Xite with Usb connector

The Sonic Core XITE hardware platform for Scope

Moderators: valis, garyb

User avatar
katano
Posts: 1438
Joined: Mon Jan 20, 2003 4:00 pm
Location: Zurich, Switzerland

Re: Xite with Usb connector

Post by katano »

I disagree, the best examples are the RME interfaces, which work flawlessly. It‘s all because of they seem to have gotten around the usb issues and developed a rock solid driver, which is key. also, even usb 2.0 has enough bandwidth to provide 96 channels auf audio…

maybe the difference is because xite is realtime?
User avatar
katano
Posts: 1438
Joined: Mon Jan 20, 2003 4:00 pm
Location: Zurich, Switzerland

Re: Xite with Usb connector

Post by katano »

User avatar
garyb
Moderator
Posts: 23248
Joined: Sun Apr 15, 2001 4:00 pm
Location: ghetto by the sea

Re: Xite with Usb connector

Post by garyb »

it. does. not. work.
making it work causes more problems than it solves.
an XITE is not even slightly related to an RME interface.

the XITE needs a connection below the operating system level, like PCI or PCIe, or true Thunderbolt, which is PCIe. if it were simply an interface, then USB would be just fine. this is an unarguable fact, until further notice.
fra77x2
Posts: 411
Joined: Sun May 03, 2015 3:23 pm

Re: Xite with Usb connector

Post by fra77x2 »

USB is a very good protocol that forces standarization. It is also cheaper

https://www.rme-audio.de/rme-usb-technology.html

And you can read here that the USb and thunderbolt protocols have merged

https://en.m.wikipedia.org/wiki/Thunderbolt_(interface)
Last edited by fra77x2 on Thu Mar 31, 2022 1:00 pm, edited 1 time in total.
User avatar
garyb
Moderator
Posts: 23248
Joined: Sun Apr 15, 2001 4:00 pm
Location: ghetto by the sea

Re: Xite with Usb connector

Post by garyb »

i don't give a fuck if USB is used or not. i am not in your way. why argue with me when you know that i have no say?
i have been involved since long before Creamware went out of business, some 20 years, and i have been privy to many ideas and attempts, but as you say, i'm not a programmer, so what do i know? well, i know that USB won't happen in the near future.

what about my approach do you not like, exactly? that i did not support the idea?

please, up until now, it is impossible. could a driver be written? yes, i'm sure that it can be. would the system work if there was a driver? nope. it's the same reason that virtual machines don't work.

Scope is not unique or different? then try runnning it with just a driver. Scope hardware CAN be just an interface, and a fine one at that, but that is not what it IS. the hardware is a stand-alone dedicated audio computer which must be in sync with the host.

did you think that Scope on USB is a new idea? did you think that it has not been explored? i remember Frank Hund and Ralf Dressel in my shop telling me about the new USB interface back in 2004 or so. yeah, that was never made, it never worked.
fra77x2
Posts: 411
Joined: Sun May 03, 2015 3:23 pm

Re: Xite with Usb connector

Post by fra77x2 »

ok i thought you said it can't work. If you say it won't happen its a different thing and i have no problem. Neither I care too much about such change because i use xite now. but perhaps in the future something changes

I meant you can't be explicit about it if you are not aware of the exact limitations
User avatar
garyb
Moderator
Posts: 23248
Joined: Sun Apr 15, 2001 4:00 pm
Location: ghetto by the sea

Re: Xite with Usb connector

Post by garyb »

sure! maybe something will change.
User avatar
valis
Posts: 7306
Joined: Sun Sep 23, 2001 4:00 pm
Location: West Coast USA
Contact:

Re: Xite with Usb connector

Post by valis »

Keep in mind that S|C is reliant on upstream providers for the product they produce, including the binary bits that glue things together (not to say that Holger can't code in C, but noone codes and builds a system end to end in this era). If Analog Devices didn't pose a similar issue even for UAD, there would not have been a fixed buffer size with their own DAW (Scope can change latency, but not while connected to the DAW! and it vastly complicates the XTC/PDC issues and solutions required).

It's also helpful to look back at the 99-2002 era when we were stress testing the 32bit PCI busses to see how our solution fared against 'normal' soundcards. I have an RME Multiface 1 + PCI card slot solution that I owned alongside my Pulsar1, which was only 4 dsps, and when doing only the masterverb test on 4 DSPs without any ASIO drivers (testing what a DSP implementation that uses the PCI bus does in comparison to the RME doing 'normal' ASIO, in other words) the card was already using double the PCI bandwidth of the RME when I had 18 channels active.

That tells us that a 6 dsp card doing a few reverbs, a mixer, and summing ASIO certainly is equivalent to trying to fill a system with RME cards in terms of PCI bandwidth, so filling a system with Scope cards has always been a challenge. What happens when there's a dropout to Scope? It's far far worse than just a buffer error in the DAW as with ASIO drivers alone, and in the worst cases it's a BSOD.

