ISO REGEN launch thread! (product web page up; photos, etc.)

[JohnSwenson]

7 hours ago, Em2016 said:

Hi Alex @Superdad @JohnSwenson

 

Has John heard the absolutely final production build too?  Or has he left the final approval to your expert ears !

 

Only asking because of his movements with looking for a home etc and wondering if he's had a chance to hear the final thing. Perhaps it was difficult for him (logistically) to hear the final thing, so thought I'd ask.

 

Only asking out of general interest and as a John S fan, no other reason.

 

Cheers, Sean

 

No, I have not been able to listen to the final production board. I have no separate DACs with me at this point. The only thing I have is a SqueezeBox Touch which has an integrated DAC and headphone amp. (oh yeah the LPS-1 to power the SBT). There is no way I can listen to the board at this point, all I can do is plug a USB stick in and see that it does in fact work as a hub and transfer data.

 

When we moved I had to really streamline things and take the minimum amount of stuff, basically what would fit in the car. It will be July before I can actually hear this.

 

John S.

[JohnSwenson]

On the questions about what I have tested, my testing has been before we moved, and primarily concerned with getting it to work with all sources and DACs. This product was very difficult to get out, we found all kinds of strange interactions between the isolator chip and the hub chip, and in isolated systems period. The last several months before we moved I was spending almost every moment packing, but Alex kept on bugging me to get the ISO REGEN working, so I spent as much time as possible on it. Well I got it done the day before I packed up the listening room, so I have not had hardly any time to actually listen to the final design.

 

So any questions about what configuration sounds best etc, as far as I am concerned, here is my answer "I have NO idea". My testing in the last several revisions has all been about getting it to work, it now works, so you all out there can figure out what configuration sounds the best. You are going to be able to hear what this really does long before I get to.

 

As to theoretical analysis of specific configurations, this is only as good as what we know about a system and unfortunately we don't know the ultimate reasons that different things sound better, so it is hard to come up with any "configuration A is always going to sound better than configuration B" .

 

For example the one EVERYONE is asking (and not just here, I have a bunch of PMs and emails on this) is "what about the microRendu and ISO REGEN". First see above, I HAVE tried it but only to see if it actually works, it does.

 

From a theoretical standpoint, SQ improvement from an ISO REGEN comes from two different things, signal integrity and blockage of leakage loops. The ISO REGEN does have better SI, so from just that it should sound better. As far as leakage goes it is a more complicated. A microRendu when powered from an LPS-1 should be blocking leakage loops already so there should not be an overall improvement because of the isolation of the ISO REGEN, BUT there are other things involved here. How are you going to power the microRendu? One configuration is to power both from the same LPS-1. As long as the DAC does not take a lot of current this should work, BUT you are now adding the current spikes of both devices at the output of the LPS-1. Now the LPS-1 has very low impedance (one of the best out there) but it is NOT perfect, the current spikes from the load do cause some small voltage noise on the output. Putting both devices on the same LPS-1 will increase the noise seen by the ISO REGEN. Will this degrade the sound? Who knows. It may, it may not.

 

Then there is all the talk about the Ethernet, there are a lot of people talking about SQ changes with different Ethernet configurations. At this point I don't know what is going on. The way the Ethernet inputs are done is SUPPOSED to block leakage loops, but some of the reports seem to indicate that this may not be true, but we don't know for sure one way or the other. So IF there is some leakage current through the Ethernet input, then having the extra isolation of the ISO REGEN is a GOOD thing. BUT if you feed both microRendu and ISO REGEN from the same LPS-1 you are disabling that isolation so you won't get the isolation advantage, but there may not BE an advantage. I don't know.

 

So there are several combinations:

1) Same LPS-1 on microRendu and ISO REGEN

2) Separate LPS-1 on microRendu and ISO REGEN

3) LPS-1 on ISO REGEN and something else on microRendu

 

If you have the money and want the configuration that has the highest probability of giving the best SQ, go with #1.

 

BUT #2 or #3 may sound just as good and going with #1 is just spending money you don't need to spend.

 

Unfortunately there is no way I can tell you which it is going to be. You guys out there are going to have to do some experimentation and figure out what where it lies.

 

Remember that Alex has a 30 day money back guarantee, so you can get two LPS-1s and try different configurations. If you decide there is no advantage (or the advantage is so small it is not worth it) you can get a refund for the second LPS-1.

