Electronics for Serious Audiophiles
The Legato USB-SPDIF Asynchronous Converter Technical Information and Other Stuff
We realize that some of you who may be potential buyers are a tad concerned about the fact that you have to use a BNC-RCA adapter. Yes, the Legato will perform at its full potential if used with a DAC that has a BNC connector. Until we introduce one, some of you may be stuck with using an adapter with your DAC.
It is important to point out that you must use the BNC-RCA adapter at the DAC end. We have made nuemrous measurements, to verify how much the adapter affects things, on both ends. The only end that makes sense is the DAC end.
"But what if I already have a cable, with RCA connectors?" Why can't I just stick an adapter on my cable, and use it?"
The cable, provided with the Legato, has been designed with this possibility in mind. There are some sound technical reasons why this is so. We have decided to share some of those here. Granted, this discussion may be more technical than some of you want to digest. However, would you rather have someone address your concerns with technical facts, or blow smoke in your face with marketing mumbo-jumbo?
We think that you prefer the former, even if it is a bit "geeky" in its presentation. (We are nerds, remember. OK, our CSB isn't, but she made a conscious decision to hang out with us. Feel sorry for her!)
Years ago, we made a series of technical measurements, and decided that in order for reflections to not be a problem, that the return loss had to be -30 dB, or lower.
"OK, what does that mean to us?"
Return loss is simply a way of quantifying how much of a signal that is sent down a transmission line that will not be absorbed, and reflected back to the source.
Any piece of electronics will reflect a certain amount of energy back to the source. Doesn't matter what it is, who designed it, and how good it is. It is the job of the designer to minimize it. Through our research, we have conclude that the amount needs to be below -30 dB. Or, in less technical terms, around 3%. So, if you have a return loss of -20 dB, that means that only 10% is reflected back to the source.
Getting a return loss of -30 dB can be a bit challenging, especially if you want it to be so from 100 kHz, up to 100 MHz. Might be a bit more bandwidth than needed for SPDIF, but we feel that it is a goal to shoot for. This goal is a bit harder to attain when you consider that the output is transformer coupled. On the low frequency side, one has to contend with cores that are very small (by necessity), and therefore limited by the lack of magnetic material in the core. On the high frequency side, you have to deal with technical gobbledy-gook like leakage inductance and interwinding capacitance. In other words, it is really tricky.
Here is what the output return loss of the Legato looks like:
As you can see, it is easily below -30 dB, throughout its entire range. Even with the transformer! Quite a tidy accomplishment. In the mid-band, you can see that it is in the -45 dB range. Which is less than 1%.
"OK, but what happens to that return loss figure when you add a BNC-RCA connector?"
Actually, you will be using a BNC-RCA connector on the DAC end, but since we don't have your DAC here to measure, we will stick it on the Legato for demonstration purposes.
Be thankful that we don't have your DAC here to measure. Most of you would be ill if you saw how poorly they measure. But, we digress..........Take our word for it: if it was -20 dB, consider yourself very lucky. Many of you are not lucky.
Here is how adding a BNC-RCA connector on the Legato looks like:
This measurement was made with a BNC-RCA adapter, coupled with a RCA-BNC adapter, as this will more accurately take into account that your DAC has a RCA input connector.
Yes, it does cause the amount of the reflections to rise, but it makes it to almost 20 MHz before it exceeds -30 dB. And remains below -20 dB out to around 60 MHz. (Remember, your DAC will not measure this good.) All things considered, this is pretty good.
But, you are going to be using the included BNC-BNC cable with your new Legato and your existing DAC. Remember that we said that the cable helps to alleviate the increase in return loss?
Well, here is what it looks like, first with the cable all by itself:
And when you add the adapter:
Well, all things considered, this is not too bad. Yes, the return loss does rise to around -25 dB at the high end, when measured at the high end. But adding the adapter(s) does not have much effect until you are above 50 MHz.
Now, before you jump to the wrong conclusion, you must keep in mind that this cable is included at no cost.
"Yeah, yeah, yeah..............making excuses............."
No excuse. But are you interested to see what happens when you use an expensive cable in its place?
Here are the same measurements, made with our U-byte © cable. First, without:
And then, with adapters:
Clearly, the U-byte © cable is better, but it costs $200. (Remember, the other one is free.) It stays below -30 dB throughout its range, and does not appreciably affect the mid-band performance. The adapter has about the same effect as it does on the other cable.
"Yeah, yeah, yeah..............still making excuses............."
Think so? Ever see these measurements for your cable or DAC........................?????????????
Yeah, thought so!
Anyway, back to that subject about how the cable helps to offset this problem.
Let's assume that you have a DAC that has an input return loss of -20 dB. That means that 10% of the signal is not absorbed, and bounces back to the source. If the source also has a return loss of -20 dB, then 10% of that reflection (10%) is re-reflected back to the DAC. This means that 1% of the total amount arrives back at the DAC.
Is this a problem? Can be, but let's put this aside for a second.
In the case of the Legato, which has a return loss of less than -40 dB, then less than 1% of that 10% which has bounced back can be sent forward to your DAC. In this case, we are only talking about less than 0.1%
See why return loss is so important?
