Power Supply parameters - Dynamic Accuracy

[@tiagoferreira086]

This channel is so underrated... Really great educative content here, and a lot of effort and time invested to show us as clear as possible. There are channels in youtube that shows bad practices, wrong things, nonsense and fake content, and they have millions of views...

[@yt4krist0f]

Great content! Like to see this "layered" analysis of effects regarding different type of loads and static vs dynamic behaviour. It's always kind of rewarding when we can see and model "all" aspects that produce the final outcome of a system.

[@lohikarhu734]

so nice to see someone NOT using the so-called 'ground lead' !! As a scope AE with 'some years' of experience, probing has been the cause of more 'issues' than can be counted by me, from10 MHz to several GHz....

[@gkdresden]

A very important topic. A follow up video about means to achieve good power supply stability would be interesting. The best would be a comparison between single chip (linear) voltage regulators and a voltage regulator which is build with discrete standard devices and optimized for very high output voltage stability.

[@FesZElectronics]

I will try to look at stability in more detail at some point; however, I want to take a step by step approach - first to look at the problems you usually face in a real life application, how the problem can be investigated - the tests to perform and afterwards see how it can be fixed.

[@gkdresden]

@@FesZElectronics I am also looking for a device or circuit to get a good current regulation / current limit as close as possible to the voltage setting.

[@EJEuth]

Amazing demonstration of concepts, showed in real devices. Thank you for preparing it for us all.

[@stuckathome5896]

What you see at 8:20 is probably a ground bounce (from the capacitor current) and not an overshoot of the regulator.

[@pieters286]

Thanks, as always high quality!

[@m1geo]

Very interesting! I'd like to see these tests on an switching power supply!

[@alaricsnellpym]

Good to see the current probe again :-)

[@joejane9977]

thank you and as always excellent content

[@benatesnaola6001]

Thank you fesz! Great as allí your videos. About estability could you measure gain margin and phase margin? Or explain how it could be done in an easy way? Thanks again and keep working like this, you are amazing!

[@electronics.unmessed]

Really interesting topic! Thanks a lot.

[@kadamrohan16]

Nice Video ! Hoping to see next content on Bode plot analysis for stability of converters.

[@FesZElectronics]

At some point I will get to that as well :D

[@lorenzobagnoli314]

Very good video as always! What is the power amplifier that you are using? It seems diy, maybe you can show it in a next video, that would be very interesting! Thanks for your videos, top notch quality content

[@FesZElectronics]

You are right, it is DIY; I built it a few years ago. If I remember correctly its built around an LT1210 - that is part that is providing the high current output. Maybe at some point I will do a video about the amplifier, but its not really planned at the moment.

[@windward2818]

The load can be complex, such that a failure can cause energy (overvoltage event) on the output of the regulator. This is usually mitigated by a large TVS on the output or a diode across the regulator. You may also see a TVS on the regulator input which may also be protected by a fuse depending on how the supply may respond to a huge line transient (like a lightning event or you live in a very noisy AC environment like some cities in England). The challenge in use case testing of the PCB is in how to do the measurement. PDN low impedance testing is not easy and you usually need a very high bandwidth if you are doing preliminary EMC screening. And as I am sure you know, the design can actually be working with no operational problems other than not passing EMC. So, it is hard to understand all of the use cases when it is also PCB layout dependent. The more experienced design teams will specify features in the layout which make it easy to support the needed testing. From a mathematical standpoint the impulse step response should be ideal. This means a perfect zero rise time step response, which is not really possible. You can however get a very fast step if you use a fast high current low Rds On MOSFET with a pre-driver circuit which is driven by the signal generator. The pre-driver can supply a huge amount of current in a very short amount of time at around a ns. This is usually not needed however with most electronic loads which switch quick enough to do supply edge testing perturbations for time domain stability testing. Your custom current probe is very noisy. I really like the fact you designed your own. My own custom current probe designs are battery operated and run off of two 9V batteries for split supplies and have a small nose on the current sense board pcb (6 layers to allow me to bury signals) to solder directly across a small SMT sense resistor (very low inductance) on the PCB using very short low inductive leads. I have options for large gains for small sense resistors up to about 5 Mhz worse case bandwidth, which is usually sufficient. The shunt resistor current sense approach can be made much more compact than making a current loop for a magnetic current probe. Making you own shunt current probe is rather easy with the newer fast instrumentation amplifiers available, and they seem to be getting better every year.

