EEVblog

[@flymypg]

After leaving the military I worked my way through uni as a technician building precision instrument prototypes. To help bring up new boards, I had a bodge-board filled with AND gates, caps, pots, SCRs and a resistive load so I could get any power-on sequence I needed by quickly twisting a few pots. The main reason the bodge-board was created was for mixed-signal circuits, getting the analog rails stable before bringing up the digital side. Also had some mixed ECL-TTL digital circuits which had their own peculiarities. But the most smoke I ever released was powering up early jelly-bean CMOS logic (4000-series?). As power triacs dropped in cost I was able to add AC supply control, eliminating AC relay contact bounce (no expensive mercury-whetted power relays on the tech bench). As technology advanced the bodge-board began to be used less for bringing up individual boards (many had on-board supplies and sequencers by then), and more for developing and testing new power supplies, switching loads instead of sources. That's when I had to start adding heatsinks and replacing the SCRs with MOSFETs (to better support dual-use). Then I added digital pots and a microcontroller that talked to a PC over a TTL serial port, so that setups could be easily saved and restored. Then an ADC and analog switch to monitor the rails.Then a spare 2-line LCD display. Then current monitoring and power trips to protect the attached board and supplies. Basically, whenever I learned something new in class, I tried to apply it to my bodge-board. As graduation day approached and I was due to leap from the Test group to the Software Engineering team, one of my last tasks as a technician was to document my bodge-board, put it in a pretty die-cast case, then with great ceremony hand it over to the tech who would be taking over my bench. After that, software and firmware claimed all my time at work, and my hands-on hardware work was mainly hobby stuff at home. A decade later my experience with power sequencing became vital when developing a tiny avionics system for a micro-satellite. While it was intended to operate primarily in LEO (Low Earth Orbit), we wanted it to also work in Molnyia and GTO (Geosynchronous Transfer Orbit), which meant surviving multiple transits through the high radiation levels of the Van Allen belts. "Real" satellites with such needs shut down their main electronics while traversing high radiation areas, and rely on a smaller rad-hard processor to do the minimum work needed during transit. We had to use our COTS (Commercial Off-The-Shelf) micro for this. Non-rad-hard electronics have many ways to freak out in high radiation fields. Fortunately, most power electronics are relatively rad-hard as a side-effect of their manufacturing process. Our chosen way to cope with radiation effects was simple in the extreme: Cycle power. We monitored the current to our system using minimal inherently rad-hard COTS circuitry, and when it spiked we'd turn off power long enough to let the conduction channels punched into the silicon "heal", then restore power. We also included a very crude WDT (WatchDog Timer) that was also made using inherently rad-hard COTS components. The failures encountered while evolving my bodge-board provided some key insights for designing both the trip and power cycling circuits, particularly for getting stable off and sequence times without clocked circuitry. (And I also had to learn about what kinds of capacitors were inherently rad-hard.) Under the most extreme conditions, the micro could need to be power-cycled several times per second. This posed an extreme software challenge: To get all necessary work done despite frequent unannounced power cycles. Key parts of the code had to be moved to assembler and hand-optimized. The final result was that the micro could perform one full processing loop within 80ms of power being applied. Subsequent testing yielded normal operation despite power being cycled at a 10 Hz rate. This was possible only by careful joint optimization of the power cycling circuitry and the avionics firmware. This is why I believe all embedded ("bare metal") software engineers need to know a fair bit about analog power circuits (not to mention digital and analog circuits in general). It's also where the most fun is, writing software that directly interacts with the outside world, and is affected by it at all levels.

[@EEVblog]

Great story, thanks for sharing.

[@rationalmartian]

Indeed, splendid story. Thanks muchly.

[@flymypg]

Wish I had a photo of it to share. It started as a 4" square piece of perf-board with terminal strips on two sides, onto which I stacked other 4" squares as I added functionality. It could sequence 4 DC supply inputs to 4 outputs, plus the later single AC port. When I put it in the case, the terminal strips were replaced with banana jacks, and the stack was spread out over the surface, with hot parts mounted to the case and one big heatsink for the whole case. Rather than make a window in the case for the LCD, I just drilled a hole for the wires and mounted it on stand-offs. Gotta keep some ugly when making pretty.

[@spagamoto]

That's really neat. Did you get it launched? How did you deal with bit-flips?

[@EEVblog]

I've got practically no photos of anything I've ever worked on at work pre-mid late 2000's.

