I did not use any 3.3→5V level conversion strategy.
My logic circuit (74HC164) is directly powered by 5V
If "3.3 to 5V level conversion" is necessary, please give me a solution, thank you
@CNOracle
No level translation for "HCT". If AD1862 at 192 kHz is working on my PCB and not on yours, and 96 kHz works on both, problem can be in the PCB routing. I see a "broken" GND under some of data lines on your PCB. Compare it with my PCB which has solid GND layer and not crossing data lines (if you need crossing some data lines - use resistor as a bridge over another data line even if it is a compromise). Solid GND (and not crossing data lines) is very important for high speeds like 192 kHz (BCK here is 12.288 MHz, in comparison with 48k where the BCK is 3.072 MHz). This can produce bad timing and data not aligned. Maybe my PCB looks very basic and simple (due to THT components), but it is just simple enough and not simplier. You can follow the proven (or use 4-layer PCB with 2 solid GNDs for more complicated routing) 😎
No level translation for "HCT". If AD1862 at 192 kHz is working on my PCB and not on yours, and 96 kHz works on both, problem can be in the PCB routing. I see a "broken" GND under some of data lines on your PCB. Compare it with my PCB which has solid GND layer and not crossing data lines (if you need crossing some data lines - use resistor as a bridge over another data line even if it is a compromise). Solid GND (and not crossing data lines) is very important for high speeds like 192 kHz (BCK here is 12.288 MHz, in comparison with 48k where the BCK is 3.072 MHz). This can produce bad timing and data not aligned. Maybe my PCB looks very basic and simple (due to THT components), but it is just simple enough and not simplier. You can follow the proven (or use 4-layer PCB with 2 solid GNDs for more complicated routing) 😎
Last edited:
Hi Miro,
I think the last part of your last post refers to my 4 layers and 4 layers the genral way. Some thougths :
- Yes your board is fine and measure good. It is enough measurement for 16 bits. it is well made for a 2 layers. But it can be enhanced.
- what matters is: not interrupted below the data lines. With 4 layers you reduce the loops of the signals in the heigth plan.
- 4 layers : you avoid the +5V current cross 3 data lines like on the Rev 1.3 board the J3/J4 jumpers or J9. you avoid two dataline crossing with the 22R resistor on R1 R2 R3 that also cross a power supply track.
-4 layers : you avoid different impedance matching on the data line that have sometimes on the V 1.3 close side ground on some length aera, sometimes not. The whole has to be computed. The signal integrity is better and crosstalk enhanced.
Your advices are perfectly fine about CNOracle. I think CNOracle should not play with the "moats" without knowing exactly what he want to achieve, can be worse otherwise.
But It can not achieve what more layers are adding and confort of options. IMHO. Is it worthing the cost more and the long time spent to test, protos, chase the little tricky details. ? It is up to individual expectations imo.
Don't get me wrong, your design is okay for a 2 layers. It sounds good, I tested several. It is generous too. We disagree on very few things only (I will not have designed the 78/79 reg that way. I'd like you have tested the reg gerbers you generously sent me for a reg I have sligthy enhanced (you designed for the pre of your friend). I could not did a better two layers. Great project. Thanks again for that and all the help you provids with the Eagles files.
the excellent result of your project pushed my motivation to try to do better with long time spent and 4 layers and more involved monney and testing.
A journey I begunn long ago with Painkiller 4 layers boards he sent to me.
Thanks again for your marvelous project, it is sounding very better than all that Ebay/Ali designs and it is free : Bravo. My strong words and big mouth don't prevent my great admiration about you and this marvelous thread. Critics when existing not to said it is bad, just motivated to go beyond, giving tips. I strongly people to try the 2 layers if not knowing what the AD1862 can achieve and only spend few monney more for a 4 layers if wanting only a voiced project by a fanatic too ! 🙂. If the pcb is a part of the ounding result, selected BOM is the third part, the last being the front end and DAC chips.
I made a thread, not to promote anymore my 4 layers I am proud about, not to waste the spirit of the thread for the sensible moralist lurkers and bad tongues whom certainly don't know about all the penibility to post a design and prefer a beer or a cigarett packet or spent a lot of monney in cheap commercial Ebay/AAli projects made with no soul and listening Demning wheel involving. So does it worths to spent a liitle more ? some found it is worthing the small monney added that will not cover all what i spent too ! 🙂. Thanks the board found few homes already.
