Do you need thick copper layers in PCB for high currents? Are you sure? | Steve Sandler
Вставка
- Опубліковано 6 чер 2024
- What is important when designing boards and power supplies with high currents? Explained by Steve Sandler
Links:
- Steve's LinkedIn: / steven-sandler-022a7210
- Steve's Online Course: marketplace.fedevel.education...
- Steve's company: Picotest www.picotest.com/
Chapters:
00:00 What is this video about
00:32 Current state of the art
09:20 Why inductance is the biggest challenge (1)
19:15 Minimizing the inductance (1)
20:47 Thinner dielectric for even lower inductance (1)
23:35 Why inductance is the biggest challenge (2)
39:03 Decoupling the ASIC
45:50 Minimizing the inductance (2)
50:55 Thinner dielectric for even lower inductance (2)
53:23 Carrying the current (ANSWER on the question in title)
59:24 Stability - Both input and output (control loops)
1:10:45 Million dollar chip
1:17:55 What VRMs can supply this kind of currents?
1:26:20 Is GaN the answer?
1:31:58 Testing gets really tough
1:37:34 Thank you and contact
------------------------------------------------------
Would you like to support me? It's simple:
- Sign up for online courses hosted on our platform: marketplace.fedevel.education/
- Sign up for my Hardware design and PCB Layout online courses: academy.fedevel.com/
- You can also support me through Patreon: / robertferanec
- Or sign up for my Udemy course: www.udemy.com/learn-to-design...
It is much appreciated. Thank you,
- Robert
40 years into my own electronic design company , to be constantly reminded how little I know and how the devil really is in the detail !!!. Thanks to your guest for sharing a lifetime of knowledge . Normally I watch videos at *1.5 speed but in this case I will rerun where it will be a few minutes - pause - and then many many more to digest , then rinse and repeat like enjoying a fine wine . Thanks Robert and Mr Sandler
VRM? Viewers coming to this video cold will want to know that "VRM" stands for "Voltage Regulator Module", meaning voltage regulator circuitry located very close to a high current load like an ASIC. These typically don't just regulate, but rather convert an input voltage like 12V down to some low voltage with high current capability, like 1.something, or the 0.7V mentioned in this video.
Some of the first VRMs on mainboards were on the socket 7 boards of the first Pentium line, to support the lower voltage of the newer MMX Pentiums 😂
Oh and also on the 486 Pentium Overdrive Processor was a VRM!
never thought of the power design in high power parts... the minimization and mitigation of the parasitic values of the power stages, and the board itself... its truly an art!
For those struggling to understand why inductive behavior is so problematic, they are covering powering mostly computing ICs, devices that have abrupt and sudden transitions on the demanded current (millions of transistors switching on and off).
If your delivery network is inductive in nature, it’s like having an inductor between your power supply and your IC. You’ll have big voltage spikes on it.
Inductive loads by contrast will be much more forgiving in this aspect since they’ll be the limiting factor in current change.
Yeah, I don't know why they struggled so much and talked all around the topic without stating the underlying problem: If a chip has a sudden increase in demand for current, if there's inductance in series in the power distribution network (PDN) (on VCC or Ground side), that inductance will respond to the instantaneous current demand with a large voltage drop, and thus a large sudden dip in the supply voltage to the chip. This could be a dip in VCC, or lift on the ground side, and these compromise the chip's ability to operate, as each stage sees extraneous steps disrupting the intended signals.
Then when the chip has a sudden _decrease_ in current demand, that will cause the PDN inductance to dump a pulse of current into the PDN, seen at the chip as a momentary overvoltage on VCC or under voltage on ground. Again disrupting the intended signals.
Adding to this is that the combination of L and C resonates, so that individual current steps not only result in voltage pulses, but pulses that ring.
@@Graham_Wideman The problem is only with the title, one might have assumed first that it is for a lab PSU or other (higher voltage) PSU of some kind.
But the thumbnail already telling the truth (1kW, 2kA, which suggests it is few volts).
The video itself is great, and who designs digital systems at this level is certainly do not need explanation why inductive beahviour of the PDN is a problem: you can encounter tbis issue with a few magnitudes lower frequency level, so in simple microcontroller kind of jobs. Just the distances and inductances are bigger, but you can get a very good sense of how easy is to mess up a design up to failure. Like it said in the video, this design mindset is important at every level, just the numbers in the equation are different.
Thank you, Robert! Your videos are an amazing source of knowledge and wisdom.
Academic understanding is wonderful, but your videos deliver the practical real world side which makes them invaluable.
Excellent video. More power electronics videos, please! I think the length is perfect.
Very good video!
My ears perked up when Steve mentioned controlled-ESR capacitors - these are mentioned by both Ott and Archambeau as the cure for inter-plane cavity resonances, but sadly they never caught on and are difficult to find. A little surprised that three-terminal capacitors were not mentioned, as manufacturers like Murata tout them as being able to vastly reduce the number of required decoupling caps due to their extremely low inductance.
