EEVblog® Electronics Community Forum
EEVblog => EEVblog Specific => Topic started by: EEVblog on August 23, 2023, 11:00:01 pm
-
Part 20 of the uSupply USB power supply design. An update on where the project got to and why.
A look at the finished unit, PCB and schematic.
https://www.youtube.com/watch?v=FKC6lnTKH-c (https://www.youtube.com/watch?v=FKC6lnTKH-c)
-
I haven't watched it yet, but thanks for the follow up to this project...
Certain projects are really subject to the tides of life... :-DD
-
Cool retrospective. I think many of us can relate to the story of this almost-a-product project.
IMO the design was indeed too complex to make it a viable product for its intended audience (and thus corresponding price range) - including the firmware itself AFAIR. But quite a bit was learned along the way, with the custom LCD and custom transformer, how to implement USB PD, and more.
But hey, third time's a charm. I think the next uSupply can be the one that you'll actually sell. ;D
-
Strangely, this video has the biggest thumbs down ratio of all my recent videos except the LK99 one.
Are people pissed off they can't buy it?, or don't like projects?
I don't get it. It's my most requested video, and countless people cite the uSupply design series as one of the best things on my channel. :-//
-
Yea, she's gotten a bit old in the teeth, didn't she? Your first publication about this project is more than ten years ago I think.
Interestingly, the thing I would change is the custom LCD... make that either an e-Paper or even a color TFT, they are cheap now.
You've mentioned open sourcing it but I can't find it on the 'official' github, is there a different one to github.com/eevblog?
-
You've mentioned open sourcing it but I can't find it on the 'official' github, is there a different one to github.com/eevblog?
Sorry, gitlab:
https://gitlab.com/eevblog
-
Strangely, this video has the biggest thumbs down ratio of all my recent videos except the LK99 one.
Are people pissed off they can't buy it?, or don't like projects?
I don't get it. It's my most requested video, and countless people cite the uSupply design series as one of the best things on my channel. :-//
Maybe it is because the video is about actual electronics design? >:D IMHO it is a good recap of the project (judging from going through the video quickly) so I'm as surprised as you are.
-
Strangely, this video has the biggest thumbs down ratio of all my recent videos except the LK99 one.
Are people pissed off they can't buy it?, or don't like projects?
I don't get it. It's my most requested video, and countless people cite the uSupply design series as one of the best things on my channel. :-//
I've watched almost all of your videos (seriously) and I thought this one was one of the better ones. It's not easy making an accessible, easy to watch, and entertaining video on electronic design.
-
Strangely, this video has the biggest thumbs down ratio of all my recent videos except the LK99 one.
Are people pissed off they can't buy it?, or don't like projects?
I don't get it. It's my most requested video, and countless people cite the uSupply design series as one of the best things on my channel. :-//
I loved the video.
Especially the part about using a digipot (or alternatives like using DAC and opamps) for switchmode voltage (and current!) control definitely deserves its own video, if you ever feel like doing one. I mean, even hobbyists like myself can nowadays use TI Webench to get good suggestions on chips and circuits to use, but to adapt one for digital control is "black magic". (I only know enough to know that the loop size, actual bandwidth of the digipot or opamp, and capacitance and inductance in the feedback loop, will affect the stability and functioning of the switchmode controller.)
A particular example I have, is to digitally adjust a boost controller from 5V to 7V-12V, to control three-pin voltage-controlled fans. (I've recently showed my PWM fan controller designs using ATtiny85, but I also have a bunch of pretty good 3-pin voltage-controlled ones with tachometer output.) The three-pin fans can be controlled by PWMing their 12V supply just fine, although one may need to adjust the frequency and add some filtering to avoid additional noise. The step-up-based one would allow efficient use of 5V USB wall-warts. (My ATtiny85 ones do not simply control the PWM or duty cycle based on user control; the user controls the fan RPM directly, with the ATtiny85 managing the duty cycle as needed, to get the fan to that speed. The purpose is to basically let the user control the airflow directly, even when the fan ages. Not a common need by any means, but very useful for the rare cases you do want/need one.)
-
Strangely, this video has the biggest thumbs down ratio of all my recent videos except the LK99 one.
Are people pissed off they can't buy it?, or don't like projects?
