Hi all,
I am looking at a way to switch different CD digital outputs to a D/A converter without degrading the signal. I am looking at RF coaxial switches at the surplus market, but the ones I found are all 50 ohms characteristic impedance, while an AES/EBU connection should be 110 ohms IIRC.
Would there be a signal degrading if I used the 50 ohms switch? Do you have another solution?
Thanks,
Jan Didden
I am looking at a way to switch different CD digital outputs to a D/A converter without degrading the signal. I am looking at RF coaxial switches at the surplus market, but the ones I found are all 50 ohms characteristic impedance, while an AES/EBU connection should be 110 ohms IIRC.
Would there be a signal degrading if I used the 50 ohms switch? Do you have another solution?
Thanks,
Jan Didden
11 Mhz may be a bit too enthousiastic:
"The [AES/EBU] data transmission is sent as a Bi-Phase mark mode to minimise the DC content of the data stream. A logic '1' is represented by a transition in the middle of the data bit and a logic '0' doesn't have a transition. The end of each bit time also has a transition regardless of the data level. This method of encoding requires that the data clock is two times the data rate. For 48kHz sample rate data the clock would therefore be 6.144MHz (2 channels x 32 bits per sub-frame x 48kHz)"
Jan Didden
"The [AES/EBU] data transmission is sent as a Bi-Phase mark mode to minimise the DC content of the data stream. A logic '1' is represented by a transition in the middle of the data bit and a logic '0' doesn't have a transition. The end of each bit time also has a transition regardless of the data level. This method of encoding requires that the data clock is two times the data rate. For 48kHz sample rate data the clock would therefore be 6.144MHz (2 channels x 32 bits per sub-frame x 48kHz)"
Jan Didden
OK,
At that frequency your wavlength is somewhere between 30 - 50 meters.
A 50 Ohm switch in a 110 Ohm line will cause reflections back toward the source... in theory these reflections should die when they return to the source amplifier/driver (if it has a 100 Ohm Zout). So your signal between the driver and the switch (and down stream for that matter) is the original with the reflection superimposed.
On paper it doesn't sound to well... but in reality 6Mhz is really pretty slow, the reflections are not 1 to 1 in magnitude (not even close), and digital is very forgiving.
Is there a real price advantage with these surplus switches?
As an alternative I'm sure there are silicon switches for this purpose. You would terminate each line with 110 ohms; feed this to a switch, and reamplify with 110 Zout and be on your way down stream. If the distance between your switch and the DAC is very small, you could avoid the transmision line completely.
Do you have a complete description of the signal standard or its "name"?
At that frequency your wavlength is somewhere between 30 - 50 meters.
A 50 Ohm switch in a 110 Ohm line will cause reflections back toward the source... in theory these reflections should die when they return to the source amplifier/driver (if it has a 100 Ohm Zout). So your signal between the driver and the switch (and down stream for that matter) is the original with the reflection superimposed.
On paper it doesn't sound to well... but in reality 6Mhz is really pretty slow, the reflections are not 1 to 1 in magnitude (not even close), and digital is very forgiving.
Is there a real price advantage with these surplus switches?
As an alternative I'm sure there are silicon switches for this purpose. You would terminate each line with 110 ohms; feed this to a switch, and reamplify with 110 Zout and be on your way down stream. If the distance between your switch and the DAC is very small, you could avoid the transmision line completely.
Do you have a complete description of the signal standard or its "name"?
poobah said:OK,
At that frequency your wavlength is somewhere between 30 - 50 meters.
A 50 Ohm switch in a 110 Ohm line will cause reflections back toward the source... in theory these reflections should die when they return to the source amplifier/driver (if it has a 100 Ohm Zout). So your signal between the driver and the switch (and down stream for that matter) is the original with the reflection superimposed.
On paper it doesn't sound to well... but in reality 6Mhz is really pretty slow, the reflections are not 1 to 1 in magnitude (not even close), and digital is very forgiving.
Is there a real price advantage with these surplus switches?
As an alternative I'm sure there are silicon switches for this purpose. You would terminate each line with 110 ohms; feed this to a switch, and reamplify with 110 Zout and be on your way down stream. If the distance between your switch and the DAC is very small, you could avoid the transmision line completely.
Do you have a complete description of the signal standard or its "name"?
Hi-quality HP coax switches (DC to 18GHz) are < 50 $, which is a steal really, but as I said they are 50 ohms.