[@andyknowles666]

I find this channel so fascinating, every video just shows me exactly how little i know about the low level concepts of electronics, I'm glad that someone is happy to demonstrate all this and talk about it, keep the videos coming! :)

[@EEVblog]

Thanks.

[@gandsnut]

Our host demonstrates that wonderful ability to make blindingly sophisticated topics accessible to us electronics hobbyists whose brains hold no more than 25,000 neurons. Sadly, now I understand my brilliance was only a kindness extended by my Mom, when she called me 'son'. Brilliance... 'son'.... hello, is this mic on?

[@pekkagronfors7304]

Never heard of power supply sequencing. Learning all the time... Thanks.

[@nazarlive2137]

I am one of the most keen fans of the channel and very, and learned from your experience a lot I am thankful for your ban

[@yendak]

Did you ever do a video on what to consider when picking a Mosfet for a specific task? Or is there a website out there with good guidelines on which datasheet values are really important?

[@garthhowe297]

This was a great video. I had never heard of the concept of power supply sequencing before... very interesting. Thanks very much!

[@rikvdmark]

Great video. Definitely keeping this in mind :) Thanks Dave!

[@keithnoneya]

Very interesting Dave. Love the simple solution to. I guess if you actually had a sequence requirement you could build a simple timing circuit to power them on in the right order and time. Really cool video. Thanks for the education. Best Wishes n Blessings. Keith

[@EEVblog]

It can be done with simple RC timing if needed.

[@slademcthornbody9060]

Is there a free trial of Davecad available?

[@EEVblog]

Just pirate it

[@VKuzmenkov]

You can pick some limited version on a local stationery supply store

[@dezzier]

Lmao

[@ElmerFuddGun]

Don't fall for the DaveCAD hype. It is totally overrated and only has the bare minimum of features. No ruler function, no colours, no undo or even erase function. It is really a sad CAD. I'm sure you could design your own version with better features in a matter of seconds. Just try, I know you can do it! (Sorry Dave, just being honest!)

[@phalcon23]

Cranium Assisted Drafting?

[@edhalferty]

FPGA documentation usually specifies very strict sequencing (and ramp-up times!), and it's hard as a hobbyist to know what is absolutely essential and what's just recommended.

[@EEVblog]

A lot of the modern ones don't care about sequencing these days which is good, but it can still have impacts on inrush current at power-on. But ramping usually still matters. Ignore at your own peril!

[@MattHollands]

Lattice don't even bother conforming to their power sequencing in their iCE40Ultra dev board that I have been using: "The power supply sequencing does not conform to the NVCM boot requirements as specified in DS1048, iCE40 Ultra Family Data Sheet. The user may encounter intermittent boot success and/or higher than specified startup currents when attempting to boot from NVCM."

[@n2n8sda]

Great video, between this and Mr Carlsons power supply video I think i'm set to tackle a repair!

[@willynebula6193]

yes get into how to select the right mosfet dave

[@raviteza8]

We had similar problems where we were blowing up Mosfets (at 800VDC) driven by optocouplers. If the 24 V DC (which then was converted to +20/-5 for Mosfet, 3.3V for optocouplers etc) was switched off before the 800 V was off.. puff magic smoke appeared :D... We quickly realized that this was due to the fact that the optocouplers were active high and were held low by another inverter gate on the secondary side to achieve a active low gate signal for the Mosfets. So when the Optocoupler's primary 3.3 V switched off, the secondary 3.3 V was still on and was essentially shorting all the mosfets ( since the 800 VDC was still on).. Here is where one needs power sequencing.. But still in an event when the power supply fails, we still couldn't guarantee that the mosfets don't blow up. So we got rid of the optocouplers( was an old design, dont know why one would use optocouplers these days). Other way to use optocouplers in such a case would be to buffer the circuits with enough capacitance and there by acheive a sort of power sequencing. :)

[@rinaldo732]

5 people didn't turn on in the correct sequence today

[@mikewillis1592]

Couldn't you have used the negative supply to switch on the MOSFETs for the other three? It might need a few more bits. One of the reasons sequencing is so important is when the negative voltage is needed as gate bias to switch off a depletion mode MOSFET. Quite a few microwave RF transistors require negative bias on the gate to be present before the drain supply or they will self destruct. A simple solution is to use a P-MOSFET in series with the drain supply, switching it on with the negative rail. Well designed equipment like that scope will have this protection built in because as you have found, PSUs can fail. My experience is that most equipment failures are related to the PSU.

[@244jts]

Mike Willis Let's not forget, that a power off sequence might also be an issue. Negative supply will be at treshold voltage, when others will go out.

