Yes your right, i couldn't find a good quality crystal at 11.2896, now i remember. Its been about a year and a half since I set up the FIFO...
And a FOX Expresso ? Not good enough in relation to a Crysteq ?
Finally what is the best frequency to choose for the TDA ? Does it have to be reduced after like John (ECDESIGNS) used to before gave up for an another dac chip (and come back now with new idea but with no precise inputs yet) ?
Is a FIFO the unique solution ? I read than on such a chip a simple async usb and a good clock near the TDA could be more than sufisant but than create a non tubee outputstage stay the "tour de force" (the big difficulty) !
Anyway P Rodric showed it is possible and accurate for classic music and the Cen/Sen as said to be good also.
On the paper I like too the Abraxalito way (not for the 1541 in particular) for i/v : passive, then a passive filter with smd (don't remenber the slope...), then a buffer like the AD561, but many exist (Pedja used to output with the opa861 as a buffer also in the AYA2/3), BUF03 seems good also (look at George Hifi thread...)... But I'm not worry about the output stages, many shematic and many prices exist for all the people....
Off the top of my head the FOX Expresso phase noise plot didn't look anywhere near as good as the Crystek 957, I think this is why Ian Jin updated the firmware on the FIFO because he was aware that there were no comparable crystals to the Crystek that were as easily obtainable.
Cen all the way for me, very impressive sound - ultra simple too, i like that about it.
The only thing I would change with my configuration of the CEN is parallel another matched pair of jfets to get the impedance on the 1541 down. At the moment im at about 12ohms, which is pretty darn close to exceeding the output compliance.
Last edited:
Eldam, Ryan, Shane,
in any case you have to use some cables for digital signals. I wouldn't use S/Pdif anyway and for semplicity ignore for a while the simultaneous mode. Talking about I2S you have 3 signals and 3 cables: BCK, WS and DATA.
I suggest to use exactly the same cable and the same length of each cable, but in a different way: MCLK is the same as BCK, so MCLK is placed on the dac board, as close as possible to the dac chip.
The only difference is that about master clock of the system. I prefer a slave source rather than a slave dac. This because the DAC works at its conversion frequency with a great improvement in its internal logic jitter.
I have not yet tried this configuration, since I don't own a 1541 board, so I asked Ryan if he could try this approach (in theory).
I'm not sure, but I believe the source of CD-77 is slaved from the DAC (maybe via S/Pdif). I remember I did debate with Thorsten about the MCLK and he suggested to slave the transport with S/Pdif transmitter and receiver (CS8412 and so on).
Well, if we can avoid S/Pdif using I2S, we can do that simply slaving the clock of the source from the BCK of the DAC.
In the case of the FIFO buffer to use with a 1541 board that accept I2S, I suggest:
- a good MCLK at 5.6448 MHz as the DAC BCK, placed as close as possible to the dac chip (my oscillator board measure 1" x 1.5").
- remove the reclocking board from Fifo buffer
- send back the MCLK to the main Fifo board, that will operate at 5.6448MHz (128fs). I asked Ian about this configuration and he confirmed that should work, providing 2 words of 64 bit each (16 bit of music and 48 bit of trailing zero). So the 1541 will operate in NOS at 176kHz discarding the data exceeding (trailing zero).
Ryan, in your actual system the Fifo stage operates at 22.5774MHz, while the DAC operates at 2.2884MHz, so the MCLK of the system is divided by 8 to feed the DAC chip (Dac is slaved from the Fifo). In theory each division improves the phase noise of 6dB, pratically adds some jitter due to the timing error of the active devices used to do the division.
You could surely slave the QA550, but I believe it works at 11.2896MHZ, so you could anyway divide by 2 to feed the dac chip.
Andrea
in any case you have to use some cables for digital signals. I wouldn't use S/Pdif anyway and for semplicity ignore for a while the simultaneous mode. Talking about I2S you have 3 signals and 3 cables: BCK, WS and DATA.
I suggest to use exactly the same cable and the same length of each cable, but in a different way: MCLK is the same as BCK, so MCLK is placed on the dac board, as close as possible to the dac chip.
The only difference is that about master clock of the system. I prefer a slave source rather than a slave dac. This because the DAC works at its conversion frequency with a great improvement in its internal logic jitter.
I have not yet tried this configuration, since I don't own a 1541 board, so I asked Ryan if he could try this approach (in theory).
I'm not sure, but I believe the source of CD-77 is slaved from the DAC (maybe via S/Pdif). I remember I did debate with Thorsten about the MCLK and he suggested to slave the transport with S/Pdif transmitter and receiver (CS8412 and so on).
