mmc: UHS Voltage switch and operating frequency question

Shawn Lin shawn.lin at rock-chips.com
Wed Aug 2 03:07:39 PDT 2017


On 2017/8/2 17:20, Jerome Brunet wrote:
> On Wed, 2017-08-02 at 10:54 +0800, Shawn Lin wrote:
>> On 2017/8/2 0:56, Jerome Brunet wrote:
>>> On Tue, 2017-08-01 at 08:38 +0800, Shawn Lin wrote:
>>>> On 2017/8/1 0:20, Jerome Brunet wrote:
>>>>> Hi Ulf,
>>>>>
>>>>> I am working on adding signal voltage switch callback to the meson mmc
>>>>> driver.
>>>>> While testing, I noticed that a few cards fail to exit the busy state
>>>>> after
>>>>> the
>>>>> voltage switch.
>>>>>
>>>>> After tinkering with the driver a bit, I noticed that increasing the
>>>>> clock
>>>>> frequency from 400kHz to 1MHz solve the problem. Strange, isn't it ?
>>>
>>> Thanks for your reply Shawn !
>>>
>>>> AFAICT, most of the internal circuit of card is running against the
>>>> input clock from host. So if you increase your clock, cards will finish
>>>> thier logic switch faster.
>>>
>>> Yeah that would be my understanding as well
>>>
>>>>    Note that the spec does state that 400KHz is
>>>> the max freq for indentification mode. But the fact is that almost all
>>>> cards could accept higher frequecncy.
>>>
>>> Agreed. but I'm not talking about the frequency used during identification.
>>> The step I'm referring to is after CMD41, during CMD11. After CMD11, the
>>> spec
>>
>> I also notice you said "a few cards fail to exit teh busy state after
>> the voltage switch". That implies that card didn't finish voltage switch
>> at all, so it shouldn't argue for increasing the clk as the spec only
>> say "cmd11 invlokes voltage switching squence. If it is *completed*, the
>> card enters SDR12(default)." and "Completion of voltage switch sequence
>> is checked by high level of DAT[3:0]"
> 
> I understand increasing the frequency does not make much sense, but it makes it
> works. I'm trying to understand why ... hopefully it will lead to a real
> solution.
> 
>>
>> So these two cards are probably still in id mode or some un-documented
>> state, I guess. No hints for that per the spec but you should try to
>> make sure the behaviour of your contrller follows the whole requirement
>> of SD spec "4.2.4.2 Timing to Switch Signal Voltage".... especially
>> to see if your 1.8v is stable within 5ms...
> 
> Unfortunately section "4.2.4.2 Timing to Switch Signal Voltage" is blank in the
> simplified (freely available) spec. I could not get my hands on the full spec
> yet.
> 

okay, the full spec seems confidential for SD association members.

> Unsettled regulator was one of my first guesses. The regulator setup on my board
> needs some time to settle when switch from 3.3v to 1.8v, around 70ms.
> I made sure to delay the return of the host voltage_switch callback until the
> voltage is stable at 1.8v.
> 
> This probably is a bit long but, at this stage, the clock is gated so the card
> should not see the difference, right ?
> 

No. I saw some failure cases for that, so that's why I asked you to
check this.

I can't share the full spec, but I think I could paste some bits.

Per section, 4.2.4.2:

"1.8v output of voltage regulator in card shall be stable within 5ms.
Host keeps SDCLK low at least 5ms. This means that 5ms is the maximum 
for the card and the minimum for the host. "

So that is mandatory for HW behaviour, not software relevant....



> So on return of "mmc_set_signal_voltage(host, MMC_SIGNAL_VOLTAGE_180)", the
> voltage should be stable at 1.8v and we will the get the 10ms (even if spec says
> 5ms apparently) before re-enabling the clock.
> 
> 
> Again, Thanks for helping out Shawn !
> Cheers
> Jerome
> 
>>
>>
>>> says that the frequency is up to "default speed or SDR12" so 25Mhz >
>>> If id mode is limited to 400kHz, it is clearly a different stage.
>>>
>>>>
>>>> For you issue, please try to hack mmc_set_uhs_voltage to increase the
>>>> delay there to see if that could solve your problem.
>>>>
>>>
>>> Ohhh believe me, I did :) however these 2 cards are quite stubborn
>>>
>>>>>
>>>>> I'm don't know MMC that much but is it possible that some card require a
>>>>> minimum
>>>>> operating frequency to enter UHS mode ?
>>>>>
>>>>
>>>> We shoule never give a hypothesis beyond the scope of spec.
>>>
>>> I don't get what you mean ?
>>>
>>>>
>>>>> The simplified spec (Part 1 - Physical Layer) says that before CMD11, we
>>>>> should
>>>>> have had CMD41 (Init Command) . After CMD41, the card should be
>>>>> operating at
>>>>> Default Speed or SDR12.
>>>>>
>>>>> I believe that in both case, the frequency is "up to 25Mhz" ? This would
>>>>> be
>>>>> the
>>>>> maximum, but is there a minimum ? 400kHz seems pretty low compared to
>>>>> 25Mhz
>>>>> ...
>>>>>
>>>>
>>>> 400KHz is good for work unless your IP has limitation for that.
>>>
>>> Ip seems to be fine at 400KHz with every other cards, including uhs ones.
>>> The 2 "annoying" cards switch to UHS mode fine on a laptop.
>>>
>>> It the combination of the ip and the cards which seems broken, whatever the
>>> delays I put here and there.
>>>
>>> For a reason I don't understand, increasing the minimal frequency from
>>> 400Khz to
>>> 1MHz make these cards work very nicely.
>>>
>>>>
>>>>> I hope you will able to shed light on this :)
>>>>>
>>>>> Best Regards
>>>>> jerome
>>>>> --
>>>>> To unsubscribe from this list: send the line "unsubscribe linux-mmc" in
>>>>> the body of a message to majordomo at vger.kernel.org
>>>>> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>>>>>
>>>>>
>>>>>
>>>>
>>>>
>>>
>>>
>>>
>>>
>>
>>
> 
> 
> 
> 




More information about the linux-amlogic mailing list