how add ESD surge supression mod to firewire audio interface?

[dreamcat4]

Recently aquired a couple of focusrite saffire pro40 firewire audio interfaces... they are both in fully working order now.

The thing is that this device is notorious for being succeptible to power surges on the firewire bus killing it's TI firewire chip. For example if you do not power on the equipment in the correct sequence.

This also means it isn't possible to do device hotplugging etc. And you have to be very careful not to forget when power cycling both the PC and the interface. Since the interface needs to also be power cycled too (quite often).  And lest you not forget, they put a big red label next to the firewire ports (photo attached).

Therefore was wondering if it would be possible to have a go at modding one of them. The goal being to improve whatever is the input protection. In the hopes to make the input more resilient to transient spikes at power on.

My thinking is that perhaps when this thing was originally designed, they didn't realize or just didn't have access to better (newer) surge protection devices. Which might have come out since then.

Doing some research didn't bring up any existing mod(s) that i know of. However I did find an application note by TI referenced from this exact TI firewire interface chip, which linked to their range of protection devices. Namely:


ti 4 channel
===================
tpd4e02b04 = +-3.6v, 0.25pF
tpd4e05u06 =   5.5v, 0.40pF


however there seems to be a more reverred / more highly regarded device which is this one:

eaton / cooper / bussman
============================
41206ESDA =    30v, 0.15pF

and it's specs are better. Including being rated to the +30v continuus (which is the vcc+ of the firewire bus)


So I checked the package dimensions, which is 3.2mm with 0.8mm spacing between the pins. And this seems to match exactly the same width of 2 existing devices already there on the PCB, marked on the silkscreen as "L1" and "L2". Which themselves measure a nominal resistance of 2 ohm.

It seems that there is (at the very top of the photo). A ceramic cap between the +30v firewire vcc coming in thru the cable. And the local case ground.

However the GND line (pin 2) on the firewire cable itself is left unconnected to anything on the PCB. This would make sense, being a piece of analog audio equipment wanting to isolate the ground from the computer.

In fact that +30v vcc line from the firewire only seems to be used as a return path for transients protection. And nothing else. Since the firewire signals are diff pairs (similar to usb2 with its D+ and D-, just except there are 2 differential pairs instead of only 1).

So that seems to make some sense their design choices. However I am actually wondering if the devices marked L1 and L2 are really simply just inductors. As the silkscreen says... well i suppose they must be given they are in series on those data lines, with no escape path to ground. So then it's just to round off the edges of transients?

With the CR simple filter network to the sides, which is what is actually returning that path to the +30v firewire vcc pin1 line?


So assuming this is correct, then I would probably want to install the 41206ESDA surge supression device... right on top of / butting against next to those inductor terminals? with the other side bridged to the local case ground? since the package footprint / spacing lines up. Does that make sense here?

See it might be a bit easier if could instead make the return the path to that same +30v firewire vcc trace. Which is already there running horizontally across. But the datasheet for the eaton surge suppressor (41206ESDA) says the escape path should go to GND. So given this difference between the circuit and the datasheet... am not entirely certain which is best. Not really my area of expertise!

Guessing should contact the manufacturer and ask them. Eaton / Cooper / Bussman.

Anyhow thought it might be cool to post the photos / details here. And see if anybody else finds it of any interest. Just as far as I could get with it

See I probably will hear people saying "firewire is old hat" etc. But the thing is great. And is quite affordable / repairable / servicable actually. Unlike today's replacements. So the idea of adding the ESD protection kindda seems worthwhile to me. These things could last many more years if well maintained.

[dreamcat4]

huh... or maybe the Texas Instruments devices are a better idea here? Because in their application note they show a different wiring configurations. For example the 2 devices they say are appropriate for firewire bus... they give them as an example for HDMI on page 10 in the pdf here (attached below).... Which is a bit different to the Eaton device (only 8 pins)

[NoMoreMagicSmoke]

This is a problem with Firewire in general. Turns out it is nowhere near as hot pluggable as claimed. The connectors have too much slop which can allow the +V pin to contact the signal pins which can damage the PHY. There are some whitepapers floating around the internet that document this well. It's not an ESD problem. It's a garbage connector design problem which there is no remedy but to treat the connector as NOT hotpluggable.

