Question: Power consumption and clock sources for Low Power mode In G4 Family (HSI, HSE ,HSE Bypass)

[BarryWhit]

Hi,

Since there is no obvious to ping an ST rep directly from an ongoing thread, I'm posting these questions in a new thread in the hope that ST will respond with answers. 

These arose during discussion in another thread:

How Can I Provide a 1mhz HSE Bypass Clock For Low-Power - Without Wasting More Power Than I Save?? 

Questions:

1. In "Low-Power RUN" mode, the current consumption at 1Mhz given by datasheet (refA)
is 255uA for fHSE, and 820 uA for fHSI. The power consumption with HSI is noticeably higher.
What is the reason for this difference in power consumption?

2. Does "Low-Power RUN" mode support the use of an HSE *crystal*, or only HSE Bypass mode (i.e. an external clock).

(The following questions assume an HSE crystal is supported in "Low-Power RUN" mode):

3. Do the datasheet tables for "Low-Power RUN" (see refA), which detail the current consumption
for "HSE" assume the use of HSE bypass (i.e. external clock), or are they valid when an
external crystal is used?

4. If use of an HSE crystal is possible in "Low-Power RUN" mode, what is the current
consumption in that configuration?

(refAa) DS12589 Rev 6, "STM32G431 Datasheet", Table 21 "Current Consumption In Low-Power run modes (continued)", pp 77/196.

[Simon.T]

Hello @BarryWhit ,

First of all, if you want to have a dedicated support with ST employee you can create an OLS ticket through this link. 

When I'm giving a table number, I took the DS12589 Rev 6 as reference.

 1. What is the reason for this difference in power consumption?

The consumption is higher due to the HSI consumption itself (see Table 42) and also all the clock divider linked to this clock. 

 2. Does "Low-Power RUN" mode support the use of an HSE *crystal*, or only HSE Bypass mode (i.e. an external clock).

Sure the Low-Power RUN support both HSE configuration.

3. Do the datasheet tables for "Low-Power RUN" (see refA), which detail the current consumption
for "HSE" assume the use of HSE bypass (i.e. external clock), or are they valid when an
external crystal is used?

All the datasheet power consumption data using the HSE as clock source are in bypass mode. 

 4. If use of an HSE crystal is possible in "Low-Power RUN" mode, what is the current
consumption in that configuration?

It mostly depends of your crystal oscillator parameters, in the Table 40 of the datasheet you can have an idea about the crystal power consumption that you need to add.

Best regards,

Simon

[Andrew Neil]

---

@BarryWhit wrote:

Since there is no obvious to ping an ST rep directly

---

You can '@' mention them?

This is the public forum; if you want private support, use the support link from the ST homepage:

Note that you can give feedback to ST about the forum & its operation:

[waclawek.jan]

Hi @Simon.T ,

> The consumption is higher due to the HSI consumption itself (see Table 42) and also all the clock divider linked to this clock.

The HSI consumption according to Tab.42 is 155uA typ 190uA max, whereas the difference we are talking about is almost 500uA.

Do you say, that using HSI always implies consumption increase of cca 500uA (due to "all the clock divider linked to this clock")? What's the point of giving 155uA as typ. HSI consumption, then? I'd call this misleading.

JW

[BarryWhit]

Thank you Simon, that's helpful. You did not include the support link but Andrew Neil supplied it. I didn't realize that option was available - I thought you just had to jump up and down in the forum until someone noticed...

Follow-up questions:

1. [retracted] Why is there no information on current consumption with an HSE crystal? (Answer Given Above: There is, but it is separate from the run mode current consumption tables. See Table 40 "HSE oscillator characteristics" in the DS)

2. Does the power estimator in CubeMX rely solely on datasheet values or does it contain a more extensive table/model? For example, it doesn't force you to use HSE bypass for HCLK<48Mhz, but the datasheet doesn't include data for that. Yet, an LP-RUN step in the estimator *is* limited to HSE Bypass and HSI modes (Just like the datasheet tables).

3. Based on your answer, should I assume there is a substantial power penalty for using a clock divider even with HSE Bypass (for example, providing a 2 Mhz external clock and dividing by 4)? Does it even matter if I use them or not (i.e. is /1 less wasteful then 1/n)?

[BarryWhit]

ping @Simon.T , I'm still hoping for answers to my follow-up questions. Thank you.

[Simon.T · Accepted solution]

Hello @BarryWhit ,

First of all, sorry for the delay.

2. Usually the data for datasheet and the Power Consumption Calculator (PCC) are coming from the same source. For sure the PCC has more ore less extended value, for example run consumption at different voltages, and this tool need to be taken as an help. 

3. The penalty should be negligible for HSE. When I talked about the divider consumption it was an example, for HSI it seems that there are an over consumptions due to the low power mode, this is why the data are given with HSI and HSE bypass. @waclawek.jan  the value given in the Table 42 is only the HSI consumption (when you are switching on or off the clock), it can be useful when you want to use the HSI as kernel clock in stop mode.

Best regards,

Simon

[BarryWhit]

Thank you Simon.