All,
Lately I've been trying to investigate the different types of
spdif coax output topologies. I want to modify or replace
the one in my DV-45A (used as a transport with a dAck)
to work the best with the CS8414 receiver input that has
a 75-ohm resistor from input to ground (impedance matching),
and 0.01uF cap in front of the RXP and RXN pins. (as recommended in http://www.cirrus.com/en/pubs/proDatasheet/8413-4.pdf
page 35)
I've seen some topologies that use hex inverters to buffer the signal, some that use pulse transformers, inductors in series, different resistor values that 'combine' to make 75-ohm-ish output inpedance, etc.
I recently stumbled across this one:
http://www.benchmarkmedia.com/appnotes-d/cdjeep.asp
Has anyone used this circuit to replace a typical hex-inverter
buffered circuit? I am trying to figure out what the schottky
diodes are used for. I don't see how the output voltage would
be true 0.5Vpp SPDIF. Then again, does it have to be for
the CS8414 to work properly?
My DV-45A uses a transistor in an emitter-follower config for buffering, and then a 68-ohm resistor, followed by a 47uF electrolytic cap, and then to the phono plug with a 100k resistor to gnd. If my dAck already has a cap in front of the RXP and RXN
inputs, is a dc blocking cap necessary in the digital out circuit? Maybe transmitting the signal via coax cable with dc on it is not desireable?
I have the feeling that when using an external dac, the digital out circuit of the transport has a noticeable influence on the sound. I do hear some differences with different transports.
Thanks for all your input,
Vinnie
Lately I've been trying to investigate the different types of
spdif coax output topologies. I want to modify or replace
the one in my DV-45A (used as a transport with a dAck)
to work the best with the CS8414 receiver input that has
a 75-ohm resistor from input to ground (impedance matching),
and 0.01uF cap in front of the RXP and RXN pins. (as recommended in http://www.cirrus.com/en/pubs/proDatasheet/8413-4.pdf
page 35)
I've seen some topologies that use hex inverters to buffer the signal, some that use pulse transformers, inductors in series, different resistor values that 'combine' to make 75-ohm-ish output inpedance, etc.
I recently stumbled across this one:
http://www.benchmarkmedia.com/appnotes-d/cdjeep.asp
Has anyone used this circuit to replace a typical hex-inverter
buffered circuit? I am trying to figure out what the schottky
diodes are used for. I don't see how the output voltage would
be true 0.5Vpp SPDIF. Then again, does it have to be for
the CS8414 to work properly?
My DV-45A uses a transistor in an emitter-follower config for buffering, and then a 68-ohm resistor, followed by a 47uF electrolytic cap, and then to the phono plug with a 100k resistor to gnd. If my dAck already has a cap in front of the RXP and RXN
inputs, is a dc blocking cap necessary in the digital out circuit? Maybe transmitting the signal via coax cable with dc on it is not desireable?
I have the feeling that when using an external dac, the digital out circuit of the transport has a noticeable influence on the sound. I do hear some differences with different transports.
Thanks for all your input,
Vinnie
Vinnie R. said:
Vinie
Yes, that quality is of importance, but not only determined by the output circuit
The diodes are for protection
I do not like parallel inverters
A galvanic isolation at the CDP output helps
Maybe you get inspired by
http://www.epanorama.net/links/audiodigital.html#spdif
differntial drive of the 8412/14 is also benneficial
http://members.chello.nl/~m.heijligers/DAChtml/dactop.htm
may give some hints
best regards
Hi Guido,
Thanks for your reply and links.
For the circuit shown in this link:
http://www.benchmarkmedia.com/appnotes-d/cdjeep.asp
Why would you want to use schottkys for protection? Doesn't the cap offer protection? Also, the output is not going to be 0.5Vpp spdif, right? Is not using a buffer okay if the coax cable is <= 1m ? Is simpler better in this case?
Why the two 150ohm resistors instead of one 75-ohm resistor?
As far as using transformers for isolation, I find that they slow down the rise/fall times of the spdif waveform and worsen the SNR. The nice squareness to the waveform becomes more pulled apart (going through inverters seems to this as well).
I can't seem to find enough threads discussing the details of the spdif out circuit and modding it. If anyone has done work on this, I'd love to hear about it.