[NoMoreMagicSmoke]

Another issue is a late Vg. https://www.ti.com/lit/an/slea072a/slea072a.pdf?ts=1659057271304&ref_url=https%253A%252F%252Fwww.google.com%252F

[tooki]

This is a problem with Firewire in general. Turns out it is nowhere near as hot pluggable as claimed. The connectors have too much slop which can allow the +V pin to contact the signal pins which can damage the PHY. There are some whitepapers floating around the internet that document this well. It's not an ESD problem. It's a garbage connector design problem which there is no remedy but to treat the connector as NOT hotpluggable.

There are three different FireWire connectors: 4-pin FW400 (unpowered), 6-pin FW400, and 9-pin FW800. Which of the latter two are you referring to? Each of the three connector designs is radically different, so “a problem with FireWire in general” is not true as such.

[Audiorepair]

I have replaced several blown audio soundcard TI firewire IC's of the 6 pin variety.  In most cases the firewire sockets have been quite worn and "sloppy".

I had always assumed that they had been blown because the customer inserted the plug in upside down (this is possible with a warn socket)

Now I am more swayed  by the TI Late Vg event issue pointed to by NoMoreMagicSmoke.
Thanks for that.

I would have to agree that at least these 6 pin devices should definitely NOT be considered hot swappable.
That's an expensive repair.

[tooki]

I worked as a Mac consultant during FireWire’s peak, and there certainly weren’t widespread problems due to hot swapping. It was designed with that in mind, after all. But I can’t rule it out entirely.

What is true is that, on devices which didn’t provide any additional mechanical protection to the jack’s shell (see below), a rough user could force the plug in 180° rotated, by splitting the jack’s shell apart. It would not surprise me if such a jack would then be far more prone to off-angle insertion, allowing a short to happen. The “sloppy” jacks you refer to sound like this could be a similar effect.

In contrast, the FW400 jacks Apple installed on the aluminum Macs, where the opening is milled out of the solid aluminum of the computer’s enclosure, provided excellent connector support.

[NoMoreMagicSmoke]

This is a problem with Firewire in general. Turns out it is nowhere near as hot pluggable as claimed. The connectors have too much slop which can allow the +V pin to contact the signal pins which can damage the PHY. There are some whitepapers floating around the internet that document this well. It's not an ESD problem. It's a garbage connector design problem which there is no remedy but to treat the connector as NOT hotpluggable.

There are three different FireWire connectors: 4-pin FW400 (unpowered), 6-pin FW400, and 9-pin FW800. Which of the latter two are you referring to? Each of the three connector designs is radically different, so “a problem with FireWire in general” is not true as such.

The 6-pin FW400 with power.

[NoMoreMagicSmoke]

I worked as a Mac consultant during FireWire’s peak, and there certainly weren’t widespread problems due to hot swapping. It was designed with that in mind, after all. But I can’t rule it out entirely.

What is true is that, on devices which didn’t provide any additional mechanical protection to the jack’s shell (see below), a rough user could force the plug in 180° rotated, by splitting the jack’s shell apart. It would not surprise me if such a jack would then be far more prone to off-angle insertion, allowing a short to happen. The “sloppy” jacks you refer to sound like this could be a similar effect.

In contrast, the FW400 jacks Apple installed on the aluminum Macs, where the opening is milled out of the solid aluminum of the computer’s enclosure, provided excellent connector support.

I work as an engineer designing equipment that used the 6 pin FW400 ports and we used to experience a handful of failures every year out of about 100-200 devices a year. The failure was never a full catastrophic failure, but always glitchy connections. This glitchyness likely would not be an issue for many applications but was a major issue for us due to using the firewire connection for hard realtime communications. After dealing with this issue which always required a controller swap (bad phy) I did a LOT of research on firewire connections and found a number of whitepapers by other companies (like TI) that discussed phy damage due to imperfect insertion of the cables (angled connections not 180 flips). We implemented a zero hotswapping of firewire cables and our issues disappeared overnight. We did have a few failures in the 6-7 years that followed but every failure was 100% attributable to the cable being hotswapped shortly before the failure.

[dreamcat4]

thanks so much all you guys for these responses

unfortunately i only just discovered this thread wasn't in fact dead. just read them all. the forum didn't notify me (had to manually click the notify button....) 

ok so my primary concern is not to intentionally be hot plugging etc. but instead if the firewire cable ever gets accidentally ripped out while the thing is on. or if i ever forget, or the computer decides to turn itself on at a critical moment (after turning it off, and the thing then clicks back on, and yes this has happened before).

So all in all trying to err on the safer side.