 

If you decide to do the trials, please let us all know the results, I'm very interested to find out how this comes out.

 

John S.

 

[JohnSwenson]

1 hour ago, Energy said:

@Superdad

 

Alex. Like with what John says regarding losing galvanic isolation if the MicroRendu and ISO Regen shares the same LPS-1 power supply, does the same go for if the ISO Regen and Singxer SU-1 shares one?

 

I know the ground from the ISO Regen will make contact with the SU-1's USB Input, however after XMOS, the signal goes through digital isolation chips and pulls new power from the internal power supply. If this power supply is also run from the LPS-1, doesn't both sides of the boards become linked due to shared ground?

 

Also, the ISO Regen takes 6-8V whereas the SU-1 takes 5V. Using a split power cable, how can I properly power both units with one LPS-1? Do I need to use a voltage drop resistor for one device when the LPS-1 is set to 7V?

 

And finally, Is the 90 degree rotated USPCB your most recommended configuration for use with the Singxer SU-1? I tried looking at it multiple times and doubt that horizontal placement of the ISO Regen behind the SU-1 can be made.

 

Cheers,

Timothy

Yes if you do manage to power the ISO REGEN and the SU-1 from the same LPS-1 you will be bypassing the isolation in the SU-1, but you still have the isolation in the ISO REGEN.

 

Alex has previously responded to the question about voltages and the SU-1, make sure to find that post for the details.

 

John S.

[JohnSwenson]

54 minutes ago, Energy said:

 

Now to my understanding, the ISO REGEN already regulates clean power from the Linear Tech. LT3042 voltage regulators to feed the Singer SU-1's USB Input (XMOS). With the isolation barrier working due to the use of separate PSU's, wouldn't it be best to use the LPS-1 to power the Singxer SU-1?

 

For someone intends to use a converter like the Singxer SU-1 after the ISO REGEN, wouldn't the improvements of the LPS-1 feeding the ISO REGEN be less significant than if it was used on the SU-1?

 

Cheers,

Tim

WOW fantastic understanding of the situation, it really nice to know that people are starting to "get" this stuff.

 

All three block leakage loops from the computer to the DAC. #3 still has a leakage loop from the SU-1 PS to the DAC, so I would not recommend it. Both #1 and #2 have no leakage loops. So it comes down to #1 or #2. My gut feeling is that #1 might be slightly better, but it is close so either one could be the best. Really it is going to come down to trying them in real systems.

 

John S.

[JohnSwenson]

For all the people wanting to use an ISO REGEN with a SU-1 and an LPS-1, I just talked to Alex about this. He has done it, the recipe is:

 

Set the LPS-1 to 5V, use a Y-cable to power both at 5V. (see the DIY DC Cable thread on how to make your own) The VBUS output regulator will be undervolted so it won't actually be regulating (but the regulator in the LPS-1 is the same as in the ISO REGEN). The measured voltage is 4.95 V on the VBUS coming out of the ISO REGEN, this is fine to drive the upstream side of the SU-1.

 

This will preclude the isolation in the SU-1, but the isolation in the ISO REGEN will still be working.

 

If it were me I would get the external power kit from the manufacturer and cut off the internal cable, leaving pigtails on the both connectors. Then use a piece of 4S6 starquad to replace the cable you just cut off.

 

John S.

[JohnSwenson]

1 hour ago, mikicasellas said:

 

 

Hi John,

 

What "power kit from manufacturer", where can i find it?

 

Best

The kit is talked about in the DIY DC Cable thread, page 4.

 

If anyone wants to discuss this please do it in that thread rather than here.

 

John S.

[JohnSwenson]

7 hours ago, Em2016 said:

Hi Alex @Superdad (or John @JohnSwenson)

 

I asked about eye pattern measurements of the ISO REGEN vs original USB REGEN. Completely understand if you're not yet ready to publish measurements but it would be nice.

 

But in the meantime you've mentioned that the ISO REGEN has much better USB signal integrity than the USB REGEN (and everything else).

 

How about output jitter performance? Is ISO REGEN's output jitter performance better than USB REGEN? You did mention jitter issues during the design phase associated with this new jittery isolator chip I recall.

 

If USB signal integrity is better and output jitter is lower and output noise is lower (maybe related to jitter) than the original USB REGEN and leakage currents are blocked (like an Intona) then this is the holy grail.

