That TI USB Host Controller is very interesting because it is designed to take a 48MHz clock as standard, although it can take any frequency up to 50MHz.  The NEC chips can only take 24MHz crystals (although a clock can be used) which could be important considering these chips all 'clock-up' to 480MHz.  The TI can do this with half the frequency change to the NEC.

You will see that the crystal used on the Silverstone board (in Peters photos above) whith the TI chip is in fact a 48MHz.

Possibly important.  Peter (or anyone else for that matter), if you have one of these boards with the TI chip lying around I would be most interested to hear your impressions compared to the Silverstone with the NEC chip.

Cheers,

Anthony

Hi Nick,

I just sent you an email with the photo of the card.

Cheers,

Anthony

Hi Coen,

I run the Silverstone on only the 5V pins in the Molex.  I have never connected the 12V wires to the plug.

Like you, when the card is running and I unplug the molex the card becomes invisible to the o/s (but I don't have 12V connected at all).  My deduction is that the 5V is used for more than just the 5V for the USB plug.  I think that it might run the oscillator or something else.

Cheers,

Anthony

A noise sub-forum does make sense good Peter.  That would save me having to search through the other boards for noise stuff.  There will be plenty more talk about noise coming up too by the sounds of it.

Cheers,

Anthony

I'll have what you're drinking Peter.  No such thing happened here...not even close.  Are you able to remove the condom and see if it reverts?

Anthony

I concur.


Peter, you have just changed to a PCIe-USB card because it sounds better than your mobo port.  What makes you think it can't get better again?

Anthony

Hi guys,

I have done several in/outs with the Silverstone card and have isolated it by simply removing the bracket from the card when I want it isolated and then replacing the bracket and screwing it into the chassis when it is not to be isolated.  It takes a while between re-boots and fiddling with screws (Peter, I even dropped the chassis screw onto the mobo at one stage and held my breath when it was time to restart).

Regardless of whether the bracket is in place or not I have a hum through my speakers that is only evident when I put my ear within about 10cm of them...it is standard transformer noise...and is evident in both the woofer and the tweeter.

DC offsets seemed to be the same in both test scenarios.

What did I hear?  Not much difference.  Sometimes I think that with no bracket that there is a little more clarity to the sound but that does not seem to be a stable observation.

Then, instead of powering the Silverstone from the SMPS I power the Molex from my 5v lab LPS.  Nice improvements...blacker black, more flesh (ringing, decay) and separation on the cymbals, the double bass is easier to follow etc..  This is when the Silverstone is bracketed and non-bracketed.

Cheers,
Anthony

This made me laugh Peter.  Personally I remove the screw from the cross-member (through which you put your screw-driver) and cast it aside before I screw in the USB card.  Three screws instead of one I know, it's a pain!!

Cheers,

Anthony

I remember you saying that Paul.  My V1 card responds to linear but not in a 'wow' sort of way...more a "notice it when it is not there" kind of way. 

Ok...I've can't recall if I have tried the Silverstone without 5V connected to the Molex (I don't connect the 12V at all).

I am pretty sure the 3.3v goes straight to the USB Host Controller IC, but I could be wrong.  I guess I will find out when the PCIe riser turns up and I can cut the power from the PCIe socket.

Anthony

Quite sure.  It even says so on the Silverstone box.

Hi Coen,

I have just had a look at a USB board design that I have here (from Renesas) and you are right, it does seem as though the bracket is connected to both ground layers in that particular design.

Regards,

Anthony

Alain, have a look here for the pinout of the PCIe socket.  We are only interested in pins 1-18 on both the A and B sides.

Cheers,

Anthony

Hi Alain,

I don't think that there will necessarily be two voltages required (there might be 3).

The USB Host Controller Chip (NEC 720201) requires two voltages:  3.3V and 1.05V.  Obviously the 1.05V supply will need to be regulated from either the 3.3V or 12V coming from the PCIe bus or from the 5V feed from the Molex Plug (the Molex can also provide 12V but I have not seen a USB card that uses it).  On top of the 3.3V/1.05V for the IC, 5V is needed for the power output of the USB plugs themselves, so there are at least three voltages required on the USB card (3.3V/1.05V and 5V).

But the PCIe bus only provides 12V or 3.3V.  I would think that most well designed USB cards will send the whole 3.3V PCIe feed straight to the USB Host Controller IC and the 1.05V the IC needs will be derived from the 12V rail of the PCIe (and possibly the 5V for the USB ports as well).  The Silverstone card is a little different in that it NEEDS 5V from the Molex plug to operate, which suggests to me that that 5V is used for more than the voltage in the USB plugs, it may be used to run the crystal or derive the 1.05V rail or something else, which may be part of the reason that it 'sounds' better than other cards.  Who knows?  Hopefully some of this will become clear when that PCIe riser cable turns up.

So, I would definitely hold off on a high priced LPS because I don't know where this is going to lead, if anywhere.

Regards,

Anthony

Hi Juan,

I would hold off on the Pardo for now.  As you would have read in the other thread I am about to start to play with isolating the USB card from the pcie bus.  If that works then you will be up for another supply because the pcie bus only supplies 3.3v and 12v to its inhabitants (not 5v like the Pardo supply) so your new supply may become quickly redundant.  Plus there are better supplies about than the Pardo.

Cheers,

Anthony