Sorry for the long post, it was just to sligthy balance your last post and my previous ones.
warm regards,
cheers.
I think the last part of your last post refers to my 4 layers and 4 layers the genral way. Some thougths :
- Yes your board is fine and measure good. It is enough measurement for 16 bits. it is well made for a 2 layers. But it can be enhanced.
- what matters is: not interrupted below the data lines. With 4 layers you reduce the loops of the signals in the heigth plan.
- 4 layers : you avoid the +5V current cross 3 data lines like on the Rev 1.3 board the J3/J4 jumpers or J9. you avoid two dataline crossing with the 22R resistor on R1 R2 R3 that also cross a power supply track.
-4 layers : you avoid different impedance matching on the data line that have sometimes on the V 1.3 close side ground on some length aera, sometimes not. The whole has to be computed. The signal integrity is better and crosstalk enhanced.
- 4 layers : you avoid some power line to cross each others : a little everywhere on the V1.3
- 4 layers you avoid the feedback loop of the I/V to be crossed by a power line
- 4 layers : you can figth better stray capacitance & inductance and have a better layout for oscillating speedy op amps.
- the THT components are not the problem, though you have no choice with 2 layers for the current path. With a 4 layers you can play with the layers paths through the PTH. SMD are more critical to reduce the RF by shorting the indudtance of close ICs decoupling.
- 4 layers is not making more complicating routing, it makes it easier.
- I am more about to use the 4 layers for "ground/Vref for the same reason of routing, surface copper enhancing with stitching. Also it permits to play on heigth plan to make a thicker pcb core and two thinner prepreg.
- 4 layers permits to make a pcb that can be compatible AD811, opa861, all fast VFAs w/o socketing, and all sort of tricks that can't be seen.
- better output to plug third party I/V boards ot takes a close current output near the dac chip for tubes lovers.
- adding external pad to try non unity gain excellent op amps and provide external fine compensation
Your advices are perfectly fine about CNOracle. I think CNOracle should not play with the "moats" without knowing exactly what he want to achieve, can be worse otherwise.
But It can not achieve what more layers are adding and confort of options. IMHO. Is it worthing the cost more and the long time spent to test, protos, chase the little tricky details. ? It is up to individual expectations imo.
Don't get me wrong, your design is okay for a 2 layers. It sounds good, I tested several. It is generous too. We disagree on very few things only (I will not have designed the 78/79 reg that way. I'd like you have tested the reg gerbers you generously sent me for a reg I have sligthy enhanced (you designed for the pre of your friend). I could not did a better two layers. Great project. Thanks again for that and all the help you provids with the Eagles files.
the excellent result of your project pushed my motivation to try to do better with long time spent and 4 layers and more involved monney and testing.
A journey I begunn long ago with Painkiller 4 layers boards he sent to me.
Thanks again for your marvelous project, it is sounding very better than all that Ebay/Ali designs and it is free : Bravo. My strong words and big mouth don't prevent my great admiration about you and this marvelous thread. Critics when existing not to said it is bad, just motivated to go beyond, giving tips. I strongly people to try the 2 layers if not knowing what the AD1862 can achieve and only spend few monney more for a 4 layers if wanting only a voiced project by a fanatic too ! 🙂. If the pcb is a part of the ounding result, selected BOM is the third part, the last being the front end and DAC chips.
I made a thread, not to promote anymore my 4 layers I am proud about, not to waste the spirit of the thread for the sensible moralist lurkers and bad tongues whom certainly don't know about all the penibility to post a design and prefer a beer or a cigarett packet or spent a lot of monney in cheap commercial Ebay/AAli projects made with no soul and listening Demning wheel involving. So does it worths to spent a liitle more ? some found it is worthing the small monney added that will not cover all what i spent too ! 🙂. Thanks the board found few homes already.
Sorry for the long post, it was just to sligthy balance your last post and my previous ones.
warm regards,
cheers.
