Wakibaki,
Let's try to clear up your point here.
First - the power spectral density, vn for 75ohm is 1.1147nV/sqrtHz.
(Using your formula)
Vrms for 175MHz bandwith is vn*sqrtB, 14.75 uV.
Now, what is the jitter introduced by this noise floor in a square signal typical for our transmitter?
The Hiface is doing 2.5Vpp in less than 2nsec. That is > 1.25V/nsec slew rate, but let's stay with 1.25. That is 1250mV/nsec.
In figure 7, cited by You the introduced jitter in function of the slew rate is given.
It shows 500psec jitter at 100mV/nsec slew rate. It is inversely proportional with slew rate, so for 1250mV/nsec it is 500psec*100/1250 = 40psec.
This the introduced jitter when 50 mVrms noise is present.
From Figure 6 cited by You we know that jitter is directly proportional with the noise Vrms value.
We have 40psec jitter introduced by 50 mVrms noise, then at 14.75 uVrms (calculated for 75ohm) noise level we have 40psec * 14.75/50000 = 0.0118 psec, 12femtosec jitter.
If we attenuate by 6db, then we half the slew rate, so the jitter introduced this way will raise to 24femtosec?
Can we agree in this? when attenuating 6db, we introduce +12 femtosec jitter?
Also, with 10dB, we are introducing +24 femtosec extra jitter?
And all this is Gaussian distributed noise.
Let's try to clear up your point here.
First - the power spectral density, vn for 75ohm is 1.1147nV/sqrtHz.
(Using your formula)
Vrms for 175MHz bandwith is vn*sqrtB, 14.75 uV.
Now, what is the jitter introduced by this noise floor in a square signal typical for our transmitter?
The Hiface is doing 2.5Vpp in less than 2nsec. That is > 1.25V/nsec slew rate, but let's stay with 1.25. That is 1250mV/nsec.
In figure 7, cited by You the introduced jitter in function of the slew rate is given.
It shows 500psec jitter at 100mV/nsec slew rate. It is inversely proportional with slew rate, so for 1250mV/nsec it is 500psec*100/1250 = 40psec.
This the introduced jitter when 50 mVrms noise is present.
From Figure 6 cited by You we know that jitter is directly proportional with the noise Vrms value.
We have 40psec jitter introduced by 50 mVrms noise, then at 14.75 uVrms (calculated for 75ohm) noise level we have 40psec * 14.75/50000 = 0.0118 psec, 12femtosec jitter.
If we attenuate by 6db, then we half the slew rate, so the jitter introduced this way will raise to 24femtosec?
Can we agree in this? when attenuating 6db, we introduce +12 femtosec jitter?
Also, with 10dB, we are introducing +24 femtosec extra jitter?
And all this is Gaussian distributed noise.
Last edited:
I can follow your whole argument. I cannot disagree with anything in this post except your opening statement.
I did not insist that a DAC behaves like an RF device.
I used the analogy of the RF device requiring properly terminated coax to illustrate that signal splitting and cable termination go hand in hand.
Status, Waki & Joseph,
Thank you
Last edited:
With reference to the scenarios shown in my post #53, here are the simulation waveforms from my simulation software.
I would just like to point out that it is a new toy we got recently as SI is becoming ever more problematic for high speed PCB design (I don't use RF as we are not trying to broadcast, just use very fast clocks on our digital interfaces, i.e. DDR ram interfaces).
Also I think that the reply to my posts was not in the spirit of DIY audio
, we are all here to learn, The advice and guidance I am reading up on is by people like Howard Johnson, Eric Bogatin, Ralph Morrison, Lee W Ritchey and Ralph Bruning who I believe are also experts in this field, so while I may not be a hundred percent correct in some of my assumptions they are not totally wrong and are based on reading and studies on material by some of the acknowledged experts in this field.
