The parallel universe of FireWire hubs | Hacker News Reader

The parallel universe of FireWire hubs

oldvcr.blogspot.com

I recently directly connected my 2020 MacBook Air to my iMac G4 in target disk mode… and it worked.

I had to use a Thunderbolt 3 to Thunderbolt 2 adapter, a Thunderbolt 2 to FireWire 6-pin adapter and then FireWire 6-pin to FireWire 9-pin adapter. I was shocked that it worked so well.

Apple might be the Mayor of Dongletown, but the dongles they do make are usually very good.

They stay away from complicated ones like port replicators. I'm convinced nobody can make a stable one of those for USB-C/ThunderBolt.

You make a very adulterous point.[1]

[1] https://youtu.be/CcyXRsJFjuc

> I had to use a Thunderbolt 3 to Thunderbolt 2 adapter, a Thunderbolt 2 to FireWire 6-pin adapter and then FireWire 6-pin to FireWire 9-pin adapter. I was shocked that it worked so well.

This same setup also works for other FireWire peripherals, like camcorders, for what it's worth.

> a Thunderbolt 2 to FireWire 6-pin adapter

How in blazes does that one actually work?! Is it a PCI FireWire interface chip or does Thunderbolt have a FireWire alternate mode?

FireWire standard mandates implementation of a fully capable controller behind every FireWire port. So it's a FireWire Card (sic) which talks PCI over TB at one end, with a single FW port.

This is why FireWire was always very light on CPU utilization. It always did its work by itself. There was no WinModem magic behind.

A Firewire CDRW was the only way I could reliably record CD's on my 2000's laptop. The USB ones were either 1.1 which barely had enough bandwidth, or the CPU usage went so high...

As a bonus I could daisy chain a 40GB external HD off of it!

I still have a Maxtor USB/FW400 external 1TB drive. It's both handsome and worked wonderfully on a FW port. It also has daisy-chain capability.

It worked so well for so many years.

Please, share a picture of the dongle chain :)

Not OP, but imagine something like this https://postimg.cc/N53tdRDc

(Which was just getting from standard serial port to some weird nonstandard RJ45 thingy for some weird SINIX terminal concentrator with the available technotrash(USB2Serial-> gender changer, etc))

:-)

This gives me hope I will still be able to pull a bunch of digital 8 tape content off a cam with firewire output still. Even if I need to get a mac, it might still work!

Now if someone can come up with a similar solution to get an ultra-wide scsi external HD to usb or whatever, I'll have a few great projects to get into.

For what it's worth, I used to do exactly this with one or two programs on Linux. I think I had the most success with dvgrab. It even can control the Digital8 camera and rewind and whatnot. It should work with any Firewire card that works on Linux as well.

I wonder if you can stick one of those old Firewire to SCSI adapters on the end of that

Two things FireWire got right:

1. Visibly obvious directionality

2. 12V instead of 5V. Dunno what insane person thought carrying 5V a distance was a good idea, but maybe this has to do with the proliferation of cheap SMPSs in any size/combination.

> Visibly obvious directionality

I like how it's the same family of connectors as the Game Boy link cable https://en.wikipedia.org/wiki/Game_Link_Cable#/media/File:Ga...

Add to that: hardware offload and DMA.

USB is limited by your CPU, FireWire gets pretty close to max speeds at all times, even when your cpu is anaemic or busy.

In fairness, that did create security issues. But if you're using your iommu correctly then yeah it's excellent.

With the rise of Thunderbolt in affordable consumer devices, I've seen quite a few tech enthusiasts get excited about the benefits of doing DMA transfers

Of course, modern IOMMUs are much better than the ones back in the day,bbut it's still as if the world has forgotten FireWire completely.

Though to be fair USB was designed to be cheap to implement. firewire's designers didn't take that compromise and it has performance benefits, but it isn't necessarily a wrong decision.

The 12V output of FireWire is convenient!

I built a DIY FireWire-to-car-socket adaptor, and use it with a 2-port USB car charger to get an extra two USB charge-only sockets on my laptop.

Just a warning: the Thunderbolt adaptor actually outputs 16V not 12V. Thankfully most 12V devices (including iPods) accept higher voltages, usually up to 24V or 30V.

USB finally carrying non bullshit-ly low voltages is a huge obvious upgrade in usb-pd. I'm ecstatic to hear about recent extended power range modes that go all the way up to 48v @ 5a (240w). that usb-c connector continues to impress.

can you clarify what you mean by obvious directionality?

At first I read this as upstream vs downstream, which USB-A and USB-B connectors did illustrate, but usually not for much purpose. What the other poster means is cable orientation. Especially the FW-400 plug was very obviously asymmetrical so it was easy to plug in the correct way.