Who knows. I'm guessing that many, seeing a new video about the uSupply, may have expected "good news" (ie.: finally making it to your store), and seeing it was "just" a retrospective with you definitely stating that it will never be a product as is, they may have reacted as spoiled kids not getting their toy.
Likewise, during the (short-lived) "hype" on LK-99, many seemingly got pissed about anyone that would try debunking the thing, whatever you would actually say. It could be seen even on this forum.
I don't get it. It's my most requested video, and countless people cite the uSupply design series as one of the best things on my channel. :-//
Yeah I don't know. I noticed that people seemed more inclined to be negative lately than they used to, on YT but not only.
-
Likewise, during the (short-lived) "hype" on LK-99, many seemingly got pissed about anyone that would try debunking the thing, whatever you would actually say. It could be seen even on this forum.
That is likely because people are fed up with the umpteenth nonsense debunking video from a creator trying to score cheap points.
-
Likewise, during the (short-lived) "hype" on LK-99, many seemingly got pissed about anyone that would try debunking the thing, whatever you would actually say. It could be seen even on this forum.
That is likely because people are fed up with the umpteenth nonsense debunking video from a creator trying to score cheap points.
So you think that demonstrably silly promotional video on their website didn't deserve to get called out?
-
Likewise, during the (short-lived) "hype" on LK-99, many seemingly got pissed about anyone that would try debunking the thing, whatever you would actually say. It could be seen even on this forum.
That is likely because people are fed up with the umpteenth nonsense debunking video from a creator trying to score cheap points.
So you think that demonstrably silly promotional video on their website didn't deserve to get called out?
I suspect that Nico is referring to the fact that *everyone* piled on the thing with flashy and catchy thumbnails and clickbait titles, which becomes somewhat tiring after a while (something similar to what is going on with LTT right now).
-
I suspect that Nico is referring to the fact that *everyone* piled on the thing with flashy and catchy thumbnails and clickbait titles, which becomes somewhat tiring after a while (something similar to what is going on with LTT right now).
Yeah, that's why I did something different. I wasn't going to make a video until I saw that demo video and laughed my arse off.
I didn't see anyone else even mention it.
-
Likewise, during the (short-lived) "hype" on LK-99, many seemingly got pissed about anyone that would try debunking the thing, whatever you would actually say. It could be seen even on this forum.
That is likely because people are fed up with the umpteenth nonsense debunking video from a creator trying to score cheap points.
So you think that demonstrably silly promotional video on their website didn't deserve to get called out?
I suspect that Nico is referring to the fact that *everyone* piled on the thing with flashy and catchy thumbnails and clickbait titles, which becomes somewhat tiring after a while (something similar to what is going on with LTT right now).
That and there is so much nay-saying and negativity on Youtube already from self appointed 'experts' which go around with claims like 'if you have the tiniest bit of engineering knowledge, you can obviously see in 0.1 second that this system can never work' but fail to deliver a good analysis. It is just hyperbole piled on hyperbole like gossip magazines for women.
-
I am always suprised that project videos perform so poorly, but to be honest it probably makes sense that the usual tiktok kid can not maintain their focus for longer than some 60 seconds short clip.
Does Youtube differentiate in their stats between performance amongst actual subsribers vs. random viewers? I would assume that less click-baity project videos might still perform reasonably amongst actual subscribers.
I personally would love to see more project videos just like the usupply. It just gave a great overview of electronics design, pcb design, mechanical design considerations, custom lcd/keypad, and could have even more potential regarding UI/UX design etc.
But I guess this is just not for the random folks on YT
Anyway, love the work David 2 put into this. He seems to be a good engineer and totally got that engineering mindset that so many people in the industry are missing. 10/10
-
I completly understand you don't want to sell the current design, it sounds like the BOM cost alone would push the unit in the $ 250 range.
You really 'gilded the lily'
Here is my suggestion what I would do if I would redesign this unit;
1) Get ride of the display and touch pad, I know they are beautiful and functional, but you really don't need those you have your phone...
2) Make it wifi enabled or (bluetooth if you think that is better)
3) Get ride of the transformer
4) Add battery function so it can run of that and use usb-c to charge and/or run the unit from
5) Use an off the shelve heatsink
6) use a wifi module with powerfull cpu's
7) Use an off the shelve enclosure, I like the idea of the original design
In short keep it very simple and functional.