I really would want to keep this as simple as possible to use for listening A/B tests of different CD players digital output. If I used silicon switches everyone would say that they would mask any differences
.As to the signal specs, is this what you are looking for:
"The [AES/EBU] data transmission is sent as a Bi-Phase mark mode to minimise the DC content of the data stream. A logic '1' is represented by a transition in the middle of the data bit and a logic '0' doesn't have a transition. The end of each bit time also has a transition regardless of the data level. This method of encoding requires that the data clock is two times the data rate. For 48kHz sample rate data the clock would therefore be 6.144MHz (2 channels x 32 bits per sub-frame x 48kHz".
Jan Didden
janneman said:
I really would want to keep this as simple as possible to use for listening A/B tests of different CD players digital output. If I used silicon switches everyone would say that they would mask any differences
Jan Didden
The advantage of an on-air type switcher is that switching is glitch-free but I can see how its relatively complex nature would be a disadvantage in your scenario.
Being an RF kinda guy, the idea of H-P switches sounds neat. I would not do it, though. I would switch after conversion.
Or....you could make a nice linear buffer of some sort...which the '841x needs anyway........and do the switching after it.
Hate to disappoint you guys, but "digital", as in SPDIF, is not that forgiving.
At least not if you want it to sound good. Passing data and "sounding good" are not the same.
Jocko
Or....you could make a nice linear buffer of some sort...which the '841x needs anyway........and do the switching after it.
Hate to disappoint you guys, but "digital", as in SPDIF, is not that forgiving.
At least not if you want it to sound good. Passing data and "sounding good" are not the same.
Jocko
Its not the frequency of the signal that needs to be considered, but speed of the transition edges, which are normally less then 10nS (100MHz+), this is where correct termination is critical.
RF switches don't normally have very good isolation - say 30dB to 60dB at the frequency of interest. With poor isolation, the un-switched signal will cross-talk with the switched signal degrading jitter performance. If RF switching is used, then 3 switches need to be used, 2 to "Ground" each unselected input, and one to switch the wanted signal - however at 100MHz and above, this only improves the isolation to say -80dB which is still very poor in Audio Jitter terms.
I would recommend to convert each input (correct terminated) into CMOS, then use a Digital MUX, again grounding the unwanted input before the Main MUX switch – use single Gate MUXs’ such as the excellent Fairchild 7NSZ157.
To obtain good isolation very careful RF design techniques need to be used on the PCB, Separate RF (and LF) filtering on the PSU rails to EACH Logic Gate (MUX) and SDPIF to CMOS level translators…..
At best, only 70dB to 80dB isolation between inputs can be obtained due to RF leakage on the PCB and within the casework unless crazy “RF plumbing” and decent RF screening is used. This level of "Digital Cross talk" will be very audioable with DAC designs where some method of reclocking is not implemented.
In conventional designs, it’s all too common to hear the audio quality degrade when 2 or more digital inputs are active to a DAC.
John
Well, to be brief........as one RF guy to another....
My RF background says "less than 300 MHz......impedance should not matter." The problem is SPDIF has to recover the clock from the datastream. When you do, you will find that it is not a clean recovery. it has artifacts that are corelated to the data.
Try listening to the recovered clock when you are playing music. You will start to understand why.
So......reflections......that stuff that bounces back and forth...can arrive at a point in the waveform that will affect the decision point. Yes, a 1 will still be a 1....a 0 a 0. Just off a little in time. So now you have FM to deal with along with the other crud.
So.......back to RF.........input rho has to be as low as possible. You can easily hear the effects......down to around 28-30 dB or so. At that point, things don't make a big difference.
Most SPDIF circuits are in the 10-20 dB range. On the circuits that I build the rho is in the range of sticking a 50 ohm barrel in 75 ohm coax.
Back to jitter...........typical SPDIF circuit is 1 nSec or so. Yuk. Really good CD player is 10 pSec or so. Big difference. You can hear it.
'Nuf for now........I'm typed out.
Jocko
My RF background says "less than 300 MHz......impedance should not matter." The problem is SPDIF has to recover the clock from the datastream. When you do, you will find that it is not a clean recovery. it has artifacts that are corelated to the data.
Try listening to the recovered clock when you are playing music. You will start to understand why.
So......reflections......that stuff that bounces back and forth...can arrive at a point in the waveform that will affect the decision point. Yes, a 1 will still be a 1....a 0 a 0. Just off a little in time. So now you have FM to deal with along with the other crud.
So.......back to RF.........input rho has to be as low as possible. You can easily hear the effects......down to around 28-30 dB or so. At that point, things don't make a big difference.
Most SPDIF circuits are in the 10-20 dB range. On the circuits that I build the rho is in the range of sticking a 50 ohm barrel in 75 ohm coax.
Back to jitter...........typical SPDIF circuit is 1 nSec or so. Yuk. Really good CD player is 10 pSec or so. Big difference. You can hear it.
'Nuf for now........I'm typed out.
Jocko
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.