[art]

Let's see how far I can get, without any sophomoric interruptions, ok?

There have been a lot of 'scope shots, and other stuff, proclaiming what attenuators can and can not do. Mainly on the DIY sites. Part of the problem is while they used real-world products, it is not much a of stretch to say it may not be the best source, to use for this.

Here is what attenuators do:

1.) They attenuate the incident signal, by the amount of the pad.
2.) They attenuate the first reflection, by twice the amount of the pad. This is because this reflection has to make 2 round trips, so it gets hit twice.
3.) Some folks think it will fix ringing, overshoot, and other design problems. It will help some, but counting on them to be a magic bullet........well, maybe you ought to address those problems, without worrying about pads. It can do things, to make a crappy output look better, but it makes more sense to fix the problem, rather than hope a pad will fix it.

OK, enough sarcasm. On to the pics.

First, we need to know the set-up. Output is some SPDIF gizmo, with a fast rise time. (Under 1 nSec. Important.) Output goes to a Tek 475 'scope, through the shortest possible means, which is a 75R BNC female-female. The 'scope has a Tee, and yes, it is 50R, with one end going to the female BNC, and the far end going to a cable. The 'scope has 20 pF or so, of capacitance, so it will have a slight effect on the waveform. Let's ignore that. It really doesn't make enough of a difference to get excited over. At least not for this experiment.

The cable is a 5 m or so cable. For the first picture, the far end is unterminated. After the signal rises, at about the 1.5 div mark, it goes up, about the same amount as the incident wave. What happens is the reflection coefficient is +1 (because it is an open circuit), so you add +1 to the original 1, to get 2. Yes, that may sound simplistic, but that is how it works.

At around one more division, there is a slight negative spike. This is the first reflection! As it hits the source, there is a negative spike, as the reflection is terminated. Part of it bounces back, and you can see that as slight positive spike. This is the second reflection. Eventually, it is absorbed, and it dies out.

This is what it looks like:



OK........everyone have some idea what reflections look like? If not, let me know what needs to be clarified. I have been doing this for around 40 years, so it is routine.

Here is what happens when you terminate the far end:

(The vertical scale for the unterminated shots are all 1/2 vertical scale. This is so the whole waveform can fit on the screen.)



"Why is there a glitch?"

The glitch shows the cable is not really 75R. It is 2 ohms or so higher. OK, before someone freaks out, you have to understand something about coax cables. Almost all of them are spec'ed to be +/- 3 ohms. Not just at the end, but how much it can vary, along its length. This is within spec. Good cables are spec'ed at +/- 1.5 ohms, but it comes at a price. You are not likely to run across many. There are precision ones, spec'ed for +/- 0.75 ohms. They are only found in test labs, so don't ask where you can get one.

But, you will notice there is no first reflection. Since the load almost perfectly matches the source, and the cable, it is absorbed. At least to the point where what little makes it back to the source, is non-existent.

"Yeah, but what happens when you hook it up to a real DAC?"

Glad you asked!

OK, here is one. Contrary to what some know-it-alls will say, you need more than a ohmmeter to get the impedance right.

Yes, it has a RCA. Just like most of yours. What happens, once it is inside? Well, it isn't mine, so I am not going to take it apart, and nit-pick the guts. Let's just agree it is a lot like what you have.

Here it is...............!



Still like RCA jacks?

"Why is there a small step down, right before that big spike? What caused that?"

I inserted an variable attenuator, in the line. Right before the DAC. Had to run a very short cable, from the attenuator, to the DAC. That is the short step down. (Yes, its impedance is lower than the main cable. Closer to 75R, maybe even a tad lower. I am doing this at the house, so I don't have access to the TDR, to give a 100% accurate answer. But, I think it helps show the effect cables and stuff have.)

OK, since there is an attenuator, and what I'm sure you guys are dying to know.....................

2 dB:



4 dB:



6 dB:



8 dB:



10 dB:



15 dB:



20 dB:



Gee, when we get to 20 dB, you don't see the DAC! All you see is the glitch, caused by the short cable.

"My DAC won't lock, with that much attenuation."

Not the point, of the exercise. The point is to show what attenuators do. Not whether or not it will work in your system.

"Well.............well.............we don't want to use one of those stupid 15' cables. Why don't you show us something we use?"

Fair enough. I think since we have an idea what reflections look like, and since I can not hook up the DAC, with only a 1 m cable, I am just going to show open and terminated conditions.

1 m, terminated:



1 m, open:

(Remember, 1/2 scale)



Yuk, eh? Look at all that nasty ringing, that wasn't there, with the long cable.

Let's magnify the horizontal scale, 10x, just to get a close look:

The terminated condition:



And, the unterminated (open) condition:



Not something you want to come home to.........

Ok, here is the key, to this:

We went from a 5 m cable, one that is so long, almost no one wants to use it, to one that is 1 m, which is what everyone wants to use.