Those are just my 2 cents, please don't take this personal.
-
DAVE here are a few design questions:
1 why were 2 power elements used? aka why not limit current in the same FET as the current limitation?
2 why is the SMPS only .5MHz ? you could get a way smaller one at 2MHz
3 Was the custom heat sink truly needed? wouldn't a Al plate have down the job? Asking since cheap ish AL PCB could have done the trick .
4* why are you still using such an old Altium version?
-
DAVE here are a few design questions:
1 why were 2 power elements used? aka why not limit current in the same FET as the current limitation?
2 why is the SMPS only .5MHz ? you could get a way smaller one at 2MHz
3 Was the custom heat sink truly needed? wouldn't a Al plate have down the job? Asking since cheap ish AL PCB could have done the trick .
4* why are you still using such an old Altium version?
1. 2. Don't recall, sorry.
3. Because it was cool, pun intended.
4. I have AD23 installed. Accidently used AD17 which is also installed.
-
I don't get it. It's my most requested video, and countless people cite the uSupply design series as one of the best things on my channel. :-//
It's more of a "how to not" video than a "how to" like the ones on designing a custom LCD. That's is probably harder to appreciate if you aren't actually in the business of making products and have experienced products that just never came together.
-
Strangely, this video has the biggest thumbs down ratio of all my recent videos except the LK99 one.
Are people pissed off they can't buy it?, or don't like projects?
I don't get it. It's my most requested video, and countless people cite the uSupply design series as one of the best things on my channel. :-//
My guess is that you've tantalised some people with a nerdy streak and then told them they can't have one.
-
I don't get it. It's my most requested video, and countless people cite the uSupply design series as one of the best things on my channel. :-//
It's more of a "how to not" video than a "how to" like the ones on designing a custom LCD. That's is probably harder to appreciate if you aren't actually in the business of making products and have experienced products that just never came together.
Yes, probably.
Videos about failed projects are actually rather popular - that's a genre on YT - and some Youtubers even make almost exclusively such videos. But then they show a "spectacular" failure, so that entertains people.
You know, a bit like acting like the EE clown.
Here, that's not really a failed project - the project went to completion and works - at least technically. The failure part is more the story of a few questionable decisions and failed market, more of a management/marketing issue than a technical one per se, and even if Dave explains the points, it probably takes some knowledge and experience to even understand why the project didn't become a product.
-
Dave explains the points, it probably takes some knowledge and experience to even understand why the project didn't become a product.
Not really, ultimately I just lost interest in it.
-
It was a good video, I enjoyed watching it. I like project videos, fundamentals fridays, new part / datasheet deep dives, teardown / repair videos etc. It's what attracted me to your channel.
-
I did like the video, as it was a light talk about various 'engineering choices'.
Even if the uSupply as it is won't see the light of day as final product, I think it could be an interesting base on more videos, e.g. about its evolution history (i.e. why some product changes) or how it could be optimized to potentially be further evolved to become a real product.
-
Curious about the dual FETs as well, originally I thought it would be for power dissipation but they are not used in that way. I'm sure there was some justification. Ultra-safe against overshoot maybe.
Also with a tracking pre-regulator, might be able to get away with no heatsink at all.
Not many negative youtube comments so hard to tell what people didn't like.
-
Curious about the dual FETs as well, originally I thought it would be for power dissipation but they are not used in that way. I'm sure there was some justification. Ultra-safe against overshoot maybe.
I don't recall. It might be in the design notes somewhere.
-
I do believe the reason for those downvotes are not in the video itself, but elsewhere; maybe something Dave has said on social media has riled up some activist types?
One reason I still cannot post under my own name is situations like this: when you find out there is an issue/problem/unhappiness/negativity, but no way to find out exactly what or why. That still bugs me to no end. Using a pseudonym gives sufficient buffer for me to realize it's just online social pretend stuff, and move on; if it was real, somebody would describe it.
-
I do believe the reason for those downvotes are not in the video itself, but elsewhere; maybe something Dave has said on social media has riled up some activist types?
One reason I still cannot post under my own name is situations like this: when you find out there is an issue/problem/unhappiness/negativity, but no way to find out exactly what or why. That still bugs me to no end.
Ask yourself: who is the activist in the end? You for expressing your opinion or the people reacting to your opinion?
