibly transparent;
I recently compared this player to the passive I/V player and the cd player with single ended common base I/V and this is another level.
It's a matter of music now whether I use this TDA1543 player or the TDA1541A-S2. The latter is great for jazz and easy listening, the TDA1543 I really prefer when things have to rock.
One happy chap :D
PS
I took the pictures just before I finished the boards, so I could take 'm outside in daylight. So some parts are not there yet (like the integrator cap)
PPS
Peter Daniel, how far are u with these boards :cool: |
|
|
| Peter Daniel |
I finished the output board and I'm half way done with this board. I have a feeling that this will be the first one I'll try.
What can I use for Tent clock (if I don't have one)? I have a pretty good Raltron VCXO running at 24,704 Hz; would it work here? The other choice could be Kwak Clock, I guess, as I have one too. What are you using for smd regulators?
I see a small cap installed on servo chip. What's it for?
PS: I would've finish much earlier, but I was too busy with the amps. |
|
|
| Elso Kwak |
| quote: | Originally posted by rbroer
Basically the same I/V stage as for the TDA1541 posted earlier, just some adjusted values,
Like 2 LED's for setting the output on the TDA1543 output to aprox. 2.4V
This time it's a one board solution containing:
TDA1543 with REG101-5 smd LDO regulator,
Tent XO with REG101-5 smd LDO regulator,
74VHC374 for reclocking all I2S lines; note the double reclocking which even allows one to hook up an external asynchronous clock ;)
3 legged pre-regulator (one choosen for lowest noise, thanks Guido)
Of course lots of RC filtering, smd bypass caps (not shown on schematic, many 220u/10V Oscon's.
The TDA1543 has a combo of 1000u/10V Rubycon ZL (best bass), 220u/10V Oscon (HF filtering, creates "peace" in sound).
The I/V part uses DALE/RIKEN for Riv, and Philips KP for Riv.
The caps for filtering the TL431 references are 10u/10V Oscon this time.
The board only needs unregulated power supplies and I2S as inputs and outputs a clock signal for my cd player and the DC coupled analog outputs.
:cool: |
Hi Rudolf,
Looks nice this PCB.
Any reason why you don't use the CFB (Skizlai) in the input?
I don't see any use for current sink T5. It may present a high impedance to T6 but the IV resistor is parallel to it and is only 1k3.
In the same way I don't see any use for T11. The T6 circuit is a common base amplifier and thus presents a low input impedance.
T6 is a current source in itself. The current through T11 is already the sum of currents determined by T1 & T6. You can get a conflict of current sources and sinks in your circuit.
I don't like servos and preregulators, you already know.
You use two flip-flops in series but if you use the same clock as for the player only one is needed, as the clock is not asynchronous. Also I doubt if there is any use for reclocking in a one-box player as you don't have any data related jitter and no PLL recovered clock. My research has shown that you gain very little by using an 11.2896 MHz clock for reclocking just as reclocking the SPDIF output does not bring anything. Kusunoki the originator of the reclocking idea was using a 50 MHz clock. Still on my schedule is trying the 125 MHz clock from Vite. Sonically results are very dependent on the clock used. In my current set-up with I2S Direct I don't use any reclocker.
:cool: |
|
|
| tschrama |
Hi,
Looks very nice indeed! But I think that a CFP for the input can give a improvement. It lowers input impedance 10x.
As for the servo.. why a OPA627? You would need it's speed, it is not very quiet. Maybe the DC/drift spec are the reason? That's a very expensive servo....... :D
Greeting,
Thijs |
|
|
| rbroer |
| quote: | Originally posted by Peter Daniel
I see a small cap installed on servo chip. What's it for?
|
It's just a socket with integrated decoupling cap between 8 and 4, here it won't do anything since power is on 7 and 4, but I just happened to have only these ones :cool:
| quote: | Originally posted by Elso Kwak
|
T6 is NOT a current source Elso :bawling:
I suggest studying the circuit a bit better Elso, you're mixing up constant current sources with (folded) cascodes. :cool:
In fact, the board is layed out for CFP everywhere, but I had oscillations. Also Terry pointed out, they wouldn't sound natural. In the end I got it working with CFP, but must concur with him, somehow it doesn't sound right, natural.