They can reduce inductance, but ESR is intended too low. Also hard to fit between bumps
Thanks a lot Robert, for making such kind of videos. Also thanks for asking questions, you were asking really those kind of questions which a designer always thinks but gets very difficult to find the answers. I really love your videos, keep posting these kind of videos.
We are so lucky to have you and your guests as our teachers!
Love this video by Steve and Robert. One of the best available talking about trend and challenge of high current VRM solution for ASIC.
Very good video. I have learnt so much from your videos. Thank you Robert.
This was an amazing explanation on why lower ESR isn't always a good thing. (while, intuitively, it has always felt to me like lower == better)
I've been making boards with GaN, like 100A+ on a 3inch x 3inch board.
You really do need 2oz copper minimum and the inductance for power and gate loops is extremely important.
For those trying to understand why low inductance is important, a GaN transistor switches 100A in like 3ns. Even if your loop inductance is only a few nH, you can still do your Ldi/dt calculations and see that you get a high induced voltage which can kill devices easily during transistion.
Where do u work?
kill devices?
Great video, thanks for sharing!
Great video, Robert! Super cool topic.
Great interview!
Thank you so much for this very interesting video! Absolutely great and well made, so informative, you did a great job!
Mindblowing stuff. I am going to have to rethink how I understand power delivery.
Great stuff. I've used Vicor VTM "current multiplier" modules - they are amazing parts.
Great Robert , thanks to to Steve , could you also make video on high current PCB with busbar , how to layout it .
This reminds me of the 80s when Data General made a computer that I used that came standard with battery backup because it would go into power fail about once a week and the battery inverters held up the power rail when the supply dipped too low.
This is super interesting. I know next to nothing about these regulators but PC overclockers actually deal a lot with them and how well they work determines how high they can clock before a CPU becomes unstable. There is a lot of talk there about how many phases they are using, about loadline calibration, etc.
I just understood the problem of low esr caps, self resonance and underdamped LC filters. I was very surprised to know more about this topic than Robert, because all the videos I watched on this just recently 😂 I highly recommend the omicron videos on yt about the topic, input LC filter and so on.
Not a fan of short videos that can't fit the required material for learning. Great video Robert!
Neither am I.
I think I have something similar in clock distribution, with a rise time of 300ps for 12 outputs in 50ohm the power needs a lot more consideration.
I Love it your video
Does the voltage level affect the ratio of power going through copper vs the power going through the field around the conductor?
I tried a 8 layer PCB with high power and well it worked electrically, but made popping noises as the inner layers got too hot. we went with power traces on the bottom and adding a thermal pad to a heat sink
You live and learn, my friend.
That sounds more of a via issue. Either that or your fab totally screwed up and used FR4 that had adsorbed water.
@@theondono It was a high power RF project and the FR4 PCB was starting to turn to charcoal. It took a bit to convince management that it was even possible which is why we started with a FR4 board, but with time we simplified the design to a 2 layer Arlon PCB and standoffs to a FR4 board. That then allowed the thermal pad on the bottom to transfer the heat to the water cooled heatsink while the FR4 board handled the low power stuff.
@@LaserFur You DON'T use FR4 for RF!!!
@@nameredacted1242 you can for prototypes. the purpose was to prove something so that the real PCB could be developed.
I expanded the awareness of my own ignorance. A lot.
All this GaN, Diamond transistors are good, but they are pretty expensive for now, cause not many manufacturers make them. Also they need more complex drivers cause they have higher band gap. And what i understood so far, engineers can create any design, even for million amps, but the question is for what cost and purposes :)
Cloud computing. HPC, data center, graphics cards, AI, supercomputers, etc.
Your phone charger , laptop supply is most certainly already gan powered. Your EV car OBC is SiC powered . It’s not that just a few companies make them. It’s that currently SiC makes more sense , and gan foundries have low yield because of silicon impurities contamination
Hello, Robert! I'm working on a PLC for educational proposes, and I've been following your great content since I started this project, currently using EasyEDA as the main software to produce and then manufacture the board. However, more recently I'm facing some problems with the software, when I make any little changes on schematic and update the PCB I lose all my progress on the PCB design. Do you know any way or command I can use to prevent from this?
Moores law also lives in power electronics
Didn't get why bigger capacitance solves a problem of a low ESR, cause by multiplying capacitance we lower an ESR even more.
Video length just right
It should be clear that an inductance in the power supply line has to be avoided because it creates a voltage drop and therefore in a lower supply voltage where it is needed so that the ICs can’t operate anymore because the voltage is too low
It feels like him and Rick Hartley are saying opposite things, but ending up with similar designs.
He's saying you want to maximize impedance while Rick wants to minimize it. Rick says the field travels in the dielectric so copper thickness doesn't matter beyond dissipation, he says that he uses 14mils of copper on each layer.
I'd like to watch a conversation between them.
This sounds totally crazy to me, I never thought it was possible to get so many A with very little V. For example 2kA*0.7V = 1400W, I don't know any component that could handle that. Even to get 10W resistor you need a huge wirewound one, cannot imagine how a 1KW component could even look like
Look at household electric baseboard heaters. 240V, ~5kW.