-
I do believe the reason for those downvotes are not in the video itself, but elsewhere; maybe something Dave has said on social media has riled up some activist types?
One reason I still cannot post under my own name is situations like this: when you find out there is an issue/problem/unhappiness/negativity, but no way to find out exactly what or why. That still bugs me to no end.
Ask yourself: who is the activist in the end? You for expressing your opinion or the people reacting to your opinion?
Are you serious?
Reacting to an opinion by downvoting a completely unrelated video by the same person, is a behaviour pattern that labels you an asshole.
Ask yourself: Who is the asshole here?
-
On YouTube I see 1670 thumbs up and 69 down at 37000 views.
No idea what the normal should be, but that ratio doesn't look bad to me.
-
I do believe the reason for those downvotes are not in the video itself, but elsewhere; maybe something Dave has said on social media has riled up some activist types?
One reason I still cannot post under my own name is situations like this: when you find out there is an issue/problem/unhappiness/negativity, but no way to find out exactly what or why. That still bugs me to no end.
Ask yourself: who is the activist in the end? You for expressing your opinion or the people reacting to your opinion?
Are you serious?
Reacting to an opinion by downvoting a completely unrelated video by the same person, is a behaviour pattern that labels you an asshole.
You are missing the point. I wasn't reacting to downvoting the video but to your statement that you are not going to share your opinion because people might react to it.
About downvoting a video: who cares! Unconstructive criticism
-
I do believe the reason for those downvotes are not in the video itself, but elsewhere; maybe something Dave has said on social media has riled up some activist types?
One reason I still cannot post under my own name is situations like this: when you find out there is an issue/problem/unhappiness/negativity, but no way to find out exactly what or why. That still bugs me to no end. Using a pseudonym gives sufficient buffer for me to realize it's just online social pretend stuff, and move on; if it was real, somebody would describe it.
I understand that, and it's all due to the very wide exposure we get on online social networks, something that's very new in our history and that we certainly have never had a chance to evolve to accomodate it yet (if we ever do.)
People can get very negative and insulting - compared to what they would do "irl", sometimes as far as death threats - partly due, I think, to the disinhibition that comes with being behind a screen, and partly to the "mob" effect (if a few people start insulting someone online, then it is very likely to snowball). The mob effect has always existed, but obviously when the potential mobs go from a few/a few tens of people to potentially millions of them, this is a completely different story.
With such large numbers, from a purely probabilistic POV, you're bound to run into a few people that just want to rip you apart for absolutely no reason. That's just probabilities.
Thing is, online, many of us tend to act as though we had little to almost no inhibition, while inhibitions are what has made social interactions possible at all without society literally collapsing and everyone killing everyone else.
Some people have no inhibition of this kind due to a damage in - I think - the frontal cortex. This usually makes people who act as though they had no morals and no limits. Horrific.
The "virtual" nature of online interactions seems to trigger that to some degree.
So anyway, I don't blame people for protecting themselves from that. It can be devastating.
-
Hello
I am absolutely sorry for bumping this thread but I kept watching the video and the series and just wonder what is the target price range you deemed suitable for this? also, mind asking about bom estimated price according to the schematic of this video?
Again sorry for bumping this but this project is the most interesting topic in eevblog content to me.
thanks !
-
I am absolutely sorry for bumping this thread but I kept watching the video and the series and just wonder what is the target price range you deemed suitable for this? also, mind asking about bom estimated price according to the schematic of this video?
Sorry, it's been so long I forget.
-
I am absolutely sorry for bumping this thread but I kept watching the video and the series and just wonder what is the target price range you deemed suitable for this? also, mind asking about bom estimated price according to the schematic of this video?
Sorry, it's been so long I forget.
I mean a ballpark number not necessarily accurate, to get an idea. like maybe 150$ selling price? or does it have to be 100$ or less?
I can estimate the BOM based on the video but some stuff cannot be determined like transformer, custom case, custom heatsink, etc...
Another thing that interests me is how did you get it to be stable in its LDO loops? for me I have a similar design but uses 2 N-mosfets which I got it to be working in LTSpice but did not use any equations or compensation calculation.
So can we expect this project to get a comeback or is it confirmed dead?
-
AFAIR the supply is using LDOs and thus has no extra design of the control loop. Only the CV part is added externally - one of the problems in the design.