Why is this relevant to Xite? Well, we know someone at some point helped build tools that facilitate replacing the PCI bus connection with PCIe, we don't know where in the mix between S|C's intellectual property (stuff they had to make to get their products to work), Analog Device's IP (which they license or give access to when using their chips as a developer), or potential 3rd parties...well we don't know exactly what's involved in simply saying 'replace it with a USB connector' now do we? But given the realtime needs and the fact that a 4 dsp PCI card would have used about the equivalent bandwidth of the higher end RME interfaces that work via USB right now, and then consider how much more powerful an Xite is compared to a Pulsar1 and think about your babyface comparison again.

Oversimplified examples from a consumer or user perspective do no good when dealing with a niche product like Scope. "USB" ASIO soundcards are so cheap now the iRig is under $100 and works on everything you connect it to. However those have the benefit of a large enough development path over enough companies that things are markedly improved compared to 20 years ago. Given that the other major user of A|D chips (UAD) has now decided to go native...I suspect most of the progress on the Analog Devices side has come from outside of our industry, which is where we got our tools from originally anyway (telephony and etc).
fra77x2
Posts: 411
Joined: Sun May 03, 2015 3:23 pm

Re: Xite with Usb connector

Post by fra77x2 »

--------

Ok, now if my "interface" can record 100 channels and your "non-just-interface-but-whole-computer" can record 40 channels and pass around here and there 60 signals, then both are the same. From the perspective of data transfers there is no difference, we can abstract both as "just-interfaces".

Scope required normally great bandwidth to do this and yes they were ahead of their time. That is the reason it used PCI. After PCI become obsolete PCIe replaced it. To not spare your time nowdays a USB2 connection can outperform in bandwidth a lot of tried-and-failed systems (hello communication sat limit systems). USB3 is even better. And USB4 has merged with thunderbolt. So the original question that asks whether a xite with USB connector is possible, is yes it is feasable. Whether it could happen is another question that personally I do not really care getting an answer.

Anyway a system that allows the user to get a signal and pass it through a dsp reverb then send it to the CPU and the resending it back to the dsps and then again to the CPU and introducing on each movement of data 5 samples of latency is a mediocre system in terms of engineering. This was marketing's bad choice. This is the reason I use the sdk and not the normal software. I do not want my signals go crazy and have this identity crisis.

If you read the articles about thunderbolt in wiki you will see that the industry (intel, apple etc) has the same concerns about connectors. We want standarization, simplicity for the user, the same cable for everything and all devices to use the same connectors. This is sane, reasonable and the way to go. If you do not like the name "USB" and you have associated with mouses and keyboards you have to update your info. It is just a communication protocol that forces standarization. The benefits of that are too many to point out here.
User avatar
valis
Posts: 7306
Joined: Sun Sep 23, 2001 4:00 pm
Location: West Coast USA
Contact:

Re: Xite with Usb connector

Post by valis »

Bandwidth != latency, and yes true thunderbolt is PCIe and I think currently works with Xite? USB-C modes compatible with TB transport for video and the like did not, at least from what I saw here. That's all down to having the chipset support natively, and a (currently alpine ridge) TB controller inline. I'm all for future versions of USB being more compatible, but there's still more for direct PCIe attach than just the connector type. On a positive note, Gigabyte makes a card that will even work on current AMD chipsets if needed, so the tech is at least gaining some market penetration.

None of that matters if some critical piece of glue needs to be updated either upstream or via 3rd parties for whatever reason, which was my primary point.

And I was of course comparing bandwidth usage while pointing out that the latency and need for continuous connection also seem a consistent theme with Scope.
fra77x2
Posts: 411
Joined: Sun May 03, 2015 3:23 pm

Re: Xite with Usb connector

Post by fra77x2 »

About latency with WASAPI I get 132 samples - 44100 (3ms) with a laptop soundcard...
User avatar
Bud Weiser
Posts: 2684
Joined: Tue Sep 14, 2010 5:29 am
Location: nowhere land

Re: Xite with Usb connector

Post by Bud Weiser »

I´m too old waiting for sucess of such USB connector project.

I´m still happy w/ the older XITE-1 PCIe card and it´s DVI connector, so the cable can be locked and is impossible to remove accidently.
Any USB connector (just only the connector !) is mechanically crappy consumer electronics.
I hate USB connectors !
In one of my laptops, one of these is broken and I cannot replace because of unavailability of the original replacement part,- not to forget to mention laptop dis-assembly and re-assembly isn´t the easiest and doing solder jobs on laptop mainboards isn´t too.
(Same rules for the damn tiny audio output connector)

And yes, I also wished the PCIexpresscards were lockable in the expresscard34 slot.

For the time being I´m waiting patiently for the announced SCOPE 7 update and wonder if the beta testing period started already,- did it ?

Finally,- all I understand is, when a system urgently requires connectivity and communication below OS level,- USB will never work because it is NOT below OS level.
Correct me if I´m wrong, but IIRC, only PCIe and network is below OS level,- Mac and PC.

When Thunderbolt merged w/ USB, it might not be below OS level anymore,- but I´m not sure because of lack of knowledge.
Wasn´t there rumour Apple planned ditching Thunderbolt in favour of USB (-C) and upcoming designs ?

