• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required as is 20 years of participation in forums (not all true). Come here to have fun, be ready to be teased and not take online life too seriously. We now measure and review equipment for free! Click here for details.

SOTM sCLK ex multiple clock board hackery

Jinjuku

Addicted to Fun and Learning
Patreon Donor
Joined
Feb 28, 2016
Messages
625
Likes
255
#1
How does taking carefully placed clocks, optimized traces, proper isolation, and adding this piece of 'engineering' with long zip lines not just make things worse? I wouldn't be surprised to see system instability:

upload_2018-1-17_23-52-37.png


We are talking about going from optimized traces measured in inches or fractions of an inch to possibly a foot or more.

Is there anyway this makes sense?
 

DonH56

Technical Expert
Technical Expert
Patreon Donor
Joined
Mar 15, 2016
Messages
2,978
Likes
3,538
Location
Monument, CO
#2
Since the tX-USBexp board's claim is to provide excellent power isolation and a low-noise clock I am not sure why you would buy that board to hack. Nor do I see how replacing the system and Ethernet clocks helps anything but I don't claim to be an expert. These days I thought most USB issues had to do with how well the DAC was isolated from the USB power/ground and clock signals so this seems like an expensive solution for a cheap board, not sure it's needed?
 

Jinjuku

Addicted to Fun and Learning
Patreon Donor
Joined
Feb 28, 2016
Messages
625
Likes
255
#3
What would the longer wire lengths do for clock accuracy? My understanding when PCB's are laid out that there is a very particular approach and often determined by Intel and AMD as a reference design.

They want the best experience for end users with their product so it stands to reason they've done all the optimization required and you aren't going to do better then Intel/AMD at this sort of thing.
 

DonH56

Technical Expert
Technical Expert
Patreon Donor
Joined
Mar 15, 2016
Messages
2,978
Likes
3,538
Location
Monument, CO
#5
PCB clock traces are not all that great most of the time, Intel or not, but it isn't that hard to get them more than good enough. Cost is a big driver (even for prototypes and reference designs). Accuracy is determined by the source, which is usually either a packaged oscillator or crystal circuit, and either way can be more than clean enough to meet spec. The usual issues with clock traces are noise and crosstalk coupling in the board (and devices) plus loss and discontinuities in the traces since clock distribution networks tend to get large and things like PCIe edge connectors are not well-matched from an RF standpoint. Adding an external source using RF cables may well provide a cleaner clock but I just don't really see the need. Plus most of the clocks are differential so you need twinax cables (or a pair of RF cables if you are loosely coupled). And it does not solve the problems of loss and crosstalk to other components on the board, just brings in a dedicated clock for the USB board (which seems like all you would need, but maybe you have to replace them all to maintain synchronization among clocks -- not my area of expertise).

As for the connectors, I would guess the hack involves finding a suitable place to hardwire a cable. I have seen some boards with test points to SMP connectors but usually those are not populated and only used during initial test and validation.

Still seems like a solution in search of a problem; is there a link to a description of what this hack brings to the table, i.e. what improvements it claims to provide?
 

Jinjuku

Addicted to Fun and Learning
Patreon Donor
Joined
Feb 28, 2016
Messages
625
Likes
255
#6
Hmmm. I don't understand how the new clocks are injected into the motherboard. Are there motherboard with this kind of easy provision?
The only thing I can think of is you de-solder the old and then solder the wirelead.
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
20,539
Likes
25,140
Location
Seattle Area
#7
The PC motherboards use many layers. It is hard for me to imagine there are surface wires on top layer to just hack into for what is being sold to average Joe. It has to only work with very specific motherboards which they don't mention at all on their web site.

They come to audio shows. When I see them next, I will ask. :)

For now, everything Don said is true. :) :)
 

Jinjuku

Addicted to Fun and Learning
Patreon Donor
Joined
Feb 28, 2016
Messages
625
Likes
255
#8
The PC motherboards use many layers. It is hard for me to imagine there are surface wires on top layer to just hack into for what is being sold to average Joe. It has to only work with very specific motherboards which they don't mention at all on their web site.

They come to audio shows. When I see them next, I will ask. :)

For now, everything Don said is true. :) :)
This is something they should be requiring the mainboard being sent to them for re-work. I could only imaging how many ruined mainboards because of this.
 

DonH56

Technical Expert
Technical Expert
Patreon Donor
Joined
Mar 15, 2016
Messages
2,978
Likes
3,538
Location
Monument, CO
#9
I checked one of our board stacks: the top layer signal traces (we use for clock and high-speed signals) are 0.13 mm wide. Better have a good soldering system. My guess is they do like I do when I have to wire in a clock (or whatever) and find a component (like coupling caps, series resistors, or termination resistors) I can tap into to add the cable. And it is still a PITA, and you have to worry about lifting the pads during assembly and mechanical integrity of the cables so you do not accidentally rip them (and the pads) off the board (we usually hot glue and tape the cables). This whole scheme seems a little scary... In the best case, the boards may have test pads for SMP sockets or similar and zero-ohm resistors they can swap around to disconnect the on-board clocks and add the cables (we do that for clocks on our eval boards since we have to run with different clocks for some tests).

Still like to see some data showing real improvements gained...
 
Top Bottom