For me the main point to learn from this project is not to use a voltage regulator to build a lab supply. A lab supply looks simple at first, but is more complex than one may think.
One complication is that a lad supply is supposed to work with variable load - in terms of control theory this is a robust regulator. Equations from classical control theroy can be a start, but to test all load cases and the cross over between CC and CV mode simulations (both in time and frequency domain) is now the way to go.
Overall I would consider the project a failure - more an example of how not to build a lab supply.
-
AFAIR the supply is using LDOs and thus has no extra design of the control loop. Only the CV part is added externally - one of the problems in the design.
For me the main point to learn from this project is not to use a voltage regulator to build a lab supply. A lab supply looks simple at first, but is more complex than one may think.
One complication is that a lad supply is supposed to work with variable load - in terms of control theory this is a robust regulator. Equations from classical control theroy can be a start, but to test all load cases and the cross over between CC and CV mode simulations (both in time and frequency domain) is now the way to go.
Overall I would consider the project a failure - more an example of how not to build a lab supply.
I think maybe you did not follow the last videos, since he ditched lt3080 long time ago and now reverted back to using a transistor design (p-mosfet). He used 2 mosfets in series, one for cc and one for cv which may make the control simple since each one is responsible for just one aspect.
I made a similar design in ltspice but I used n-mosfets instead, with good opamp and added compensation of resistors and caps, i tested it with various loads and seems to work fine. still i want to do some ac simulation to get phase margin but thought of making it work first. I tried from 0uF up to 3000uF, high and low loads, variable loads, transitions between cc and cv, etc... and similar cases i could think of.
I suck at maths so hard, so I was trying to weasel out of doing or learning it :-DD thus i thought simulation in ltspice with lots of different compensation values is a better alternative (and later on phase margin simulation). mostly the current measurement cc loop is was the problematic but now stable (as it seems at least).
however, is there a better way to make it stable than simulation? you mentioned equations and so on, which exactly one can look into.
I was interested to check what was the scope of this project and which price target it was.
-
Having separate power stages for the CC and CV part is an odd solution: it adds const and asks for issues (overshoot) in the transition between the modes.
For the simulation a very useful case to look at is the output impedance in the frequency domain (current source as load). This way one can check if the output is too close (or even more phase shift to cause stable oscillation) an ideal inductor and thus case issues with capacitive load and at which capacitance the issue would start.
The CC mode analog is a test with a voltage source as load.
One would still need to check the CC-CV transition and back in the time domain. This is because of large signal issues from nonlinear / saturation effects that the frequency domain simulations don't show.
With the custom case the costs would likely be relatively high - still depends o lot on the volume and source for some parts.
-
Having separate power stages for the CC and CV part is an odd solution: it adds const and asks for issues (overshoot) in the transition between the modes.
For the simulation a very useful case to look at is the output impedance in the frequency domain (current source as load). This way one can check if the output is too close (or even more phase shift to cause stable oscillation) an ideal inductor and thus case issues with capacitive load and at which capacitance the issue would start.
The CC mode analog is a test with a voltage source as load.
One would still need to check the CC-CV transition and back in the time domain. This is because of large signal issues from nonlinear / saturation effects that the frequency domain simulations don't show.
the classic way of doing it is the 2 diodes oring but this one seems easy as well. i believe if the loops are stable it won't be much of a difference but you know more than me.
when you say output impedance in frequency domain you mean generate .ac analysis by putting 1v ac source at feedback loop of cc opamp and then cv opamp, then generate the bode plot and check phase margin? I designed mine to have 0-20v 0-2 amps so i need to put 1 amp current source as a load instead of a 10 ohms resistor then do bode plot?
testing with inductor loads is something i still did not do, i did not even do bode plot yet but did lots of testing with output capacitances and rapidly changing loads with changing current/voltage settings. I get very clean transition and behavior, no ringing or anything.
so you suggest i put like 100uH as a series load at the output followed by 1000uF then do bode plot?
as for cv cc transitions, i checked as mentioned but what more robust testing i can do? I mean when to expect the circuit to be good and working?
the only thing i still did not add is the anti-windup function, still do not know the best way to do it. designs i see here use a pnp with its base driven by CV opamp output and its emitter to cc output and collector to cc negative. i did not understand the way it works.