About reclocking in a cd player, my experiences are quite different, it does quite a lot IMHO, especially when using a very low jitter clock like the Guido Tent XO. My preference goes to his oscillator, least veiled sound IMO, but YMMV of course ;)
It does make sense of course, a clock feeding a decoder chip will propagate through many internal gates, not improving jitter performance at all.
I've implemented synchronous reclocking in cd players 3 times now, and each time there was a very noticable improvement, tighter bass, better PRAT
Servo works in current mode; simulations indicate no THD effect, just for the sake of it I used good resistors. I've listened with and without the servo opamp, there's no difference I can notice.
The servo is there for safety, there is some drift due to temperature changes.
I had several requests for boards for the TDA1541 output stage, and told y'all to wait till I finished fine-tuning the TDA1543 board. This has finished now, so I'll probably order some new boards soon. I'll implement a few changes and make provisions for the "super-pair" this time, replacing the CFP.
Y'all got my email, closing time two weeks from now (both pcb's). |
|
|
| Elso Kwak |
| quote: | Originally posted by rbroer
T6 is NOT a current source Elso :bawling:
I suggest studying the circuit a bit better Elso, you're mixing up constant current sources with (folded) cascodes. :cool:
In fact, the board is layed out for CFP everywhere, but I had oscillations. Also Terry pointed out, they wouldn't sound natural. In the end I got it working with CFP, but must concur with him, somehow it doesn't sound right, natural.
About reclocking in a cd player, my experiences are quite different, it does quite a lot IMHO, especially when using a very low jitter clock like the Guido Tent XO. My preference goes to his oscillator, least veiled sound IMO, but YMMV of course ;)
It does make sense of course, a clock feeding a decoder chip will propagate through many internal gates, not improving jitter performance at all.
I've implemented synchronous reclocking in cd players 3 times now, and each time there was a very noticable improvement, tighter bass, better PRAT
Servo works in current mode; simulations indicate no THD effect, just for the sake of it I used good resistors. I've listened with and without the servo opamp, there's no difference I can notice.
The servo is there for safety, there is some drift due to temperature changes.
I had several requests for boards for the TDA1541 output stage, and told y'all to wait till I finished fine-tuning the TDA1543 board. This has finished now, so I'll probably order some new boards soon. I'll implement a few changes and make provisions for the "super-pair" this time, replacing the CFP.
Y'all got my email, closing time two weeks from now (both pcb's). |
I beg your pardon, Rudolf, you take me for dumb? :rolleyes:
I posted even a similar circuit with CFB input on 07-05-2003:
http://www.diyaudio.com/forums/show...1512#post201512
:devilr: |
|
|
| Elso Kwak |
Rudolf,
Replace current source Q4 in my design by a resistor.
Then Q5 and this resistor are a current source on their own. Anyway my circuit is better without Q4 I believe... |
|
|
| rbroer |
No, Elso, I don't think you're dumb :)
Here's a more readable schematic, basically the electronic circuit posted earlier.
Current sink T1 sets the bias current through T8,
Current sink T5 sets the bias current through T6,
Current source (T11) provides:
(1) current required by current sink T1
(2) current required by current sink T5
(3) zero level bias current (sinking) for TDA1543, typically 50% full scale, 1.15mA
Any "left-over" DC current can only go through the Riv.
Hence the current sinks is adjusted for zero DC flow through Riv, ergo 0 Vdc offset over it.
Since there are some thermal effects, the DC servo slowly sources or sinks a tiny adjustment current at the DAC, to keep 0 Vdc at the output, it's influence is limited to about +/-100uA
Now I'm not gonna argue with you Elso about "your designs" being better, I'm sure they are :cool:
This works great for me, never had such a great sound from my cd players. In the end, that's what I care about. :) |
|
|
| rbroer |
In the next (and final :)) version I will add the "super-pair" as suggested by JCOX. Might as well give it a try, like I did for the CFP on my current boards.
If you look at the photo, you seen some unused places for transistors and SMD resistor for the CFP.
Will post my findings in due time when I receive some new pcb's.