And.............

This is with a very fast source. It has a rise time, around 0.8 pSec. The problem is...........

A lot of you guys have transports/sources that are closer to 3-4 pSec. And what is the significance of that?

That first reflection...............guess what.............guess where it ends up.............yeah, somewhere in the transition portion of the waveform.

I will let you guys decide how attenuators will (or will not) help, with really short cables, on typical HC logic sources.

"Can we ask you one more question? Where should we stick the attenuator?"

Good question. I would stick it, on the end that has the poorest return loss. And since no one knows that (except dorks, like me!), that isn't much help.

Let's put it this way: you now know what it does, when you stick it on the DAC end. Maybe tomorrow, I will do the same tests, at the source. The results may be skewed, as the source has a very low reflection coefficient. So, if someone asks nice, and/or I find the time, I will run a short version.

Hope this clarifies some of the myths and other confusion.

Pat

[art]

Oh, forgot to add one minor detail...........

The termination was connected through a 50R barrel. I did this, on purpose, so that it would be easier to see where the termination is. It would have been hard to see, with a 75R barrel.

Pat

[Speedskater]

What time base are we seeing?  Or what is the frequency of the square wave?

[art]

It is a 44 kHz SPDIF signal. I forget what setting, but it was near the end of the scope's range. Probably 0.02, or 0.05 uS. Probably the latter.

Pat

[Speedskater]

What's the fundamental pulse frequency of a 44 kHz SPDIF signal?

[art]

The long pulses will be 44.1 kHz x 32 = 1.4112 MHz.

Pat

[stormsonic]

Nice. Thank you, Pat.

What time base are we seeing?  Or what is the frequency of the square wave?

Theoretical max. speed is 300.000 km/S or 0.3m/nS (nano-Second). Inside cable signal is slowed down.
Velocity factor for 75R coax is 0.67 for RG-59. Speed of signal inside coax cable is 0.3 m/S * 0.67 = 0.201 m/nS or 4.975 nS for 1 meter.
If lenght of the cable used is 5 m, then reflection must travel through cable twice, this is 10 m.
10 m * 4.975 nS = 49.75 nS. Reflection on the pic is visible 1 divider after signal launch, then divider used must be 50 nS or 0.05us.

[art]

Not sure about that...............there are also short pulses, at 2.822 MHz. I'll have to look at the 'scope, when I get home.

Wouldn't hurt to measure the cable, as I am guessing on the length. Plus, that 'scope is so far out of calibration, which is why it is at my house, and not the lab.

Pat

[Speedskater]

The long pulses will be 44.1 kHz x 32 = 1.4112 MHz.
Pat

I thought that there were some additional housekeeping and error correcting bits in the signal?

[art]

Yes, lots of bits. No error correction.

http://www.cirrus.com/en/pubs/appNote/an22.pdf

Pat

[art]

OK, looked at the 'scope settings. It is 0.05 uSec. Did not check to see how well it is calibrated. Not sure that is germane, but I am sure someone will ask.


Hooked up a device, with a HC output. I'll post some pics, when I get time.

Pat

[stormsonic]

yes, it should be 0.05 uS or 50 nS. I'm using 4m RG-59 for measurement and got reflections ~ 40 nS later. Capacitive termination.

[art]

Someone wanted to know when I was going to show the same pics, but with a more "conventional" driver.

OK, the source used is something a lot of you guys have (and like to modify), with a HC logic driver. So, a much slower rise time, than the unit used in the previous examples.

OK, open circuit:



And with a termination:



The thing that I first notice, other than the exposure and brightness are not the same, is that a lot of the fine detail, in the waveform, is gone. This is what happens when the waveform is slower. You really wouldn't know the BNC barrel, that connects the termination, is 50R. It does show up, with the faster driver.

Anyone who has paid attention to my numerous postings will recall my stating that using a faster output allows for a shorter cable, but you need to control impedance, over a wider range.

You will see, once the DAC is hooked up, the small step, that is caused by using a very short cable from the variable attenuator, to the DAC, is no longer noticeable. It is still there: it just can not be resolved, with a slower driver. If all you have is a Tek 465, it will suffice, as a TDR, but to really see what is going on, you need something that works into the GHz range.

So, the DAC, hooked up, with the attenuator at 0 dB:



5 dB:



10 dB:



15 dB:



20 dB:



Next, we have a series of pics, but using a short (1m) cable. First, with the cable open, on the far end:



Hmmm.......almost hard to see the glitch, in the rising waveform, that indicates it is unterminated. So, let's put up the terminated one, for comparison. (Again, the unterminted has half the vertical scale, as all the others.)

Terminated, with short cable:



Well, looks fine. Let's see what happens when the DAC is hooked up. (Same way as before. Yep, you won't see the short cable.)



You can barely tell there is a cable, of any kind, at the input. The slow rise time obscures it, almost to the point of it not being detectable.