 

Appreciate if you can share some technical stuff on the above, not really in comparison to any other products (I was naughty for mentioning Intona there) but specifically against the original USB REGEN.

 

Cheers, Sean.

 

This one is tough. In order to try and measure this sort of thing I bought some older (then) top of the line test equipment from ebay, I can get great eye patterns with this, it is good for looking at the waveforms but useless for jitter measurement. The jitter from any of our devices is lower than what these devices can measure. The waveforms from the ISO REGEN look REALLY good. Unfortunately getting screen dumps is quite a complex task and very time consuming so I don't do it very often. I don't have any saved eye patterns so I'm not going to be able to post anything for quite some time. None of this equipment has USB interfaces that would make it easy, they work off of GPIB interfaces, or in the case of one of my scopes, a floppy! The scope take FOREVER to write an image file to the floppy, then I have to put it in a USB floppy drive to get it into my lab computer. BUT the image format is very old so none of the current programs will read it, so I have to run a series of conversion programs to get it into something that a modern browser can handle. Its a lot of work.

 

As for jitter measurements I have been looking for test equipment that can actually measure the jitter on actual USB data streams and have not found any that will get us into the area that our stuff works at - for any price. The ones that exist are designed for standards compliance not characterization at very low jitter levels.  So I'm designing my own. Unfortunately I can't start working on that until the new lab is set up, and then I will have lots of things to work on, so realistically it is most likely going to be a year before I have that device up and running. THEN I can actually measure the USB jitter. Oh yeah, the existing differential probes aren't good enough for this so I will have to build my own as well!

 

John S.

[JohnSwenson]

I've been thinking about whether I should make this post for several hours, but I finally decided to go ahead and do it now.

 

Alex and I are overjoyed with the enthusiasm shown for the ISO REGEN, but at least I am getting overwhelmed by questions asking us if a particular combination of components is going to sound better than a different configuration. 

 

The problem here is that this is all conjecture. I don't have ANYTHING to actually try things out and Alex only has a very small subset of all possible components out there so we can't actually set the components up and listen.

 

So it all has to be based on analysis of the components involved and MANY factors relating to how these devices interact with each other. This is not trivial to do, it is a lot of hard work. Most of the components people are asking about I do not own or have used extensively, so I have to go on the net and spend a couple hours researching them to try and make a decent guess about the circuitry inside. THEN do an analysis of possible leakage currents, sensitivity to them, USB signal integrity, sensitivity to said SI and a whole host of others. It takes all of this to make a semi accurate attempt at what the interactions are going to be. This takes me several hours for each system and is a lot of mental work.

 

After doing several of these today my brain is hurting and I'm wiped out. I'm just not going to be able to do much more of this going forward.

 

Both Alex and I want to take care of our customers and give our best at answering your questions, but this is just getting overwhelming for these types of questions.

 

So unfortunately I'm not going to be able to respond to such questions for several days (I have to recharge the mental batteries). Please don't think I'm try to "put off" our customers, I just can't give these questions the resources they deserve.

 

Please remember that nobody has been able to actually listen to most of these combinations, so it is all a guess anyway. I know everybody wants to know how all possible combinations are going to sound before buying anything, but the reality is: nobody is in a position right now to actually test it, and guesses are not terribly accurate and take a lot of work to make them any better than random chance.

 

The reality is that until you all start getting your devices and start trying them in various combinations, no one will really know.

 

Remember Alex has a 30 day money back guarantee so you can try different approaches and return what doesn't work.

 

I love to talk about "my baby", what it is and how it works and interesting interactions of this technology, it is just the requests for us to say which specific setup is going to sound better than another that I'm going to have to not respond to for awhile.

 

Thanks for your understanding.

 

John S.

 

 

[JohnSwenson]

3 hours ago, SWL3600 said:

I'm not clear what the red switch is for? Vbus? I don't even know what Vbus is....

 

I will be inserting the new ISO Regen in place of my USB Regen, which is powered by the LPS-1. Will the voltage setting stay the same? How should the 'Red Switch' be set?

 

Chain will be:

 

Laptop

Curious Regen-Link 

ISO-REGEN 

USPCB adaptor

IFI Micro-link USB-Spdif converter 

DEQ 2496

Maverick Audio Tubemagic D2 DAC 

 

Thanks!

