Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > comp.sys.raspberry-pi > #9781

Re: rPi and USB 3.1 w/ type C connector

Show all headers | View raw

On 10/7/2015 2:06 PM, Theo Markettos wrote:
> rickman <gnuarm@gmail.com> wrote:
>> On 10/7/2015 6:11 AM, Theo Markettos wrote:
>>> Switching to the Synopsys USB 3.0 or 3.1 IP will cost substantially more
>>> in licensing fees than the USB 2 IP which is a mature product
>>
>> Can you substantiate this?  I couldn't find anything indicating this.
>> Why do you say "Synopsys USB"?  Surely they aren't the only provider of
>> USB IP.
>
> The Pi uses the Synopsys Designware USB 2.0 HS OTG core (dwc_usb_2_0_hs_otg).
> I've worked on drivers for it.  I don't know how much the licensing fees are
> (and if I did know I couldn't tell you) but USB 2 is mature, USB 3.0/SS is
> mainstream and USB 3.1/SS+/Type C is bleeding edge.  Typically mature products
> are cheaper than new models.
>
> Changing IP vendor involes changing IP ecosystem, tool setup, verification
> solutions, licensing setup and contracts.  You don't do that without a good
> reason.  Mix and match of IP cores is possible, but life is a lot simpler
> and less risky buying from one vendor.

Fine, but you miss the point that USB 3.1 *will* be incorporated in new 
designs.  This is not a rPi issue, it is a Broadcom issue and *will* 
happen.  The market will use these new CPU chips and they will be 
competing in a market with the older chips.  I can't see the inclusion 
of USB 3.1 will significantly impact the price of the Broadcom CPU chips 
other than potentially for the first 6 months while there is a scarcity 
in the market of USB 3.1 solutions.


>>> The USB 3 IP may be targeting a smaller process (22, 28nm?).  The PHY is
>>> substantially different.
>>
>> Other than the PHY, IP typically doesn't target a process.  Even if it
>> does, a smaller process is usually less expensive for the chip.  That's
>> why they continue to shrink the feature size.
>
> On-chip USB is all about the PHY: it's possible to have an external PIPE PHY
> chip but you don't want to do that for cost, pin count, board cost and space
> reasons.  The rest of the controller is synthesisable logic which, as you
> say, is less process dependent.
>
> The Synopsys DesignWare USB-C/USB 3.1 PHY IP webpage mentions it targets
> USB-C 3.1 SS/SSP PHY, Type-C - TSMC 16FF+LL, North/South Poly Orientation
> which is TSMC 16nm FinFET.  There is no public information on what other
> processes they support.
> https://www.synopsys.com/dw/ipdir.php?ds=dwc_usb_3_1_phy
>
> The DesignWare USB-C/USB 3.0 PHY IP documentation page:
> https://www.synopsys.com/dw/ipdir.php?ds=dwc_usb_3_0_phy
> has a long list of processes including 16nm, 28nm, 40nm, 55nm, 65nm from the
> likes of TSMC, GlobalFoundries, UMC and SMIC.  So that looks like a much
> more portable target BUT only the TSMC 16nm FinFET supports Type C.
>
> None of the Synopsys USB IP cores mention video - I'm unclear how much PHY
> support it needs.

You seem to be saying a lack of information implies negative information.


>>> The above assumes SuperSpeed (5Gbps) or SuperSpeed+ (10Gbps).  It's possible
>>> to do USB 3.0 or 3.1 at only 480Mbps, which is just other spec changes.
>>>
>>> Doing 480Mbps/USB 2 over Type C is an option.  But you still need to route
>>> the video into the type C connector: that's an extra chip unless an SoC
>>> respin is on the cards.
>>
>> What extra chip?  Video and USB all come from the SoC, why wouldn't it
>> be routed inside the SoC?
>
> /If/ you respin the chip, it's all inside the SoC /but/ you need a respin
> ($million) /and/ you need a USB type C PHY that supports video ($lots, exact
> amount not public)
>
> /or/ you don't respin the SoC but you need an external chip (unit cost ~$5
> or whatever)

I have no idea why you think there won't be new SoC devices with USB 
3.0.  It's not a question of "respinning" an old chip.  They are 
constantly designing new chips.  The CPU used in the rPi 2 is not the 
same as the chip in the other rPis.  It was not a "respin", it was a new 
design, as will be any chip with USB 3.1.