Now, if you have a fairly short cable, that only takes the signal 1 nSec or so to get there, the re-reflected pulse gets there 2 nSec or so after the initial pulse. So, it is possible that the reflection can arrive at a critical point in the waveform: the transition point.
1% of the original signal, arriving at the wrong time, may not sound like much. Granted, it isn't. But that doesn't mean that it won't muck up the sound. There is a good chance that it will have some effect. Not enough to generate errors, but enough to add some jitter.
So, our approach addresses the problem in 2 ways. One, the Legato has a very low reflection coefficient, so it will not allow very much of any reflected signal to find its way back to the DAC. Even if it has a reflection of 10%, 20%, or even 40%.
The other little trick is the included cable is longer than most. All that length, which may seem needless, actually helps. The length of this cable has been selected to work with the Legato. Taking into account the rise time of the Legato, it is possible to design a cable that will delay how long any reflection can arrive at the DAC, which would be able to degrade the signal.
So, very little reflected energy, and it arrives at a point where it can't really muck up the sound. Simple as that.
We had a thought after reading the initial draft of this text, that perhaps we could include some data on how similar products measure. So, just for grins, and something to jaw about, here is how two competitive prodcuts measure.
You will note that we said competitive products, and not competitors' products. These two products are very popular products that use a different scheme for getting music off of your computer, and to your DAC, via SPDIF. (They also work at 44 kHz, and not high-rez!) So, we really don't compete against them directly, as they use a totally different method.
Now, before any of you start to get worked up, and want to send hate mail to our CSB (she has enough problems dealing with us as it is; please don't make her job any worse!), we would like to point out that we admire the guys that turned out these products. Granted, if they had employed us as consultants during the design phase...........for only a few pennies more.............we could have helped it a fair amount.
But, alas, when you make things in the quantities that these were made in, the bosses don't want to hear diddly about "a few pennies more". In any case, don't think that we are "slamming" these products, or the folks who designed and manufactured them. They are really neat products, that pack a lot of performance and features, into a neat little package. And one that does not cost an arm and a leg. So, this is not an attempt to deparage them. Merely a way of pointing out the extreme lengths that we went to in the design of the Legato.
Here is what the output return loss of "mystery product #1" looks like:
And "mystery product #2":
Well, well, well.............isn't that interesting? One stays below -30 dB return loss out to only 2 MHz. The other, a bit more............it makes it to 4 MHz.
Both of these products use RCA outputs, and are not transformer coupled.
One more example of ways that we refuse to take short-cuts, can be shown here:
We could use an inexpensive, PCB-mount BNC connector, instead of the expensive one that we use. You know........the ones that you can buy for $1 or $2 from any number of popular electronics distributor. And terminate it, directly at the point where it enters the PCB. (Oh, wait........that isn't how it is terminated in the real world!) That is what is shown in the above plot. Not too shabby, only really hurts above 20 MHz. But we have chosen to use one that costs 4-5 times as much. In a product that costs $500. Not $5000, $500.
If it was a $5000 product, we already know what connector that we would use, and it costs around $25.
Still not convinced...................?
"What about a RCA connector? How bad are they, all by itself?"
Well, here is what a BNC-RCA and RCA-BNC look like, terminated by a precision 75 ohm load:
You will notice that it looks pretty much the same, as when measured at the Legato output. OK, not any worse, all by itself. But this is without any wire, or PCB traces. Won't look that good in a real application. Remember, the Legato has a very small reflection coefficient. In any real application, you would also have to take into account the wire and/or PCB traces from the RCA connector. A poor implementation in that regard will show up.
Now, granted, this may all be overkill. That is a real possibility. But we weren't about to settle for just a "decent" product, when we could make an "exceptional" product for only slightly more. The Legato may be priced as an affordable audio product, but it was designed as if it were a serious audio product. Try one, and if you don't agree with us, we stand by our "30 days, no questions asked" policy. Try one: you will be surprised.
OK, assuming that you have not fallen asleep yet, there is one other technical detail that we would like to highlight.
In order to make a top-notch clock (oscillator), you first must have a high-Q crystal.
"What is 'Q', and why is it important? We get the reflection part............please don't tell us any more about that."
Q, which is short for quality factor, is simply a fancy way of quantifying bandwidth. The only wrinkle is that you take the frequency of the device, and divide it by its -3dB bandwidth. It is important, because the narrower the bandwidth (higher Q), the sharper the frequency distribution of the oscillations will be. A low-Q crystal will have a lot of noise, on either side of the clock signal. In geek speak, this noise is called phase noise. A higher Q crystal will do a better job of filtering out the noise that is present in the circuit, and will therefore have less phase noise. Which is significant contributor to the total amount of jitter.
Let's look at some crystals that you might find in a similar product. A lot of them will contain a crystal that measures like this:
This comes out to be a Q of 70,00. That is just so-so. It is not hard to get a crystal in this frequency range that is 100,000, or better. If you are lucky, you may find a unit that has a crystal that measures like this:
Well, that is better, at 95,000. But not still good enough for us. We use a crystal that measures like this:
Now, that is more like it! This crystal has a Q of 205,000. Even more impresssive, when you take into account that it is an inexpensive part, that is easy to find. (Wonder why no one else uses it?)
