Yup. Plus a much better pinout for Vcc/Vss for less ground/supply bounce and the actual AUC logic which is interesting in itself.
My goals are low noise across the board.
Thor
It's actually simple. I already suggested the structure.
Input, two 16 bit serial to parallel shift registers (SR) are loaded with the correct data, BCK to each SR is stopped when the data is present correctly in the SR.
Output, two 16 Bit parallel to serial SR load this data on the falling edge of LE. MSB inversion with a simple inverter in the MSB line.
This is actually similar to the jitter killer logic part in the Marantz DA-12 and my PM-75 DAC/AMP. In principle added FIFO Chip's could be inserted between the SR's (no point mostly).
The output BCK is divided from 64 X WCK IIS input BCK to now be 16 X WCK and clocks out our 16 bit during the WCK cycle.
We simply place the LE pulse derived from the WCK edge suitably using a latch and MCK/BCK from the IIS bus and use that to drive the stopped clock logic for the input SR , the transfer of the data from input SR to SR and the latching of new data into the TDA1541.
Actually, it is.
Thor
I tend to use smd only, so most it will be. Trough hole only if something is not available in smd, and caps that still have some better specs across the board in TH packaging (and here i tend to use modern tantals if it fits the requirements).
Since the issue with forum here is only single private messaging, i don't mind having a trello board started on this project (easier for start that most project management tools). PM me your email addresses, and i'll get it rolling.
Since the issue with forum here is only single private messaging, i don't mind having a trello board started on this project (easier for start that most project management tools). PM me your email addresses, and i'll get it rolling.
Well, I want (P)ECL Latches as reclockers because they add the least jitter to the MCK.
I used PECL drivers in the AMR CD-77 (MC100EPT22 IIRC) with RC and around 500mV or so PP around 1.3V with a powersupply that did the thermal compensation.
A "Raw" PECL driver has 0.8V PP on each output. This is a bit much, ideally we would like 250mV PP with excellent thermal compensation.
Also, PECL driver output is centered around ~2V DC so it's high for the TDA1541 Inputs. Finally ECL is incredibly fast, so for TDA1541 we do need to limit slew rate on the signal to ~ 10nS/V.
So some RC networks are unavoidable. Diode clipping with RC slew limiting is likely the best choice.
Thor
Using modern tools this can be minimised, having multiple pairs of eyes on the job helps spot mistakes.
Intentionally.
The Audio DIY community IME roughly breaks down into two groups (of course in reality it's a spectrum). One group are experienced electronics pros moonlighting in audio, like for example Mr. Soekris and others.
The rest buy PCBs preferred assembled, puts them into a box made somewhere else and calls it DIY.
They cannot transpose and wire a trivial simple circuit on veroboard and always ask "do you have a PCB for that?".
For the first group, Veroboard, Kapton tape and copper tape for something on the TDA1541 complexity level is fine.
For the second group, "modules may have some utility, but modules invariably mean even extremely simple and primitive things are messed up and you have to provide serious support, especially when selling modules.
So I would suggest a full system PCB similar to what is offered on e-bay/AliExpress but done right for a change.
You seem to misunderstand, I was not specifying any MCK frequency or ratio, just to have independent clocks at integer multiples of 44.1kHz and 48kHz and I guess the necessary dividers to ensure that both 512X and 1024X clocks can be fitted.
Oh my, so much magical thinking. And so few facts.
There is no difference between Amanero and XMOS if correctly set up.
In china the Amanero footprint is now used for off the shelf boards having other functions.
There are BT systems (QCC with LDAC, aptX HD and possibly even lossless) on an Amanero footprint and there are SPDIF inputs with TI, AKM and CS/Wolfson.
That is the reason why I suggested MULTIPLE footprints.
And with an Amanero footprint nobody stops you using the real thing.
But not necessarily for other input options.
Not display, just a few LED's for "connect", "signal streaming" and sample rate and input selected.
They are useful to troubleshoot a system of after plugging everything in there is silence.
Ok, on my own build I am using an OLED sample rate and format display operating of the IIS WCK. It even gets its own transformer!
You might have noticed that I actually agreed with this view.
