1V regulator

Any technical questions about the Epiphany chip and Parallella HW Platform.

Moderator: aolofsson

1V regulator

Postby kcarnold » Wed Nov 06, 2013 4:53 am

I almost pre-ordered, but wavered on it and missed the cut-off :( But in the mean time, I noticed that the 1V power regulator for the Zynq is rated even lower than the Zedboard, and people have already encountered problems with that (see "1V Power supply under-powered" on their forums). We were planning on actually using the Zynq for some real processing, so if you're tweaking the board design, it would be really helpful if you could sneak in a regulator boost too.

(For reference, the Altera SoCKit puts an 8A regulator on their FPGA fabric. Granted the FPGA is not the star of this show, but it would still be nice to be able to use all it's got.)
Posts: 1
Joined: Wed Nov 06, 2013 4:49 am

Re: 1V regulator

Postby msegado » Wed Feb 05, 2014 12:02 am

Any plans to change this in a future revision?

(FYI, here's the thread that kcarnold mentioned about the Zedboard's bigger supply being underpowered, plus a more recent one:)
http://www.zedboard.org/content/1v-powe ... er-powered
Posts: 3
Joined: Tue Feb 04, 2014 11:41 pm

Re: 1V regulator

Postby tnt » Wed Feb 05, 2014 9:18 am

From a previous thread this seems unlikely to happen just because there are no PMIC as small as the ones used now that could provide that much current.

However the board now has a proper connection point so you can feed 1V externally if needed.
Posts: 408
Joined: Mon Dec 17, 2012 3:21 am

Re: 1V regulator

Postby aolofsson » Wed Feb 05, 2014 1:34 pm

As tnt noted, given all the constraints, and the fact that only a small number of people needed more than 1.5A on the 1P0V rail, the on-board regulator is not going to be made any beefier.

We did make a change to the gen 1.1/1.1 board due to feedback from folks with intent to really beat up on this board. The through-hole in question is TP13 and the picture below shows that this power might just be good enough to feed more current to the Zynq. Since this is very much an experimental feature, we need help testing it. Since the board runs quite hot with 5 Watt, cooling is going to be a an even bigger concern as you use even more of the Zynq device.

NOTE: You MUST ground the REG_EN1 through the PEC_POWER connector input if you are going to drive power directly to TP13 from an external supply. If you don't, you WILL break the board.

1P0V_rail.png (99.19 KiB) Viewed 11927 times
User avatar
Posts: 1005
Joined: Tue Dec 11, 2012 6:59 pm
Location: Lexington, Massachusetts,USA

Re: 1V regulator

Postby mark03 » Wed Feb 05, 2014 3:58 pm

I read through the ZedBoard community discussion on the spontaneous resets. While I am still concerned, I think it should be noted that we still don't have a smoking gun. Despite repeated pleas by the board designers, no one was willing (or paying enough attention?) to step forward with (1) their Verilog/VHDL design, and (2) the corresponding XPE (Xilinx Power Estimator) calculation. Those things would be the most basic, fundamental prerequisites to take this issue seriously and open a real investigation, yet they still don't have them ~6 months after the claim was initially made.

Granted, it's hard to imagine what else could explain the symptoms they were seeing, but it's possible that their PL design was really pathological in some way or another. I'm hopeful that this won't be a huge issue on my Z7020 Parallella... fingers crossed! And if it is, at least we can provide the data to Xilinx that the Zedboard community couldn't.
Posts: 21
Joined: Wed Aug 14, 2013 12:00 am

Re: 1V regulator

Postby msegado » Sat Feb 08, 2014 2:33 am

Wonderful, thanks all! Yes, keeping the PMIC as-is but making the 1V0 rail accessible seems like a good design move - thanks for including this =)
Posts: 3
Joined: Tue Feb 04, 2014 11:41 pm

Re: 1V regulator

Postby solardiz » Tue Jul 15, 2014 1:01 am

aolofsson wrote:NOTE: You MUST ground the REG_EN1 through the PEC_POWER connector input if you are going to drive power directly to TP13 from an external supply. If you don't, you WILL break the board.