[@mikewillis1592]

That's a good point but needs quite a bit more design effort to ensure foolproof operation - this is what the specialist chips do. The simple solution usually works in my RF transistor example because having negative gate bias present with no drain supply is not usually a problem.

[@punker4Real]

13:06 that is because of the floppy drive. those motors draw some power up to 2 amps that explains the voltage sag

[@rubusroo68]

i never even knew this was a "thing" good video thank you

[@supernumex]

do you mean "ground it to turn it off" instead of turn it on? 17:00?

[@sjwsbetaskiller6218]

Yes, he goofed it.

[@Motorman2112]

It's probably a low-side switch, not a "goof".

[@lmiddleman]

While it's true that overshoot would be bad, the non-monotonic rise of the 15.7V is also concerning. Many multi-rail circuits with sequencing requirements also specify rise time ranges, and that the rails must rise monotonically.

[@CarlThom]

We do a lot of RF amplifiers at work and they are sensitive to power supply sequence. Negative gate voltage normally before positive drain voltage. I've lost one or two (quite expensive) RF amps by messing up the sequence.

[@Daveyk021]

I wanted to see you fix the power supply.

[@tlangdon12]

So a video on selecting a suitable Charge Pump so you can generate the voltage needed by the N Channel MOSFET gate would be great...

[@Jilocasindragon]

Hey Dave. When I first read power supply sequencing; I thought you'd talk about chaining power supplies, especially in cases you would want higher current. Is there any chance you could do a video about that topic and what sort of lab PSUs you need?

[@tom5566772008]

Dave are you planning to do an review on the R&S RTB 2004?

[@RobertKohut]

Awesome!

[@alexwang007]

awesome! just what i needed. i am designing a headphone amp, and it consists of digital and analog chips at then same time, Which made me wonder about the sequence i should be supplying the power..(dvcc:3.3v,1.8v avcc:+-5v)

[@BrackenDawson]

Why not use the 5V rail to trigger the three FETs and turn it on last? Is it the negative voltage being cumbersome?

[@neuralnetwork653]

Dave, you have 3 channel on your DP832, and you can switch them on simultaneously. Why do you need an external circuitry?

[@maximumiq158]

Use a simple transistor bc547 to replace the switch and pull base high from 15.7v output. So when it's base is pulled high by a manual switch to 15.7v input the 15.7v output confirms that, and overrides the manual switch to prevent bouncing. of course there should be some extra resistors to the base. Otherwise a good reminder on the topic. Also be sure the mosfet gate source reverse voltage is within psu limits. 3.3v vs 15.7v is not much in this case, but if you also had a 24v this could burn some mosfets.

[@davestech6357]

You also need to be wary of in rush current. Some power supplies have ntc thermisitors to limit the in rush current or purposely long or small pc traces for added resistance.

[@EEVblog]

Sequencing FPGA supplies is often done do that in-rush current isn't too large, and not an actual issue with the chip needing it.

[@robyoung9463]

could u use "solid state relays" instead ??

[@God-CDXX]

can I use a 555 timer to drive my mosfets or is the chip to slow

[@learnelectronics]

Jamie Phillips I used a 555 clocked at 500k to drive a MOSFET in my diy boost converter. worked great

[@EEVblog]

No problem.

[@DAVIDGREGORYKERR]

can you use four latching relays all driven of a VN66AF transistor mounted on a heatsink.

[@no-defun-allowed]

Five more videos to go, Dave.

[@EEVblog]

TOO. MUCH. PRESSURE.

[@no-defun-allowed]

EEVblog *applies current*

[@TheRealXandes]

i never would've thought about sequencing....

[@FerretAD]

I wouldn't either. I thought you'd just turn power on, and everything would be hunky-dory! Interesting video Dave.

[@EEVblog]

FerretAD The rabbit hole goes deep...

[@PlasmaHH]

"Thou shalt check voltage" ... sequencing... if you are doing repairs, it should definitely be on your checklist, although I had it only once in my lifetime being the cause of an issue, its a thing you almost never figure out when you don't directly look into it. The device acts up, but everything you measure appears to be fine...

[@Professzore]

Sorry for the lame question. At this point I just guess that -5 V on Rigol PSU made by simply switching + and - side of the outputs, right? And since channel 1 and 2 are separated, channel 1 gnd and channal 2 +5 volts are connected without any problem. Am I right?