Anyway these are only initial simulations to compare the basic scenarios as post #53 shows and are for comparison of the ways of splitting a cable (though the cable length is only 600mm total, 100 before the split, 500 after) and I have only used a basic transmission line model for the cable.
The files are 3x25r (red=drive, yellow=receiver) for the top scenario, and 2x150r (red=drive, purple=receiver) for the bottom scenario. As can be seen both methods provide a satisfactory waveform at the receivers, though the 2x150r scenario has a lower voltage swing from the same drive!
As stated these are basic simulations and provided for a visual reference, I would like to model them more realistically and dig out some of my old cable TV bits and pieces (including splitters and terminations and some 75ohm cable) to correlate the simulations with reality. So if anyone has some driver and receiver info that are used on SPDIF interfaces I'd appreciate the info.
Caveat: These are basic simulations for comparison purposes only between the two methods discussed earlier. The tool requires a quite steep learning curve, but once that is achieved the results do match reality.
I would just like to point out that it is a new toy we got recently as SI is becoming ever more problematic for high speed PCB design (I don't use RF as we are not trying to broadcast, just use very fast clocks on our digital interfaces, i.e. DDR ram interfaces).
Also I think that the reply to my posts was not in the spirit of DIY audio
, we are all here to learn, The advice and guidance I am reading up on is by people like Howard Johnson, Eric Bogatin, Ralph Morrison, Lee W Ritchey and Ralph Bruning who I believe are also experts in this field, so while I may not be a hundred percent correct in some of my assumptions they are not totally wrong and are based on reading and studies on material by some of the acknowledged experts in this field.Anyway these are only initial simulations to compare the basic scenarios as post #53 shows and are for comparison of the ways of splitting a cable (though the cable length is only 600mm total, 100 before the split, 500 after) and I have only used a basic transmission line model for the cable.
The files are 3x25r (red=drive, yellow=receiver) for the top scenario, and 2x150r (red=drive, purple=receiver) for the bottom scenario. As can be seen both methods provide a satisfactory waveform at the receivers, though the 2x150r scenario has a lower voltage swing from the same drive!
As stated these are basic simulations and provided for a visual reference, I would like to model them more realistically and dig out some of my old cable TV bits and pieces (including splitters and terminations and some 75ohm cable) to correlate the simulations with reality. So if anyone has some driver and receiver info that are used on SPDIF interfaces I'd appreciate the info.
Caveat: These are basic simulations for comparison purposes only between the two methods discussed earlier. The tool requires a quite steep learning curve, but once that is achieved the results do match reality.
AndrewT,
Why do you question the quality of the termination with the AD811?
According to the quoted application note, it is as good as it can be?
On the input there are no problems at all. At the output, any driver will have a frequency dependent output impedance?
I doubt the original SPDIF driver would be any better?
A CMOS output driver is more inconsistent in it's ouput impedance.
And if the problem is the isolation, just apply two independent chips, on the same input termination - every high speed line buffer/distribution amp works like that.
As You see, I really value the resistor divider as well - but I think the dist. amp is just as ~good termination wise..
I think the resistive divider has an edge for some other reasons.
Though really has that 6dB "transparency" for the reflections..
Ciao, George
Why do you question the quality of the termination with the AD811?
According to the quoted application note, it is as good as it can be?
On the input there are no problems at all. At the output, any driver will have a frequency dependent output impedance?
I doubt the original SPDIF driver would be any better?
A CMOS output driver is more inconsistent in it's ouput impedance.
And if the problem is the isolation, just apply two independent chips, on the same input termination - every high speed line buffer/distribution amp works like that.
As You see, I really value the resistor divider as well - but I think the dist. amp is just as ~good termination wise..
I think the resistive divider has an edge for some other reasons.
Though really has that 6dB "transparency" for the reflections..
Ciao, George
Your figures look good to me JosephK (apart from this last one), despite your reference to the Hiface, which is mentioned nowhere in this thread and is a decidedly nonstandard SPDIF transmitter. I could disagree on detail, but the increase in jitter due to attenuation is unlikely to be > 1pS.