Make a big mother board, add DAC daughter boards for the really common DAC's (PCM56, TDA1541 etc.) found in cheap old players and sell the tested PCB's, customer savages the Chips from the CDP and plugs them into the sockets, configures jumpers and voila, Bob's your Uncle and Donny POTUS.
But that is rather different from an optimised TDA1541 only design.
I'd even go as far as prescribing a CPLD or FPGA to be absolutely Universal, all the way to modules with ES9038PRO.
In some ways this "universal kit set" is more interesting to me than the TDA1541 only option.
But it should probably be a totally different thread and has little to do with my work towards a optimum extreme DIY TDA1541 DAC.
Thor
Last edited:
Yes, i do think dac should be as a stack on psu. I design stacks more often than not, as it saves space, cost, ease of soldering. Heres one example as stack of three, psu, dac and jlsounds board. This one was squeezed quite a lot, 10 regs, full dac in 10x11cm case.
Hmmm. The fundamental principle is such that if you use all 4 cycles of the DEM system and DEM clocck jitter is low enough, there are no inaccuracies, beyond the effect of the ~10...20nS glitches at each DEM switching.
With 44.1kHz you need at least 3.6uF to get the worst case ripple to 1/2 LSB and 1uF on the next two pins, so your values are not sufficient.
As you sync DEM with WCK and on top create bit pattern that WCK related, you obviously get correlated errors.
Free running creates non-correlated errors, with sufficiently high frequency for the DEM oscillator, they become noise.
Try locking to BCK. It may change your results a lot.
Correct. HOWEVER the DEM system is designed to null that error on average.
Chip's were graded based on measured performance. What causes these is debatable. With the right cap and right effective internal resistance the BCK noise induced from TTL input levels will force the oscillator into sync with BCK. I think chip's with this undocumented SYNC measure better.
This is not that difficult anymore. BUT, the point of DEM is to make this unnecessary.
Agreed. You can use C0G ceramics, in terms of leakage, stability and other parameters they are about equal to the best film capaictors.
One caveat more caveat.
The narrow glitches when the DEM system switches state are a real problem. We may assume them to be constant and induce a fixed small error, but as frequency rises the error increases and the risk of the error not being as fixed. So we want capacitors with good HF behaviour (in the 10-100MHz frequency range) and short current loops for the DEM decoupling.
Hence my take to use 4 X WCK as the lowest frequency that completes a whole DEM cycle per sample and keeps the DEM frequency low and to size the DEM capacitors such that their capacitance is sufficient to hold over (or filter out) the worst case DEM ripple to < 1/2 LSB.
Capacitor calculations here:
https://www.diyaudio.com/community/...ate-nos-dac-using-tda1541a.79452/post-7775446
Thor
If a "universal" DAC system (that can also be completed with TDA1541), kind of "Audio Lego" is desired, please open a separate thread.
As noted, I'm actually quite interested in that as a separate project.
I like Zoran's core logic a lot for that.
I can contribute to power supplies (ultra low noise, drift etc.) though I'd still recommend shunts and also an updated version of the "universal tube stage" we had at DIY-Hifi supply and stuff on specific DAC IC's I'm familiar with.
Here in Thailand I can buy dozen's of CD players with old R2R chip's for less than a decent meal out each. There is a job for a semi retired beach bum, recycle electronic junk and make custom DAC's with a nice solid hardwood base carved by Thai master craftsmen, with Toobz sticking out and all that. Go whole hog with globe 6SN7 & 300B and 5U4 rectifiers plus 300B & 6SN7 tube regulated power supplies, dual mono. Would look more like a 300B PP amp than a DAC. 555
Thor
Nah, that will be soonish as another thread, but since speaking of this thread, it was specifically targeted at tda1541a dac. Stacking dac board on top of a "motherboard" makes it easier to route things, save cost, use two tda chips instead of one if one likes to, perhaps go bal, makes it easier to stack different output stages that one may prefer. Contact resistance is miniscule. Modular analog stage per se, may be the better wording. In general modular if we include multiple chip route.
