Attenuators & Jitter Reduction

[jkeny]

There was a unanimously positive reaction to the attenuators at Fran's get-together on Thursday & I thought I would copy here some of what I have written on other forums about them: http://www.audiocircle.com/index.php?to ... #msg796176 (there are users impressions in the pages following that link)

Digital attenuators are used all the time in your UPC cable box on your house. They use various levels of attenuation (dBs) to bring the signal down to within spec going into your house. If used in digital audio they are usually found built in to the output stage of devices again to bring the signal down to within spec.

Apart from whatever benefits might be conferred by bringing this SPDIF signal into spec (some DACs don't like a high signal & some DACs prefer a high signal), the real benefit of these attenuators is that they doubly reduce the signal strength of any reflections generated between digital transport & receiver. They doubly reduce the reflection because the reflection has to pass through the attenuator twice before it gets back to the DAC where it can generate jitter. Now, to paraphrase a Bruce Forsyth catchphrase "What do reflections make - jitter" :)

As mentioned they confer a better advantage (in jitter amelioration) than a high end SPDIF cable because they also kill any reflections generated outside the cable in the transport output stage & the DAC input stage - something a SPDIF can't do.

They are in-line adaptors that can be connected at the start or end (or both?) of your digital cable. Cable reflections are one of the many sources of added jitter in every digital system. Reducing the strength of these reflections should result in a reduction in the jitter generated by the transport to DAC connection.

There are many situations where this should result in better sound. I would think that they will improve any SPDIF connection as I don't believe there is one made that is reflection free? They may also be an effective way of reducing the sonic penalty usually incurred in using a BNC to RCA adaptor. So this could mean that by using these attenuators an RCA input on your DAC should now be about the same sonically as a BNC input.

Minicircuits is the cheapest place I have found on the internet for these attenuators at $12 each - there are other places that will charge you >$30 for eactly the same thing - I've even seen them for >€100 in RS which I thought must be a mistake! There is a UK branch - Fran can say what the shipping costs, etc are.

They come in 3, 6, 10, 15, 20dB versions - the 3 & 20dB ones could probably be ruled out as too low & maybe too high an attenuation. What we want to aim for is enough attenuation to suppress signal reflections BUT not too much to reduce the signal voltage to too low. You will know if you have done this as the DAC will not lock to the SPDIF signal.

Obviously if you reduce the signal level too much & drop outside of the level at which your receiver will lock, it isn't of much use. If your DAC receiver chip is one of the CS84** ones - these are known to like higher input voltages then the SPDIF spec of 500mVpp but I don't know how the reduction in reflections versus the overall reduction in signal voltage will work out. All I can say is that these same receivers are used as DACs to transports that output the SPDIF standard of 500mV & that's all I'm suggesting the attenuators should reduce the signal to.

Why buy expensive digital cables when these simple & cheap devices can be used to achieve a better solution in certain circumstances noted above. It's a better solution because of the fact that a good digital cable if well made (expensive!) will have controlled impedance from end to end & will be a close match to the desired 75ohm. Outside of the cable at the DAC end or the transport end the cable cannot ameliorate any impedance anomalies that are encountered so these will give rise to reflections travelling down the cable.

[Diapason]

Interesting stuff, jkeny, and it all makes a lot of sense. I was another who was very impressed by the improvement brought to the sound.

Have you tried putting 3dB attenuators at both ends of the cable rather than a single 6dB? Might be interesting.

[jkeny]

Interesting stuff, jkeny, and it all makes a lot of sense. I was another who was very impressed by the improvement brought to the sound.

Have you tried putting 3dB attenuators at both ends of the cable rather than a single 6dB? Might be interesting.

Never tried two 3dB. For the squeezebox it may be right, for the Hiface we can get away with a much higher attenuation, 10+6dB is feasible. It is dependent on the transport's output SPDIF signal level & the DAC's SPDIF receiving chip operational point. For instance the commonly found CS84XX SPDIF receiver chips works best at the high raw output levels of the Hiface (i.e at TTL levels) but the Sabre DAC clips at this level.

[Fran]

I thought I should mention that these attenuators will have no effect on the volume level that you hear at your speakers.

