Power on Reset

Look at the MAX809 series. 809 is low on reset. Several people second-source this.

Most por inputs are low at powerup, high to run. The TUSB part appears to be low for reset.

John,

Thank you for the response. This help greatly.

May I verify that I got it correctly? I believe that if I use an MAX809, giving 5v0 at Vcc, Grounding the GND pin, and sending the RESET pin to the RESET pin of the TUSB2046B chip, also putting a 100k resistor on this RESET between the MAX809 and TUSB2046B to hold reset low due to any leakage that may take place, as with Figure 2, Page 6, at

On the "Selector Guide" Table on page 5 of that same document, do I want a low reset threshold, MAX809Z, or a high threshold, MAX809L?

Thanks again for your help, Ben

And the problem with a resistor and a capacitor is....?

it will work most of the time, but some ICs are picky about slow rising reset

-Lasse

Really? You need to ask this? What there is a power glitch and the power supply voltage dips below proper values, but not to ground? An RC circuit will typically hold up and let the device be corrupted. This is why many customer service folks tell you to unplug a device for a number of seconds, to let crappy reset circuit completely discharge.

A proper reset chip is a good thing to use in any product that has to work reliably.

Rick

Soft edges, brownout hangups, uncertain thresholds.

(sometimes) slew rate.

The TUSB needs a 3.3 volt supply, no?

If you run the TUSB gadget from 3.3 volts, you could use the 809T (3.08 volts reset threshold) or S (2.93) maybe. S might be good.

Got it, thank you. I appreciate your help, I have learned a bit about Power-on Reset circuits today, thank you.

Ben

How about a TPS77833 regulator, it has a POR output.

_/ +-----------+-o/ o+--------+----------+ | | .-. | | | | 5k | | | | + .1u --- '-' --- 3.3V --- +----------+ RESET PIN - + | | | ||-+ | | ||

Only if the chip does not have a hysteresis input. Most do.

You simply need to calculate the raise time to make sure the hysteresis window falls with in the spec's of the reset time.

But, it never hurts to add a little speed I guess and set a time gap externally with a square pulse.

Jamie

Products that don't POR cleanly are a total pain in the ass. However, POR is a common place to skimp.

POR is really a big deal if you have some part of the circuit that can draw significant power without control circuitry functioning. A SMPS is a prime example, where you could be "charging" the inductor while powering up. This alone is one reason why SMPS controller chips are far better to put in a product than those roll your own power supplies, which usually skimp or POR and/or UVL.

It is possible that the power supply can itself can be rising slowly, or even dropping as parts of the system turns on and load the supply. Typically chips under evaluation (pre-production) have to handle any voltage from 0 to abs max without exhibiting unusual behavior.

I've gone as far in designs to include POR, UVL, and a watch dog timer should the product be clocked by one-shots. Pre-production evaluation should include shorting any two pins together, to ground, or to the supply voltage, without causing the chip to go haywire. This is to avoid the situation where the tech or design engineer poking around with a scope probe.

Well, because it only works "most" of the time, which is a bit like having a spouse who is "mostly" faithful.

Das ist absolutely verboten for any application where decent reliability is required, and, sadly, a typical newbie mistake.

The only application where I'd consider it is a battery-powered type where the user will see it not working and will cycle power if it doesn't start up right, and a bad startup won't kill something like non-volatile memory.

I'm retrofitting this onto an $8K-per-unit product that will occasionally start up with an output enabled that does some very, very bad things. Typical sort of setup, a microcontroller controlling stuff via a bunch of HC latches- the latch states are more-or-less random after power has been off long enough, and the micro is counted on to reset them during initialization- so on a less-than-clean startup (eg. blip during power-up) you can get the outputs to stay on continuously because they left out a 50-cent reset chip (one occurence will pay for a truckload of chips). Shockingly bad design. Oh, and unplug the PLCC micro (or let it fall out of the socket) and sometimes you'll get outputs stuck on. I would have put both a proper hardware reset circuit to both the micro and latches, AND an external watchdog timer in place (both of which are in something like an ADM805). And more, but that gets into system level stuff.

The OP should pick a chip that is guaranteed to assert "reset" for any conditions under which bad things could happen, and is guaranteed not to come out of reset until the chip can operate to spec (usually it's a threshold with hysteresis and a time delay of some milliseconds for clocks to spin up, PLLs to stabilize and that sort of thing). On-chip reset circuits, when present, are often inadequate for the task. The second criteria is usually well specified, but it's often up to the engineer to define the first criteria-- for example at what worst-case _minimum_ voltage could a non-volatile memory be corrupted or at what worst-case minimum voltage (over temperature and unit-to-unit variations and vibration etc.) could a relay pull in to turn on a heater. Semi manufacturers will usually not tell you the guaranteed to "not" operate conditions (relay manufacturers usually do).

Other than the xx805, I've often used Microchip's MCP10x series, which is 3-pin supervisor- a precision voltage detector and time delay of a few hundred msec. Simple, cheap, and works fine in many cases. There are better choices if you need really low power consumption or other features (eg. Seiko for low power). The MAX809 (eg. On-semi) that JL pointed to is a really popular choice that's pretty good all-round,

Digikey lists about 35,000 different part numbers of supervisors so obviously there's a recognized need! ;-)

--sp

Best regards, Spehro Pefhany

On a sunny day (Thu, 25 Oct 2012 13:33:18 -0700) it happened "Benjamin David Lunt" wrote in :

The cheap way would be an electrolytc to +, with a resistor to ground, maybe add diode too against reverse input.

Argh, terrible. No hysteresis, no time delay, and assert / release reset levels are not controlled. Reset is never asserted if the supply comes up slowly, a short blip down that disrupts the digital stuff may not generate a reset pulse.

In fact, if the reset input is a ST, this might actually be worse than an RC directly on the reset pin.

About a dime in quantity. Check it out, if it works, spec it, sleep well at night.

Best regards, Spehro Pefhany

The switching levels of Schmitt trigger inputs are never specified in such a way as to be useful. Not only are they uncontrolled (in terms of guarantees of chip operation or benign non-operation) but they change with power supply voltage. With all that unspecified flipping back and forth they should be called Schmitt Romney triggers. (ducking)

A bulletproof POR circuit does need an absolute voltage reference in it.

If you can guarantee that the power supply will always be either zero, within spec or transitioning in a specified way between the two that's true. For every other (almost every real) situation, you need a proper POR circuit. Digital circuits can get into untold mischief in brownout situations, for example.

Time + tightly specified absolute voltage levels for on/off.

No. A schmitt threshold tracks Vcc, so a brownout dip could mess up the chip's logic but not cause a reset.

A good por chip has absolute voltage thresholds.