The red switch on the ISO REGEN defeats the isolation, it shorts the the upstream and downstream grounds.

 

The switch on the USPCB turns the 5V (VBUS) off for the downstream device. Both of these are for very unusual circumstances that most people will never encounter so can pretty much be ignored by most people.

 

One of the purposes of the ISO REGEN is to isolate the grounds between the upstream and downstream ports. (this also includes isolating signal and VBUS) This blocks leakage currents (not enough space here to talk about leakage currents, see the hundreds of posts I've made on the subject) and will be a significant SQ improvement for most people.

 

BUT we noted in our tests that some systems that have already gone through extensive isolation techniques that the isolation in the ISO REGEN causes the entire audio system to be so isolated from the outside world that static charges build up on the whole system which every now and then discharge through some path (unknown at this point) and when they do the data traveling to the DAC gets interrupted causing a click or pop in the audio.

 

The switch is there to short the grounds which will get rid of the ticks and pops.

 

We have found that there is usually another way to fix this problem, which is to make sure that one and only one place in the entire audio system is grounded. This prevents the buildup of static charges and does NOT cause any leakage loops and ground loops. This is the preferred method to get rid of these ticks and pops, the switch is the backup in case there is no way for you to ground some part of the system.

 

Any device with a three pin power plug is going to automatically be grounding the system (DAC, preamp, poweramp etc) so if you have something like that currently in your system you probably don't have to worry about that.

 

You can also use the switch diagnose the problem. If you add the ISO REGEN into your chain and start getting ticks and pops, try flipping the switch, if they go away you know you have a system that needs grounding at some point. So then ground something and flip the switch back to the ISO position and the ticks and pops should be gone.

 

So to recap most people can ignore the switch, it is only need in a small number of systems.

 

John S.

[JohnSwenson]

4 hours ago, rickca said:

I hope you don't mean the kind of loud, potentially damaging pops you can get with PCM/DSD transitions.

No They are not very loud. In load music you may not even notice them.

 

John S.

[JohnSwenson]

Several days ago I mentioned that I had no way currently to listen to the ISO REGEN that Alex sent me, well that has changed!

 

I was very annoyed at not being able to listen to this so I got on to Amazon and orederd an inexpensive DAC so I could try out the ISO REGEN. My sister had sent me an Amazon gift card, I had to use it , right?

 

I finally settled on the Micca OriGen+ DAC. It came in two days ago, I hooked it up to the SBT and it was recognized right away, no problems. I was very relieved because I was concerned the LPS-1 would not be able to power both the SBT and DAC. Well it worked fine. I wanted to let it burn in for a day before listening much, so I just listened enough to properly set levels etc and walked away.

 

Yesterday I started listening to it plugged in directly to the SBT. It was kind of interesting, it has better bass, but it wasn't as "open", the subtleties, intonation etc were not quite as good as the SBT.

 

Then on to the ISO REGEN. Well I couldn't do it, I just had one LPS-1 and I didn't have a different supply to run the SBT. So I cobbled together a Y cable with both 2.1 and 2.5 on the ends (using the Canare starquad of course), and set the LPS-1 to 5V (for the SBT, it is famous for getting fried at higher voltages), but that meant the ISO REGEN was running at 5V, which means the VBUS regulator feeding the DAC was not regulating.

 

Well I plugged it all in and turned it on and the SBT just started playing music. With the USB REGEN I had to do an extra reboot to get the SBT to figure out that the DAC was behind a hub, but with the ISO REGEN it just worked right off.

 

Well it was a big difference, the sound was so "smooth", edginess that I didn't even know was there, was now gone. But the ambiance, reverberation etc was not where it should be. So I left it alone for several hours, came back and was blown away!

 

It is still smooth, but not in a bad way, the artificial edginess is all gone and the music is just there. This is phenomenal performance from a $109 DAC. The music is just there, there is no "sound" to listen to. I'm pretty darned impressed with what this little thing did with an inexpensive DAC..

 

Alex said the production version was significantly better than the prototype, he was right. I'm not quite exactly sure why, the prototype had an ESD chip tack soldered on with very tiny wires, and it is now connected with proper impedance traces, that is about the only thing I can think of.

 

Well anyway, back to listen to music.

 

John S.

[JohnSwenson]

Alex made it to Hawaii!