What strikes me here, is that the device in question (in my first post here, the saffire pro 40)... well it isn't actually powered by the firewire bus. So if i can disconnect or remove the power at the other end. Or use a different cable. Or modify the cable or it's termination. To get rid of the +Vcc and Gnd only, wouldn't that help to protect the device?

Because looking at where the signal comes in on the PCB to the TI firewire chip. It just seems to need those differential pairs to carry the signal, if that is enough without a ground reference?

Or maybe I should just keep the ground connected (so 5 pins instead of all 6?) ?


Also do you guys see a point that i could still order those TVS protection diodes? Because I have to order some other things (and for other projects) from digikey. And they seem to be pretty inexpensive and cool looking devices.

But my main question here is about modifying the firewire cable / link. Actually have already some spare 6 pin firewire 400 cable for trying that out on here.

See these 2x saffire pro 40s is my only firewire devices, it's not like I have any other types firewire devices to also plugin or use those same cables. It's not going to cause any difficulty for something else.

Anyhow at least will get the reply notifications next time. Thanks so much for these previous responses BTW. Really appreciate it.

[dreamcat4]

but what i am really beginning to question myself here, is back on my 1st post. In the PCB photo. Where that the silkscreen labels L7 L6...

if those existing packages are in fact actually inductors or not. Or if they are already some TI tvs protection diodes. And merely the silkscreen is trying to be 'near enough' to reflect that. In terms of supressing high inrush currents. Without actally labelling them as D7 D6 (as diodes).

So that is where my understanding falls down. Since I don't have access to the schematics of these Saffire Pro 40 product mainboard. Am kindda left guessing here.

But maybe the example circuit in the datasheet of this TI firewire interface chip can clear that up. Assuming they were following whatever official examples from Texas Instruments hopefully.

[fzabkar]

In "PCB_top.jpg" you show that a thick trace is connected to the 30V firewire supply, yet TI's datasheet shows that it should be connected to the "outer shield termination". :-?

What does the firewire Vcc actually do in this application other than tell the device that a host is connected? The TSB41AB2 datasheet states that the CPS pin of the IC can be tied to ground via a 1K resistor if the application does not need to detect host's firewire supply, otherwise this pin should be wired to the firewire supply via a 400K resistor.

If the device doesn't need to see or use the firewire Vcc, could one then just open the Vcc pin at the host end? This would prevent any damage due to "jiggling" in the connector.

[dreamcat4]

yeah i never understood that. right from the very beginning, the 1st moment that i buzzed it out to discover that the horizontal bar was connected to +30v. It had me quietly questioning that.

Clearly we can see the 56 ohms, and the 5.1k corresponds to that diagram provided by TI in the datasheet. So the whole thing should be connected to GND here. But what GND ?

See my assumption was that being a pro audio device they were trying to avoid tying to GND for reasons of introducing unwanted noise on to the local GND of the audio interface. Which then might end up getting presented on the analog channels. As unwanted computer noise or whatever.

That was the best explanation I could come up with here.

The other GND (if not local GND)... would be whatever the signal GND pin on the 6 pin firewall cable (which then is coming directly from the PC GND).

So I think this is the crux of it. Whichever GND focusrite would use in following that diagram, it would either be injecting noise from the local onboard firewire interface chip here (as in the one in the same photo). Or otherwise it would be the GND on the cable, which (presumably?) might then upset the delicate high speed firewire signal. Injecting transients from the computer onto those diff pairs or whatnot.

If this is incorrect then please correct me here. Because really I am just guessing what they intended by diverging in this manner from the TI example circuit.

But the logical alternative is then to use the +30v firewire power instead. Which (again i am assuming) is more protected on the host PC side from such computer noise. Being a special voltage only derived and used specifically for firewire card and nothing else.

See I could potentially cut that trace going to the +30v pin. And then following the TI circuit instead re-route it to become a GND. However I hesitate to do this because it might be detremental in some other regards, for maybe introducing unwanted noise. But I could be wrong! (but then the focusrite engineer who designed this thing would also be wrong, which i am much more doubtful of than myself being wrong). If you get my drift here

[fzabkar]

I suggest you check your measurements. Something is not right.

FW802B, Agere, Low-Power PHY IEEE® 1394A-2000 Two-Cable Transceiver/Arbiter Device:
http://www.bulcomp-eng.com/datasheet/FW802B.pdf

[dreamcat4]

ok i can check. maybe not next time i re-open the device. because its all put back together now. and takes some significant dissassembly. but it should already be possible to check with that pcb_bottom image. if i didnt make a mistake. which pins it goes to on the firewire port. i will be happy to double check again. since i didnt believe it the first time, 2nd time or 3rd time i checked (few weeks ago now, so will have a fresh eyes).