However, one has to be careful. Just because we buy a big bag of crystals does not mean that all of them will be suitable for our use. In any large batch of crystals that we purchase, there will be a handful that are not up to our level of quality. Here is one that we do not deem suitable for use in the Legato
Not that bad, at 80,000. But not much better than the mystery crystal you might find in another product, from another manufacturer.
Sure, we could forego the added expense of testing each and every crystal that goes into the Legato, but we don't. We could not only save that expense, but use all of the crystals that we purchase. Instead of throwing some out. (OK, we are notoriously tight-fisted............we probably won't throw them out, but find some way to sell them to hobbyists. For probably not much more than the cost of postage.) But you can rest assured that this crystal will not find its way into your Legato
Remember.........this is a $500 product, not a $5000 product. Where do you think that you can purchase a product, as such a fair price, that has this much attention to detail built into it?
Let us know, because we may try to get a job there!
They said it couldn't be done...........
"What did they say, and when did they say it?"
You're right: they said nothing. Just a cheap tactic, to see how closely anyone reads this stuff. (And why.)
Anyway, there may, or may not, have been, some changes to the Legato. For instance, here is the phase noise plot, of both the clock, and the S/PDIF output, of a Legato, grabbed at random.
Did I say random? I really meant to say the first one, of the latest production run. Could have been the second one............or the third one, but, no...........it really was the first one, and, yes, it was random. Even if there really were only 3 in the batch, it could have been random. Since there are more than 3, and..............oh, never mind. Here, the plots:
"Uh, care to explain what the hell we are looking at, wise guy?"
Sure. The blue plot is the phase noise of the clock. There is a marker, at 1 Hz, that shows the phase noise, at that frequency. It is around -88 dBc, or so.
"Is that good?"
Well, one of our competitors is proud that theirs is -70 dBc, at 1 Hz. Considering a lot of other brands are only -60 dBc (if they are lucky), yes, I suppose you could say it is good.
Which makes ours, at almost 20 dB lower, killer!
"Well, ok, nerd boy, but what does that mean in terms of how good it sounds?"
I just told you: killer!
Seriously, it means your digital system can now sound almost as good as your reference analog system. If it doesn't, won't be our fault. We held up our end of the bargain.
"Yeah, ok............but, what is the other trace? Why is it lower? And why is the noise floor higher?
The output, of the Legato, is reclocked. Which means the output jitter should be the same as the clock jitter. Which, it essentially is. The jitter number shown, which is the jitter, measured from 1 Hz, to 1 kHz, is less than 1 picosecond. Which is pretty damn good! (A lot of gear is spec'ed at a jitter frequency that starts at 1 kHz. Not stops at 1 kHz. Some folks go down to 100 Hz. Some may go as low as 10 Hz. Can't think of any that go down to 1 Hz. So, there!)
You have already noticed the noise floor is higher, for a reason. The input signal is 10 dB lower, in level. That accounts for a lot of it. Plus, there are tons of sidebands, starting at slightly over 200 Hz, from the signal center frequency. (Which is why it stops at 1 kHz. Nothing but sidebands, and noise, above that frequency.) The "RMS Jitter" is calculated, by the s/w of the fancy gizmo we use, by setting markers at the frequency range of interest. Which, to us nerds, is 1 Hz to 1 kHz.
Since the output frequency is 1/8th the clock frequency, it should be 18 dB lower, than the clock jitter. Which, it roughly is. The number is a tad higher, due to that rise, in the noise floor. But, it is essentially the same as the clock jitter. Which means the cheap reclocking circuit actually works, and is doing its job.
We could put up the plots of the output signals, of some of our competitors. But, why embarrass them any more than we already do? Unless they start to blather, on the forums. They do so, at their own risk! So, you may want to warn them. Maybe not.
"Yeah, well, you are just cherry-picking a good one, off the line."
Yeah, you are right. We did. We didn't pick the best one!
"You have better.................??
Yeah, there are. By a few db. And some, a few dB worse. But, that is typical of what we presently produce.
"Oh. But aren't you just cherry-picking crystals?"
Well, kinda, sorta................yeah. We confess. Buying a big giant bag of crystals...........sorting through every damn one of them...............figuring if we buy a big enough bag, that the cost per crystal is not too bad. Even after we allow for tossing 75% of them away.
So, there you have it: the secret of why it sounds so damn good. And proof. Of course, some folks might demand a double-blind listening test, and blather on how you can not hear jitter, at less than x microseconds, etc. (We know otherwise.) Just ignore them.
Or tell them "It takes a company, like Analog Research-Technology to make digital audio sound like analog." Not that they will believe you. As long as you, our customer does. We do give you 30 days, to find out if we are telling the truth, or just full of techno-babble.
So what the hell are you waiting for??? The Veloce, to be ready?
"As a matter of fact: yes!"
Well, stuff that grow out of the Veloce development found its way into the Legato. So, only natural some of the new stuff will find its way, into the Veloce.
Hopefully, January, 2012. Stay tuned. In the meantime, we need the money. Buy now!