 

John S.

[JohnSwenson]

3 hours ago, rickca said:

John, can you please explain why you decided to have a switch to defeat galvanic isolation?  For which devices are we likely going to need to use it?

 

Alex already pointed out that this also gives us the ability to evaluate how much GI contributes to the sound quality.

It is not a "device" specific issue, it is a system configuration issue. There is no way to give a hard and fast "this exact configuration" will have an issue. The exact precise mechanisms are not known so there is no way to make precise proclamations about this.

 

There is a higher probability there might be an issue when people have systems that have already had extensive isolation work done such as devices run off batteries, isolation transformers, optical interfaces etc.

 

And in the situations where we have found this happening adding a ground to exactly one point in the system has alleviated the situation, so to be accurate there has been no known situation where the switch was actually required, although it does help a LOT in figuring out what is going on so the grounding can be applied.

 

I am NOT going to name specific devices involved in the situations where we found this happening because it is NOT device specific, a configuration where this happened had the issue with many different pieces of equipment in the same configuration.

 

The upshot is that in all the isolation going on in a system just make sure you have one and only one point of ground in the system and you should not have the problem.

 

John S.

[JohnSwenson]

4 hours ago, Kafy said:

Hello,

 

I have a question for John Swenson. 

John, if you have time, can you tell if the JS-2 and MMK for a Mac Mini are still usefull if I add an ISO REGEN and LPS-1 ?

 

If not, I could remove the MMK from the Mac Mini and use the original power supply to be able to use the JS-2 to power the network switch on which the Mac Mini is connected to. 

 

Thank's in advance for any help. 

I don't know. The ISO REGEN and LPS-1 will block the leakage loop from either the JS-2 or the original supply, but I'm not sure about the the signal integrity issue. The ISO REGEN seems to be more resistant to SI issues on it input than the USB REGEN, but exactly how much is I really don't know yet.

 

So my guess is that there might be  some difference in SQ with the two supplies but only you can determine whether it is important to you.

 

John S.

[JohnSwenson]

3 hours ago, SWL3600 said:

Just to be clear......

 

If I'm replacing my USB-Regen with an ISO-REGEN.....powered by the LPS-1,the voltage setting of 7v will stay the same, correct?

Yes. The only difference power wise between the two is that the ISO REGEN takes about 40mA from the VBUS (on the USB B connector) to power the upstream side of the isolation chip.

 

All the other issues about voltage and dissipation etc are the same as before.

 

John S.

[JohnSwenson]

21 hours ago, thyname said:

Hmmm... For now I have microRendu and SU-1 (ordered the DC conversion kit already), as well as LPS-1 currently powering the microRendu. Ordered the ISO Regen, but that does not arrive until June (did not get into the first launch batch).

 

I was hoping I could use existing LPS-1 for both SU-1 and uRendu via Y cable, but it looks like this would not work, as I need to set the LPS-1 at 5v, and uRendu needs at least 6 (range is 6-9). Which means that I may need to buy the second LPS-1 now, or just go back to powering uRendu with its iFi iPower. Or maybe I buy a (cheaper) sBooster. Don't know....

This has been discussed further up thread. It does in fact work. You can set the LPS-1 to 5V and power both the ISO REGEN and SU-1 off a Y-cable.

 

The only part of the ISO REGEN that needs higher than 5V is the 5V VBUS regulator. If you feed the ISO REGEN 5V this regulator will not be regulating and will drop the 5V a little bit (to 4.95V). Since the output of the LPS-1 is the same regulator as in the ISO REGEN VBUS you are not really loosing much in this scenario.

 

John S.

[JohnSwenson]

16 hours ago, rickca said:

Alex, you said that when you and John evaluated the Crystek CCHD-575, you quickly decided it was well worth using in the ISO REGEN.

 

Have you experimented with more expensive clocks?  I'm trying to understand whether there's a point of diminishing returns even if you had no cost constraints.   

No, we have not tried better clocks than the 575, the next step up in lower phase noise needs an OCXO. Note that inexpensive OCXOs do NOT have lower phase noise than the 575, you have to go to very expensive OCXOs to better it. And it is not just the cost of the OCXO itself. An OCXO takes a lot of power to run the oven and other circuitry, this will also add cost to the system. With the right OCXO we can probably still use the LPS-1 to power the board, but then you would have a very hard time powering anything else from the same one.