... but i think the more important realization for me here is that there are no inductors included anywhere on that  ti provided example circuit diagram. none at all.

this then increases the likelihood that those 4 channel component marked as l7 and l6 are in fact not inductors at al. but instead is the package of some ti brand of tvs protection diodes bank. (which ti instead reference elsewhere later on in other application notes pdf rather than the original datasheet).

and when i really think about this for longer the more it makes sense that is what it really is. since how small and tiny must the inductor be to fit into that tiny smd package. wheras the dimensions of a tvs are actually exact same size of this thing.

then since there actually always was already had a tvs there, since the very beginnin. then there is the added  understanding thanks so much you guys. that its presence cannot possibly in any way protect from this type of a failure mode. coming from the incorrect hotplug insertion. at an angle

since the time duration of that event. with the misaligned pins. that is going to far exceed the duration of any short lived surge transient. which is what the tvs is meant for.

although i do wonder why if the original factory tvs cannot have a clamping feature of 30v why this damages those differential lines? the data lines themselves are not meant to be so high voltage then? i should just check the datasheet now... ok so that is vod the differential input voltage. indeed is listed as 176mv minimum an 265mv maximum. presumably goes into a very high speed fet. or something... and that is getting blown up. with this mis-plugging and crossing the +30v pin. and any tvs cannot stop it.

correct / incorrect?

[fzabkar]

VT6308P/VT6308S PCI 1394a-2000 Integrated Host Controller:
http://bitsavers.trailing-edge.com/components/viaTechnologies/VT6308P_PCI_1394a-2000_Controller_200811.pdf

[dreamcat4]

I suggest you check your measurements. Something is not right.

yep! measurements checked now 

as i'm sure you already know. I clean gone dun messed up. Was smart enough to figure out that I had to flip the pins over from the underside PCB photo. To where they came out on the firewire 400 port....

NOT smart enough to realize that I was looking at a cable pinout from the internet. Which then flips the port to the cable. (and therefore... then cancels out and puts them back to being on the same side).

What was the saying? 2 negatives don't make a right?

However to flip the poliarity twice times in a series configuration... gets you back to the original polarity. Something like that.

So anyhow I checked and indeed the horizontal bar isn't +Vcc (measures +12v coming in on the host PC card). It measured 0v when powered up and connected. And then the other pin (I marked in green as being n/c on the port)... that is in fact the +V bus power coming in from the cable. Which is not connected.

So sorry for this mess-up. I feel so stupid. Should update the photo. Cross out the mistake. Stupid

 

[dreamcat4]

OK then. So on to the different cables. Presumably this means that:

fw800 (4-pin)
============

* If i can use a fw800 (4-wires) connector on the host PC side. Then this is best for hotplugging safely? Since it only has the 4 differential lines as 2 differential pairs.

I measured from my firewire host PC card, that the diff pair B (= 'B+ && B-') rests are +3.2v dc above ground. Wheras the diff pair A ('A+ && A-') is resting bitch faces it hangs at 0v.

At least when nothing is connected at the other end of the firewire cable (client side). Just except for my multimeter test leads.

The reason I bring this up is what happen if the connector is angles. And pair A crosses to pair B. If that matters or not to cause any damage.


But regardless anyhow this is the safest possible type of cable to hot plug.



firewire 400 cable (6-pin)
==================

* This cable was measured to give +11.6v (dc) coming from this cheap chinese no-name host controller PC card. Which is considerably less than the +30v (but perhaps that is something negotiated in the protocol spec IDK).

* It makes sense to modify the cable at the host PC end, to insulate pins. Or block this +Vcc bus power pin. Since to the downstream device it does not need to use that pin for sensing if it works on the firewire 800 cable. Which i need to double check, but i believe that it does from my memory using a previous cable.

* I question: it does not matter? (at least no reason i can think of). Not any obvious advantage to disconnect the GND pin. On this 6 pin firewire. At least I don't clearly know myself. Perhaps there might be a benefit one way or the other. But we are dealing with a very low voltage digital differential signal here (only 0.36v +-). Which is about 1/10th of that +3.2v resting voltage on 'diff pair B'.

So with this modification that should be enough to safely hotplug. Or at least be less worried about the cable being ripped out from the back of the device! ...at some sideways angle enough to short the power pin.