Well, if we can avoid S/Pdif using I2S, we can do that simply slaving the clock of the source from the BCK of the DAC.
In the case of the FIFO buffer to use with a 1541 board that accept I2S, I suggest:
- a good MCLK at 5.6448 MHz as the DAC BCK, placed as close as possible to the dac chip (my oscillator board measure 1" x 1.5").
- remove the reclocking board from Fifo buffer
- send back the MCLK to the main Fifo board, that will operate at 5.6448MHz (128fs). I asked Ian about this configuration and he confirmed that should work, providing 2 words of 64 bit each (16 bit of music and 48 bit of trailing zero). So the 1541 will operate in NOS at 176kHz discarding the data exceeding (trailing zero).
Ryan, in your actual system the Fifo stage operates at 22.5774MHz, while the DAC operates at 2.2884MHz, so the MCLK of the system is divided by 8 to feed the DAC chip (Dac is slaved from the Fifo). In theory each division improves the phase noise of 6dB, pratically adds some jitter due to the timing error of the active devices used to do the division.
You could surely slave the QA550, but I believe it works at 11.2896MHZ, so you could anyway divide by 2 to feed the dac chip.
Andrea
Hi,
I really hope two project can suceed : AYA GB, if people fellows can help each others (and if P. Rodric agree) will allow for a maximum of people (as this the goal of the original question "Any good kit today") to have enough quality. A not brand new design but a kit used to be sold and reprint will give a good result for maximum people with minimum risk and level with a owner manuel assembly. CFT is doing the work but ha already top material at home. People may thanks to him.
On the more fanatic side: for the moment SET3up is working to something.
Here people are merging to a I2S DAC board with a good quality :uFL connector. It seems a need exist. I can just understand the half oy what you TALK ABOUT (Am I not the good willed Mr Mustach Chewbacca ?) but let's go if a pcb for good diyer can exist and can to be print of a web plateform or be edited in a small GB : it's fantastic also for people to have a good fundation for the tweaks around (source, output) with a SOTA or almost SOTA coreboard (good layout, goos PS, output connection for output daughter board or stacking board; UFL connector to improve quality of I2S connection).
A question to Andrea : if a four pad emplacement near the DAC for the crystal, does it need too a serie smd space for a smd resistor (resonances) or smd ferrite bead (also resistor attenuator if needed) before the dac clock lead ? It need also a powersuply so all the layout for the good ldo 5 leads reg (3.3 V for the Crysteq ?) and the smd caps and/or inductors/ferrite beads around ?
I really hope two project can suceed : AYA GB, if people fellows can help each others (and if P. Rodric agree) will allow for a maximum of people (as this the goal of the original question "Any good kit today") to have enough quality. A not brand new design but a kit used to be sold and reprint will give a good result for maximum people with minimum risk and level with a owner manuel assembly. CFT is doing the work but ha already top material at home. People may thanks to him.
On the more fanatic side: for the moment SET3up is working to something.
Here people are merging to a I2S DAC board with a good quality :uFL connector. It seems a need exist. I can just understand the half oy what you TALK ABOUT (Am I not the good willed Mr Mustach Chewbacca ?) but let's go if a pcb for good diyer can exist and can to be print of a web plateform or be edited in a small GB : it's fantastic also for people to have a good fundation for the tweaks around (source, output) with a SOTA or almost SOTA coreboard (good layout, goos PS, output connection for output daughter board or stacking board; UFL connector to improve quality of I2S connection).
A question to Andrea : if a four pad emplacement near the DAC for the crystal, does it need too a serie smd space for a smd resistor (resonances) or smd ferrite bead (also resistor attenuator if needed) before the dac clock lead ? It need also a powersuply so all the layout for the good ldo 5 leads reg (3.3 V for the Crysteq ?) and the smd caps and/or inductors/ferrite beads around ?
Maybe the solution is much simpler: if the board provides 3 u.fl connector (SMD) everyone is free to use the BCK coming from the source as well the MCLK coming from the master oscillator (4 u.fl cnnector if the simultaneous mode should be an option).
The oscillator board and its power supply could be placed inside a metal box (to shield from RF interference and from moving air), providing 2 SMA connectors, one for the dac chip and the second for the source/ASRC.
About signal attenuation we have the same problem as taking the BCK from the source.
P.S. In reality I own a 1541 board (the Red Baron), but no time to build it.