[@PERILEX]

Dave, are you going to review the R&S RTB2000 any time soon? :D

[@Sevalecan]

3.3V seems like a low Vsg. The datasheet I picked up shows these typically cutting out before that, am I missing something?

[@fuzzy1dk]

you are right, it worked this time but 3.3V is borderline and outside spec

[@i8BBQ4Lunch]

Dumb question, why not use a 4 pole switch? You video convinces me that even power supplies can be over my head.

[@kirkpennock2997]

I liked it.

[@jeffmerlin2580]

If all grounds are common, it would be far more easy to swich the negatve with only one N-channel MOSFET. Perfect timing as well. :-)

[@Graham_Wideman]

Nope. If you connect all the negatives together, you've established a voltage between each positive rail, and those positive rails are then applying their voltage differences to the system under test.

[@mirkomueller3412]

Jesus - i never thought about that... but now i know.

[@felipe34k]

5462D repair!1! want to see that!

[@RicoElectrico]

There are PMICs like ones from AMS that can be extensively programmed via dedicated software and do exactly that - rather complex power sequencing. Honestly, is there any reason for sequencing power rails instead of sequencing RESET signals?

[@rocketman221projects]

Some chips will let the smoke out if you don't sequence the power rails.

[@RicoElectrico]

This is begging for trouble whenever something fails ;)

[@Graham_Wideman]

Dave gave at least one answer as to why sequencing the rails is important: wrong sequence can lead to latching up the SCRs that are formed by various silicon layers that are normally reverse biased, but would be forward biased when unanticipated voltages are across them as a result on one supply being on while another is off. Any such turned-on SCR draws a large current and burns out the IC.

[@mikeselectricstuff]

Need to use Shift to turn a PSU output on - seriously ?

[@FalcoGer]

4 connected switches. seems easy enough. or if you want to be fancy, use an array of relays and maybe a bit of debouncing logic with an rs latch and some FETs for the actual switching.

[@JohnniOstergaard]

Why not use a single Relay?

[@EEVblog]

Because most products don't use a mechanical solution for something like this, and that's why I wanted to show a solid state solution and mention the chips that are available to do this. A single relay also doesn't do sequencing and has contact bounce. But sure, for just this one hacked test thing a multi pole relay would have done fine.

[@JohnniOstergaard]

Okay, I see what you mean, thanks.

[@richardgoebel226]

Wow, if you have that much visible bounce doing it that way \then a 4 pole toggle switch would be outrageous.

[@phalcon23]

what about multiple relays, debouncer circuit, could be controled with microcontroler?

[@Graham_Wideman]

@@richardgoebel226 The noise you saw was not really bounce. It was because he jiggled the plug after a second or two.

[@chenartahir1865]

I have 15 samples of artificial batteries. I want to measure the changes of voltage with time of each for continuous 2 weeks. How can i measure all in same time with one oscilloscope? Is there any system i can connect to the oscilloscope?

[@omitsura]

@3:27 What candy you're eating?

[@stevenbliss989]

So why did you pick P-Channel mosfet if you want low rDS? :):):)

[@douro20]

-5.2v...got ECL?

[@maxsnts]

Could you have used the -15v as the "switch"? So that when you turned -15 ON, the gates would be pulled down and it would all be in sync?

[@Laogeodritt]

Most MOSFETs have an absolute max Vgs value of 20V. Beyond that, the gate oxide may undergo dielectric breakdown-so using -15V as the gate signal may very well have have destroyed the +15V7 switch. In applications that deal with voltage rails greater than 20V, you would want to use a gate signal that's higher than ground, mayve Vgs=15V for example. Even worse, for situations where the switched voltage can be unpredictable, you might actually want to track the gate voltage to the source in some way so that Vgs is fixed to ~15V.

[@maxsnts]

Even then, since the gate take no current, he could have broken down that 30v amplitude (+15v->-15v) using a simple voltage divider and connect the gate to that right? Maybe i missing something fundamental here.

[@duanecook4227]

What would be wrong with just having a single return line splitting to all the power supplies. Then just switch the single return. Only thing I can think of that would be bad would be any mains earth reference in the oscilloscope.

[@Graham_Wideman]

No, this won't work. Answer attached to the several other comments saying this same thing.

[@AlexandreKandalintsev]

I'm surprised to see that much bouncing. Power mosfets should have very high gate capacitance. I think a gate resistor with higher resistance could solve the problem.