The idea is that they attenuate the height of the "ons" and "off" that make up the spdif stream. So for eg, if the height of the normal stream is a unit of "1", then after the attenuator goes in, it might be "0.75". Once your DAC can "sense" or lock on to that height of signal, the DAC can interpret the series of "on" and "offs" correctly and you get your translation into analogue and the sound at your speakers.

So you can imagine the scene where you have a transport that puts out a stream with a height of 0.5. and DAC that needs a height of 0.4 to lock, and your attenuator drops the signal to 0.35 - you will have silence (or static noise as it nearly detects, then loses the lock)! However, if say your transport puts out height =2 and the DAC can lock on to height = 0.2, then you have a big margin to allow the use of these attenuators. I have had this experience using a 10dB attenuator with my squeezebox - wouldn't lock, but if I used a 6dB one, it locked no problem.

The other thing is that these attenuators are lab devices, precision made, and are only available in BNC format:



and



So if your transport or DAC doesn't have that type of connection (and most won't) you need to factor getting a bnc to RCA adaptor:



and



Maplin do these AFAIK - but its probably cheapest to get them off ebay.


So with all this talk of attenuation, does the level of attenuation actually matter (once the DAC can lock?) That's what I don't know yet - does a 6dB sound better than a 10dB?. As far as I can tell, the cleaning up of the sound that these things cause, is not due to the reduction in the height of the signal, but more that the impedance of the transport-cable-dac all gets matched up much better. John can probably step in here and explain this better but as I understand it, where a mismatch occurs, you will get a reflection that interferes with the DAC deciding is the signal meant to be on or off - the result is that mismatches in timing occurs - which is what jitter is (I think). So with the attenuator in place, the DAC has an easier job of deciding whether that incoming signal is an "on" or "off". and so does its job much more accurately.

I can tell you, they sure work in my system and I don't intend sending them back!!


Fran

[jkeny]

Yes Fran, agree with all you say except that the attenuators do not make the line impedance correct - they remove any reflections that might be on the line from any impedance mismatches.

Let me explain - if we imagine the SPIF signal is a wave in water - if there are no objects sticking up out of the water, there will be no reflections of part of the wave. Now in the digital world, if the signal doesn't see a 75ohm impedance from start to finish, then a reflection of some part of the signal will be created. Why is this of concern? Well this reflection will bounce back off the transmitter & come back to the DAC. If this reflection arrives at the DAC (& is of a high enough energy) at the moment the DAC is interpreting a 1 bit it will cause a mis-timing of this 1 - this is jitter. Jitter affects the sound - the demonstration of the attenuators was an example of what lower jitter sounds like & it was very noticeable.

The attenuators doubly reduce the energy of any reflections on the line. How? Well because the reflection has to pass through the attenuator twice, once on the way back to the transmitter & again when it bounces back towards the DAC. So this is why it is so effective. Possibly having two on the line at each end would be more effective - some experimentation id needed?

BTW, a note about SPDIF cables & cable length. A really good SPDIF cable will be 75ohms all along it's length (requires precision engineering & assembly) & will be terminated properly. This will help to avoid reflections happening BUT outside the cable one finds the output stage of the transmitter & at the other end the input stage of the SPDIF receiver. Apparently it is pretty common for there to be mis-terminations in both of these stages. So your really expensive SPDIF cable will do nothing to reduce the reflections generated in these stages. The attenuators will.

SPDIF cable length will have an effect on when the reflections hit the DAC as the signal takes time to travel back & forth along the cable. So if it can be arranged that the reflection doesn't fall at the DAC decision point then no jitter will occur. There are two approaches to this (& two camps that argue about this). The camp that says short SPDIF cables are the way to avoid jitter is correct - below a certain cable length, the reflection will bounce back & arrive at the DAC before it has reached it's decision point. A foot long cable or shorter will probably do this. The other camp say that long cables will cause the reflection to hit the DAC after the DAC's decision point. This is partially correct as some calculation is needed to decide the correct cable length. At a certain cable length jitter will be avoided but increase the length & the reflection will hit the DAC's next decision point because the delay has fallen on this point. What isn't a good cable length is the in-between length i.e 1 metre - the typical length commercial cables come in!