 

Both Alex and I are very much interested in producing items that are very high performance but still low enough in cost that a fairly large number of people can afford them. This is one of those cases where my gut feeling here is that spending the money on a better clock will give better results in a DAC rather than in an upstream device.

 

John S.

[JohnSwenson]

4 hours ago, Em2016 said:

 

Hi Alex, not sure if this is a stupid question but is the ISO REGEN output port BC1.2 compliant, for powering something downstream? Or is the output current limited to 0.5A? 

 

 

I know you don't want to discuss every product under the sun so you can ignore the following context, but I'm asking about an iFi iDSD Dac (charging the battery in Turbo mode can only be done with B1.2 compliant ports).

 

Cheers, Sean

 

The downstream port is setup as a standard port only, the VBUS regulator we are using cannot supply enough current to support a CDP port. If a device initiates the handshake to turn the port into a CDP port it will not get the answering signal from the ISO REGEN port.

 

The VBUS on the downstream port is actually current limited to 1A, so if a DAC tries to pull say 700mA it will work. It's not official spec, but I would rather give the DACs a little leeway rather than hard limiting to 500mA.

 

John S.

[JohnSwenson]

12 minutes ago, thyname said:

No 5 volts?

If you feed it 5 volts the internal parts will work fine (they run at less than 5V so their regultors will work fine from 5V). But there is a possible problem, the VBUS coming out of the A jack is 5V, thus there is a very high quality 5V regulator feeding the VBUS pin. All linear regulators require a higher voltage in order to work correctly, thus feeding 5V to a 5V regulator means it is cannot work properly. The result is that the regulator does not improve the quality of the input power and actually drops the voltage slightly.

 

Whether this is important depends on what you are powering the ISO REGEN from and where the VBUS on the downstream port goes. If the ISO REGEN is powered from an LPS-1, the power is already very clean so this issue is not all that big a deal. If it is a lesser power supply, it MAY be an issue. If the DAC you connect to the ISO REGEN uses the VBUS to power its circuitry using 5V may be a significant issue. If the DAC either doesn't use VBUS or just uses it for a handshake, it probably doesn't matter.

 

So the upshot is: don't use 5V unless you have a really compelling reason to do so. If you do use 5V you may be short changing your system, but that depends on what the DAC is doing and what you are getting the 5V from.

 

John S.

[JohnSwenson]

1 hour ago, scan80269 said:

I can only imagine the amount of email you get for both shipped and unshipped products...

 

Perhaps a FAQ put together with help from the more geeky members would be useful.  There's so much HW out in the field with so much variation it's virtually impossible for anyone to be a know-it-all, except possibly John and yourself...

 

Earlier today I posted about an idea of getting an LPS-1 to output 5.25V to power both ISO REGEN and SU-1, and was hoping John or you could weigh in, but then I realized tampering with an LPS-1 would void its warranty.  I guess John's answer to those wanting to do this dual powering is good enough.

 

Changing the voltage on an LPS-1 requires re-writing the firmware in the FPGA, not a trivial task. I think that would be one of the more ambitious reverse engineering tasks. 

 

John S.

[JohnSwenson]

9 hours ago, greenleo said:

The 3k price point was not true.  John mentioned that the price could be similar to that of the current LPS-1 with 12V/1A output.  Probably the discussion was in the Uptone website rather than in here.  I cannot remember the exact content.

 

However, I can see 12V demands in other applications which is both Video and Audio related and that could bring business to Uptone.

I think the $3k price point was for an Ultracap ATX supply! Going to a 12V 1A version is fairly easy, going to higher current is MUCH harder and costly.

 

John S.

[JohnSwenson]

9 hours ago, greenleo said:

This has been discussed before.  Two LPS-1 in series means two energiers and will the cabling more clumsy.  Let all alone another cable is needed to contact tnbe 2 LPS-1.

You CAN use a single feeder supply for two LPS-1 units, it just has to have enough power to handle two at the same time. There is no way the Meanwell that ships with the LPS-1 can handle two. The Meanwell is a 22watt unit, so you will want a unit the can supply greater than 44watts. (same voltage range, 7.5 to 12V).

 

John S.

