Hardware Projects » M.2 (NGFF) WWAN modem USB breakout board

I think the decision to pull it to 1.8V was taken in reference to some specific modem where the manufacturer stated it to be don that way? Maybe mschramm can comment.

I guess in the worst case we should have a jumper (either physical 1x3 jumper) or some jumper SMT resistors that can be used to switch.

Thanks for the feedback!

When designing this breakout board about four years ago, we reviewed the PC-SIG specs as well as several modem hardware integration manuals; amongst them were Fibocom L860, Huawei ME906s, ME936s, Sierra EM7565, and recently Quectel RM500Q. Despite the PC-SIG remark about pulling it up to V _host (usually 3V3), they all allow the system integrator the usage of either power rail of 1V8 or 3V3 (btw, same is true for BODY_SAR, GPS_DISABLE and RESET (partly for WAKE_ON_WWAN)). Furthermore, it's noted in the integration guides that even leaving this pin open is also sane, as they internally pull it up to the 1V8 domain (and add a diode internally, cathode towards WWAN_DISABLE pin). As the signal is low-active, a voltage above 0,7-1,26V (depending on manufacturer) is considered HIGH, which always will be achieved by a 1V8 rail and 10k PU to it.

So far, no problem with usage of that power rail has been reported. We might revise this detail for an upcoming production run.

Hi!

Thanks for the rapid response.

we reviewed the PC-SIG specs as well as several modem hardware integration manuals

Excellent due diligence!

Despite the PC-SIG remark about pulling it up to V host (usually 3V3), they all allow the system integrator the usage of either power rail of 1V8 or 3V3

I couldn't find this remark in the PCI-SIG spec, do you have a reference? PCI-SIG section 3.1.12.3 calls out W_DISABLE1 must be 3.3v explicitly, with a 100-200k pull-up resistor.

same is true for BODY_SAR, GPS_DISABLE and RESET (partly for WAKE_ON_WWAN)

Looking at Section 3.1, Table 3-11, BODY_SAR (DPR), GPS_DISABLE, RESET and WAKE_ON_WWAN are 1.8V. Additionally, W_DISABLE2 is listed as 1.8V (vs. W_DISABLE1, 3.3V)

As the signal is low-active, a voltage above 0,7-1,26V (depending on manufacturer) is considered HIGH, which always will be achieved by a 1V8 rail and 10k PU to it.

Anecdotally, I have observed that 1.8V for Pin 8 seems to be fine (for example, with an FN980m and an lt4120). Based on the the feedback here, it seems like this behaviour is the same for a lot of other devices. I think in practice this is fine, even if it might not be PCI-SIG standard.

marwalte wrote:

Despite the PC-SIG remark about pulling it up to V host (usually 3V3), they all allow the system integrator the usage of either power rail of 1V8 or 3V3

I couldn't find this remark in the PCI-SIG spec, do you have a reference? PCI-SIG section 3.1.12.3 calls out W_DISABLE1 must be 3.3v explicitly, with a 100-200k pull-up resistor.

My "they" referred to the modem's hardware integration manuals mentioned, not the PCI-SIG spec.

As the signal is low-active, a voltage above 0,7-1,26V (depending on manufacturer) is considered HIGH, which always will be achieved by a 1V8 rail and 10k PU to it.

Anecdotally, I have observed that 1.8V for Pin 8 seems to be fine (for example, with an FN980m and an lt4120). Based on the the feedback here, it seems like this behaviour is the same for a lot of other devices. I think in practice this is fine, even if it might not be PCI-SIG standard.

That's right, same here: despite their deviation from PCI-SIG spec, this rather appears to be the 'normal' case. - Redesignation and RFU pins and not fully following PCI-SIG standards are unfortunately widely spread amongst the manufacturers, e.g. there are even miniPCIe modem cards which source a voltage on the sink pin 1V5, while other name this pin VBAT_3V3_IN ... While the latter can be a mess, I don't see a problem with the 1V8 rail usage mentioned here above.

My "they" referred to the modem's hardware integration manuals mentioned, not the PCI-SIG spec.

Thanks for the clarification, this makes sense.

I don't see a problem with the 1V8 rail usage mentioned here above.

Yup - in practice, this seems to be the case.

Thanks for following up!

mschramm so what is the resulutin? Close this issue as we think 1.8V is fine? Add soem kind of placement option for different resistors to different rails? ...

Leaving the ticket open while not further [intending to do / doing] anything doesn't look all that great...

Again sifted through several modern M.2 FF modem datasheets, among them Fibocom FM150 and Quectel RM510Q: they all integrate an internal PU to the desired rail for W_DISABLE# . This implies that we even could strip the external PU. Those manufacturers who pull up to 1V8 (mostly w/ 100k), have the diode towards W_DISABLE input when integrators follow PC-SIG and use 3V3 as PU rail. This even allows driving those signals with either 1V8 or 3V3. - So let's stay here with 1V8.