Tweaking an ESI Juli@ XTe card

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I raised this with the guys at FA. They responded:

The ground return goes back to the psu. You always want to take noise away. The clock should not be grounded at the signal end. It is bad practice to connect the ground up at the clock end. What you should do is use a coax for the signal with the ground connected at the clock but not at the destination end.

I'm still trying to understand this. (I sent them the link provided by Nige2000.)
 
Member
Joined 2003
Paid Member
On clock removal, I tried a couple of things last night.

The trick is that the long SMD leads under the clocks have to be sufficiently heated to loosen their solder joints before one tries to move the clock. I don't think my hot-air station would do this.

Best was using two soldering irons coming in from opposite sides of a clock. Prepare by tinning each terminal with a medium-sized blob of solder to ensure good heat transfer.

Then touch and hold the irons on each and apply a very slight pressure to move the terminals. When the solder is sufficiently softened, they will begin to shift. You can then GENTLY push them until you have cleared the pads underneath.

I was also able to do this with one iron moving one side at a time just a slight distance, then switching over to the other side, heat and move it a slight distance, repeat until the clock is loose. But this method has a greater danger of damaging the pads.

Trying anything to move the clocks before the solder is loose WILL damage the pads.

And as I said, a hot air station is not the right tool for this as the SMD leads extend a long way under the clocks AND are insulated by the plastic base. It would be very hard to heat them up sufficiently with a station.

My 2 cents on this.

Greg in Mississippi
 
I raised this with the guys at FA. They responded:



I'm still trying to understand this. (I sent them the link provided by Nige2000.)

Sorry, but this is so wrong.... The power should have its own twisted pair connection and the clock signal should go to the board via twisted pair or coax with its return (ground) soldered as close to where the clock signal is as possible on the board. To recommend anything else is sheer stupidity and shows no understanding of how critical clock signals or signal integrity (or EMC engineering). Having the clock return go via a long torturous route via the power supply is stupid....
EMC and signal integrity are two sides of the same coin, with that in mind read this rather humorous look at how NOT to do things from one of the Guru's regarding this sort of thing.... The first point he brings up is the pertinent one.

The 10 Best Ways to Maximize Emission from Your Product

The instructions put up by Nige2000 say the same as I do, as do instruction for the Tentlabs clocks....
 
I'm inclined to agree with you but then I am a novice at this sort of thing and not really in a position to debate it hard with the folk from Fidelity Audio UK.

I also find it somewhat intriguing that the LC Audio instructions appear to show no GND connection at the clock side. Yes the shield is connected on the device but doesn't appear to be shown connected on the clock.

I'm now wondering what, if anything, I can do - particularly given the damage caused by removal of the oscillators...

PS: in contrast to the LC Audio instructions, FA also said to remove the resistor.

(Re clock removal, yes two soldering irons at the same time is likely the best way to go. I, however, only had one.)
 
I'm inclined to agree with you but then I am a novice at this sort of thing and not really in a position to debate it hard with the folk from Fidelity Audio UK.

I am though:)
To be perfectly honest I find most of these add on clock board worrying, clock signals should be short (as short as humanly possible) which is why on most good PCB design the clocks are placed as close as possible to a devices relevant pins or if this is not possible some thought of correct termination of the transmission line. Taking the clock of-board is never a good idea, but if it is done it should be done with the uttermost care, preferably with some buffering to drive the long lines created by being off-board, a bespoke clock buffer being the best solution.
The clock daughter card must have a thick low impedance supply lines especially the Ground to minimise any chance of the two boards grounds being at different potentials. The clock signal itself should be transferred preferably by co-ax and miniature rf connectors or similar, the signal having its own ground connection connected to the main PCB as close as possible to the point where the clock signal enters the board.
To do anything less is more likely to create a clock that performs worse than the one being replaced and may also add more noise to a system due to EMI, negating the point of the modification...
A bit more on signal propagation....
http://www.x2y.com/filters/TechDay0...log_Designs_Demand_GoodPCBLayouts _JohnWu.pdf

Your PCB can be re-worked with care, have a look at this...
Circuit Board Repair and Rework Guides
http://www.intertronics.co.uk/resource/guide1.pdf
 
Guys, I'm following this with interest.

In case it's any help, when I've removed oscillators from boards I've done it by melting the solder on one pin and then immediately (and very slightly/gently) levering the same end of the clock up with a thin scalpal blade - or other very thin thing! Then go to the other end and repeat. Try to keep the heat to a minimum but it works for me.

Re the grounding points for the clock signal. I agree that theory dictates the closest point to where the clock signal attaches to the board and also that sending it via the psu is nuts. However, a couple of points: firstly the closest point on the board may not be the closest electrically...I've had to fiddle around trying to find the best ground point to get motherboards to boot and sometimes it's not been the closest physically . Secondly, I'm not convinced that it's the improved clock signal that is contributing to the improved sound that we're hearing. When I've replaced clocks on motherboards and I've measured the clock signal before I've removed it, I usually find a perfect sine wave...literally textbook. However, often the clock signal that I replace it with is certainly not perfect. Not sawtooth but more like sine with kinks in it! Anyway, the point is (and I seem to remember nige2000 saying similar somewhere) that it's the improved power supply to the clock that is actually the major contributing factor here rather than the signal itself.