[@EEVblog]

I was very deliberate in trying to make it happen so we could see the effect

[@Txepetxcc]

I think power mosfets should have as low gate capacitance as possible, to reduce gate driving losses, isn't it? Gate losses are one component of the switching losses in SMPS power supplies. I think that each time you open or close a mosfet, you charge and discharge that gate capacitance. Therefore, you would like it to be low, to reduce the charge you need to transfer at each switching event in order to change the gate voltage (charge or discharge gate capacitance to the required voltage to turn on or off the channel). Then, as you pointed out, if you increase gate resistor value, then the time constant of your gate loop is increased, and that gate capacitor charging time increases, and slows down turn on / off. But still I think the gate capacitance is too low to reduce contact bounce only by that way. But you an add external discrete capacitors to slow down the gate voltage build up. Also, doing that, it might happen that due to manufacturing tolerances, each switch might have a slightly different gate threshold voltage. So slowing down the rising edge might produce more sequencing problems due to slightly different turn on times for each power device.

[@AlexandreKandalintsev]

That's all valid as well :). I think I'll try make a simulation on this.

[@spudhead169]

Couldn't you have used one of the rails to pull the gates of the mosfets low via an npn? That way you could put that rail straight through and the other 3 on the mosfets. Then all you need to do is switch that rail on last and they'll all come on at the same time.

[@NivagSwerdna]

It would be good to eliminate the bounce with a 555 or other astable... but also... aren't those swanky power supplies also current limiting (they would have finite output capability anyway) so could some of the funny business actually be them regulating for the initial surge as bits and pieces charge up?

[@cosmicmatrix6238]

Well Bob's my uncle!

[@DonaldSleightholme]

❤️

[@russellthompson3201]

Dave: I am rather ignorant with electronics. It seems to me that all those PS should have a common ground. Why can't you just put a switch in the ground lead before breaking out to the various PSs? Your MOSFETs still mimic the switch bounce problems.

[@Graham_Wideman]

If you attach all the PSU grounds together, you've created a voltage across all the positive (and one negative) terminals. So between 15V and 3.3V you have 11.7V, and so on. Those voltages would all be applied across the system under test. Not what you want.

[@jimmybleron4700]

haha no eating in the laboratory^^

[@kentvandervelden]

Wow, very interesting. Who would have thought something that seems so simple would have so much to it.

[@rageagainstthebath]

Contact bounce after 500ms? Is this a joke?

[@Graham_Wideman]

No. Perhaps you misunderstood what was makin intermittent contact. It wasn't a switch.

[@1969yaniv]

Hi, I think there is a mistake on the N-Channel circute. the switch to the gate is placed incorrectly, isn`t it?

[@sjwsbetaskiller6218]

No, it's OK.

[@omitsura]

Dave, could you get away with this setup instead? 1. Grounds are disconnected 2. connect +ve voltages to targets 3. Wire up all power supply grounds through a mosfet to the scope and just switch one mosfet?

[@Graham_Wideman]

No. Because even though you haven't connected the grounds to the system under test, you have connected the grounds to each other, which means you have a voltage between the positive supply rails, and thus across portions of the system under test.

[@EscapeMCP]

Couldn't you have used the 15V rail to switch the other mosfets on?

[@rsattahip]

Don't mean to sound sarcastic, but why not just fix the power supply first. To power all up at once plug them into one extension cord.

[@Feedback_Jack]

Power supply heat sink is looking incredulous about what's happening

[@FalcoGer]

just put a 8 connector switch across the power rails and flick it.

[@metheone4]

Happy endings with your oscilloscope ? Lets hope MrsJones doesnt watch your vlogs... *lmao*

[@MCEngineeringInc]

Why couldn't you have a 3 pole relay pull all rails at the same time

[@sjwsbetaskiller6218]

Maybe he doesn't have proper relay (but have MOSFETs)? Plus, there will be contact bounce!

[@FennecTECH]

To eliminate bounce id just use a box with a bunch of relays

[@sjwsbetaskiller6218]

Stupid. Relays also have terrible contact bounce!

[@pvc988]

I am too drunk to to watch whis stuff.

[@mbaker335]

Please do not display a schematic and then immediately cover it with a datasheet. Boy is that annoying. Interesting vid otherwise.

[@gotj]

FIRST!

[@EEVblog]

FIRST to be 11 hours late, congratulations, you win the internet.

[@gotj]

Wahaha, the internet is now mine. Big thumbs up Dave, great vidjeo!

[@God-CDXX]

if that is homer in the X-ray the brain is too big

[@milamiglia2437]

Don't repair old junk. It's so not worth it. Move ahead.