I will send you the ltspice file after 2 weeks when i return home. I made something i did not see anyone doing which is irrelevant and I can revert from it, which is instead of taking the voltage feedback resistor (90k and 10k for exactly /10) directly from the output, i took it from an opamp buffer so that it does not draw any current since i want it to be precise down to mA.
The same with current sense amp as both its inputs are buffered with opamp follower first then fed into regular current sense diff amp. yes 1mA error can be ok but i was picky and will have spare opamps anyway. tested this and really worked well and sometimes in current mode even showed less spike than regular.
With the custom case the costs would likely be relatively high - still depends o lot on the volume and source for some parts.
everything is relative.
i mean if the full bom including everything is about 20-30$ which should be the case approximately, then it can be sold at 100$ easily and maybe 150$.
assuming making 1000 of it, the case and custom parts won't have much of a negative effect.
this is why i would like to know the opinion of dave and others about the reasonable cost for such a device, and if this will ever be a thing at some day. you could say that i grew up with this project.
-
Simulation for the output impedance (in CV mode) is with a current "source" as the load. The current source would have an DC background part (one would have to use a few values (like 0 , 10% , 99 % full range) here ) and the standard AC amplitude of 1 . To "measured" AC voltage (in V) at a the output would than give the output impedance in Ohms.
A typical voltage regulator has the impedance going up with frequency from some 1 Hz to maybe some 10 kHz or 100 kHz. This range usually also has a near 90 deg. phase and thus looks like an inductor.
If this simulated inductance with a capacitor at the output one gets a resonance with ringing after transients. If the phase shift is more than 90 degree there would be oscillation.
For a good power supply one wants the phase shift to be well less than 90 deg. for the higher frequency part. It is only when the absolute value of the impedance is low enough (e.g. < 1 mohm) that near 90 deg. and even a little above can be OK as this would than be only a problem with no longer realistic capacitors at the output.
One can already use the simulated output impedance to find a suitable outout capacity (and RC elements). This is to dampen the higher frequency end.
The time domain simulations for the CV CC transitions should cover the typical (like 10-90% load) cases and also some extremes (e.g. 1% loading, worst case (from Z out simulation) and maybe 1000 µF cap (with some realistic low ESR) at the load, low and and high set voltage.
What is still acceptable is a good question. A point that gets obvious from the impedance part is that with some load capacitance there will be ringing to a certain degree.
An extra buffer at the output, to reduce the output loading from the FB divider is done with SMUs. With a lab supply this is usually overkill and can cause more harm than it helps. The current regulation is usually a bit lower priority and for the measurement one can subtract that part.
For the current measurement there is rarely a need for extra buffers, as the current shunts are usually (except for an SMU) relatively low resistance. In addition the loading from a difference amplifier could usually still be seen as parallel to the shunt and thus be included in the calibration.
-
I mean a ballpark number not necessarily accurate, to get an idea. like maybe 150$ selling price? or does it have to be 100$ or less?
I can estimate the BOM based on the video but some stuff cannot be determined like transformer, custom case, custom heatsink, etc...
Sorry I have no interesting in revisiting this.
So can we expect this project to get a comeback or is it confirmed dead?
It's dead. A have a couple of working prototypes and that's it.
-
Simulation for the output impedance (in CV mode) is with a current "source" as the load. The current source would have an DC background part (one would have to use a few values (like 0 , 10% , 99 % full range) here ) and the standard AC amplitude of 1 . To "measured" AC voltage (in V) at a the output would than give the output impedance in Ohms.
A typical voltage regulator has the impedance going up with frequency from some 1 Hz to maybe some 10 kHz or 100 kHz. This range usually also has a near 90 deg. phase and thus looks like an inductor.
I am not sure I got it correctly but please check my attachment and see. I put current source doing 1 amp (max for design is 2 amps) with ac amplitude of 1, also having what I will put as fixed output capacitance, usually just 100nf, 100pf, 10uf ceramics + 2x10 or 22uF polymer caps + optional 10 to 22uF elec cap with series resistance of say 2.2 ohms or 4.7 ohms as a general damper. Then AC simulation of 1 hz to 1 MHz.
Now where to probe and what to look for?
the ac analysis I am used to in such stuff are putting voltage source of 1v ac in feedback loop then probe output/input, then check for phase margin.