This would seem to tally with projects like Ian's FIFO where the clock supply massively affects the sound coming out the other end.

Hope this helps,
Crom
 
Hi Crom, could you explain what you mean by closest electrically?
If you look at how a signal travels and the lines of both the E and H field, you are trying to keep that relation ship the same or as near as possible for the complete path the signal travels. Because of the layouts I do I always think of every signal as a balanced or differential signal, so every signal I route I am thinking of how its return current with travel and am I adding any discontinuities...
The whole clock power delivery system is critical, including decoupling, now a Audio DAC is relatively critical, when you get into some of the communication stuff it can get quite hairy... Its always worth a look around and see whet they are up to.
 
its prob a bit ot but suppose the theory is still relative
crom and myself have been replacing pc motherboard xo/clocks/ps
to aid sq
i realise it makes little sense but it really really works

we have found it particularly fickle in regards gnd routing
this could be down to a lack of understanding/knowledge on my behalf

its certainly something myself and crom would be extremely interested in due to the benefits we have found so anything we can learn would be greatly appreciated

id like to have a better understanding about the resistors and caps in the crystal circuit and what to remove or modify when replacing with an external clock

on the motherboard there is 25mhz and 32,768khz and if using a usb pcie card an additional 24mhz clock on that

the idea sprouts from this guy

SHOPPING AREA: Clock Upgrade Module for Main Board

ive had a play with some of his stuff i dont think its optimised either

i think theres much more that can be done
 
So he adds a clock with a 40cm power lead, a 35cm xtal lead (un-shielded at RF frequencies) with no thought to correct termination and a ground loop for the clock return that is at least 75cm... And this is an improvement!
All I can say is have a look at point 2 again:
Appendix B: The Ten Best Ways to Maximize the Emission from Your Product - Electromagnetic Compatibility Engineering - Ott - Wiley Online Library

Words fail me:eek:

lol

its actually coaxial but yea even I with limited knowledge thought wtf


the scary bit is that even this shows improvement over stock configuration

ill give your links a good read later thnx
 
Its coax with the screen connected at one end only! So the signal couples to it inductively and capacitivly, have a look up how antennas work I detect some similarity...

"connected at one end only"

Which would seem to tally with what FA said, no? Except that they said GND connected at the clock but not at the destination end. In both cases the coax shield is connected at only one end...

EDIT: sorry was catching up with the posts in this thread and initially thought Marce was responding to my comments that the LC Audio implementation also seems to show the coax shield is only connected at one end.
 
Last edited:
I am though:)

;-)

clock signals should be short (as short as humanly possible) which is why on most good PCB design the clocks are placed as close as possible to a devices relevant pins or if this is not possible some thought of correct termination of the transmission line. Taking the clock of-board is never a good idea, but if it is done it should be done with the uttermost care, preferably with some buffering to drive the long lines created by being off-board, a bespoke clock buffer being the best solution.

Agreed and one of the advantages of the Microclock is its small size and ability to be stuck directly to the Juli@ with very short leads to the crystal pads (preferably not destroyed when removing the initial crystals)


The clock daughter card must have a thick low impedance supply lines especially the Ground to minimise any chance of the two boards grounds being at different potentials. The clock signal itself should be transferred preferably by co-ax and miniature rf connectors or similar, the signal having its own ground connection connected to the main PCB as close as possible to the point where the clock signal enters the board.

I think the only bit being debated is connection/use of the shield, no? or are you suggesting the shield should be used like a Faraday cage (connected only at the clock end) and there be another connection for GND?


thanks for the links, I will take a look
 
So he adds a clock with a 40cm power lead, a 35cm xtal lead (un-shielded at RF frequencies) with no thought to correct termination and a ground loop for the clock return that is at least 75cm... And this is an improvement!
All I can say is have a look at point 2 again:
Appendix B: The Ten Best Ways to Maximize the Emission from Your Product - Electromagnetic Compatibility Engineering - Ott - Wiley Online Library

Words fail me:eek:

When I looked at his clock upgrade I had exactly the same reaction. (Same with the stuff from Jaehong Lee where the crystals are replaced but little thought is given to the way they're powered.) But remember he plays to a relatively inexperienced crowd over at Computer Audiophile.

I dipped my toe in this water by purchasing the cheaper and smallest of the FA clock modules...
 
"connected at one end only"

Which would seem to tally with what FA said, no? Except that they said GND connected at the clock but not at the destination end. In both cases the coax shield is connected at only one end...

EDIT: sorry was catching up with the posts in this thread and initially thought Marce was responding to my comments that the LC Audio implementation also seems to show the coax shield is only connected at one end.

Which is if you have any involvement with design of digital circuitry and EMC know that this is stupid and is not how you shield things but how you make an antenna....
The coax shield should be connected at both ends and carry the return current for the clock signal, so the add on board will have 2 connections to ground... one for the power supply twisted paid and one for the clock signal to provide the return path....
I am seriously thinking of using that site as an example for my training sessions as a perfect example of how not to do things with clock signals..... There is no real technical information I can find for any of the products (most of which should be on a certain thread in the Lounge).....
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.