Sorry, you need to enable JavaScript to visit this website.

Microzed Power Supply Enable Issue

Zedboard forums is currently read-only while it under goes maintenance.

Unsolved
5 posts / 0 new
mcbridejc
Junior(0)
Microzed Power Supply Enable Issue

Hello,

I wanted to report an issue that I've encountered with the Microzed board, in hopes that it can be improved in future revisions or that it can save others some time. I believe this is a bug in the hardware design, but please let me know if I've misunderstood something. I'm using Rev F of the board.

The 1.8V switcher is enabled by the power good (PG) output of the 1.0V switcher. The PG pin is open-collector, so it must be pulled up. This is done by R42, which pulls up to 1.0V, with R88 as a pull down making a voltage divider. The nominal value of PG_1V_10 when high is 1V*5k/(1k+5k) = 830mV.

This net drives the EN pin of U15. The datasheet for TLV62130RGT (http://www.ti.com/lit/ds/symlink/tlv62130.pdf) states that the minimum "high" input voltage for this pin is 0.9V, so the microzed is not meeting it. It works most of the time, but while debugging some intermittent problems with a USB device, we found that the 1.8V switcher was occasionally kicking off and collapsing for a short period of time (but long enough to cause problems communicating with the USB phy). We definitely had one board that had more problems than others, so there is also some board-to-board variance.

We've worked around this problem by blue-wiring the pullup to our 5V input voltage. I would suggest that microzed designers do the same in future revisions, perhaps with a zener to prevent the net from rising above the 7V max allowable voltage in the case where a higher input voltage is used to power the board.

tggzzz
Junior(0)
Caution: this comment is

Caution: this comment is worth what you paid for it! (I've only looked at this topic for 5 minutes)

Your analysis of 830mV vs 900mV looks sound. However, I wonder if it would be simpler to merely remove R88 completely...

I don't understand why R88 is present at all. The TI datasheet's examples only show the equivalent of R42 with two different values: 100k, plus unspecified "R3" in fig 50.

I don't see the pulldown effect of R88 having any benefit. Alternatively, increasing R88 to 10k would enable Vpg_1V_10 to be >900mV

Since the MicroZed has R43 = 1k, it looks like Avnet are using PG to discharge the output capacitors, as in fig 50. In this case the peak current would be ~1mA, which is well within the 10mA noted in the paragraph describing fig 50. (And less than the Iol=2mA required to ensure Vol_pg < 0.3V)

Removing R88 will continue to allow the discharge operation and will not, as far as I can see, affect the power enable operation.

I suspect the suggestion of a Zener is unnecessary: the enable pin will tolerate an input up to Vin, and that can be +17V.

Comments are welecome.

mcbridejc
Junior(0)
Changing R88 is better solution

Regarding the zener: it is actually the PG pin that cannot tolerate greater than 7V.

Regarding R88: You make a very good point! I suspect R88 is there because the PG pin is High-Z when the regulator is disabled (see pg 16 of datasheet), so it won't drain the 1V rail, and potentially the 1.8V could remain on when the 1V was shut down. And in fact, I need to go back and look at my fix, because I suspect that the other power rails are now running even when the PWR_ENABLE to the microzed is not asserted. So I think the solution I originally proposed is a bad one.

Changing R88 values to be larger so that the voltage divider value is always >0.9 (even when the 1V rail is a bit below 1V and for the worst case resistor tolerances) is probably a better solution.

tggzzz
Junior(0)
Datasheet speaks with forked tongue?

No, it looks like TI (or is this really an LT device?) have set out to be deliberately unnecessarily misleading - I assumed that the A suffix was a speed bump or other bug-fix revision, not a fundamentally different device that "ought" to have a different number.

P2, footnotes(!) "(3) ... While TLV62130 has PG=High Z, TLV62130A features PG=Low, when device is in shutdown through EN, UVLO or Thermal Shutdown."

The Zener: if PG=17V then isn't the 1V0 output voltage 17V? If that's the case we would have other problems!

Alternative to increasing R88 would be to halve R42, so the current would still be <2mA, and the discharge time reduced a bit as well. I prefer removing/increasing R88; practically removing it appears to be a simpler bodge.

TroutChaser
Moderator(18)
Thanks for the observation

Jeff,
 
Thanks for the observation, we are always working to make our products better. Our engineers followed up on your input and we will be changing the MicroZed Bill of Materials to reduce the value of R42 to 240 ohms which, as you pointed out, will increase the nominal value of PG_1V_10.
 
Each MicroZed has been functionally tested prior to shipment and we have not observed any failures, even in the units we tested for the extended temperature range, so we don't anticipate anyone having issue with the current boards.
 
If anyone does experience problems please contact me directly.
 
Regards,
 
-Gary