Anyway, there is a lot more I could say about digital cables, like why they can sound different if reversed but the above is probably more than anybody expected or wanted!

Hope this gives some technical reasons for how these attenuators work & why they are so effective!

[frd1996]

Interesting reading guys. I will have to try them out :)

[jkeny]

frd1966, glad you found it interesting - here's some more that might be of interest.

Have you ever found a digital cable sounding different when used in one direction than if turned around & used in the otrher direction? Ever wondered why?

Here's my theory & thoughts - most cables have impedance problems on them which give rise to jitter! Lets say this impedance anomaly is 1/3 of the way down the cable, then the reflection generated by it will arrive at the DAC at a certain time. Turn the cable around & now this anomaly is 2/3 down the cable & the reflection will arrive at the DAC at a different time. Of course there are usually more than just one such impedance problem on a SPDIF cable. The timing of the reflections generated by these imperfections will be different depending on which way the SPDIF cable is used. Each direction of the cable will sound different - which is better is up to you!

So you see SPDIF cables do have directionality but there is no one way which is better than another so if you find a SPDIF cable with a direction arrow on it, it's pure bunkum.

The attenuators kill the reflections & hence remove this question.

[DaveF]

very interesting thread jkenny. I've designed a few SPDIF receivers and have been involved in one or two DAC designs over the years but nothing too high end. The reflections I would say is only a problem if the SPDIF stream is used to directly clock the DAC or that the extracted clock is used in the digital filtering. Even then it would depend on how 'immune' the DAC is to jitter. As you mentioned somewhere above, it's all down to the implementation and results and observations may vary from player to player and cable impedence too.

[jkeny]

very interesting thread jkenny. I've designed a few SPDIF receivers and have been involved in one or two DAC designs over the years but nothing too high end. The reflections I would say is only a problem if the SPDIF stream is used to directly clock the DAC or that the extracted clock is used in the digital filtering. Even then it would depend on how 'immune' the DAC is to jitter. As you mentioned somewhere above, it's all down to the implementation and results and observations may vary from player to player and cable impedence too.

Hi DaveF,
Can you expand a bit on what you mean by " if the SPDIF stream is used to directly clock the DAC" & "how immune the DAC is to jitter". I'm not being contentious just trying to fill in my blank spots (unfortunately, I can't fill in my bald spot :)).

[DaveF]

very interesting thread jkenny. I've designed a few SPDIF receivers and have been involved in one or two DAC designs over the years but nothing too high end. The reflections I would say is only a problem if the SPDIF stream is used to directly clock the DAC or that the extracted clock is used in the digital filtering. Even then it would depend on how 'immune' the DAC is to jitter. As you mentioned somewhere above, it's all down to the implementation and results and observations may vary from player to player and cable impedence too.

Hi DaveF,
Can you expand a bit on what you mean by " if the SPDIF stream is used to directly clock the DAC" & "how immune the DAC is to jitter". I'm not being contentious just trying to fill in my blank spots & learn.

The SPDIF stream as you may know is a single bit stream that contains the data and the clock frequency 'encoded' within it. At the receiver end, this clock needs to be extracted from the stream. ALL digital bitstreams will contain jitter. It cannot be avoided, even if reflections are present or not. Again, how much of it there is, is down to implementation and the designer. The extracted clock will in effect be derived from this incoming stream. Therefore it will inherit any jitter already on the SPDIF stream. Using multi-stage PLL's can reduce this jitter but it can be tricky. So if this 'jittery' clock is now used to clock a DAC it will drive the DAC with subtle timing variations which is not ideal.

Some DAC's can be clocked separately using a dedicated high precision crystal, therefore not inheriting any jitter from the SPDIF stream. But you will then have transfer the normal data from the SPDIF clock domain to the DAC's own separate clock domain. This presents its own problems and there are various ways to deal with such as a type of 'flow control'(which I posted about before on the old forum: dCS would use such a scheme over firewire I'm sure) or use Asynchronous Sample Rate Conversion or use low jitter locking PLL's. It can all get a bit theoretical from this point on. Normally this stuff is designed into the DAC chip itself so its not something than can be tinkered with at board level.

Just a few thoughts. I've written that off the top of my head so Im sure I've left out some of the finer details.