The actual connectors for USB A are bullshit. It only goes in two ways, but you still have to get it wrong twice before you can connect it.

Even with micro and mini USB it's not easy to slot them into the receptacle without squinting at the male end.

Mini-USB I agree, but at least Micro-USB allows you to run your fingers over the male side to feel for the prongs. It's unfortunately still a guessing game which way the female side is though.

Imagine how many cumulative hours, or human lifetimes, have been wasted just from people trying to plug in USB-A cords repeatedly.

USB Mini-A was a little more obviously keyed: http://www.penguin.cz/~utx/hardware/USB_Mini/

Sorry for this approach. Lammy could you please contact me about an issue related to another topic discussion (CS3). I really need some help and I think you could. Email: oje.grafico@gmail.com

You got mail.

240w should take a pretty thick cable, and serious connectors.

USB-C already carries rather complex wires for high-speed data transfers. Combined with 48v insulation, the cables are going to become even thicker and stiffer.

This random table I found [0] says 23 gauge, or just over 0.5mm as the minimum size for transmitting 5 amps over short distances. But that’s probably for a solid conductor, so braided would be a bit thicker I guess.

Still, it’s not that thick, since only one pair of the 6(?) needs to take the current.

[0] https://www.powerstream.com/Wire_Size.htm

Yes. But a reversible connector beats them all.

Took USB a fair while to get there though.

5v pairs better with Li-ion though

One thing I really miss from FireWire devices, particularly of the storage variety, is how rock solid the connection was. It never wavered or blipped out, even when carrying out multi-hour transfers. Compare that to USB storage, which depending on the device can periodically blip out for no apparent reason, even if you’re using reputable cables and all that.

Thunderbolt is better than USB in that regard, but it’s still more flakey than FireWire was.

On 2 different Windows XP machines, the 1394 controller would reset and the drive would disappear. So maybe that was part of the FW adoption problem. But mostly USB2 was fast enough for consumer drives, unless you were sitting there with a stopwatch.

I believe it's slowly replacing MIL-STD-1553 data bus in aircraft.

https://en.wikipedia.org/wiki/MIL-STD-1553#%3A%7E%3Atext%3DM...

According to Wikipedia the F-35 uses IEEE 1394b

I still have a bag of FireWire 400 and 800 cables and a number of adapters, plus one external drive that has both USB 2.0 and FireWire 800 ports... only have used them once or twice in the past five years when helping someone offload data from an ancient Mac still in operation.

It seems those old Macs may still be in service in niche applications, most notably in audio processing chains at radio stations or recording studios.

I think it's interesting how long the service life of that gear has been. And as long as it's air-gapped, I don't see much reason why someone couldn't keep running some of those tools until they die.

A lot of the specialized software used for that is customized enough that the OS8/9 system UI doesn't bleed through in the actual editing software in use.

I even found an actively used Jazz drive a couple years ago!

> It seems those old Macs may still be in service in niche applications, most notably in audio processing chains at radio stations or recording studios.

Highly recommend looking at the tour of fat-boy slims recording studio for an example of this kind of airgapped system in active use for decades: https://youtu.be/qLjgXPDzeZo

(Though he uses an Atari ST, which is known for being a musical powerhouse)

The best FireWire Hub by far was Hubzilla.

http://images.pcworld.com/howto/graphics/115629-2206P164_3B....

I have no idea how I missed that one.

Firewire is such a cool bus, it’s a shame it fell out of favor. How many buses have support for up to 80% of the bandwidth being Isochronous and bandwidth guaranteed? I’ve been playing with a vk9000 firewire analyzer to reverse engineer some anchor to cameras, it’s been quite fun.

I would say that the fact that firewire is a bus is part of the reason why it fell out of favor. It was designed in a time when packet switching was though to be expensive and to cause unacceptable latencies for A/V applications. Just few years after firewire was designed both of these issues were found out to not be as relevant as they at a time seemed. In todays world the firewire's “logical bus from spanning tree” design seems like overly complex waste of bandwidth.

Edit : ancient cameras aka Sinar digital backs

FireWire (or at least OHCI-1394) had another fun feature: It exposed memory from the machine over the bus.

https://www.kernel.org/doc/html/latest/core-api/debugging-vi...

[EDIT: This was 100% feature, not bug]

This (Direct Memory Access - DMA) is still a feature of Thunderbolt (which is really just an evolution of FireWire), and thunderbolt is coming to USB4.

This is one (of many) areas in which Thunderbolt absolutely slays USB3: the ability to chain devices. Granted Thunderbolt anything is more expensive, but the performance is incredible.

This would be great in theory. In practice, major devices are hamstrung, like Apple not supporting MST which would enable monitor daisy-chaining.

