Hi

Well you have done the easy stuff, the hard stuff starts now!

I have to have a proper think about this!

First guess,  the boot up message is hard coded from the sata driver to say that the sata is not powered, because it wasn't designed to be powered in the plug. 

I cannot find a sata "power ok" signal that the kernel would use. There are a couple of sata status signals mentioned in the marvel data sheet, but who knows?

However looking more carefully at the plug schematic and the
marvell data sheet, what you have connected is called the Sata analogue voltage.

We may have missed another missing component or link, need to pore over the schematic again.

What kernel and uboot  are you trying.   You will probably need to  look to the openrd-base kernel to get an idea of the patches used on that device.  I haven't had time yet to see if the sata driver patches have been incorporated in the present kernels

I am off on a family visit to see my new grand daughter today,  so cannot do any research.  There is a 794 page marvell functional specification doc on the net, not the short sales data sheet. It is probably worth a read.

I cannot post the document, one it is MArvell's and 2) it is too big! bu this is the title:- FS_88F6180_9x_6281_OpenSource.pdf


cheers

Patrick

Edit  added detail on document

Hi

You are correct.  The filter system provides the 3.3V to the sata subsystem.

When I first got my Plug, I happened to look all the way through the dmesg output of boot up.  One of the entries said, (from memory) something like "Sata subsystem not powered up".

If you look at the various sheets of the V1.3 schematics,  (they are a bit confusing, since only parts of the CPU dealing with the title on that sheet are drawn out). you will eventually find that the sata subsystem, the usb subsystem and the ethernet subsystem all have individual power input pins.

I suppose the main reason is that the OEM's  using the Marvell  computer on a chip,  only need to power up the parts of the CPU and subsystems that they are going to use in their designs, thus saving on total power consumption in their design.

It is also easier to filter the HF noise out of the 3.3V supplies by putting in several small filters. Filters that have a "large" dc current, (amps), mean that you need a physically large inductor, to prevent it saturating.  Remember the cpu is clocked at 1.2 GHz, there is going to be a lot of noise generated from the clock and cpu  switching.

cheers

Patrick
 

Hi,

A 100 ohm  resistor is no good.  You will drop all your volts across it!

Unfortunately Global did not release the BOM for the V1.3 Plug. However if you look at the BOM for the original version,  a spec and part number  for a similar ferrite is:-

60 ohm IMPEDANCE 25% tolerance 100MHz 3A dc 0603  TCMA 160808-U0060.

I could not find a supplier for this one  but chose a Mouser / Kemet part Z0603C600APMST,   this is a 60 ohm imp 2.5A device. This has a dc resistance of 0.1 ohm.

The BIT7 is a series inductor, used in a "pi" filter to suppress HF noise in and out on the 3.3v rail to and from the sata port.  The ferrite inductor is PART of the filter and should not be omitted.

I have attached a pdf of the Mouser ferrites for you.

best of luck

Patrick



 

Hi,

You do not appear to be measuring the 3.3V at the correct pad !

The 3.3v is connected to the top of  C197. ( Not fitted as standard)

C197  is  situated on the Marvell CPU side of the PCB to the right hand side of the CPU   The other side to your picture!

The other pads of the 3.3v filter C78  C79 C196 are NOT connected to 3.3V until the ferrite BIT7 is installed

However BIT7 pin 2 should have given 3.3 V is also connected to C197 hot side .

ON MY PLUG PCB, measuring between the black pin of the 5v power connector and the two pads of BIT7 I get the CORRECT 3.3v on the lower pad, nearest the edge of the pcb (your green pin1)

BTW,

This Esata part from Mouser fits exactly to the pcb   Mouser p/n 649-10067172-70CLF   manf # 10067172-70CLF

Finally Soldering the components.

The components are VERY VERY difficult to solder. They are minute 0402 size apart from the ferrite (0603),  and the difficulty has been increased since Global Scale used  a solder paste mask which included the unfitted components. This means that the pads already have big bumps of solder on them that have to be removed before one can solder the caps and ferrite in place, otherwise you get tomb stoning!

If you have not hand soldered 0402's before, do NOT start to learn on this pcb.  You need a proper rework soldering station, a microscope (or other optical aides) and a great deal of patience.

I am waiting for delivery of another 0.2mm radius tip bit for my iron, my existing bit has worn down too much. I have soldered the 4 serial caps in the esata lines, but need  the new bit now.

Unfortunately it has been held up in the UK postage strike for over a week!
 

Those are very harsh words, and I think over the top. What if Global Scale did not receive your email?

I suggest you write to Global Scale Snail mail,  politely, detailing your problems. I am sure Global scale will sort out your troubles

I have always found them responsive to emails, although slow.

My order took approx two weeks to complete, which I was satisfied with. 

In future I will be using the UK  based distributer, since they are so much cheaper, (I am in the UK and the shipping and duty charges from Global Scale in the states are exorbitant).

By the way. Global Scale are not only a distributer, but the designers and manufacturers of the Plug computer.

Marvell only make the cpu and glue chips!

regards

Patrick

Hi,

What we are after is a through hole pcb right angle connector similar to the one  on the attached pdf.

This can be purchased from Mouser in one off quantities, but shipping from US to Europe for me is a bummer (again)


(edit spelling!)

@andy

Re missing components to get esata working

Hi

Don't forget that you also need to connect the POWER to the esata part of the processor.  If you look on the schematic  powerplug v1.3 sheet 7/11 cpupower you will see a power filter set of components connected to esata cpu pins N6, N8.  (These are not fitted on my system as shipped).  The components are C78, C79, 100nf, C196, C197, 1uF, and a ferrite BIT7, 60 ohm. You will have to add the signal TX and RX Caps as well as you stated.