Last edited:
@diyiggy 4-layers can be better, but 2-layers is much more DIY friendly and it is what you see == what you get 🙂
Note that thicker core dielectric in 2-layer boards creates greater spacing between signal and return paths, thus no capacitive crosstalk. Inner layers in 4-L are often very close together due to thin prepregs, which increases stray capacitance and this can be problematic for the I/V. I have experience, that internal power planes can bring switching noise into analog path. Here I learned, that 4-L is not often better for DAC design (mixed analog-digital). But if you tested it and it works great, than it is fine 😉
Note that thicker core dielectric in 2-layer boards creates greater spacing between signal and return paths, thus no capacitive crosstalk. Inner layers in 4-L are often very close together due to thin prepregs, which increases stray capacitance and this can be problematic for the I/V. I have experience, that internal power planes can bring switching noise into analog path. Here I learned, that 4-L is not often better for DAC design (mixed analog-digital). But if you tested it and it works great, than it is fine 😉
I know that. Mine measure quieter than your's, it's all about the design. I don't use power planes with 4 layers on the AD1862, it will need more layers to add quietness by their use. But i don't care about the measurement only. the numbers are good enough with mine and your's either.
More complex only makes sense if it is paying off with the sound. And damn we use 16 bits mostly 🙂 ! I am happy of the result yes : whatever it is NOS, it is not fatiguing? I don't experience harming digital alliasing fatigue. Sure it is there, but I can't hear it and an activee filterr will be wrong after the op amps, I am not fan also of the passive one made with ceramic whatever it is above passband and COG, I can hear it not for the best.
I disagree, it is much better to have a close ground and it is one of the good practice. Think inductance too. When you have two 1 oz and two 1/2 oz you have more surface V ref too. How a two layers with two oz of copper could weigth more than a 3 oz on 4 layets?
It allows more direct current paths avoid power crosstalk by putting ground layers in between, and so on. And yes we are in the RF territorry 🙂
I had no problem for the I/V, But I use stray capacitance proofing there, you dopn't use and which can be harmfull too for a two layer according the op amp chose, which migth be btw not always important howver (is a lot for the AD811 for instance).
cheers.
More complex only makes sense if it is paying off with the sound. And damn we use 16 bits mostly 🙂 ! I am happy of the result yes : whatever it is NOS, it is not fatiguing? I don't experience harming digital alliasing fatigue. Sure it is there, but I can't hear it and an activee filterr will be wrong after the op amps, I am not fan also of the passive one made with ceramic whatever it is above passband and COG, I can hear it not for the best.
I disagree, it is much better to have a close ground and it is one of the good practice. Think inductance too. When you have two 1 oz and two 1/2 oz you have more surface V ref too. How a two layers with two oz of copper could weigth more than a 3 oz on 4 layets?
It allows more direct current paths avoid power crosstalk by putting ground layers in between, and so on. And yes we are in the RF territorry 🙂
I had no problem for the I/V, But I use stray capacitance proofing there, you dopn't use and which can be harmfull too for a two layer according the op amp chose, which migth be btw not always important howver (is a lot for the AD811 for instance).
cheers.
Last edited:
Try to mix not them too much or just where it belongs. thanks for us the AD1862 pins layout is really clever and helps a lot, whatever two mono dac chips makes things more complex.
You can't avoid a little of HF noise spreading on the Vrefs, but notice is very low power signals and you reduce that by design thanks to more layers and behavior of the return currents too according the frequency. your digital power referenced to analog has not the ssame behavior than the I2S inputs lines.
See the whole as an Analog unique ground (which it is) and fix high RF return path and high frequency where it belongs only.
The AD1862 is easier than some dac chips ! Oh the TDA1541A... what a beach about that ! 😉 Another dimension... (don't make like gabster wrong Pedja's copy by using huge MKP cap with added inductance 😉 )
Last edited:
I solved the tearing problem I mentioned about the 192k sound, the reason was that I used a DIP16pin IC socket
After I removed them, my system worked fine. Thank you 🙂
After I removed them, my system worked fine. Thank you 🙂
I saw some information somewhere that PSU1 has a problem with stability at +/-12v and that you need to change the 4700uf capacitors to 100-220uf. Now I can't find where I saw this information. Can anyone confirm this?
Hi,
It migth happens or not according the circuitry. Those low output capacitor value was often between 10 to 200 uF because of their bad esr to make the output of the reg less inductive when the load needs fast modulated current... iirc. Problem is the output impedance is not flat and low enough in the ears pass band.
And also for cost reason you could see in CD players for instance that the input of the reg was directly feed by the smoothing cap just after the rectification bridge...
So one can want to add a big capacitance at the output to lowisch the output impedance in the pass band aera. But growing the capacitance is reducing the esr making the output of the reg more inductive.