I am happy to see, however, that you are in agreement with my general point, that introducing an attenuator increases the jitter. This is in marked contrast to your earlier attitude
I am happy to entertain arguments based on reason. I would be happier still if you were to apologise for your former behaviour.
Given the minimal amount of jitter involved, I would agree that the resistive divider is an acceptable solution, and the most economical, if the receiving devices will tolerate the reduced driving voltages.
I am interested to discover that you find this an acceptable solution though, given your concern expressed elsewhere about reflections due to misterminations. Apparently a 75R termination at a DAC is likely to cause problems but 4 x 37.5R (37.4) resistors wired in a star are not. Perhaps if you gave the same consideration to this argument that you have finally given to my other suggestions we might find further ground for agreement.
w
As you know I am not an electronics man.
I look, I see, I comment.
The comment in post18 is quoted in full.
I have stated that the termination is good if the output impedance of the chip is zero for all frequencies.
I know that is an impossible condition.
I do not know what effect that will have on the level of miss-termination.
I have consistently reminded the readers that the query is "splitting".
Stratus came in and offered an amplifier. He keeps coming in and offering the same amplifier solution.
The rest of us seem to at least to some extent be looking at terminating the splitter correctly.
Are you having problems with the concept of input levels? Baseband devices (not modulated on an RF carrier) have specific levels they are designed to use. Passive splitting reduces the level. Once again I will tell you that you should use an amplifier to restore the level lost in the splitting process.
The individual resistor(s) on the amplifier output performs the splitting process. The 811 is NOT a perfect amplifier but it comes very close for this use. ADI states 40dB isolation between 2 outputs (up to 5 MHz) when using the circuit on page 14. 40dB is much better than your passive splitter can achieve.
You say you're not an electronics man. I agree.
G²
The individual resistor(s) on the amplifier output performs the splitting process. The 811 is NOT a perfect amplifier but it comes very close for this use. ADI states 40dB isolation between 2 outputs (up to 5 MHz) when using the circuit on page 14. 40dB is much better than your passive splitter can achieve.
You say you're not an electronics man. I agree.
G²
Last edited:
Well it was I whom dug up the circuit in the data sheet ,stratus just confirmed it's suitability for my intended use.
As stratus points out the datasheet shows two outputs with a claimed isolation of 40db between them.I also mentioned I did not want the signal attenuated as I was unsure of how the input reciever in my DAC's would deal with the reduced level.
Therefore the 811 circuit seems like an ideal solution in that it will give me a split signal with high isolation between output's and no attenuation.
Last edited:
Wakibaki,
I had put together your little input receiver circuit.
I have to admit, I was wrong - it does not show the expected bump in the reflection. I thought that initially it will go up to 100ohm, (~15% of reflection), and then "relax" back, to somewhere between 75-90 ohm.
It does not go up to 100 ohm. And does not relax back to 75ohm. It stays fixed at 94ohm. (11.4% reflection)
(As was suggested by Stormsonic)
Plus, it does not work, if assembled according to Your drawing.
I am asking You also here, if you could point us in the right direction, why it is not working?
Ciao, George
I had put together your little input receiver circuit.
I have to admit, I was wrong - it does not show the expected bump in the reflection. I thought that initially it will go up to 100ohm, (~15% of reflection), and then "relax" back, to somewhere between 75-90 ohm.
It does not go up to 100 ohm. And does not relax back to 75ohm. It stays fixed at 94ohm. (11.4% reflection)
(As was suggested by Stormsonic)
Plus, it does not work, if assembled according to Your drawing.
I am asking You also here, if you could point us in the right direction, why it is not working?
Ciao, George
Last edited:
Very funny George, I'm sure you'll get it working somehow. When are you going to prove that attenuators reduce jitter?
w
- Status
- Not open for further replies.