If you can find an Esata connector please let me know I am still looking!

cheers

Patrick

Hi Birdman,

Pardon?

DECNET never used the physical MAC address so you are incorrect in your statement. Please read the following quote:-

DECnet hosts do not use manufacturer-assigned Media Access Control (MAC)-layer addresses. Instead, network level addresses are embedded in the MAC-layer address according to an algorithm that multiplies the area number by 1024 and adds the node number to the product. The resulting 16-bit decimal address is converted to a hexadecimal number and is appended to the address AA00.0400 in byte-swapped order, with the least-significant byte first. For example, DECnet address 12.75 becomes 12363 (base 10), which equals 304B (base 16). After this byte-swapped address is appended to the standard DECnet MAC address prefix, the address is AA00.0400.4B30.

(From Internetworking Technology by Cisco Technologies, I could not find my old DEC Networks book!)

You nearly remembered correctly! but  I am writing about the manufacture assigned MAC addresses!

Please believe me when I tell you that all Ethernet devices normally have a non alterable manufacturer assigned MAC address burnt in. as  I believe so does the Sheevaplug, and since the only "ROM" available is the Efuse, it is probable that it is stored there. There are a 144 bytes so even after loading the MAC there is still space available. 

In most of the other ARM controllers I have worked on, the Efuse or similar circuits are also used to protect the firmware, for commercial reasons.

Cheers

Patrick

Hi Birdman

I am very surprised by your statement that the MAC address can be overwritten at will!

The IEEE802 MAC spec defines that the 48bit Mac address should be PERMANENTLY written into any ethernet device,  (otherwise chaos would ensue on a network!)

I know when I worked at DEC, we used to get a big block of MAC addresses from IEE802, or their agents to use to burn in our NIC Roms individually.

I have not yet found any Marvell documentation that reveals the addressing methods of the Efuse, the data sheet only states on the features page:-

• 128-bit eFuse (one-time programmable memory)

Have you any documentation that describes the use of the Efuse ROM?

regards

Patrick

Hi,

I have seen several posts pondering the purpose of the two pin connector marked J2 2.5V

I believe this is ONLY for a manufacturing process, enabling the writing of the write once only EFuse in the Sheeva cpu with the individual  MAC address.

There are 44 bytes of write once Efuse memory in the Sheeva for this purpose.  The read voltage is 1.5V and the write voltage is 2.5V

I would steer well clear of applying any external voltage to J2 in case you do something untoward to the EFuse memory.  It may be OK but  better safe than no mac address

YMMV

cheers

Patrick

 Hi Rabeeh

Many, many thanks for the schematic drawing.

I have had a quick look,  the esata  details are very interesting!  If I am reading it correctly, lots of fine rework soldering is about to occur, I need to dig out my Aoyue soldering station.

I am still puzzling over the GPIO  signals, where they have gone, I need to do a compare between the old and current revs!
The schematic was originally on A3 sheets, and my viewer is a bit unstable trying to view at 200%, keeps jumping from sheet to sheet.  I need to find a better way to look at these sheets.

thanks again

Patrick

Hi,

I checked out the latest version of rootstock and it works great!

It can take a looong time on a slow m/c though

cheers

Patrick

Hi,

I used rootstock method last week worked fine!

However,  downloaded latest version of rootstock (29) today. This has a bug


Noted  OLiver has been altering this part of code, I am trying register with bug reports on Launchpad, but am having problems. If anyone can check this and ping Oliver for me I will be very grateful.

Patrick


EDIT   Finally got access to bug reporting   Oliver has  fixed the bug and raised the curent revision of rootstock  I am trying out the new rev, but it takes several hours to build the file system . Will add reply when finished build

Hi,

For some time now, (since around July), GlobalScale have been shipping a new single board SheevaPlug, Version 1.3. This revision is based on ONE module, instead of the two in the original shipped product.  None of the hardware documentation on this website has been  upgraded to include the current shipping revision, nor has the CDRom supplied with the Dev Kit.

There are significant differences to the product, (one example is the change in the FTDI recognition resulting in problems connecting serial terminals). Other members of this forum have posted photographs of the intriguing additional connectors available on the module. Another problem is knowing what has happened to GPIO and other signals that were available on the FPC cable between the two boards.

I have made repeated requests for the new schematic and layout to be made available,  via PM, to  Rabeeth  whom I believe is a Marvell employee, and by several emails to GlobalScale sales and support.   I have not had any reply, not even the courtesy of an acknowledgement to my requests. It is as though my emails are being sucked into a black hole!

My understanding of the SheevaPlug dev kit was that it was an open product, which to me was confirmed by the original complete set of mechanical, electrical drawings and the BOM being made available.

Since there is apparently no movement to publish the current revision documentation,  I have come to the conclusion that the Sheeva Plug Dev Kit is only an open software design and the hardware design is now confidential.

Please can someone either in GlobalScale or Marvell tell me whether this is so, and if the hardware design is still open,  please tell us why the current hardware documentation, i.e. layout, schematic and BOM, are not being made available.

thank you

regards

Patrick

Hi,

Fedex charged me £6.80 charge and £9.00 vat @15% total extra £15.80

This makes the total UK cost of my Plug to be  £88.80 cost plus shipping  (paying by Paypal) + £15.80  = £104.60  Ouch!

BY THE WAY

My FedEx invoice arrived 19th Sept 5 days after Plug was delivered

There is a note on it that the charges are being "revamped", for details to look on their website.  When I look I find that the advance payment charge from 15th Oct will be 2.5% of value OR £10 minimum, so the £6.80 fee will be £10 from then!  Somebody is making FedEx recession proof!

cheers