Continuing, clock sch and iron i/v
Hires: Was concurrent with K&K about the LL1674 for i/v a few years back...I disagree with them loading the output side, sounds constricted to me. I prefer setting the load at the primary, and letting the secondary "rip" into as high an impedance as possible, like 100k and up. It remains to be seen if the Sabre has sufficiently low residual offset for this, if so I will try a differential i/v primary "load" from 50 ohms up to wherever , depending on the sonics.
Rossl: hmmm, not sure a kilobuck to hear if the mil spec oscs are better is exactly, er, cost effective DIY...
Back to affordable for the masses, the Crystek modules may make discrete oscs non-relevant, particularly above 20 MHz or so. There is that fine line between lazy and efficient. If somebody else has done the work sufficiently, why try to re-invent the wheel, unless it's fun.
Major discussion of a discrete osc probably could be moved to a new thread about "yet another clock", or something like that, but here is cliff notes:
Key point is not so what, but how.
Layout keeping wires, traces, loops small. Try a copper foil groundplane. Dedicated power supply, separate power transformer, rect, pre-reg not shown in the sch to keep grounds cleaner. Fine tuning the caps A and B to match the particular strays in the devices and layout. Typ 47 to 100 picofarads. Use vrefs as the reg for the fets, such as ref01, lt1021, lt1027. Depending on the comparitor used, maybe a power reg for the comp. Not shown in sch, bypass refs/regs with low esr, small caps, up to 30uF or so. An lt 1394 can give 5v differential clock out. The lt 1719 can be 3.3v single out. An lt1116 fed 7v can give 5v diff out, but takes more supply current. High Q, low esr fundamental xtal. Ground the case of the xtal. Try powering the input stage of the comparitor off the fet power/ground rather than the comp power/ground. They might sound different, one might sound better in a paticular implementation. Match Idss of the "bottom" fets, to keep the source offsets in circuit under a few millivolts of so at operating temp. Yes, it can be done. The cascode top fets are higher Idss to function as cascodes and are optional. The fets can be low noise, high gain such as J309, but decent results can be had with garden variety fets, the 5484/5486 were for sims of the cascode, and do not represent the views of the network or it's sponsors. Metal film 10 meg gate resistors. Optionally try wirewound source resistors, using the inductance as part of the tank circuit, tuned with caps A and B. Mount the xtal on sorbothane. Probably something else I will remember after posting...
Hires: Was concurrent with K&K about the LL1674 for i/v a few years back...I disagree with them loading the output side, sounds constricted to me. I prefer setting the load at the primary, and letting the secondary "rip" into as high an impedance as possible, like 100k and up. It remains to be seen if the Sabre has sufficiently low residual offset for this, if so I will try a differential i/v primary "load" from 50 ohms up to wherever , depending on the sonics.
Rossl: hmmm, not sure a kilobuck to hear if the mil spec oscs are better is exactly, er, cost effective DIY...
Back to affordable for the masses, the Crystek modules may make discrete oscs non-relevant, particularly above 20 MHz or so. There is that fine line between lazy and efficient. If somebody else has done the work sufficiently, why try to re-invent the wheel, unless it's fun.
Major discussion of a discrete osc probably could be moved to a new thread about "yet another clock", or something like that, but here is cliff notes:
Key point is not so what, but how.
Layout keeping wires, traces, loops small. Try a copper foil groundplane. Dedicated power supply, separate power transformer, rect, pre-reg not shown in the sch to keep grounds cleaner. Fine tuning the caps A and B to match the particular strays in the devices and layout. Typ 47 to 100 picofarads. Use vrefs as the reg for the fets, such as ref01, lt1021, lt1027. Depending on the comparitor used, maybe a power reg for the comp. Not shown in sch, bypass refs/regs with low esr, small caps, up to 30uF or so. An lt 1394 can give 5v differential clock out. The lt 1719 can be 3.3v single out. An lt1116 fed 7v can give 5v diff out, but takes more supply current. High Q, low esr fundamental xtal. Ground the case of the xtal. Try powering the input stage of the comparitor off the fet power/ground rather than the comp power/ground. They might sound different, one might sound better in a paticular implementation. Match Idss of the "bottom" fets, to keep the source offsets in circuit under a few millivolts of so at operating temp. Yes, it can be done. The cascode top fets are higher Idss to function as cascodes and are optional. The fets can be low noise, high gain such as J309, but decent results can be had with garden variety fets, the 5484/5486 were for sims of the cascode, and do not represent the views of the network or it's sponsors. Metal film 10 meg gate resistors. Optionally try wirewound source resistors, using the inductance as part of the tank circuit, tuned with caps A and B. Mount the xtal on sorbothane. Probably something else I will remember after posting...