One can already use the simulated output impedance to find a suitable outout capacity (and RC elements). This is to dampen the higher frequency end.
Since I still did not understand the first step, I won''t get this one immediately. However, I want to know which values of RC to change? you mean the compensation caps and resistors as well as output caps?
The time domain simulations for the CV CC transitions should cover the typical (like 10-90% load) cases and also some extremes (e.g. 1% loading, worst case (from Z out simulation) and maybe 1000 µF cap (with some realistic low ESR) at the load, low and and high set voltage.
What is still acceptable is a good question. A point that gets obvious from the impedance part is that with some load capacitance there will be ringing to a certain degree.
I've done lots of these but not all cases, all seem to translate very nicely. even 1mA current limit works, even 2 amps to 10mA works fine. voltage change is a lot easier as well. no ringing at all, maybe some steps and sometimes spikes but not on the load side. load always clean.
An extra buffer at the output, to reduce the output loading from the FB divider is done with SMUs. With a lab supply this is usually overkill and can cause more harm than it helps. The current regulation is usually a bit lower priority and for the measurement one can subtract that part.
For the current measurement there is rarely a need for extra buffers, as the current shunts are usually (except for an SMU) relatively low resistance. In addition the loading from a difference amplifier could usually still be seen as parallel to the shunt and thus be included in the calibration.
simulation so far did not show any error, but actually noticeable improvement in current regulation because it reduced spike. the reason is getting optimum accuracy chips and components level without any software intervention. 0.02% resistors will be used and good quality adc/dac/ref... going to town with it. opamp used will be tlv4888 as it is zero offset chopper with 14mhz and high voltage supply, perfect for this.
since i am using 2 chips (8 opamps) then i will make sure to benefit from them. getting the buffered voltages seem fast and ok but you know something i don't. the regulated current now does not have an error or less accuracy at all... 1.000v gives 1.000A, of course other components will do some effect but getting everything as clean as possible also helps. you can say i want to go to town with this one.
Sorry I have no interesting in revisiting this.
fair enough, but maybe you can give your opinion as an engineer interested in the market and such devices, what is the good range of price which seems reasonable for you?
It's dead. A have a couple of working prototypes and that's it.
my heart is crying in the corner, will need time.
again, no one have the right to ask you for more or even publish the documents of final design. i can speak that MANY people got interested in eevblog or similar content due to this particular project.
I myself loved the idea of power supplies and learned to design electronics by watching your videos back in 2009 when I was in university... and now I am selling my own products which are power supplies as well. uSupply without a doubt was the spark that made this for me.
Always thanks for the content.
-
Simulation for the output impedance (in CV mode) is with a current "source" as the load. The current source would have an DC background part (one would have to use a few values (like 0 , 10% , 99 % full range) here ) and the standard AC amplitude of 1 . To "measured" AC voltage (in V) at a the output would than give the output impedance in Ohms.
A typical voltage regulator has the impedance going up with frequency from some 1 Hz to maybe some 10 kHz or 100 kHz. This range usually also has a near 90 deg. phase and thus looks like an inductor.
I am not sure I got it correctly but please check my attachment and see. I put current source doing 1 amp (max for design is 2 amps) with ac amplitude of 1, also having what I will put as fixed output capacitance, usually just 100nf, 100pf, 10uf ceramics + 2x10 or 22uF polymer caps + optional 10 to 22uF elec cap with series resistance of say 2.2 ohms or 4.7 ohms as a general damper. Then AC simulation of 1 hz to 1 MHz.
Now where to probe and what to look for?
The point to look at is the output voltage. The scales shows volts and with 1 A of current amplitude (linear small signal based) the scale also gives ohms for the output impedance.
Usually one would do the simulation at least initially with only minimal impedance at the output. There is no need to seperately test with extra output capacitor cases. This makes the impedance mode so practical. One can still use it to test / tweak the lossy capacitance part (e.g. capacitors with series resistance) to suppress issues at higher frequencies. Quite often some lossy capacity part (like the 10 to 22 µF + 2 to 5 ohms) is needed one can see the effect quite well in the impedance.
Steps in the set voltage are usually not a relevant part. One may still want to check the trun on part that may be similar. It is not likely to fail and can thus be a relatively late step.