One shot, with an attenuator, as the reflection is so close to the rising edge. This is with 10 dB, of attenuation.



The only way to really see any of this, is to use the 10x magnification, on the horizontal scale. So, from the beginning, open circuit:



And with a termination:



(Looks good!)

Now,  the DAC, with no attenuation:



OK, now we can see something! On to the 10 dB attenuator:



Doesn't look much different, from a termination.

So, any of this mean anything?

I hope you have learned how attenuators can reduce reflections. And how much rise time and cable length factor into the reflection problem. Fast rise time: you can use a shorter cable. But, a lot more of the little squiggles show up. They get smoothed over, when you use a slower source. But, the first reflection gets harder to separate, from the rising edge.

While attenuators can help, there is the obvious problem of lowering the signal, too much. Some DACs put a small linear stage, ahead of the SPDIF RX chip. I have seen some DACs that will stand over 20 dB, with a source, of the proper level, driving it. There may be other benefits, of the isolation the linear stage provides.

Now, there are lots of folks will say "Well, yeah, I can see differences. They aren't all that much. Do I really need to worry about any of this?"

If you stop and think about all the other tweaks and tricks you guys try, how many of them yield radically obvious changes, that can be measured on any multimeter?

Yeah, just what I thought. All of this stuff is in the realm of diminishing returns. So, why should this be any different? You can try these suggestions, for almost nothing. Sure, you can buy a fancy industrial attenuation, for under $20. Which, is cheap by a lot of other tweaks. Or, you can make a pad, built right on the input (or output) jack, of your unit, for the cost of a few resistors. Yes, there may be some squiggles, that will show up on our TDR. But, with the typical HC CMOS chips, most of you use, you won't see the effects.

Of course, if you don't believe in tweaks, and have no interest in trying any, then why are you reading this?!

Enjoy, guys. Gotta get back to filing orders.

Pat

[art]

After reading all of this, I thought about a few pics, that should have been added.

OK, this is the fast SPDIF driver, and the 1m cable. Hooked up to the DAC. Shots at 0 and 10 dB.

0 dB:



10 dB:



OK, compare those to the short cable, and slower driver.

Pat

[kyrill]

Dear master Pat
Plz inform the public

the sound difference is immense
WITH attenuators

[Tyson]

So, why deal with SPDIF at all?  I've seen some DAC's that are USB-I2S and that seems like it would be a more elegant solution?

[art]

Lots of folks use it, despite its drawbacks.

So, you think I2S is a free lunch?

Or, how 'bout sending it 40'? Yes, I am making one, that length, for a customer. Maybe that helps to explain. (No, I don't know why he wants it that long. Maybe he will share it. I doubt it.)

Pat

[jneutron]

I preface my post with a disclaimer...nice work, thank you for the time and effort.

Let's see how far I can get, without any sophomoric interruptions, ok?

Well, you made it this far...oh well...   

OK........everyone have some idea what reflections look like? If not, let me know what needs to be clarified. I have been doing this for around 40 years, so it is routine.

40 years.  You must be tired.

Yes, it has a RCA. Just like most of yours. What happens, once it is inside? Well, it isn't mine, so I am not going to take it apart, and nit-pick the guts. Let's just agree it is a lot like what you have.

Here it is...............!



Still like RCA jacks?

hmmm.  From what I recall, the characteristic impedance of an rca is below 50 ohms, on the order of 30 give or take.  Yet the signal is overshooting.  I would be more inclined to suspect what's on the other side of the rca, not the rca itself.  Nano risetimes are still significantly longer that the geometry of the rca discontinuity.  edit:  that less than 10 nSec shelf just before the positive spike may indeed be the rca reflection, but the circuitry on the other side very quickly dominates the reflection.  I suspect the designer used a small wire from rca center to pcboard, or the center conductor of a coax, and that inductive loop is the cause of the high z reflection.

This is with a very fast source. It has a rise time, around 0.8 pSec. The problem is...........

Art, did you mean .8 nSec?

Cheers, John

ps...again, EXCELLENT post.  Thank you very much.

[art]

Thanks for pointing out the typos. 0.8 nSec. (It was late at night.)

As for the RCA......

Yes, it should show a negative going spike, but as you have deduced, whatever is after it is causing that spike. My guess is that this unit has the SPDIF  transformer that everyone (well, almost everyone) thinks they must use. Guess I should take the lid off, and verify.

(Stuck at home today............ice everywhere. So, something to do.)

The small downward step is indeed the cable. Its impedance is a bit less than the other piece.

Pat

[JoshK]

So, why deal with SPDIF at all?  I've seen some DAC's that are USB-I2S and that seems like it would be a more elegant solution?

I'd love to know what the draw backs to using I2S are (whatever cable format) too?  I do recall reading that sending the signal requires really fast drivers.  Is this really difficult to design?  Are reflections worse?

I do know that the vast majority of my sources and DACs still have SPDIF, not I2S.