[@t1d100]

Excellent.

[@mountiedm]

You're a nanofarad!! 🤪 Love your stuff!!

[@alexarif2835]

I would like to know the best possible way to reduce noise on the output voltage. Are low cost linear regulators a useful building block for such applications for example to build a 12 Volt low noise power supply for a DAC (digital to analogues converter)? For instance if the LMxxx isn't able to provide enough current, and I would use a transistor in connection with an LMxxx how would its parasitic or none linearity play into the noise floor of the output voltage. Alternatively how would you go about designing a low noise 12 V, 5 A supply ?

[@PainterVierax]

well, you need to filter the noise at the input of the regulating circuit and be cautious with the shielding and grounding. A linear regulator can easily filter out a 100-120Hz residual oscillation of a AC rectifier+smoothing circuit but it's not a HF filter at all.

[@udhayakumara4033]

We can eliminate this issue with bulk capacitors at the input and output of the regulator? How much importance this in real world applications? I can't see any bulk capacitors at the output stage

[@FesZElectronics]

Of course, having larger capacitors will usually help; I intentionally did not use any to highlight the effect.

[@MatsEngstrom]

As I see it bulk capacitors doesn't fix the problem. It merely hides them - the regulator itself is stil doing a bad job. And they also cause other (admittedly solvable) problems with inrush current and also problems when you want to have current limitation on the output that is fast enough to protect your circuit.

[@udhayakumara4033]

@@MatsEngstrom why should i care about which component is eliminating the issue. I need to eliminate the unwanted peak at the output, it can be achieved by some over engineered expensive regular or a simple low cost capacitor. That's why I asked where it is need to be considered. For normal voltage regulations, a capacitor and an inductor will be enough. For programmable PSU applications, we may need to care about the fast changing loads, but for normal applications a soft start feature is enough as he mentioned. You ca use an eFuse if you want.

[@MatsEngstrom]

@@udhayakumara4033 That's not a fix. It is a bandaid that forcefully trying to make a bad and/or insufficient design to kinda work by hiding the problem. And as I said this kind of "fixes" comes with their own problems. Like pulling a shedload of current at turn on (necessitating an unnecessary large fuse value) or or if put at the output makes any current limit to bee too slow. Take a cheap PSU with large caps at the output and sett it to max 10mA , turn it on and then connect a directly LED to it. The LED goes dead since the charge in the cap have enough energy to kill it. A real pro PSU from Agilent or some other non-crappy brand have no large caps at the output and will not destroy the LED. Reducing overshoots is to my knowledge mostly a matter of having a correctly calculated and tested compensation network in the control loop. Go ahead and use cheap stuff that's engineered by school kids or manufacturers that are removing part by part in the BOM until the design barely works if that is your preference. I prefer to use tools made by people that does things the right way and that I can trust works as specified under all conditions.

[@RexxSchneider]

Adding some capacitance at the input and output is normal design practice. However, bulk capacitance - i.e. large value capacitors, say 100μF and perhaps much larger - is only likely to be helpful at the input to the regulator because they will tend to reduce the change in the input line voltage, thus improving line regulation. On the other hand, capacitance added to the output can only be of value to a reasonably well-designed regulator in filtering out fast transients on the output, usually caused by abrupt load variations. Any such output capacitance will need to have the lowest possible reactance at high frequencies, and that invariably leads to smaller capacitances in the 100nF to 10μF range, depending on type.

[@afnDavid]

This needs subtitles