Andrea
But a good crystal needs to be near the chip or it's a waste if its qualitie is high like the crystech or Laptech ? If a connector near the dac it is hard to make very short, does such wires exist or are esay to do by a diyer (it's very small ! solding the plug on a pcb ok, but the jacks for the wire... Huuuuuurk !).
Maybe be more easy to put a u.fl connector + the fours leads for the crystal also and a through hole for the PS. People can use uf.l connector if external source clock or put an embeded crystal on the four leads (if such a design is possible to draw !) and put its ps on a little board as you described !
2 cents don't know if it is possible... believed the first picure of RyanJ was something near about a crystal close to the dac ?! i have to look at it again. But for sure the last little board he drawed is elegant...
Maybe be more easy to put a u.fl connector + the fours leads for the crystal also and a through hole for the PS. People can use uf.l connector if external source clock or put an embeded crystal on the four leads (if such a design is possible to draw !) and put its ps on a little board as you described !
2 cents don't know if it is possible... believed the first picure of RyanJ was something near about a crystal close to the dac ?! i have to look at it again. But for sure the last little board he drawed is elegant...
Right, but we are talking about an open solution with several options, so an u.fl smd socket and a short u.fl cable should be a good compromise. Obviously the better implementation should be placing the oscillator as close as possible to the dac chip. In that case we need a 5 pin connector close to the BCK pin of the dac: clock ground, clock signal, PS1 (15V for the sinewave oscillator), PS2 (5V for the squarer), PS ground (the oscillator needs 2 PS rail). Keep in mind we are talking about an oscillator, not a simple crystal. And the PCB I own accomodates the oscillator only, not its power supply, so the total space required on the board is large, while maybe some members won't use an external clock and a slave source.
As I said several times, I would think to a 1541 machine rather than a dac, in which case would be imperative to place the master clock and its power supply inside the main board. But this would be a much more challenging project, and seems that no one is willing to follow this idea.
BTW, if anyone is interested, I'm available to share the oscillator and to ask Laptech for a quote.
Sounds good. Well it's good to go for now! So how about you set up the gerber files as outlined in post 904 and we'll order up some PCBs through one of those shops!
Yeah I'd try it too.. I might have have missed it but what was the price on the clocks.
Andrea, I agree that a simple core board that suits most people is a good idea.
Eldam, Mustach?.. you're not kidding. I'll open a Wookie Waxing parlour.. until then, please fix the hyper drive, Orlando is waiting..
Regards,
Shane
ryanj, digital is 90% fine, as best as it could be. Just don't add another pads for thru-hole pin header. They will ruin all the effort made with u.fl.
1. Add another u.fl for 4th "unused" tda's pin - so you could connect it for simultaneous input, optional and won't harm.
2. Calculate the trace impedance - use "coplanar wave" mode, find the track width and track-gnd spacing suitable to get trace equal to u.fl impedance. I've assumed u.fl are 50ohm.
Saturn PCB Design - PCB Via Current | PCB Trace Width | Differential Pair Calculator | PCB Impedance
for 1.64mm PCB and 8mil track-gnd spacing i get 50mil track width. Tad too much.
Would be great if you'll go 1.2mm PCB thinckness in production. They are the same $$ (at least at chinese proto vendors i'm aware of), and provide lower track and via impedances.
40mil for 1.2mm PCB and 7mil spacing.
3. You might leave "mode selector" floating, so the end user will connect it to +5, -5, gnd based on required TDA's digital input mode in particular configuration.
If you doesn't leave it floating, then you can disregard the "1.". Most users will need the regular I2S only.
1. Add another u.fl for 4th "unused" tda's pin - so you could connect it for simultaneous input, optional and won't harm.
2. Calculate the trace impedance - use "coplanar wave" mode, find the track width and track-gnd spacing suitable to get trace equal to u.fl impedance. I've assumed u.fl are 50ohm.
Saturn PCB Design - PCB Via Current | PCB Trace Width | Differential Pair Calculator | PCB Impedance
for 1.64mm PCB and 8mil track-gnd spacing i get 50mil track width. Tad too much.
Would be great if you'll go 1.2mm PCB thinckness in production. They are the same $$ (at least at chinese proto vendors i'm aware of), and provide lower track and via impedances.
40mil for 1.2mm PCB and 7mil spacing.
3. You might leave "mode selector" floating, so the end user will connect it to +5, -5, gnd based on required TDA's digital input mode in particular configuration.
If you doesn't leave it floating, then you can disregard the "1.". Most users will need the regular I2S only.
Yes, please allow for the possibility of simultaneous mode. Then this could perhaps be "the" board for folks who have Ian's stuff and want to try the 1541A at its best.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.