As you talk about impedance, at how much "mil" gap is the -15V layer/trace you used from the DEM caps island ? (for our better understanding)
hearing difference feeling after listening test (listening first then listening after ? Biass avoiding ?? Dunno)
hearing difference feeling after listening test (listening first then listening after ? Biass avoiding ?? Dunno)
Last edited:
Me not either (remember I have zero DB knowledge on that, so thanks for the patience) 😎
.... In order to swap the Word Salad by the World Salad ; aka Ceasar salad for the World 🙂 (and everyone knows Ceasar Salad came from NICE - pun intented-) 😍
Last edited:
@sumotan : edit I meant pin 5 and pin 15 in the post just above, sorry for that. And I still think the layout matters if you really want to ear the 100 hz (remember inductance nature of this mod) : Mayve try the kapton + copper adhesive à la Thorsten, good way to avoid first to remake a PCB.
I am not against make one à la Thorsten but I need more guidance in order to avoid a short (tda1541 chips became rare)
I am not against make one à la Thorsten but I need more guidance in order to avoid a short (tda1541 chips became rare)
The DEM cap traces are on the bottom layer , the stackup is the standart 1.6 mm one at jlcpcb
measurements with the Thorsen mod are so bad , there is nothing to listen to , waiting for some other guys here to do the same mod and see their measurements 😳
real world is cruel , this is it ......
.
measurements with the Thorsen mod are so bad , there is nothing to listen to , waiting for some other guys here to do the same mod and see their measurements 😳
real world is cruel , this is it ......
.
To enlarge, DEM circuitry is connected between AGND and -15V. We cannot really divide DEM and current sources apart.
Current source biasing and biasing for the DEM switching all reference -15V. No current can "escape" except via the DEM filter capacitors, Iout (Aol/Aor) and AGND. All currents originate from -15V (pin 15).
Connecting the external filter capacitors to AGND means all glitch currents etc. are send into this ground node and then need to pass to Pin 15 (-15V) via the -15V Power supply and decoupling.
One question remains...
How do these narrow (~10nS) glitches happen? It appears the DEM switches are BBM (Break Before Make) leaving a 10nS or so "gap" where the current source system is disconnected. The outputs are still connected to "gnd" and thus are "pulled up". The charge in the filter capacitors will counteract these glitches.
If we connect the filter capacitors to -15V (pin 15) all glitches return to the relative reference (kelvin) point for the DEM system brief current "pull-up" still need to pull current from somewhere. The current should come from -15V and will "exit" via +5V or Iout (Aol/Aor).
Ok, lets look closer.
Here is an ECL D Flip Flop or "latch":
This is the logic inside the TDA1541, more or less. Inside the TDA1541 it is called CML and does not follow classic ECL conventions for exact voltages.
It is connected between +5V and -5V, based on my best guess. DGND doesn't really much come into anything other than inputs.
The reference to -5V is needed, IMNSHO to allow the bit switch to +5V to be fully reverse biased, when the current is directed to out instead.
Here again a bit switch:
Inputs to each switch are differential, even though the switch is SE.
The currents feeding the differential pair need to come from a source lower than 0V/AGND.
Thus all CCS's in the CML come from -5V. All differentials return to +5V.
Thus the current modulation on -5V is down to parasitic, on +5V it is again parasitic and both current modulations are very low. This is inherent to CML.
Any single ended current changes that come from en/disengaging bit switches appear on +5V with the current sourced from -15V.
Any single ended current changes that come from the input section appear on +5V with the current sourced from DGND. Here MVAL's guess of the input, I mostly agree, but I think the output is differential. See my alternate take
This logic suggests that the two key current path's we really need to consider are +5V -> -15V and +5V -> DGND. Meanwhile -5V is not that important.
OF COURSE, we cannot know for sure that there is nothing else connected at -5V, so we still should treat it as analogue Pin.
So, final logic.
1) Make a isolated, separate power supply for the DAC, daisy chain two 5V and one 10V shunt regulator. No series regulators. Making a separate +/-5V "digital" supply (isolated completely) and AGND/-15V supply might be done and might further separate digital and analogue currents, especially when running DEM at low frequencies.
2) Create the necessary power/ground planes for DGND & +5V with RF / Logic style decoupling between +5V & DGND. Digital input ground connects to DGND (PIN). This loop is for high frequencies. Attenuate/slew rate limit the digital input signals. DGND should connect to the AGND in ways that only establish a linkage for DC.
3) Create the necessary power/ground planes for AGND & -15V with decoupling based on the frequencies used, -15V is the main "analogue reference" voltage, not AGND. So every current loop is closed to -15V not AGND. Including DEM Filter capacitors. As there are always fast glitches at DEM Clock frequency switching, HF style decoupling to a -15V plane under the IC should be added.
4) Decouple +5V to -15V, HF and audio band decoupling, essentially capacitors between the +5V and -15V plane under TDA1541 and large value Audio capacitors beside the TDA1541.
5) Decouple -5V to +5V, HF decoupling, capacitors between +5V and -5V.
This can be easily handled on both vero board or a 4 Layer PCB.
Make a DGND plane on the top of the PCB under the TDA1541. Connect the plane to AGND with a wide trace going around pin 1..4 on the outside. There should not be any current flowing between AGND and DGND. Place the decoupling beside the IC next to Pin 28. Decouple Pin 28 to Pin 26.