I love that concept, and if I were a designer, I would have designed something like this since and I find it very smart and so beautiful. 👍
If I were a designer, also I would ask myself if a PSU further away from the motherboard could have benefits on the SQ, but I'm not and so I ask you...
However, the pics you posted above made me want to do it, also because I've a Marantz CD-94 MK2 from which I could get the two S1, if I wanted.
It is not that hard, though layouting asks a minimum of understanding.
I followed the Zoran idea with headers to stack different Dem boards above the Tda1541a with stil smd trace for it on the bottom layer once one has chosen his préférence. Best of two world.
I was reluctant to this because distance for decoupling can be critical, and it is John Brown input to makes the 14 DEM ultra short distance for the pins. Then Thorsten said it wasn't that important: cd77 MSB oscon are perhaps 0ne inch distance, perhaps à little more. In that scénario inductance towards the ground and decoupled pins seems not critical, so stacking adds 1 cm or two (if lythics).
I followed the Zoran idea with headers to stack different Dem boards above the Tda1541a with stil smd trace for it on the bottom layer once one has chosen his préférence. Best of two world.
I was reluctant to this because distance for decoupling can be critical, and it is John Brown input to makes the 14 DEM ultra short distance for the pins. Then Thorsten said it wasn't that important: cd77 MSB oscon are perhaps 0ne inch distance, perhaps à little more. In that scénario inductance towards the ground and decoupled pins seems not critical, so stacking adds 1 cm or two (if lythics).
Last edited:
I agree.
I'm not sure I said it that way. Also, the MSB still had a Film capacitor. The added os con was a few mm distance.
This is the MK1 2-Layer PCB for CD-77.
I would not do it this way again, this was pretty much two decades ago.
The "MK2" PCB for AMR CD-77 looked a lot like John's hardwired TDA1541 with all film capacitors soldered under the IC.
My first TDA1541 was build the same, but with a copper foil plane for everything and TH Epcos stacked film caps instead of SMD options.
Thor
Thanks 🙂
Yes, this is the magic of it, toroid is custom winded by well known manufacturer, Trafomatic. Exemplar quality, and has two shields, in between the windings as blind wind, and one outside. In tests, it had no audible influence whatsoever (it was also built with toroid spaced out in a larger case, and as the beauty of modularity, spdif was included so a stack on i made i2s to RJ adapter ). I don't think off the shelf would be as good, unless it comes shielded as this one.
Tda pcb design i have in mind will be simillar, but probably sized to fit a certain case, and now there is no need to fit such a small case. I've yet to start on it, as there is still to filter out the information here, but it will be on Thorsten decouple principle he wrote about. The psu as well, as it is eerie similar to what i had already going. So many new things, so many ideas, this will definitely be one hell of a dac 🙂
Most R-Cores from China use the same structure.
My personal build uses a mild steel.case with around 1/3rd of the enclosure divided by a bulkhead. Transformers ( 2 X R-Core + Auxiliary EI core) and other "noisy" parts of the PSU will go there.
AC input with DC and EMC Filter (Ali ready made) and finally DC passing into the Audio compartment via 10nF bulkhead feed through capacitors.
Planned power sections:
TX1 50VA R
TDA1541 & Analogue (AC 70V CT -> DC 20V)
IIS2SIM & Signal conditioning (AC 12V -> DC 5V)
TX2 50VA R
SPDIF or USB Input (AC 15V CT - DC 5V)
USB Isolator (AC 12 CT -> 5V)
TX3 10VA EI
Auxiliary Circuits (AC 12V -> 5V DC)
Auxiliary are relays, OLED etc.
The final power supply sections, that is super Capacitors for all sections except aux and final regulation go on the main 20cm x 30cm Veroboard.
Thor
They where used on the power lines and we're very large value, possibly 1uF. Extremely expensive at the time. Advantage, wide bandwidth.
Today we get 1uF 0402 in X6S and can use multiple of these super cheap capacitors to make a similar feed through structure.
Some IC's with suitably placed Vcc and Vss pins literally suggest this as each Vcc pin has two Vss pins to each side.
Thor
Indeed, remember what happened in Bab-El...
Avoid making towers to reach to the heavens.
Thor