>> Yeah, the power distribution of the rPi is poorly thought out.  With USB
>> 3.1 you can do away with the power connector and simplify the power
>> distribution.  Very likely a chip will be available to handle all
>> aspects of power for small designs like the rPi.  So the current maze of
>> power components will be reduced to a small handful with one IC for a
>> very low price.
>
> 'A chip' = $x added to the BOM.  Care to suggest a chip that supports USB
> 3.1 power delivery so we can see what the pricing looks like?

No, *a chip* could replace the existing power support *chips*.  I 
believe there are at least two and it may be three.  When a popular 
device requires multiple power supplies it is very common for chip 
makers to provide a power solution specifically matched to that device. 
  For example, there are any number of multiple rail PSU chips designed 
specifically to power FPGAs.  Same is true for many DSP applications.  I 
expect there would be such a device for a new rPi if the designers 
wanted it to happen.


>>> The bottom line is it's all extra cost.
>>
>> Bottom line is USB 3.1 with the type C connector is a *big* improvement
>> with little or no extra cost for small apps like the rPi.
>
> Changing the connector is cheap, but some of the other things you suggest
> aren't.  The Pi cost envelope is already tight.  How much extra on the
> ticket price would people be prepared to pay?

How much extra will it cost?  I think you are stressing this more than 
it is worth.  Just as other improvements have lowered the cost, I expect 
a new design which considers the large market for the rPi and other apps 
can be lower cost as well.

-- 

Rick

Back to comp.sys.raspberry-pi | Previous | Next — Previous in thread | Next in thread | Find similar | Unroll thread

Thread

rPi and USB 3.1 w/ type C connector rickman <gnuarm@gmail.com> - 2015-10-06 11:02 -0400
  Re: rPi and USB 3.1 w/ type C connector Dom <domafp@blueyonder.co.uk> - 2015-10-06 18:02 +0100
    Re: rPi and USB 3.1 w/ type C connector Jonathan Lane <tidux@faeroes.freeshell.org> - 2015-10-06 20:04 +0000
      Re: rPi and USB 3.1 w/ type C connector rickman <gnuarm@gmail.com> - 2015-10-06 22:40 -0400
        Re: rPi and USB 3.1 w/ type C connector Theo Markettos <theom+news@chiark.greenend.org.uk> - 2015-10-07 11:11 +0100
          Re: rPi and USB 3.1 w/ type C connector rickman <gnuarm@gmail.com> - 2015-10-07 11:47 -0400
            Re: rPi and USB 3.1 w/ type C connector Theo Markettos <theom+news@chiark.greenend.org.uk> - 2015-10-07 19:06 +0100
              Re: rPi and USB 3.1 w/ type C connector rickman <gnuarm@gmail.com> - 2015-10-07 15:35 -0400
                Re: rPi and USB 3.1 w/ type C connector fruit <fruit@invalid.org.uk> - 2015-10-07 21:42 +0100
                Re: rPi and USB 3.1 w/ type C connector rickman <gnuarm@gmail.com> - 2015-10-07 17:27 -0400
                Re: rPi and USB 3.1 w/ type C connector fruit <fruit@invalid.org.uk> - 2015-10-08 07:36 +0100
                Re: rPi and USB 3.1 w/ type C connector Gordon Henderson <gordon+usenet@drogon.net> - 2015-10-08 09:34 +0000
                Re: rPi and USB 3.1 w/ type C connector Rob Morley <nospam@ntlworld.com> - 2015-10-08 19:26 +0100
                Re: rPi and USB 3.1 w/ type C connector rickman <gnuarm@gmail.com> - 2015-10-08 18:40 -0400
                Re: rPi and USB 3.1 w/ type C connector Anssi Saari <as@sci.fi> - 2015-10-11 20:28 +0300
                Re: rPi and USB 3.1 w/ type C connector "A. Dumas" <alexandre@dumas.fr.invalid> - 2015-10-11 20:48 +0200
    Re: rPi and USB 3.1 w/ type C connector rickman <gnuarm@gmail.com> - 2015-10-06 22:47 -0400
  Re: rPi and USB 3.1 w/ type C connector julianmkizner@gmail.com - 2016-11-02 18:08 -0700
    Re: rPi and USB 3.1 w/ type C connector "Kerr Mudd-John" <admin@127.0.0.1> - 2016-11-05 13:01 +0000

csiph-web