Apple neglecting MST is so frustrating. My MacBook Pro could do MST with no trouble in Boot Camp which adds insult to the injury.

You could daisy-chain the old Thunderbolt Display, was that using a different technology?

There is (of course!) more than one way to pass multiple display streams through thunderbolt.

MST (multi-stream transport) is a DisplayPort thing, which Windows uses. So you have DisplayPort over thunderbolt, and (presumably?) the multiple displays are opaque to thunderbolt. I've used this with thunderbolt/usb3 hubs that understand MST.

On Macs, however, they use thunderbolt's own ability to have multiple devices. You need to look carefully for hubs that support this instead.

Or at least this is my very limited understanding, born of utter bafflement with our work docking stations.

I post this mostly in order infuriate somebody with actual knowledge. Please set me straight.

A lot of USB devices already have that capability; sure, internally it's a hub to which the device and the port is connected, but to the consumer it does not matter much.

It's a bit misleading to say SCSI is/was peer to peer. Yes, any device could address any other, but in the communication between any two devices there was a clear separation between initiator and target roles. AEN, which might have been used e.g. for targets to notify initiators of fan or power faults, was limited and rarely even implemented. To the extent that IEEE-1394 storage used the SCSI command set, it would have had (at least) the same limitations.

> but FireWire was actually meant to replace SCSI. SCSI and FireWire are peer-to-peer: any device on the bus can talk to any other device, unlike USB where each bus has at most one host and the host does all the initiation of data transfer. (USB On-The-Go still has one host and one host only; it just allows certain devices like your mobile phone to swing both ways.) The point-to-point capabilities of USB 3 notwithstanding, a USB hub has one upstream port for the host and multiple downstream ports for the devices. A FireWire hub, however, is like getting a longer internal SCSI cable; more devices simply exist on the same bus. Connecting multiple FireWire hubs just makes a bigger bus because all the ports are the same.

I'd love a comprehensive review of how host-to-host networking works under USB4, for comparison. USB4 is the first significant overhaul of USB, a move from switched links to a packet switching system, where-by various protocols such as USB3, PCIe, DisplayPort are multiplexed over the USB4 links & across 40Gbps packet-switching hub chips.

There still seems to be a fairly top-down tree architecture to USB4[1], but there is also mandatory host-to-host networking support: one of the most exciting things about the new Intel laptop offerings, to me, is that they almost all offer 40Gbps host-to-host communication over a good variety of plain-old usb-c cables (20Gbps for less speedy cables). It's an interesting capability, but how technically the packet-switching system works to implement this is very interesting to me, something I'm not clear on, especially as compared to FireWire.

I'd also love to know whether this host-to-host networking functions (& how) if there are routers/"hubs" between the different hosts (the linked [1] implies it ought work). This could all behave like USB-OTG, where one computer starts looking like a device, but I'm not sure if it's that fixed a system: does USB4 begins to permit slightly more bus-like non-top-down behaviors. Does a usb hub/router have to be entirely top-down in USB4, or can it negotiate different hosts access to different devices?

Still missing is DMA (short of USB4 PCIe transport, which again raises the host versus device question in my mind! Is there NTB support on PCIe over USB4 anywhere? That would be kicking rad) which FireWire obviously has had in it's wheelhouse for a long long time now.

[1] https://www.usb.org/sites/default/files/D1T1-3%20-%20USB4%20...

Thunderbolt is really poorly documented in general and the only documentation for Thunderbolt networking and the underlying XDomain stuff appears to the Linux patches: https://lwn.net/Articles/734019/ USB4 appears to have only minor changes compared to Thunderbolt.

As stated in the article FireWire (IEEE1394) to its variants was different (and much better) than USB.

The scariest thing was hot-swapping/plugging. FireWire had beefy power supply capabilities at the time.

We joked a lot of the 'fire' in firewire since you were easily able to 'fry' equipment (mostly the FireWire controllers).

I know a lot of non-techie people that plugged the FW400 port opposite which would usually fry your equipment.

USB4.0 is however great news imho since it merged Thunderbolt was is capable of low-latency which was the main feature (at least for audio) with FireWire.

Firewire was way ahead of its time (realizing early dma was going to be the bottleneck)! Although ultimately thunderbolt is probably the most efficient solution.

Fun little fact: FireWire 800 is used to back CAN busses on the f/a-18. Pretty danged cool!

Note to author: please increase the contrast of your text versus background, and make it a bit bigger (15 or 16px).

On topic: interesting read! I heard of daisy-chaining firewire before, but I didn't realize that they were secretly hubs, and that hubs are kind of like network switches (or -hubs), unlike usb "hubs".

Reply on news.ycombinator.com