To try to have the best of two worlds by raising the capacitance, you need to add a resistance between the reg output and the big capacitance. Say something like half an ohm or even bigger 1 to 2 ohms. The bigger the capacitance the lower the output impedance in the passband towards the bass register (while it will still inductive in the bass aera). so 4700 uF and better 10 000 uF is fine and a good idea with an added resitor between it and the output of the 78xx/79xx regs family and a lot of older reg ic. If you have still a stabilitty issue at high frequencies, according the circuitry, add a little MKT value or mkp >= 100 nF grossly.
You may experience a better sonic result.
It migth happens or not according the circuitry. Those low output capacitor value was often between 10 to 200 uF because of their bad esr to make the output of the reg less inductive when the load needs fast modulated current... iirc. Problem is the output impedance is not flat and low enough in the ears pass band.
And also for cost reason you could see in CD players for instance that the input of the reg was directly feed by the smoothing cap just after the rectification bridge...
So one can want to add a big capacitance at the output to lowisch the output impedance in the pass band aera. But growing the capacitance is reducing the esr making the output of the reg more inductive.
To try to have the best of two worlds by raising the capacitance, you need to add a resistance between the reg output and the big capacitance. Say something like half an ohm or even bigger 1 to 2 ohms. The bigger the capacitance the lower the output impedance in the passband towards the bass register (while it will still inductive in the bass aera). so 4700 uF and better 10 000 uF is fine and a good idea with an added resitor between it and the output of the 78xx/79xx regs family and a lot of older reg ic. If you have still a stabilitty issue at high frequencies, according the circuitry, add a little MKT value or mkp >= 100 nF grossly.
You may experience a better sonic result.
I have built a few PSU1, didn't encounter any issues. The LM78xx regulators should not have such problems.
Migth be or not according the circuitry but I don't think it is stability . However try this old diyer trick tip you migth be surprised. To do so, make a cut in the trace between the reg output and the capacitor in front of it and add this resistor. If not happy it is reversible by resoldering a lead to remake the bridge. Everyone is free to try or not. I force noones. Just my opinion.
Now does it make sense ? the reg is cheap and good if you do so, but raising the output capacitor increase the cost too... It has to be weigthed according the Q/P you want to involve VS a more expensive regulator.
Now does it make sense ? the reg is cheap and good if you do so, but raising the output capacitor increase the cost too... It has to be weigthed according the Q/P you want to involve VS a more expensive regulator.
Last edited:
This is information from the guy I bought the DAC from. His measurements look like the pictures. 4700uf and 100uf. Out of curiosity, I replaced the capacitors on the +12v and -12v rails from 4700uf 25v Panasonic FC to 100uf 50v Nichicon Muse KZ (that's what I had on hand) and by ear it sounds better. Is that possible? I can't measure it reliably. I have a simple pocket oscilloscope but I can't tell the difference with it.
Attachments
It's possible, if not likely, it sounds better even if there is no stability issue in the reg.
Big, low esr caps with inductive wires to the DAC board in parallel with small decoupling caps will ring. Physically big caps will pick up more EMI/RFI.
If you use big caps on sensitive points (e.g vref or low psrr circuit) you will notice how much they can colour and smear the sound, electrolytic caps themselves are not free from non-linearity
78xx regs are often said to tolerate any output capacitance, but I would like to see that actually demonstrated.
FC is low esr lytic and 4700uf is an unusually large (and correspondingly low esr) value to have on the output of one of these, so we are definitely dealing with a fringe case.
Big, low esr caps with inductive wires to the DAC board in parallel with small decoupling caps will ring. Physically big caps will pick up more EMI/RFI.
If you use big caps on sensitive points (e.g vref or low psrr circuit) you will notice how much they can colour and smear the sound, electrolytic caps themselves are not free from non-linearity
78xx regs are often said to tolerate any output capacitance, but I would like to see that actually demonstrated.
FC is low esr lytic and 4700uf is an unusually large (and correspondingly low esr) value to have on the output of one of these, so we are definitely dealing with a fringe case.
Last edited:
That can be a problem too. The datasheet specifies min logic Hi level at 0.7*Vdd, which is 3.5v at 5v supply.
3.3v is close to 3.5 but still below it. It may work. Or may not in certain conditions. It is not a good practice to leave it like that.
There are level translation ICs to fix that, or if you are going to use galvanic isolation IC - just power the clean side with 5V.