Happy DIY. |
|
|
| rbroer |
BTW,
If your amplifier already has an input capacitor, and you're implementing the folded cascode I/V circuit,
I suggest leaving the DC servo out and adjusting the DC level on Riv a few volts negative, to create some bias voltage for these input coupling caps. VCE(T6) increases as well, which is a good thing. |
|
|
| stefanobilliani |
| quote: | Originally posted by rbroer
Happy DIY. |
Thank you rbroer,
there is a lot of interesting stuff in your posts...:smash: :D
BTW can you explain me the abbreviations CFP and YMMV ?
Thanks
:cool: |
|
|
| Bricolo |
| quote: | Originally posted by rbroer
out +/-100uA
Now I'm not gonna argue with you Elso about "your designs" being better, I'm sure they are :cool:
|
I think that Elso was telling that his design is better without Q4 than with it. He wasn't telling his design is better than yours |
|
|
| Elso Kwak |
| quote: | Originally posted by Bricolo
I think that Elso was telling that his design is better without Q4 than with it. He wasn't telling his design is better than yours |
-Right-
:) |
|
|
| tschrama |
Hi rbroer
| quote: | | "super-pair" as suggested by JCOX |
I don't understand this configuration. Is their a ealier post somewhere that expleains it, or maybe could you elaborate a bit?
Regards,
Thijs |
|
|
| Pedja |
Comparing with the Rudolf’s previous version, actually nothing has changed but that folding is added – and both Rudolf and Vuki are reporting an improvement. What I found as an interesting difference spicing these circuits is that the current through the T11 is now more constant, with 4mA p-p signal current it varies +/-1uA instead of the previous +/-50uA. Spice also shows the harmonic distortion is 2.5 times (9dB) less in this case (all else being equal), but this last circuit can be significantly improved if runs at higher currents.
Elso, your last suggestion is really funny, I checked it too (put the plain resistor instead of the T11) and guess what… A lot of it again depends on biasing but the total distortion level went down for 2.4 times and general rule is: the second harmonic goes notably lower and the third harmonic slightly rises.
Pedja |
|
|
| Peter Daniel |
So, I have finally built it.
Had some small problems, but eventually it's working. There is a mistake in a first drawing: DATA should go to pin 7 and BCK to pin 3 of 74HCxxx chip. I inserted the board in place of my previous DAC so I could compare it directly.
My previous DAC consisted of TDA1543 powered by 5V supply (based on Elso's schematic) with asynchronous reclocking (also based on Elso's suggestion and using 100MHz crystal). Supply to reclocking circuit was separate (including transformer).
Rudolf's circuit is built without any changes except for a clock which comes from Raltron (24xxMhz). I used same raw supply for digital circuitry and also same coupling caps. At the moment I don't see a way to use the circuit without coupling caps and will try servo tomorrow. The offset oscilates around 10mV (also changes when music is paused).
My first impression is that somehow this setup sounds less harsh that my previous (passive I/V). It's almost like I placed S&B 102 transformers back in (at the moment I'm using switching volume control), very organic. There is definitely a better definition and the feeling of very relaxed and smooth presentation. The previous DAC was more sharp and edgy and while some discks sounded very good, some other were less listenable. With this one, everything sounds good. It seems like resolution is a bit lower, but OTOH I found some cues I didn't noticed before, so it's hard to say.
It's a bit more mellow, but it also draws the listener in and asks for attention, or makes you curious how the next disk will sound.
The mellow part could definitely be influenced by the parts choices, as I have many Cerafines and tantalum (parallel with Riken) at the output.
I like it.;) |
|
|
| Peter Daniel |
Few more observations as I tweaked the thing.
I tried servo and I prefer coupling caps. Servo takes away fine detail and introduces a touch of something artificial. The caps I'm using are pretty good (Siemens MKV), but without the caps it's totally different level. Everything becomes much more coherent, crystal clear and smooth (it's like changing from AC power to batteries). However the DC offset isn't stable enough to remove the caps permanently. The DC level at amp's output varied from 20mV to 300mV and more.