Best analogy of 1.5kW worth of heat is a household oil-filled portable electric radiator. Its output can heat up a whole room in an hour or two, in the winter.
@@nameredacted1242 but is that going through a PCB? I put 1.5W on a 3W resistor and it was impossible to touch it, super hot
@@carlosgarcialalicata Obviously NOT!!! 1.5kW of power must go to your LOAD (and THAT will ultimately just convert it into waste heat).
If you have to deliver 50kVA+ to the load, THEN your power electronics will be dissipating on the order of 1.5kW of losses in control electronics, at that point you must use liquid cooling...
Yes, it's a lot. Even normal cause today can hit 500Amps and GPU's often hit close to 1kA. Data center, super computer and HPC easily exceed 1kA.
My GPU (3090) pulls 400W at around 1V. So 400A is not that uncommon in higher end PCs. It’s not even water cooled (yet!)
If I had enough money I would research hw electronics like you, bit I dont so I need to become a sw guy
48VDC will be the best thing since sliced bread for the next 20 years in electronics. I am talking power supplies for laptops, small monitors, small TVs, printers, next generation of USB, Ethernet PoE, etc, etc, etc. It's "almost safe" in terms of shocking your body, and can deliver enough power for the next 20 years.
I think next version of USB-PD will be able to support 48V, so... you aren't wrong mate.
Yep I’d love to see 48V (or better yet 60V) become more common. I think you’re underselling it with “almost safe”. If someone manages to kill themselves with 48V, they kinda deserve to die. I’m not saying I want them to die, but I am saying they have to so that the rest of us can move on :)
@@NavinF I think the safety limit is around 50V, so I don't think we are going to see this anywhere near in consumer goods. 48V is alright and probably already pushing it, but they need a unified solution.
The more voltage the slower and bulkier ( or more expensive ) the switching.
20V is a nice sweet spot as many laptops get their main power at that voltage. Upping the current is much more dangerous, that upping the voltage, and because 100W is not that much for bigger devices, they are probably going to settle for something like 32V at max 5A. 48V is probably going to be a specialty for replacing Ethernet and PoE but only for very specific applications, perhaps to power certain Ethernet adaptors or something like that.
And sure some very high end laptops are going to go the way of 48V but, if you need 240W just use a proprietary / universal power plug.
@@NavinF I think that 60V is into unsafe territory. Read up on it. Plenty of regulation classifies 48VDC to be the upper range of "will not shock you". Voltages above that, while they do not have the AC-like "muscle twitching" effect, with wet skin and standing on wet ground, or wet clothes, will probably give you a jolt you will feel, and may affect heart operation. So I would be careful saying that 60VDC is safe, and the other statements that you made.
@@TrickyNekro I’m pretty sure the 50V limit you refer to is “low distribution system voltage” (LDSV) as defined by the NEC. That’s not particularly relevant to USB cables and it’s not even followed by all PoE implementations (passive ones often idle at 55V). Anyway the limiting factor has always been arcing damaging the conductors when you yank the cable, not safety
Hello Sir,
I have heard about your online training platform where we can sell our course. I have developed some courses. How can I sell on your platform?
Hello, please send email to info@fedevel.com
i got a marketable product idea. simple single motor esc 5-10a, combined with 2 servos and an rc reciever.
if you can build it lightweight and under $25 you'll ship lots and make the rc glider community real happy
For what they’re talking in this video, that’s low current. If you went up an order of magnitude, then it would start to sound fun 😂
it already exists, but is more than that cost. The cost of having separate pieces is not very much. so creating a custom design would result in a higher cost.
@@LaserFur the guy who designed & sold them died a few years ago. His son sold them for a while but moved on to other things. Their design was old & you'd have to ship it back to them for firmware updates.
Theres a 'flying wing hobbies' version currently but its priced at a wild $80. Thats the cost of a fully featured premium flight controller.
This is a simple spektrum ppm reciever with 2 servos and a low power dc motor esc. It should be about $10 in BOM.
This hobby company is just gouging because they can. If you made a newer design, flashable by the users, youd sell to all the lightweight glider people
@@terminsane so I am guessing it would take me $40,000 US on NRE. Then you have to divide that by the market size. I can't even get products made for less than $50 that just have a processor and some I/O. So overseas manufacturing and low NRE is important. And then someone will copy it.
@@LaserFur ohboy. One of those... Yeah i guess you need a million dollar investor to print at pcbway
So, the title has LITTLE to do with the video. Ultimately, you WILL be using lots of copper thickness (just spread over many, many layer pairs). The very basic and dumb Ohm's Law still works at DC. Obviously, the video is awesome to talk about everything that happens at AC (for digital people, switching waveforms IS AC in case you did not know!).
Same with the poorly chosen video thumbnail illustration. You won't transfer 1000A from a VRM that is soldered onto a PCB.
@@nameredacted1242 It’s not uncommon for LN2 overclockers to push 400A through aftermarket VRM PCBs that are soldered flat on top of the GPU’s PCB. I don’t see why you can’t push 1000A with enough bypass caps and cooling
Please sir make a demo vedio in altium.