-
The point to look at is the output voltage. The scales shows volts and with 1 A of current amplitude (linear small signal based) the scale also gives ohms for the output impedance.
Usually one would do the simulation at least initially with only minimal impedance at the output. There is no need to seperately test with extra output capacitor cases. This makes the impedance mode so practical. One can still use it to test / tweak the lossy capacitance part (e.g. capacitors with series resistance) to suppress issues at higher frequencies. Quite often some lossy capacity part (like the 10 to 22 µF + 2 to 5 ohms) is needed one can see the effect quite well in the impedance.
Steps in the set voltage are usually not a relevant part. One may still want to check the trun on part that may be similar. It is not likely to fail and can thus be a relatively late step.
ok so i do ac analysis with 1v ac amplitude and 1a dc value as shown from 1 hz to 1 or 10 mhz? then probe the output or write vout/iout? can you give a picture showing such graph? i was dumb enough to leave home without taking my spice file!!
so what would the graph show exactly? phase and gain like typical ac analysis? the thing i am not understanding the most is the impedance at the output since the 1v 1a value of the current source are constant.
a picture example would be fine, then i will try to look for more myself. i am learning on youtube right now about these topics as much as i can.
-
i mean if the full bom including everything is about 20-30$ which should be the case approximately, then it can be sold at 100$ easily and maybe 150$.
assuming making 1000 of it, the case and custom parts won't have much of a negative effect.
this is why i would like to know the opinion of dave and others about the reasonable cost for such a device, and if this will ever be a thing at some day. you could say that i grew up with this project.
He had a poll on the pricing: https://www.eevblog.com/forum/supporters-lounge/straw-poll-usupply-pricing/ (https://www.eevblog.com/forum/supporters-lounge/straw-poll-usupply-pricing/)
You can see based on the range his target was ~$100-150. Then add a bit for inflation, we are talking $130-170 today.
Selling 1000 is likely difficult as there are similar products on the market (eg Fnirsi DPS150).
-
i mean if the full bom including everything is about 20-30$ which should be the case approximately, then it can be sold at 100$ easily and maybe 150$.
assuming making 1000 of it, the case and custom parts won't have much of a negative effect.
this is why i would like to know the opinion of dave and others about the reasonable cost for such a device, and if this will ever be a thing at some day. you could say that i grew up with this project.
He had a poll on the pricing: https://www.eevblog.com/forum/supporters-lounge/straw-poll-usupply-pricing/ (https://www.eevblog.com/forum/supporters-lounge/straw-poll-usupply-pricing/)
You can see based on the range his target was ~$100-150. Then add a bit for inflation, we are talking $130-170 today.
Selling 1000 is likely difficult as there are similar products on the market (eg Fnirsi DPS150).
Thanks for the link, looks like even I voted there.
well 100$ seems the magical number, but 150$ is upper limit if the features are worth it.
I read that fnirsi is not as good, it is switching not linear too. still it has no battery for example where this one can easily integrate it as well as being lower noise.
but for selling it at 100-150$ the total bom must be in the range of 20-40$ or so. judging from the parts i believe he can achieve it, but we do not know the price of custom parts which can make or break it.
ditching isolation, customized binding posts, etc... can save a few.
-
Fnirsi is definitely worse, but I'm not sure what percent of people would care.
These days USB PD 20V power banks are common, not sure how much value internal battery is. Would be cool to be ultra compact though. A good 18650 would handle 50W+.
The custom parts aren't too expensive, what was the LCD $1-2 or something? But the NRE can add up to many thousands.
-
Fnirsi is definitely worse, but I'm not sure what percent of people would care.
These days USB PD 20V power banks are common, not sure how much value internal battery is. Would be cool to be ultra compact though. A good 18650 would handle 50W+.
The custom parts aren't too expensive, what was the LCD $1-2 or something? But the NRE can add up to many thousands.
which is why I asked to know for sure.
I've been planning a design of my own which will have internal battery (4 18650 in series) + pre-reg buck stage and ldo stage going down a lot + precision parts everywhere + USB-PD 20v input + buck-boost initial stage with NVDC to ensure optimum performance with and without battery + TFT display...etc. I have initial plan but I need first to get LDO stage verified after I return home. i will go with this step by step first make small board for ldo stage and go from there.