[JohnSwenson]

On 5/19/2017 at 10:44 AM, Daudio said:

 

I see that you have gotten some very good answers to this and subsequent questions, but I thought I might pipe up and answer you in a slightly different way. I certainly don't know everything about USB, but I did spend some significant time rummaging through the USB 2.0 Spec documents, so here goes.

 

 

I think of the clocks in the host computer, any internal/external hubs, and the PHY receiver chips in the endpoint device as controlling the USB pipeline, syncing the signals that enable bits to flow, whether those bits are commands, status, data, markers, or whatever. The 'pipeline' is designed to work at various rates for low-level activity, as well as different, specified, application data flows.

 

This is true for all the different USB modes. Where 2.0 Async is different is that it adds the capability for the endpoint (DAC) to watch it's input buffer and respond to pseudo-interrupt messages from the host (which come from the 2.0 Async driver code), polling the DAC to allow it to request more data, or idle until it needs more. USB is a purely host-controlled interface and endpoints can't control the bus.

 

So the DAC is sending info (need data/don't) back to the host, based on the DAC's clock gating audio data out of its buffer, through the USB pipeline (which is managed by it's own low-level clocks). Said another way, is that the Async timing info tunnels thru the existing, separately timed data pipe.

 

USB Signal Integrity is a very important factor for the electronic pieces of the pipeline, at the physical level, but isn't really a part of the Async mode.

 

 

I hope this may help you, or some others understand screwy USB a little better. If I've gotten something wrong here, I'm responsible and expect to be corrected, and we will all learn    Apologies if this is too off-topic...

 

Not quite,

The upstream side is exactly the same in both adaptive vs asynchronous. The difference is in how many samples are put in each packet. The host software assumes that the clock feeding the PHY is exactly correct and frames happen at exactly 8KHz. It figures out how many samples it needs to put in each packet to exactly match the sample rate (44.1 96 192 etc). This is never an integer number, so packets do not all have the same number of samples. It will alternate packets of different number of samples so the average data rate is correct.

 

The low host just sends these packets out. There are no start or stop commands, the host just continuously sends these packets. This is the same for adaptive and asynchrounous.

 

What asynchronous mode does is add a mechanism for the DAC to tell the host to speed up or slow down. This only happens when the DAC wants to change things. This may not be very often. The host takes this information and slightly changes it allocation of samples per packet. As far as the host is concerned it is sending data out a little fast or slow. For example the host might think it is sending data out at 44102 samples per second.

 

OK so that is the mechanism, but what do different clocks have to do with sound quality?

 

Note first that none of this is hard and fast, nailed down, concrete, etc. As time goes on we learn more about what is going on, so attempts at explanations might change over time.

 

In adaptive mode the average data rate is controlled by the host, period. The DAC has to adjust itself to match. Traditionally this was done with a device called a "Phase Locked Loop" otherwise known as a PLL. I'm sure everybody has heard of these by now. Older, inexpensive PLLs were not very good, they generated a lot of jitter. Asynchronous mode was designed so you didn't need a PLL. It gets rid of the PLL generated jitter, but NOT all the other forms of jitter in your system. So it is a useful first step, but not a cure-all "nothing else matters" deal. It is not that asynchronous mode is worthless, or a "scam", it is just the first step. You have to get this done before you can start dealing with all the other issues.

 

As an aside PLLs have gotten a lot better in recent years. I could build an adaptive interface that outperforms almost anything on the market today of any type, but it would be pretty expensive. This is primarily due to the high speed communication industry, they all work on the adaptive model, the receiver has to synchronize to the transmitter, but this takes really GOOD PLLS, the engineers have been working on this a LOT in the last few years.

 

OK so what else is there? The clock that actually goes into the DAC chip consists of jitter from several sources. Before proceeding we need to talk about phase noise. The term jitter is usually specified as a single number, but this is way too simplistic, it is actually a spectrum, this is usually shown with a phase noise graph. I'm not going to go into the gory details here, but it is a spectrum, noise level VS frequency. You need this on order to see what is going on with multiple clocks in a system. They all have different phase noise and how they interact determines what winds up going into the DAC chip.

 

So how can clocks interact? Primarily through power and ground interactions. This all gets a little complex, I've tried to explain how this works in the series of articles I wrote for AudioStream, if you REALLY want some detail, go read these.