Attachments
I don't think its possible. The high res audio spec (DVD-A, SACD) are made to not allow output through anything but Firewire, or HDMI unless they are internally decoded and outputted through analog connections. At least I know this is true for SACD and pretty certain its true for DVD-A as well.
MrMajestic said:
Ah yes, thats what I was thinking. Just wasnt sure if it would work. Stick a TP spdif module in there to get out and I might be in business
May not work. I have just tried to connect the WM8804 Twisted Pear board to the AK 4351 dac pins. Measured all the outputs and they seem fine, but the WM will not lock on.
Go to the Audiopraise website and look at the Vanity module wghich works flawlessly. Only install in a player with good quality PCBs with viable solder pads. Some DVD players pcbs are crap and the tracks will break if you are not careful in doldering/desoldering.
MrMajestic said:
Yeah, Ive seen the Vanity module, but its a "tad" more expensive
This one ?
http://audiopraise.com/vanity/overview.php
Hmm, let's make a parts cost
- XC3S100E : € 10 (unit price, I don't have the price by 1000)
- 3 regulators : € 2
- caps : € 2
- Flash : € 1.5
- other parts : € 3
- PCB : € 2
Total € 20.5
Add free Xilinx tool for instatiating FIR filter.
Resale price 345€, lol.
Perhaps I should design & market such a module, lol.
And it transmits over coax and not something reasonable like twisted pair... argh.
Just pick up I²S and DSD in your player to send to the Sabre, you need 6 high speed Optos for isolation...
LOOOOOL
Parts cost $20 means anyone "honest" would sell it for $100.
You coercing me into starting a business to sell overpriced gear ?
No, no, no, I want you to build it !!!
I²S : 32 bit x 2 x 192k => 12.288 kbps
Suitable optos :
HCPL-2430-300E : 2 channels, 20 Mbps, $11.09
HCPL-7723-300 : 1 channel, 50 Mbps, $6.24
You need 4 lines for I²S (BCK WC SCK DATA) or perhaps 5 lines if you have 2 data lines, and you need an extra 2 for DSD, one more if you want to slave the transport, some perfboard, one ferrite chole, decoupling cap, and some flat cable...
The other option is of course to use CS8406 and a pulse transformer...
Note about digital filters : a XC3S100E is actually cheaper than a DF1704 (about same price if you count the support components), and the XC3S100E includes enough gear to make a 2 ch digital filter (4 18-bit multipliers that you should combine to get 2 24x18 mults and some RAM)
I'm actually building a FPGA based system and intend to do oversampling in the FPGA too, but it is going to be more complex (8 channels, streaming audio from the PC over ethernet, etc).
Parts cost $20 means anyone "honest" would sell it for $100.
You coercing me into starting a business to sell overpriced gear ?
No, no, no, I want you to build it !!!
I²S : 32 bit x 2 x 192k => 12.288 kbps
Suitable optos :
HCPL-2430-300E : 2 channels, 20 Mbps, $11.09
HCPL-7723-300 : 1 channel, 50 Mbps, $6.24
You need 4 lines for I²S (BCK WC SCK DATA) or perhaps 5 lines if you have 2 data lines, and you need an extra 2 for DSD, one more if you want to slave the transport, some perfboard, one ferrite chole, decoupling cap, and some flat cable...
The other option is of course to use CS8406 and a pulse transformer...