Today I put the whole circuit inside the Marantz player (using the player's power supplies) and I was totally disappointed. The fine detail, relaxed presentation, smoothness, everything I liked about this DAC, was gone. |
|
|
| Peter Daniel |
I tried choosing different points for taking raw supply for digtal (small improvements, but not to the point I remembered from before) . I tried using Kwack clock (instead of Raltron clock) for reclocking (didn't noticed major difference). Connected the outboard supply for digital part of a DAC, improvement but still not what it was in the beginning. I was almost ready to remove the board and use it as a separate DAC, like I had originally, when I decided to try one more option. I connected the external analog supply (for I/V stage). This supply consisted of toroid transformer, best Panasonic caps available and MUR860 diodes.
I put the music on and finally it sounded as I remembered it from previous day, maybe even better
Which leads to conclusion that analog supply in a CD player is pretty important. BTW, those caps were removed from top line Technics transport ($ 8,000) and it says for Audio on them. Those are the only caps I found so far that don't have the plastic jackets on them, but are painted black directly on the aluminum body. Would it be done this way for no reason?;)
__________________ |
|
|
| Peter Daniel |
So, in with this player I have no choice but add a separate, outboard PS for digital and analog stage of Rudol's board.
While I was testing I2S signal, this is what I observed on Fsync line coming out from SAA7210. Any idea how to fix it?
After reclocking, the signal is fine. |
|
|
| Bricolo |
| quote: | Originally posted by Peter Daniel
So, in with this player I have no choice but add a separate, outboard PS for digital and analog stage of Rudol's board.
While I was testing I2S signal, this is what I observed on Fsync line coming out from SAA7210. Any idea how to fix it?
After reclocking, the signal is fine. |
Hi Peter,
Can you explain the problem? Some parts of the curve are hardly visible |
|
|
| Bricolo |
The 1st things that comes to my mind is the probe's scope.
I remember that Elso told that to be able to measure clock signals, you have to use self made probe made of a 50R coax cable and a 50R resistor at the end (scope side) |
|
|
| Peter Daniel |
| I uploaded new image to see better. It's not a probe problem. |
|
|
| Bricolo |
OK, I had to refresh my browser's cache to see the new image.
Is the 7210's I2S also connected to something else?
Have you looked at this signal before? Without the dac or with another one (but still i2s) |
|
|
| Peter Daniel |
I would imagine that SAA7210 had to be connected to something else, otherwise it would be useless in a player;)
I'm not sure, but I think before installing Kwack clock the signal was fine. |
|
|
| Bricolo |
| quote: | Originally posted by Peter Daniel
I would imagine that SAA7210 had to be connected to something else, otherwise it would be useless in a player;)
I'm not sure, but I think before installing Kwack clock the signal was fine. |
Sure, Peter ;)
But I asked if the 7210's I2S output was connected to something else that just the DAC.
Not the entiere chip, of course. |
|
|
| Peter Daniel |
| It is connected to SAA7220, as I need digital out. |
|
|
| Bricolo |
??
The 7220 is only a digital filter. It won't give you digital out (if you mean spdif). It makes 4x (or 8x) I2S from a 1xI2S
the spdif interface uses another chip, 1825 or something like that |
|
|
| guido |
| quote: | Originally posted by Bricolo
??
The 7220 is only a digital filter. It won't give you digital out (if you mean spdif). It makes 4x (or 8x) I2S from a 1xI2S
the spdif interface uses another chip, 1825 or something like that |
Wrong,
Pen 14 gives biphase-mark code output after interpolation, attenuation and muting. So before oversampling (obvious).
Read the datasheet.
GuidoB |
|
|
| Bricolo |
OK I was wrong.
But im my players, the digital out chip is conected to the decoder, not to the filter |
|
|
| jcx |
| quote: | Originally posted by rbroer
In the next (and final :)) version I will add the "super-pair" as suggested by JCOX. Might as well give it a try, like I did for the CFP on my current boards. |
glad to see some interest in the super pair, i would also consider a super pair for t5 as this current source has noticable output V swing and its cb junction nonlinearity could compromise the signal |
|
|
| rbroer |
| quote: | Originally posted by jcx
glad to see some interest in the super pair, i would also consider a super pair for t5 as this current source has noticable output V swing and its cb junction nonlinearity could compromise the signal |
Initially I had a jfet cascode on T5.