Thanks again,
Vinnie
Thanks for your reply and links.
For the circuit shown in this link:
http://www.benchmarkmedia.com/appnotes-d/cdjeep.asp
Why would you want to use schottkys for protection? Doesn't the cap offer protection? Also, the output is not going to be 0.5Vpp spdif, right? Is not using a buffer okay if the coax cable is <= 1m ? Is simpler better in this case?
Why the two 150ohm resistors instead of one 75-ohm resistor?
As far as using transformers for isolation, I find that they slow down the rise/fall times of the spdif waveform and worsen the SNR. The nice squareness to the waveform becomes more pulled apart (going through inverters seems to this as well).
I can't seem to find enough threads discussing the details of the spdif out circuit and modding it. If anyone has done work on this, I'd love to hear about it.
Thanks again,
Vinnie
Using two 150 ohms resistors makes a more exact 75 ohm load. Thus 4x300 ohm would make more exakt and so on.
But beware that noise goes up as the value of the resistors does. ..
I think there has been a lot of discussions about using and why not using transformers. Search the archives. Jocko and Fred seems to have(had) a lot to say about them
But beware that noise goes up as the value of the resistors does. ..
I think there has been a lot of discussions about using and why not using transformers. Search the archives. Jocko and Fred seems to have(had) a lot to say about them
Whoa.....I missed this one.....!
In theory, 2 150R in parallel will give 75R.
In practice: no. The gate driving this has an intrinsic impedance, usually between 20-30R, depending on the series. You need to factor that in.
Slowing down rise/fall times:
Some stuff that Philips makes really does this: to reduce EMI. It makes for a rotten output. A high quality transformer will not.
A lot of handwringing about SNR, CMRR, and the like gives rise to silly notions that the transformer should have loose coupling between the windings. Bad idea. They must be tightly coupled to get the leakage inductance down.
The main trick is to get the reflection coefficient down as low as possible. A "rho" of only 0.1 has an audible effect. And that is fairly low.
So......further discussions???
Jocko
In theory, 2 150R in parallel will give 75R.
In practice: no. The gate driving this has an intrinsic impedance, usually between 20-30R, depending on the series. You need to factor that in.
Slowing down rise/fall times:
Some stuff that Philips makes really does this: to reduce EMI. It makes for a rotten output. A high quality transformer will not.
A lot of handwringing about SNR, CMRR, and the like gives rise to silly notions that the transformer should have loose coupling between the windings. Bad idea. They must be tightly coupled to get the leakage inductance down.
The main trick is to get the reflection coefficient down as low as possible. A "rho" of only 0.1 has an audible effect. And that is fairly low.
So......further discussions???
Jocko
As long as you are modifying the transport ... dump S/PDIF and hack in something else. I think I whistled this tune a few times ....
See http://www.diyaudio.com/forums/showthread.php?s=&threadid=22741 if you have no idea what I mean.
See http://www.diyaudio.com/forums/showthread.php?s=&threadid=22741 if you have no idea what I mean.
Lovan said:
Yes, but one 150 ohm resistor is in series, and one is going to ground. I see a voltage divider when I look at that schematic, where the output is 1/2 the input. Why not use just a 75 ohm resistor in series?
Also, do you see a benefit of using that scheme instead of buffers, transformers, etc. Is simpler better for a short run application?
Also, why use a transformer? Being that the receiver of the dac has a cap before the RXP and RXN pins, why is more isolation needed? I'm not seeing the complete picture.
-Vinnie
[Q]Yes, but one 150 ohm resistor is in series, and one is going to ground. I see a voltage divider when I look at that schematic, where the output is 1/2 the input. Why not use just a 75 ohm resistor in series? [/Q]
You must consider the model for a voltage source in calculating impedances. In AC signals (digital signals), a voltage source (buffer) looks like a short circuit to ground. So, when looking at the output impedance of a circuit with the above description, the two 150 impedances appear to be in parallel, thus giving a 75ohm total impedance.
But, as Jocko metioned, real-life is never as "perfect" as theory and simulation - there is gate resistance in the buffer to consider also. I wasn't aware that it was as high as 20-30 ohms as he mentions, but I sure wouldn't argue with the guy seeing as he has so much experience.