:)

Bud
fra77x2
Posts: 411
Joined: Sun May 03, 2015 3:23 pm

Re: Xite with Usb connector

Post by fra77x2 »

To tell the truth Roland SE when is connected with the USB introduces 2 high frequency artifacts in my signal - a terrible situation. Without the USB cable connected it is crystal clear. So I control it with Midi. Xite's converters have very good s/n ratio.
User avatar
valis
Posts: 7306
Joined: Sun Sep 23, 2001 4:00 pm
Location: West Coast USA
Contact:

Re: Xite with Usb connector

Post by valis »

PCIe is more agnostic, it requires something connected to it to really say anything about the transport it provides. The USB port is, in a way, 'smarter' in that it's CPU driven in entirety and so can be 'programmed' in software to do what it needs on the fly, it has synchronous & async modes, can handshake on the fly to multiple downstream devices and change modes as required, and is serial rather than parallel (technically PCIe is also serial but discrete lanes are either shared, bifurcated or switched (PLX) and does not--in my understanding--really call for handling changes in the structure of the datastream(s) sent to the device quite as frequently). My terminology may be off somewhat as it's been a decade-ish since reading up on all of this to any depth.

In any case, as USB is a PIO mode thing, and RME proved that modern Intel & AMD chipset implementations are so 'close to the cpu' due to the nature of PIO (programmed-input-and-output aka entirely CPU driven for all data, protocol/transport and error handling) that it can indeed be lower latency for sample-rate data than even PCIe cards for them. Part of that was dependant on the binary blobs provided by the chipset maker that sits upstream from them, and we know this because Native Instruments reported the same results with their Komplete Audio 6 (and other) devices which used the same USB chipset and binary blob (sans any control panel or UI comparable to NI, not to mention lacking the FPGA).

When Scope has BSOD's and other errors, it's typically not an ASIO overflow (I recall seeing those, especially when changing driver settings without closing a DAW). This indicates there are other things going on than just handling samplerate data. I know from the SDK perspective we can talk about sync & async data, but without inspecting the PCI traffic and knowing what the underlying bits provided by Analog Devices to S|C are doing, we can't really know what else is going on with that traffic. Clearly something is extremely sensitive to data dropouts in a way that an ASIO buffer hasn't been vulnerable in years, and clearly this is what Gary is always referring to.

This is no real surprise to me, as I recall using TARGA capture cards to grab 3d animation frame by frame in the early to mid 90's, and even modern devices in that market can be finicky when from smaller vendors. Stringent hardware requirements and sharing issues are a real issue when trying to handle 3-4 streams of 8K data at 10bpp (or higher) with a decent alpha (not just binary clip key). Companies like Newtek are helping that in terms of realtime protocols, but Scope is hardly a rarity compared to some of the other tools I've used, and in many ways more stable once the configuration details are known.
fra77x2 wrote: Fri Apr 01, 2022 2:59 am To tell the truth Roland SE when is connected with the USB introduces 2 high frequency artifacts in my signal - a terrible situation. Without the USB cable connected it is crystal clear. So I control it with Midi. Xite's converters have very good s/n ratio.
This sounds more like some form of crosstalk or a bad square wave coming across from poorly isolated power/ground (wall warts generate this when using cheap rectifiers to turn AC into DC without any buffering to filter the edges off). Ticking or clockrate sounding changes make this especially evident. Some computer PSU and motherboard combinations make this a problem, as do many laptops (due to their own transformer's ability to isolate).
fra77x2
Posts: 411
Joined: Sun May 03, 2015 3:23 pm

Re: Xite with Usb connector

Post by fra77x2 »

To avoid a BSOD with Scope if ASIO crashes:

i. DO NOT PRESS OK when the "This application has crashed window" appears
ii. Close Scope.
iii. Open Scope.
iv. Reclose Scope/ reopen it.
v. Now you can press ok on the "This application has crashed window"...
fra77x2
Posts: 411
Joined: Sun May 03, 2015 3:23 pm

Re: Xite with Usb connector

Post by fra77x2 »

fra77x2 wrote: ↑Fri Apr 01, 2022 2:59 am
To tell the truth Roland SE when is connected with the USB introduces 2 high frequency artifacts in my signal - a terrible situation. Without the USB cable connected it is crystal clear. So I control it with Midi. Xite's converters have very good s/n ratio.

This sounds more like some form of crosstalk or a bad square wave coming across from poorly isolated power/ground (wall warts generate this when using cheap rectifiers to turn AC into DC without any buffering to filter the edges off). Ticking or clockrate sounding changes make this especially evident. Some computer PSU and motherboard combinations make this a problem, as do many laptops (due to their own transformer's ability to isolate).
I think it is the unshielded USB cable. Yes something like that.
jksuperstar
Posts: 1638
Joined: Mon Nov 15, 2010 12:57 pm

Re: Xite with Usb connector

Post by jksuperstar »

Using PCI/PCIe/Thunderbolt means the hardware is memory mapped, with the hardware directly accessible by the CPU.

USB can move data through buffers to/from specific endpoints, so both the hardware and software would have to change extensively to accommodate this shift...that'd work for the next gen hardware, but couldn't update XITE to do it.
Post Reply