 

Every digital "stream" is timed by some clock, the phase noise of that clock is kind of a "fingerprint" on that stream, it is embedded in the timing of the stream. When that stream goes into some circuitry, some noise is generated in the power and ground signals of that circuitry. Even if that stream gets reclocked by an ultra low phase noise clock, there is still noise generated in the PG signals that has the fingerprint of the clock used to generate the stream in the first place! That noise can modulate the local clock so what come out of it is some combination of the original fingerprint and that of the local clock.

 

How much of the original fingerprint shows up in the local clock depends on the local clock itself, the power network on the board, the reclocking circuitry, the board layout and a host of other things.

 

The result is that if don't get all these things just right, the phase noise on the final clock might wind up being dominated by the upstream clock rather than the phase noise of the local clock.

 

This is why the clock used for generating the USB signal has anything to do with sound quality. At this point in time nobody has generated a receiver circuit that completely ignores the fingerprint from the USB stream. So every DAC is going to be sensitive to some degree or another to the clock used to generate the USB stream. Different implementations vary greatly as to how much they let through, so the relevance of low phase noise in the USB clock is going to vary, for some it is going to make a big difference and for some it will not matter as much.

 

Where a particular DAC falls in this range is almost random. Very few designers even think about this. And even if they do test equipment to actually measure this is very expensive so nobody is actually measuring it. So it is pretty much a crap shoot where a particular DAC winds up.

 

I'm actually trying to design some of my own test equipment that will allow me to actually measure this stuff and actually trace a clock "fingerprint" through a system. But that is probably a year or two away. But once that happens hopefully there can be some significant improvement in the situation.

 

BTW this is not just USB specific, ALL the digital music interfaces have these issues.

 

John S.

 

 

 

 

 

 

 

[JohnSwenson]

5 hours ago, R1200CL said:

@JohnSwenson

What are the benefit (if any) using i2s vs USB ?

 

(except less DAC's support i2s)

The way i2S into a DAC is frequently done, the clocking comes directly from the source. The I2S interface includes a clock that goes directly into the DAC chip. Thus the goodness of the clock in the DAC is completely irrelevant. The goodness of the clock in the source is all important.

 

Now if the DAC has a lousy clock and the source of the I2S has a really good clock, feeding I2S into the DAC could very well sound much better. On the other hand if the DAC has a really good clock I2S is not fantastic, it could make things worse.

 

There are so many things that come into play with this that it is impossible to make any general statements. Every system is going to be different.

 

John S.

[JohnSwenson]

On 5/24/2017 at 9:36 AM, R1200CL said:

 

OK, based on this I understand now that actually conclusion earlier today was a  bit wrong misleading by some other true believers ?

 

So how I now understand this is that avoid using a PC upfront the ISO REGEN, as one have to assume the computer has the worst clock. 

 

A good endpoint with a propper clock is crusial upfront the ISO REGEN.  So since some of John's creations also contains endpoints of "various quality", I think i may have to purchase something from this paranoid guy that just closed his thread. ?

 

 

I'm not sure I follow this.

 

The ISO REGEN provides the greatest improvement for systems that do NOT have the best source. What feeds the ISO REGEN DOES matter, as stated in my post above nobody has yet come up with ANY interface that completely blocks the upstream clock signature, so whatever is used upstream will have some impact on the sound. But the ISO REGEN has less of this than most other devices. The result it that it is very effective with less than perfect sources.

 

While it is true that for the absolute best result possible, no matter what the cost or complexity, it needs to be fed by a source with very low phase noise clock, that does NOT mean it is useless with less than the best source.

 

That is it's primary purpose, to improve the sound you get with whatever source you have. Rather than spending gazillions of dollars on a DAC that has extremely low sensitivity to upstream clock signature, or gazillions on a source that has an extremely low noise clock signature, stick one of these things inbetween a source with less than stellar performance and a DAC that doesn't have the best USB performance and it will radically improve the sound without having to spend a gazillion dollars to do so.

 

That is my primary motivation in all this, allowing people to get remarkable sound without having to sell the house to get it.

 

As an example of this I am right now using a $200 source (squeezebox touch) feeding a ISO REGEN with an $8 USB cable, into a $109 DAC from Amazon, using a $12 USB cable, into $80 headphones , both SBT and ISO REGEN powered by an LPS1 with a DIY Y cable, and getting the best sound I have ever had. Total system cost: $975.

 

THIS is what the ISO REGEN does.

 

John S.