What makes you so sure that not disabling this output of ISL9307 literally "WILL" rather than "might" break the board? Or are you just making sure everyone understands that breakage is possible? Why do you think it would occur in practice? If things are really that bad, then grounding that pin might not be safe enough as any temporary disconnect of this pin from GND would then be fatal. I am looking at page 16 in http://parallella.org/docs/parallella_schematic.pdf (rev 1, Dec 30, 2013) and page 6 in http://www.intersil.com/content/dam/Int ... fn7931.pdf (Sep 20, 2012). You have a 10k pull-up resistor to 5V, and the ISL datasheet says logic low on that input must be at most 0.4 V. I guess this is OK, and I guess burning that chip with an external 1V supply on the output isn't actually that easy, but in case it is perhaps it'd be safer to have a jumper to disable the pull-up (and to replace it with connection to GND when the jumper is moved, as the ISL datasheet also says these enable inputs must not be left floating). (A curiosity: per the ISL datasheet, disabling the output connects it to a 115 Ohm or so resistor within the IC, so at 1V it'll produce an extra 9 mW of heat then. This is obviously benign, especially given that it replaces a portion of its normal heat production.)

A concern is that the external supply might not always be on, so it might actually draw excessive current from ISL9307 if this output is not disabled. I hope ISL9307's overcurrent protection will kick in, but indeed it's better not to rely on that.

Anyway, I've been thinking about a really dirty hack, initially for ZedBoard (which uses a MAX15021 there): what about connecting a carefully chosen and tested Schottky diode to draw some current from the 1.5V rail to the 1.0V rail? For example, the diode could be chosen such that its Vf is 450 mV at low current (say, 10 mA) and 550 mV at 1A (I think I readily have some diodes that would meet these criteria), so the target voltage would be between 0.95 V and 1.05 V. I guess a similar hack could be used with Parallella too, although we'd have to go from 1.8V or take some current from the other IC (ISL9305). In this context, my question is: would this sort of hacks break the boards for some specific reasons (that I'd be curious to hear) or are they potentially viable? I imagine that for Parallella there might not be a lot of capacity left in those ICs to deliver more power overall, and that might be a problem (depending on how much more current Zynq would want to draw). Cooling might need to be improved. For ZedBoard, it appears that the 1.5V rail can deliver a lot more current than is currently consumed (4A vs. under 1.5A?) Are there any other risks, and which ones? Note that the diode serves two purposes at once: brings the voltage down to what's desired and prevents current going in the other direction if the normally higher voltage source rail is slower to get up to its full voltage.
Posts: 28
Joined: Wed Apr 10, 2013 12:36 am

Re: 1V regulator

Postby solardiz » Tue Jul 15, 2014 3:28 am

solardiz wrote:Are there any other risks, and which ones?

Here's one: for some diodes, forward voltage drop decreases with higher temperature (although for many others it increases). Mitigation: need to check the manufacturer's datasheet for this (whether to expect an increase or decrease, but mostly not the actual values, which may vary widely), and need to test the diode at ~1A not only at room temperature, but also after the current has been passing through the diode for a long while, and need to consider the results (and choose another diode if necessary). Of course, it's sort of naive to use a diode in place of a voltage regulator, but hey it's a hack.

Here's another: from the diagrams in http://datasheets.maximintegrated.com/e ... X15021.pdf on pages 8 and 9, at least that IC would actively connect the output to GND during its PWM's low cycles, so even slightly higher than regulated voltage on the output, as it might be coming via the diode from the other output, would result in some current going back through the IC and to ground. This might damage the IC, the inductor, or/and the added diode. Mitigation: connect a 0.1 Ohm or so resistor in series with the diode, and maybe target a lower voltage range (e.g. 0.9V to 1.0V by using a slightly higher Vf diode) for when this extra hack circuit would kick in (then it'd only help reduce the voltage drop slightly, without eliminating it fully).

OK, not exactly an attractive idea overall. %-)
Posts: 28
Joined: Wed Apr 10, 2013 12:36 am

Return to Epiphany and Parallella Q & A

Who is online

Users browsing this forum: No registered users and 12 guests