Thunderbolt DAC input?

[jabbr]

Many of the pro recording boxes have moved to a thunderbolt interface. A few years ago we heard talk of thunderbolt interfaces coming to consumer DACs. Thunderbolt is essentially PCIe-on a wire so any of the DAC manufactures who used to have PCIe interfaces moving to thunderbolt? @PeterSt?

[mansr]

Just get a Thunderbolt-connected PCIe enclosure and populate with your sound card of choice.

[Ralf11]

What are the advantages over USB - other than thru-put?

[jabbr]

29 minutes ago, mansr said:

Just get a Thunderbolt-connected PCIe enclosure and populate with your sound card of choice.

I guess but look at pro audio sites: https://www.sweetwater.com/c1089--Thunderbolt_Audio_Interfaces

Not insignificant adoption.

[mansr]

6 minutes ago, Ralf11 said:

What are the advantages over USB - other than thru-put?

Lower latency, which matters for pro/studio applications but not for home use.

[Miska]

55 minutes ago, jabbr said:

Many of the pro recording boxes have moved to a thunderbolt interface. A few years ago we heard talk of thunderbolt interfaces coming to consumer DACs. Thunderbolt is essentially PCIe-on a wire so any of the DAC manufactures who used to have PCIe interfaces moving to thunderbolt? @PeterSt?

 

Since Thunderbolt is essentially PCIe put on a cable, it is actually quite straightforward for them to do. It has some additional security features for option BIOS (because these are run by the computer before OS boots) and some other small differences. But audio interfaces don't even need option BIOS unlike storage controllers that would wish to allow booting OS from the storage, or display adapters or such that are more "boot essential".

 

At pro-audio side I can see it largely replacing Firewire and not so much PCIe boards. Manufacturers targeting mainly Macs typically provide Thunderbolt interfaces (to avoid need for extra PCIe enclosure) while manufacturers targeting mainly PCs typically provide PCIe cards. All since Macs don't provide PCIe slots anymore and Firewire has gone out too (although Apple's Thunderbolt-to-Firewire adapter works great what I've used).

[Miska]

9 minutes ago, mansr said:

Lower latency, which matters for pro/studio applications but not for home use.

 

And much lower CPU overhead for high channel counts at hires sampling rates combined with lot latency.

 

PCIe/Thunderbolt can do nice busmaster DMA "zero-copy" straight from application to audio interface without involving things like USB packet handling. This makes difference when one wants to do something like 128 channels at 192 kHz 32-bit samples especially combined with low latency and lot of DSP activity for realtime effects and such. Of course this is also feature of many ARM SoCs with suitable local bus for I2S without involving PCIe (rather avoiding USB).

 

I think Firewire was much more capable of programming the interface to deal with audio transfer without having to involve CPU with the actual data (but had associated problems that often drivers worked only with certain Firewire controllers).

[mansr]

9 minutes ago, Miska said:

And much lower CPU overhead for high channel counts at hires sampling rates combined with lot latency.

Right, and also not relevant for normal playback.

[PeterSt]

12 hours ago, jabbr said:

so any of the DAC manufactures who used to have PCIe interfaces moving to thunderbolt? @PeterSt?

 

A few weeks ago I investigated that once again. Nothing to be seen anywhere.

Yeah, "soundcards" as how we see them in enclosures with USB output for a DAC, but ... 

 

Whether I would go the path of a thunderbolt interface in-DAC ? I don't think it will solve much because the real work is on the PCI side, read : would I try to make an I2S interface for PCI ? -> Never. The PHY would be OK, but the driver I wouldn't like to make. Especially not because AFAIK only one chip manufacturer deals with that (CMEDIA) and I don't think we can get any help from there. Quirks via USB would be doable, but ... 

 

12 hours ago, mansr said:

Just get a Thunderbolt-connected PCIe enclosure and populate with your sound card of choice.

 

Which would be none (explicitly investigated that too).

[Miska]

I've been personally happy with RME's PCI/PCIe cards for quite a while when I need certain features. And I guess Lynx makes nices ones too, but I don't have much experience with those.

 

For making a sound card, one can buy PCIe interfaces as IP blocks for FPGAs, possibly also Thunderbolt ones, so that part is not so much of an issue (also USB etc). But looking at many audiophile hardware companies, they are not so ready to deal with drivers and more hard core firmware work. So usually they slam standard XMOS or Atmel USB interface in front of FPGA instead of trying to integrate. This is major difference to many pro-audio companies who systematically have higher integration and their own drivers, so for them it seems to be less of an issue to deal with Thunderbolt/PCIe too.

 

[Panelhead]

 I use a Thunderbolt interface and am very pleased. The drivers for Mac were very fast. The Windows drivers seemed to be a struggle. They finally released a family of USB interfaces for Windows and gave up on supporting Thunderbolt for Windows.

 Same story as FireWire was. The Windows support broke several small pro companies when the interfaces stopped working after the endless OP updates. 

  I feel FW and TB are superior to USB even for low channel count. Maybe they are just easier to implement from the manufacturer side. I still have a couple FW connected devices that work fine on a Mac many years after support ended.

[jabbr]

19 hours ago, PeterSt said:

Whether I would go the path of a thunderbolt interface in-DAC ? I don't think it will solve much because the real work is on the PCI side, read : would I try to make an I2S interface for PCI ? -> Never. The PHY would be OK, but the driver I wouldn't like to make. Especially not because AFAIK only one chip manufacturer deals with that (CMEDIA) and I don't think we can get any help from there. Quirks via USB would be doable, but ... 

 

The advantage of Linux is that you can easily see how another PCIe -> I2S driver is written, and adapt. The FPGA is the PHY.