One very well could use just a 75 ohm in series (or 30-40 ohms to get total series impedance from what Jocko says) - but I'm guessing the output of the buffer is 5V... and they probably wanted to knock down the output a bit so as not to overload a reciever. They do this a lot in video buffers to... they typically have a gain of 2 (6dB) (I suppose to maintain signal integrity -less noise - through the amplifiers). The output then has the above 150ohm split voltage dividers. Just something I've noticed and thought I'd mention...
You must consider the model for a voltage source in calculating impedances. In AC signals (digital signals), a voltage source (buffer) looks like a short circuit to ground. So, when looking at the output impedance of a circuit with the above description, the two 150 impedances appear to be in parallel, thus giving a 75ohm total impedance.
But, as Jocko metioned, real-life is never as "perfect" as theory and simulation - there is gate resistance in the buffer to consider also. I wasn't aware that it was as high as 20-30 ohms as he mentions, but I sure wouldn't argue with the guy seeing as he has so much experience.
One very well could use just a 75 ohm in series (or 30-40 ohms to get total series impedance from what Jocko says) - but I'm guessing the output of the buffer is 5V... and they probably wanted to knock down the output a bit so as not to overload a reciever. They do this a lot in video buffers to... they typically have a gain of 2 (6dB) (I suppose to maintain signal integrity -less noise - through the amplifiers). The output then has the above 150ohm split voltage dividers. Just something I've noticed and thought I'd mention...
modification proposal for Pioneer DV-45A
Although I agree that SPDIF is quite flawed, we can't all build our own serial digital physical layer. So, I've looked into the Pioneer circuit and propose this solution. I hope it's not daunting; it should improve the output characteristics to the point that the SPDIF driver is no longer a limiting factor.
Comments are welcome (and if you can maintain the output voltage and slew rate without using the diodes or adding another transistor, I'd like to know).
http://www.pmb.net/projects/cdplayers/dv45a_spdif.html
Although I agree that SPDIF is quite flawed, we can't all build our own serial digital physical layer. So, I've looked into the Pioneer circuit and propose this solution. I hope it's not daunting; it should improve the output characteristics to the point that the SPDIF driver is no longer a limiting factor.
Comments are welcome (and if you can maintain the output voltage and slew rate without using the diodes or adding another transistor, I'd like to know).
http://www.pmb.net/projects/cdplayers/dv45a_spdif.html
Re: modification proposal for Pioneer DV-45A
Pieris,
Thanks for your research and simulations.
When I measured the DV-45A rise time of the spdif out (using a 75-ohm cable and a 75-ohm resistor across it), I do get around 26nS, which does match closely to your 25nS analysis. However, the output goes from around -0.6V to +0.6V. Somehow you have 0 to 0.88V. I think the cap on the output removes this DC bias.
I did open my DV-45A and found VERY miniature surface mount resistors and transistor
Soldering around in there is going to require delicate work. I didn't get a chance to play with changing the 2.2k base resistor to 220ohm, but I hope to do this soon.
The other option instead of trying to improve upon the DV-45 circuit is to just take the spdif from the chip that creates it and send it to a brand new output circuit. I never looked a any spdif circuits of high end players with rise times of less than 10ns, and I wonder how complicated they are. This is a good learning experience though
One more thing. In your simulated circuit, you model the cable with R8 and C3. What about the 75ohm resistor that is on the input side of the dAck? It goes from the center pin to the ground, and there is a cap off of both sides of this resistor and they go to RXP and RXN of the CS8414 chip. In your simulation, I guess you would need a 75 ohm resistor in series with C13 to model the cable, and then a 75 ohm resistor to ground to model the input of the dAck. Is this true?
Again, thanks for your insight. I find this topic interesting and would love to hear others opinions and experiences with this.
-Vinnie
pieris said:
Pieris,
Thanks for your research and simulations.
When I measured the DV-45A rise time of the spdif out (using a 75-ohm cable and a 75-ohm resistor across it), I do get around 26nS, which does match closely to your 25nS analysis. However, the output goes from around -0.6V to +0.6V. Somehow you have 0 to 0.88V. I think the cap on the output removes this DC bias.
