2020-08-31 08:53 AM
My design uses an STM32H750 with a 24MHz crystal on the HSE, and on my first attempt, the crystal is not oscillating. I’m going through AN2867 and having difficulty finding any 24MHz crystal that meets the gain margin requirement for this processor.
AN2867 says the the gain margin is calculated as Gm / Gmcrit, and provides a formula for Gmcrit based on the crystal specs. I assume Gm is a property of the microprocessor. The STM32H750 data sheet, Table 42, page 118, says Gm(critmax) is 1.5 mA/V Max. If this is the number for Gm, then based on 24MHz crystals I can find, it’s not possible to get a gain margin of 5 or more. The best I’ve been able to find is a crystal that yields a gain margin of 2.6.
Is 1.5mA/V indeed the correct value to use for Gm in the gain margin calculation?
2020-08-31 01:36 PM
Post details of you current attempt - PCB drawing, type of crystal and capacitors, any software. How do you know the HSE is not oscillating? You should also try first on a "known good" board such as Nucleo.
JW
2020-08-31 02:05 PM
Thanks for responding. I've been developing my code on a known good board with a 16MHz crystal. I know my new board's HSE is not oscillating because there's no activity on either the RCC_OSC_IN / RCC_OSC_OUT pins or MC01, which I've activated and set to HSE/1.
My question is not so much about my design as it is regarding the values to be used in the gain margin calculation for the STM32H750. Even if my circuit works, I'd like to think I was following the design criteria and not operating on the edge. The ST design guide (AN2867) says that the gain margin should be greater than 5, yet I can't seem to find a 24 MHz crystal that allows me to meet that. Closest I've found is one with an ESR of 10 Ohms, a load capacitance of 18 pF and a shunt capacitance of 7 pF, but using the Gm value of 1.5mA/V still only produces a gain margin of around 2.6.
Again, my specific question here is whether or not 1.5mA/V is the correct value to use for Gm in the calculation. Thanks.
2020-08-31 03:45 PM
> Again, my specific question here is whether or not 1.5mA/V is the correct value to use for Gm in the calculation.
It's not.
> but using the Gm value of 1.5mA/V
The value from the datasheet is g_m_crit_max, not g_m:
AN2867 says how to calculate g_m_crit for your crystal:
For the one you mentioned, I calculate this as 0.142 mA/V.
AN2867 also states the requirement is based on how the crystal is specified. (But note that the two requirements are equivalent, as g_m_crit_max = g_m / 5)
Since g_m_crit = 0.142 mA/V, this is way less than g_m_crit_max = 1.5 mA/V, so you're good.
At least that's how I read it.
2020-08-31 03:54 PM
Ah. I missed that distinction in the app note. Makes perfect sense. This would also explain why my current crystal is not working, since it calculates to a g_m_crit = 1.98, which is above the specified g_m_crit_max. Thanks so much!
2020-08-31 03:56 PM
Sorry, can't help you with the Gm calculation - but will make some general suggestions (that you would already know, but might help others who view this) - use surface mount components, position crystal close as possible to micro (keep tracks short as possible), avoid vias, avoid sharp bends in tracks.