But results from simulation showed no benefit doing this, so I left it out. I never compared with listening tests though.
I'll post my results in a couple of weeks when the new pcb's have arrived.
:) |
|
|
| Peter Daniel |
Here's update on my setup.
I managed to squeeze output caps inside player (wrapped in white tape at the back).
I cut the spaces on the board (as recommended by Rudolf) to devide digital section from analog ground plane and the improvement wasn't subtle. The feeling of being there and imaging improved.
As you see, I'm using batteries and I think it is improvement over the regualr supplies. The sound is much more coherent, more detail and better bass (also increased smoothness).
I used three 12V cells. The analog section gets the power bypassing the regulators. I plan to omitt LM 317 preregulator and connect 6V cell directly before low drop reg.
Also, I added 50ohm termination resistors (50 ohm) on the output of SAA7210 chip. According to Pete Goudreau, it was major improvement in his mods on a DAC. I also connected separate clock signal directly to the SAA7210 chip. When I listened after that change, I wasn't sure if it was an improvement. Somehow the sound lost it's immediacy and was more distant, less airy and image decreased somewhat. So I decided to introduce one more change. I disconnected the separate clock (running at 24xxxMhz and connected to reclocking circuit) and connected the main KC7 (inverted output) to reclocking circuit. Now, this made a major difference and things improved substantially.
I got the microdetail, air, space, feeling of being there, immediacy and everything became much more coherent. This seems to be the best version so far and the batteries will probably stay. When I switch back to regular power, it seems like the sound is polluted somehow.
Now I have to figure out a way to charge them in a smart way. |
|
|
| Peter Daniel |
| I tried 6V battery for reclocker and DAC, connected through resistors directly to low drop regulators. There is a small improvement (somewhat cleaner sound) but not as big as I expected. So LM317 preregulator is not such a big evil as one might think;) |
|
|
| Peter Daniel |
Here's another update to my setup.
After experimenting with more simple setups recently I decided to redo the whole thing. All the boards went off and I started with a clean plate;)
I attached copper foilt to the SAA7210 and mounted TDA1543 directly to it (pins up). This allowed me to create I2S connection to be less than 1/2". I/V stage is passive with Rikens. I attached Tent clock to the digital board as well, and the local regulators are AN800x. All bypass caps are BG N. Raw supply is taken right after capacitors (for 5V) and from 14V regulation stage (for the DAC). The regulator on DAC is 8V. Coupling caps are Siemens MKV. No reclocking.
I prefer this setup to the previous one. It seems to present less of the digital artifacts. The biggest improvement is probably made by getting rid of those OsCons;) |
|
|
| Peter Daniel |
| After attaching battery to DAC supply, there is no coming back. The battery is much better, comparing to on board supply in Marantz player. Much more of dynamics, great bass and no more glare. |
|
|
| zygibajt |
Very interesting!
Bartek |
|
|
| Peter Daniel |
It is interesting and it's simple.
You don't need PCB to do that, and sound is very satisfying. Some bypas on battery is recommended and I'm using BG NX 1000/16. Those NX caps are always sounding somewhat softer than either STD version or N, but in some cases it is actually desirable. Overall, I think, N type has the best tonal balance, but it is also too expensive. |
|
|
| zygibajt |
| quote: | | I cut the spaces on the board (as recommended by Rudolf) to devide digital section from analog ground plane and the improvement wasn't subtle. The feeling of being there and imaging improved. |
So all analog and DAC part has separate supply leaving CD-player's supply for digital and transport/LCD purposes?
Did you compare your CD 94 as a transport with CD PRO?
But you probably have to exchange coupling transformers at digital out in CD 94 to do the comparison fair,they are usually weak.
Bartek |
|
|
| Peter Daniel |
The supply to clock is taken from 5V raw supply and futher regulated by AN8005. The DAC supply is from 12V battery and regulated by AN8008. No other analog supply needed, as I/V conversion is done by 2K7 Riken resisitors. Those are the only reistors I would recommend at this point. They simply sound the best. Rref is 1k5 Riken.
I didn't compare transports yet, as my cable was too short and I didn't want to move them.;)
But I did some comparison before, and CD-Pro was better. |
|
|
|