I did open my DV-45A and found VERY miniature surface mount resistors and transistor
Soldering around in there is going to require delicate work. I didn't get a chance to play with changing the 2.2k base resistor to 220ohm, but I hope to do this soon. The other option instead of trying to improve upon the DV-45 circuit is to just take the spdif from the chip that creates it and send it to a brand new output circuit. I never looked a any spdif circuits of high end players with rise times of less than 10ns, and I wonder how complicated they are. This is a good learning experience though
One more thing. In your simulated circuit, you model the cable with R8 and C3. What about the 75ohm resistor that is on the input side of the dAck? It goes from the center pin to the ground, and there is a cap off of both sides of this resistor and they go to RXP and RXN of the CS8414 chip. In your simulation, I guess you would need a 75 ohm resistor in series with C13 to model the cable, and then a 75 ohm resistor to ground to model the input of the dAck. Is this true?
Again, thanks for your insight. I find this topic interesting and would love to hear others opinions and experiences with this.
-Vinnie
Consider using a very fast buffer to drive the cable. OPA633 is one.
Datasheet here.
With a slew rate in the order of 2.5V/ns it would contribute very little to the jitter, probably less than loading the poor output directly from the chip with a cable. Use a voltage divider on the input to get 1.4Vtt and use a 70 ohm series resistor to the cable.
Just my 2 coins.
Datasheet here.
With a slew rate in the order of 2.5V/ns it would contribute very little to the jitter, probably less than loading the poor output directly from the chip with a cable. Use a voltage divider on the input to get 1.4Vtt and use a 70 ohm series resistor to the cable.
Just my 2 coins.
Vinnie,
Bummer about the SMD parts. It will be a lot harder to make modifications. The advantage is considerable for high speed, though; leads are shorter (less inductive) and are closer to the ground plane (smaller swept area to radiate or pick up EMI).
I'm actually quite surprised that the risetime I obtained from simulation matched your empirically measured value. I guessed at the cable load (100pF || 75 ohm). I don't know where the pk-pk difference is coming from.
The cable itself does not act anything like a 75 ohm resistor (you can verify this with an ohmmeter); instead it is the characteristic impedance which, among other ways, can be expressed as a ratio of capacitance to inductance.
There's a nice explanation of this concept here:
http://www.generalcable.com.au/Technical/10.4.1.1.pdf
And probably elsewhere on this forum. All this means if you terminate it with 75 ohms there will be no reflections. The 75 ohm resistance that I've called "cable" is actually the resistor in your DAC. The 100pF capacitance is a wild guess at the cable's capacitance.
As for the cable driver, the OPA633 looks as good as any, but building your own stage requires considerable effort (and maybe a new thread). Has this topic been explored before?
Bummer about the SMD parts. It will be a lot harder to make modifications. The advantage is considerable for high speed, though; leads are shorter (less inductive) and are closer to the ground plane (smaller swept area to radiate or pick up EMI).
I'm actually quite surprised that the risetime I obtained from simulation matched your empirically measured value. I guessed at the cable load (100pF || 75 ohm). I don't know where the pk-pk difference is coming from.
The cable itself does not act anything like a 75 ohm resistor (you can verify this with an ohmmeter); instead it is the characteristic impedance which, among other ways, can be expressed as a ratio of capacitance to inductance.
There's a nice explanation of this concept here:
http://www.generalcable.com.au/Technical/10.4.1.1.pdf
And probably elsewhere on this forum. All this means if you terminate it with 75 ohms there will be no reflections. The 75 ohm resistance that I've called "cable" is actually the resistor in your DAC. The 100pF capacitance is a wild guess at the cable's capacitance.
As for the cable driver, the OPA633 looks as good as any, but building your own stage requires considerable effort (and maybe a new thread). Has this topic been explored before?
In theory, yes. In practice, there will always be reflections.
True, 75 is a lot better than 100 ohms. You would be surprised where I see stuff like this.
Diodes on either side of the line are a no-no, as they look like short circuits to the line. Ditto for Schmitt trigger inputs.
Sloppy installation can cause considerable reflections.
Along with using RCA connectors...............
Jocko
SPDIF Interface
Hi, I posted my own version with the OPA603 a long time ago. Worked fine for me but I2S Direct, if possible, is better. OPA603 is obsolete now, OPA633 seems to be suited. Thanks. Will try to paste in some links:
I2S Direct:
http://www.diyaudio.com/forums/showthread.php?postid=249220#post249220
SPDIF (Feline amenities):
http://www.diyaudio.com/forums/showthread.php?postid=227664#post227664
Back to jester course, can't stay all day here.