[PeterSt]

57 minutes ago, jabbr said:

The advantage of Linux is that you can easily see how another PCIe -> I2S driver is written, and adapt.

 

OK. A good reason to move to Linux.

(seriously)

[asdf1000]

On 8/13/2018 at 6:20 AM, Miska said:

And much lower CPU overhead for high channel counts at hires sampling rates combined with lot latency.

 

PCIe/Thunderbolt can do nice busmaster DMA "zero-copy" straight from application to audio interface without involving things like USB packet handling.

 

Is the Thunderbolt interface easier to isolate noise from a powerful computer, than USB? 

 

Ground noise, power line noise etc? The common things discussed/tackled with isolating USB inputs of DACs?

 

So even though it has lower latency and better support for high channel counts at high sample rates, does Thunderbolt  still face those same electrical noise isolation challenges as USB?

 

[jabbr]

3 hours ago, Em2016 said:

 

Is the Thunderbolt interface easier to isolate noise from a powerful computer, than USB? 

 

Using the Corning optical cable, perhaps

 

3 hours ago, Em2016 said:

 

Ground noise, power line noise etc? The common things discussed/tackled with isolating USB inputs of DACs?

 

Although power over Thunderbolt is an option, its not required.

 

3 hours ago, Em2016 said:

 

So even though it has lower latency and better support for high channel counts at high sample rates, does Thunderbolt  still face those same electrical noise isolation challenges as USB?

 

 

It may not but USB seems so very sensitive to cables so who knows.

[asdf1000]

6 minutes ago, jabbr said:

Using the Corning optical cable, perhaps

 

I imagine so. My Corning Optical USB works really well. Here's a post by Ted Smith (who also uses it):

 

https://forum.psaudio.com/t/corning-usb-optical-cable/3761/44

 

7 minutes ago, jabbr said:

It may not but USB seems so very sensitive to cables so who knows.

 

Hehe noted, I was asking here if Thunderbolt is as sensitive, in the hope that somebody knows ?

 

@Miska @PeterSt

[Panelhead]

  Maybe. The cables are active. There is circuitry in both ends of the cable. This is why the cost is so high (30.00). 

  The interface in the cable may reduce noise. It manages the data transmission. 

  Thunderbolt really was designed for high bandwidth. But noise reduction may be a side benefit.

  I have read that higher bandwidth USB cables sound better than standard speed USB. 

  I know from experience that connecting a FW cable and device to a TB port sounds better than plugging the same cable into a FW port on the same computer. The Apple FW to TB adapter is active also. 

  I am sticking with my TB interface. 

[jabbr]

2 minutes ago, Em2016 said:

 

I imagine so. My Corning Optical USB works really well. Here's a post by Ted Smith (who also uses it):

 

https://forum.psaudio.com/t/corning-usb-optical-cable/3761/44

 

 

Hehe noted, I was asking here if Thunderbolt is as sensitive, in the hope that somebody knows ?

 

@Miska @PeterSt

 

Peter is already preparing the Luscious as well as the Luscious^2 which has a double shield over the fiberoptic cable ...

[Ralf11]

7 minutes ago, Em2016 said:

 

I imagine so. My Corning Optical USB works really well. Here's a post by Ted Smith (who also uses it):

 

https://forum.psaudio.com/t/corning-usb-optical-cable/3761/44

 

 

Hehe noted, I was asking here if Thunderbolt is as sensitive, in the hope that somebody knows ?

 

@Miska @PeterSt

 

at $275 for the cable, that is less than many, many isolation techniques

[asdf1000]

28 minutes ago, jabbr said:

Peter is already preparing the Luscious as well as the Luscious^2 which has a double shield over the fiberoptic cable ...

 

Lol for real?

 

[asdf1000]

2 minutes ago, Ralf11 said:

 

at $275 for the cable, that is less than many, many isolation techniques

 

Well that's what Ted himself says in the post...

 

Many have had reliability issues with it though. Luckily not me though (yet).

[jabbr]

23 minutes ago, Em2016 said:

 

There's a follow-up to Lush that will be an optical USB cable like Corning, or that's a joke?

 

Stranger things have happened than a doubly shielded fiberoptic cable ( SPF 1,000,000? ) but I'll let Peter speak for himself -- don't want to pre-announce anything ?

[asdf1000]

3 minutes ago, jabbr said:

 

Stranger things have happened than a doubly shielded fiberoptic cable ( SPF 1,000,000? ) but I'll let Peter speak for himself -- don't want to pre-announce anything ?

 

Ha, you're making fun of him for jumping on the JSSG 360 bandwagon with Lush^2.

 

For others, the Corning is not a 100% fiber optic cable though. It has very thin power and ground lines but Ted explains above.

 

[jabbr]

1 minute ago, Em2016 said:

 

Ha, you're making fun of him for jumping on the JSSG 360 bandwagon with Lush^2.

 

For others, the Corning is not a 100% fiber optic cable though. It has very thin power and ground lines but Ted explains some benefits above.

 

Actually the Thunderbolt cable is, I believe, purely fiberoptic in the middle and each end is separately powered (the transceivers)

 

The USB cable is as you say -- there is a voltage drop because its thin and so the receiving end has a DC/DC converter to power the transciever and supply 5V out. A bit odd...

 

One thought at the time (I have one also) is that the resistance along the long and thin ground line, reduced ground flow and hence noise.

[asdf1000]

2 minutes ago, jabbr said:

One thought at the time (I have one also) is that the resistance along the long and thin ground line, reduced ground flow and hence noise.

 

That's what Ted Smith explains in the post I linked above..

 

Sounds fantastic in my system, from a 5Vdc LattePanda running HQP NAA at DSD512.