9 pin firewire cable
===============

* I do not like this cable
* I do not wish to use this cable

* However this is the socket for my Apple Thunderbolt <--> Firewire adapter
* On my adaptor (or with my old and used 9 pin cable here)... This cable is not fitting securely into the socket, it wants to become halfway semi-out
* It is difficult to see or know how far where the pins are still properly engaged.
* Maybe it can even to fall out completely. And so easily with little force not to even notice
* It also wobbles very much angle from side to side.

Most Critical Question: Is this the specific connector you guys were referring to as being the cause of this issue???

Because the only other Firewire <--> Thunderbolt bridging device, ever made (that I can find)... it was the "OWC 12 Port Thunderbolt 2 Dock". And this also has the same 9-pin connector on the back.

So (for a laptop connectivity, instead of desktop PC)... then It looks like I very much have to use this port. At least for any reasonable economic cost. Without getting a full pci-e card eGPU thunderbolt dock. In which case then can use host PC card and avoid the 9-pins connector.


* It is more difficult to insulate / block the terminals on the 9-pin cable. Because they are so much finer pitch.
* However i guess the same applies here... the client device (saffire pro 40). It does not even connect the bus power pin to anything on the PCB. It can only cause a damage.

And I have no reason to use any other firewire devices (ever). Just only this client device. None other.

[fzabkar]

I have no experience with Firewire, but I would start by examining the CPS pin (24) of your TI arbiter chip. If there is a 400K resistor between this pin and the Vcc pin (1) in the connector, then the chip must be sensing Firewire Vcc. Otherwise, if there is a 1K resistor to ground, then it must be ignoring Firewire Vcc.

If Firewire Vcc is not required, and if you don't mind hacking the Firewire controller card in your PC, then cut the appropriate trace, or remove the polyswitch which protects this supply. I would be very surprised if there were no protection.

This card makes it easy. It has internal/external selection jumpers for the Firewire Vcc (and no protection):

[dreamcat4]

Very much thanks.

Actually (you did not know this). But I cannot use that firewire card with the VIA firewire IO chip on it. They are not suitable for my target device. For better compatibility with the audio interface... it must have a TI firewire chip. As per linux driver support / FFADO wiki page.

Anyhow. This is my host PC card from ebay china (https://www.ebay.co.uk/itm/304003560763). It does not have such a good PCB! However it does have TI firewire chip, which matters the most. And also because all of the other known TI chipset firewire PC cards... they are much more expensive than this one (or a VIA one). So did not want to pay 3x, 4x more cost.

Anyhow you can see here my plan in the screenshots. To just remove / snip the pin on the connector itself. Seems best way for that one.

[fzabkar]

Snipping the Vcc pin from one of the ports seems like a plan, provided that your device doesn't need it.

Edit:

If your card is powered from 12V, where is the 30V coming from? Is 30V what you see when connected to an Apple host?

[dreamcat4]

Well clearly this one cannot ever make +30v... because it's some cheap chinese rubbish. And whatever the 1394 spec says... i am assuming maybe it has a +12v mode. That +30v is optional IDK.

https://firewirestuff.com/techspec400.html

For +30v i simply don't know how that is achieved. But either way, we might wonder actually how safe is crossing a +12v? (If 30v also risks a damage).

The other thing to realize... we were speaking more generally about the hotplug issue. But in this case also hope to investigate in future laptop connectivity... which then is 9-pin connector most likely.

[fzabkar]

https://firewirestuff.com/techspec400.html

"When the computer is on or the power adapter is connected, the power pin provides a maximum voltage of 12.8 V (no load) and 7 W power per port. Maximum output current for both ports combined is 1.5 A and is controlled by a self-resetting fuse."

It looks like many vendors ignore this part of the spec.

Just out of curiosity I checked whether various cheap Firewire cards had any protection. Some do, most don't, and some replace the polyswitches with zero-ohm links.

https://i.ebayimg.com/images/g/2-8AAOSw3LVb2PY5/s-l1600.jpg
https://i.ebayimg.com/images/g/vuwAAOSwEB5h7aNb/s-l1600.jpg
https://i.ebayimg.com/images/g/t3oAAOSwU-NjETKB/s-l1600.jpg
https://i.ebayimg.com/images/g/V1wAAOSwb11iFREI/s-l1600.jpg
https://i.ebayimg.com/images/g/OgAAAOSwjF1jERrw/s-l1600.jpg
https://i.ebayimg.com/images/g/kwMAAOSwZSRjD4sO/s-l1600.jpg