Note about digital filters : a XC3S100E is actually cheaper than a DF1704 (about same price if you count the support components), and the XC3S100E includes enough gear to make a 2 ch digital filter (4 18-bit multipliers that you should combine to get 2 24x18 mults and some RAM)
I'm actually building a FPGA based system and intend to do oversampling in the FPGA too, but it is going to be more complex (8 channels, streaming audio from the PC over ethernet, etc).
peufeu said:
As is uaually the case, you do not factor in the cost of programming and build. The Vanity is out of box solution for upsampled CD, DVD, and DSD to 176.4k conversion, with selectable filter. From listening it, I would say the the upsampling and filters have been done properly, unlike some of the chip asrcs.
It also provides a relocked output.
OK, it is not cheap, but then not everybody wants to or has the ability to, do it well.
Until I listen to the Sabre, I would not know what it is about, or whether the claims are true. Certainly the literature is full of hype but the demo board ($375) material does not tell you what has been implemented.
If you know please tell us.
No, I don't know what's in the Saber board, but I would like to find out. When my DAC test platform is done I definitely plan to include the Sabre in the list of its victims.
Also about the price of this Audiopraise gizmo : I would calculate amortization of 1 month of designer time over the number of sold gizmos ; if they sell 50 the price is justified, if they sell 1000, well...
Question for Dustin :
So, the Sabre has an ASRC in it.
What happens if I use the same clock in the data input and in the low jitter clock input ?
Clarification :
Suppose I have a source of digital audio data which can either run on its own clock or accept an external clock (this can be a transport, soundcard, whatever).
I input the digital audio data from this source in the Sabre.
Now, if I make the source use the DAC's internal clock (the one the Sabre uses for its low-jitter circuits), the ASRC becomes useless then.
So, what does the Sabre do ? Does it still use the ASRC, or does it just use a FIFO ?
Also about the price of this Audiopraise gizmo : I would calculate amortization of 1 month of designer time over the number of sold gizmos ; if they sell 50 the price is justified, if they sell 1000, well...
Question for Dustin :
So, the Sabre has an ASRC in it.
What happens if I use the same clock in the data input and in the low jitter clock input ?
Clarification :
Suppose I have a source of digital audio data which can either run on its own clock or accept an external clock (this can be a transport, soundcard, whatever).
I input the digital audio data from this source in the Sabre.
Now, if I make the source use the DAC's internal clock (the one the Sabre uses for its low-jitter circuits), the ASRC becomes useless then.
So, what does the Sabre do ? Does it still use the ASRC, or does it just use a FIFO ?
fmak said:
Are you sure its not locking?
The LEDS on the board only apply to the SPDIF signal if you are looking at those. So do not use those as an indicator of locking for PCM input. They only apply to SPDIF input.
I have yet to find any PCM signal it could not lock onto.
I would need to see pic of your setup etc to help you trouble shoot, but I am quite sure the WM8804 can do it.
If you have 5V logic you may need to level shift to 3.3V since the Wm8804 is a 3.3V part, but I am not sure. I have only used it with 3.3V logic. but so so far every PCM source I have tried has worked perfectly.
I just tried my DVD2910 with DVDA and it worked like a champ.
Russ White said:
-----------------------------------------------------------------------------------
I am not sure Pics will help as the source is a satellite receiver with the AK4351. The following may help you make an assessment.
AK4351 LRCK 48 kHz 1.6Vdc bias; BICK around 2.8 MHz 1.6Vdc; SDATA typical eye pattern signal 1.6Vdc; MCLK 12.2896 MHz around 1Vdc.
I coupled these to the Twisted Pear board LRCK, BCLK, DIn with and without blocking cap, MCLK and all I get is noise from the AK analog output and silence from the 8804 spdif output. For some reason my external dac indicates lock on at quad speed (??). The WM PCM input terminals have similar dc biases with no signal input and therefore I assumed that it was ok to couple. Anyway there has been no damage.
Would apprecaite your help
Either I have made a mistake or there is a timing issue?
Fred
Is there any indication when the chip will be available at all?
And am I correct in thinking that the volume control can be adjusted on a per-channel basis? If so, this woud be a great replacement for the DAC's inside de DCX and the output board, and I would be very interested to see more of the datasheet of the thing!
Great stuff!
And am I correct in thinking that the volume control can be adjusted on a per-channel basis? If so, this woud be a great replacement for the DAC's inside de DCX and the output board, and I would be very interested to see more of the datasheet of the thing!
Great stuff!