Hi, I posted my own version with the OPA603 a long time ago. Worked fine for me but I2S Direct, if possible, is better. OPA603 is obsolete now, OPA633 seems to be suited. Thanks. Will try to paste in some links:
I2S Direct:
http://www.diyaudio.com/forums/showthread.php?postid=249220#post249220
SPDIF (Feline amenities):
http://www.diyaudio.com/forums/showthread.php?postid=227664#post227664
Back to jester course, can't stay all day here.
Fred has a totally different opinion on that matter IIRC. Don't want to search for the thread. I had not the best experiences with BNC I admit but that was carried out with 75 Ohm cable and standard BNC connectors. The 75 Ohm BNC connectors were hard to find then. I stick with RCA or even better, no SPDIF at all
You are correct about Phred.....
He manufactured a product that was specifically designed to be used in a typical SPDIF arrangement. He made a conscious decision that in order to sell his product, that it had to be 100% compatible with existing units.
He went to great lengths to build a precision network to counter most of the effects of using an RCA connector. I do not know how many guys reading this thread have the equipment to measure something like that.
BTW.......Phred has less hair these days......hint, hint.
Jocko
He manufactured a product that was specifically designed to be used in a typical SPDIF arrangement. He made a conscious decision that in order to sell his product, that it had to be 100% compatible with existing units.
He went to great lengths to build a precision network to counter most of the effects of using an RCA connector. I do not know how many guys reading this thread have the equipment to measure something like that.
BTW.......Phred has less hair these days......hint, hint.
Jocko
IIS direct
Do people routing I2S in lieu of SPDIF pull out the Crystal receiver and inject their I2S signals there? Or is this usually only done with homemade DACs?
Elso, I looked at your old post, and want to know more about the "Asynchronous Reclocker" that is referred to but not explained. Can you refer me to the schematic and explanation?
Do people routing I2S in lieu of SPDIF pull out the Crystal receiver and inject their I2S signals there? Or is this usually only done with homemade DACs?
Elso, I looked at your old post, and want to know more about the "Asynchronous Reclocker" that is referred to but not explained. Can you refer me to the schematic and explanation?
Re: IIS direct/ASR
The schematic is the easy part to answer as you can mail me through the forum system and I will reply with the schematic attached.
The second part is much more difficult to answer. I guess it has to do with the fact that any residual jitter attributed to the clock in the ASR is not data related as in the SPDIF. You can look at the ASR as a very short FIFO. Of course AS reclocking is not perfect. I prefer to prevent the disease jitter rather than to cure it, hence my I2S Direct scheme. You can also pose the theoretical question if it is possible to recover all the data jitter free from a jittery source like the DIR. I think Guido T. thinks yes, I believe not.
I have only practical experience with a Philips CD650 and a CD931 using I2S Direct on a NON-OS TDA1543 DAC.
It also worked with a Sony CDP-X333ES and the same DAC but the TDA1543 replaced by the TDA1543A
Audio Alchemy had an I2S system but with a connector system totally hopeless for digital signals. (Mini Din jack). There are others with better connectors. I cannot remember the names, been all day on the forum.....
Hi Pieris,pieris said:
The schematic is the easy part to answer as you can mail me through the forum system and I will reply with the schematic attached.
The second part is much more difficult to answer. I guess it has to do with the fact that any residual jitter attributed to the clock in the ASR is not data related as in the SPDIF. You can look at the ASR as a very short FIFO. Of course AS reclocking is not perfect. I prefer to prevent the disease jitter rather than to cure it, hence my I2S Direct scheme. You can also pose the theoretical question if it is possible to recover all the data jitter free from a jittery source like the DIR. I think Guido T. thinks yes, I believe not.
I have only practical experience with a Philips CD650 and a CD931 using I2S Direct on a NON-OS TDA1543 DAC.
It also worked with a Sony CDP-X333ES and the same DAC but the TDA1543 replaced by the TDA1543A
Audio Alchemy had an I2S system but with a connector system totally hopeless for digital signals. (Mini Din jack). There are others with better connectors. I cannot remember the names, been all day on the forum.....
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.