Make a AGND plane under the TDA1541 that continues out from below the TDA1541 between pin 14/pin 15 towards the power supply.
Stack a -15V plane on top of the AGND plane and run together towards the power supply. Decouple Pin 5 (AGND) the -15V plane and it may be desirable to have a second set of decoupling Capacitors at pin 15.
RF DEM decoupling capacitors under the IC, pin to -15V plane. LF decoupling capacitors (electrolytics if used) ideally Star to -15V.
Not really too complex.
Thor
That is a terrible PCB. Sorry. And all decoupling is wrong.
There is no "Thorsen mod".
I do not suggest any kind of "mod" to existing designs, but the principles to use for a clean sheet design. This is about understanding how things work, not "X says this is the way" and then randomly trying single elements out of context.
If you return the DEM filter capacitors with polar capacitors fitted to - 15V, the + of the capacitor points to DEM Pin, - to -15V. Get it wrong (like the capacitors are normally fitted) and nothing works because the capacitors become diodes.
Thor
At least if we talk about science you have to push the experiment to the listening to corelate.
Yes you right, inductance everywhere but the improvement in SQ is ear opening. lol
Its a good lesson though to think about diying a pcb to incorporate what works.
Maybe you like noise (or how it is distributed in its harmonics ?) . Always listening first and measure after is my preference (or never measure, lol ! but it is biased of course). Something is telling me that the nature of your loudspeaker (beaming highs on anarrow window) + age of both the ESL panel and your ears, makes the treble is dying very early and so you are saved from the odds harmonics at some point. Also your output traffo maybe breaks things as well ! Me I want harmonics to have the ligth slam of the midbass that needs that harmonics much higher. As sdaid Hiraga to make it simple : you bass are muddy and fat, tune too your trebles... where alas we have all this interpolation comeback (and also lower in the high mid) because we listening to NOS. But how to keep the timming structure that NOS guys seems to prefer over complex filter while the treble is not a least flat enough and rigth high enough in the ears band is certainly a concern or because phase delayed cause too much steep filter,... but dunno at the end !
Didactic point about DEM as well, also for readers like I am with very basic understanding : https://www.audialonline.com/community/topic/tda1541a-dem-clocking/
Didactic point about DEM as well, also for readers like I am with very basic understanding : https://www.audialonline.com/community/topic/tda1541a-dem-clocking/
Last edited:
Dont know bro, it just sounds really more realistic. Using Seas Fa22 now & with its high sensitivity even at loud volume
I hear no nasties. As far as PS is concern Aya's design makes it much more simple to apply Thorsten advice & ideas
I hear no nasties. As far as PS is concern Aya's design makes it much more simple to apply Thorsten advice & ideas
I don't see how, his layout is quite the opposite of what is saying T. All the current are shared on the top and bottom island between the TDA pins all tied to pin 5, the usual method. I would not risk a tda1541A trying T. mode on P. board, at least in my shoes !
What new loudspeaker ! So you have time 😛! Try my tweak, that good Sim DEM, before to damage the spare board.😎
What new loudspeaker ! So you have time 😛! Try my tweak, that good Sim DEM, before to damage the spare board.😎
Last edited:
You dont have to be sorry , my stackup works perfectly , decoupling is perfect and the all thing sound wonderful , my measurements are great for such setup ( full passive without filtering ) , waiting to see yours 😎
after all Philips engineers and many others were not that wrong 😏
If your ideas are not to complex to be made , let's do it and see 😎
.
Last edited:
The amr cd-77 was working the famous " wrong way " , what I wait for and I thing I am not the only one is the " new way " , that became the ONLY way , to be honest I thing he should have done it first and tell after , fact tells more than " word salad " like he said 😉
.
.
Last edited:
He has a proven track of a very well made device in this chip category, commercial, high end tier. I belive that his knowledge of circuit and pcb design surpasses ours here. So i'd rather pay attention and learn, rather than stroke my ego if i were you.
it is not a matter of ego , I know he knows about digital stuff , and he may be right this time too , and dont worry if there is something to learn I will , this is what I am here for too 🙂
but admit that the man is difficult to follow and the way he communicate to us is borderline
.
but admit that the man is difficult to follow and the way he communicate to us is borderline
.
@ThorstenL
I haven't read the whole thread or all your posts, but I've read quite a few of yours especially because I find your way of writing engaging and instructive at the same time.
And that's saying a lot, why did I read them with pleasure even though I only understood, let's say, 1%?
That's why putting my "Like" to your comments wouldn't too many make sense.
But I'll put some ones anyway to thank you for your dedication, your availability and your competence.
Thank you! 👍
I haven't read the whole thread or all your posts, but I've read quite a few of yours especially because I find your way of writing engaging and instructive at the same time.
And that's saying a lot, why did I read them with pleasure even though I only understood, let's say, 1%?
That's why putting my "Like" to your comments wouldn't too many make sense.
But I'll put some ones anyway to thank you for your dedication, your availability and your competence.
Thank you! 👍
It only matters that you like what you hear. Perfection in audio is a matter of opinion not of fact.
Well..... when the words matter !
(I find Thorsten very patient & tempered)
Last edited: