Low ripple low noise small power supply design

[gnuarm]

oscilloscope wise.. getting my dream DS1054Z will cost me 461$ from amazon but without customs which may get 100$ more... while from a proxy service ubuy will cost me totals including customs 421 JOD which is 594$ not so huge but not so little. Isolated Owon will cost 150$. the DOS1102 guy still didn't respond and I am getting frustrated.

The only think available locally is this: https://mikroelectron.com/Product/PICOSCOPE-2204A-MADE-IN-UK/ which is 100 JD (145$ or so) for just 10 Mhz with 10mv per div.

For what you are doing now that is just fine.  You say you want the noise and ripple as low as possible and so need a good scope to measure it.  But to sell a product you only need the noise and ripple low enough to work in the application.  i think that is going to allow noise and ripple at levels easily seen by the 10 MHz Picoscope.  The idea of getting the levels below 1 mV are not useful really unless you had sensitive analog circuitry which does not include standard video signals. 

[Vovk_Z]

what is you suggestion then?

It was at post #318.

[VEGETA]

what is you suggestion then?

It was at post #318.

you said:

I think you using C38, C38 1nF caps wrong. Remove them from gates, and put them between 1-2 and 67 legs of LM358.
...
And you may try lower R16, R17 a bit (22-47 R). But this is not the main problem.

1nF caps removed from v1.1 and gonna physically remove them from current boards. putting it at opamp 1-2 and 6-7 legs didn't do much without 220uF at final output rail. I tried putting 22nF instead of 1nf at opamp 1-2 and 6-7 but didn't work too.

lowering R16 and R17 is not that important as you said, not the problem.

so now we need a solution other that bulk elec. caps at output, if you have one.

[Vovk_Z]

Is there a last schematics?

[VEGETA]

Is there a last schematics?

i have sent it to you via PM.

i didn't lower 100 ohm resistors though.

[BrianHG]

The only think available locally is this: https://mikroelectron.com/Product/PICOSCOPE-2204A-MADE-IN-UK/ which is 100 JD (145$ or so) for just 10 Mhz with 10mv per div.

I don't know... Never heard of picoscope.  I would start a thread here on EEVblog and ask around first.  There exist many true lemons when it comes to these tiny devices.

[Vovk_Z]

as for your previous question on requirements, i wanted low noise low ripple psu for Sega Dreamcast. I would be happy with less than 10mV total ripple+noise.

If less than 10 mV noise is ok then you don't need any linear post filtering at all. Just a small LC output filter could be fine.

[gnuarm]

as for your previous question on requirements, i wanted low noise low ripple psu for Sega Dreamcast. I would be happy with less than 10mV total ripple+noise.

If less than 10 mV noise is ok then you don't need any linear post filtering at all. Just a small LC output filter could be fine.

I agree.  This project is what happens when you don't have proper requirements.  Better is the enemy of good enough. 

I believe the progression was,

• I need a cleaner PSU output.
• I need a capacitance multiplier
• Add op amp feedback to that capacitance multiplier
• Change the NPN to an nFET because it's lower impedance

Nowhere in this process was the imagined result compared to what was needed.  If some is good, then more is better and too much is just enough.

Not trying to harp on things, but I'm working on a volunteer ventilator project with virtually no requirements and no understanding of requirements analysis.  A revision 1 of the pcb was completed in a couple of months.  Revision 2 is at the 9 month mark.  I think I'm suffering from PTSD.

[gnuarm]

The only think available locally is this: https://mikroelectron.com/Product/PICOSCOPE-2204A-MADE-IN-UK/ which is 100 JD (145$ or so) for just 10 Mhz with 10mv per div.

I don't know... Never heard of picoscope.  I would start a thread here on EEVblog and ask around first.  There exist many true lemons when it comes to these tiny devices.

For USB attached scopes Picoscope is top notch.  That's why a 10 MHz unit is $145 rather than well under $100 like the crap devices.

I remember back in the day of analog scopes the real separation between the good scopes and the so-so scopes was their triggering.  I think that is still true.  Unfortunately it is harder to specify the triggering specs, not that it would matter much.  Makers lie about all the other specs, triggering would just be another one to lie about.

There's a thread here somewhere for a new scope company that might actually be producing a decent, low cost product.  Over a year ago the guy was posting about a new lower end product like this one.  I think now he is selling a better spec devices for a price like this one.  Here is a price sheet the guy posted.
https://drive.google.com/drive/folders/178z_YfgW_qpdFXZxuJiwVYOVIO74Y6BA

Here is one of the threads where Loto is mentioned. https://www.eevblog.com/forum/testgear/how-to-promote-an-usb-oscilloscope-to-individual-engineers-i-am-the-loto-instru/?all]
[url]https://www.eevblog.com/forum/testgear/how-to-promote-an-usb-oscilloscope-to-individual-engineers-i-am-the-loto-instru/?all[/url]

[VEGETA]

update: I managed to have an offer for Hantek 6022BE very lightly used for 100 JOD (140$). this is the only thing available to me now. it is the same as that 10Mhz picoscope.

the hantek doesn't support ac coupling, i guess it is important to have ac coupling for our use.

which one should I get? since they are the same price.

[gnuarm]

I don't know what you mean the Hantek is the same as the Picoscope.  I have a 6022 Hantek that I never got to work because the software would not run.  They have no support and their products are universally deprecated.  I have the version with the logic analyzer built in and paid only $80 for it. 

If I knew where it was I would send it to you. 

My recommendation is to get pretty much anything other than the Hantek.

[VEGETA]

I don't know what you mean the Hantek is the same as the Picoscope.  I have a 6022 Hantek that I never got to work because the software would not run.  They have no support and their products are universally deprecated.  I have the version with the logic analyzer built in and paid only $80 for it. 

If I knew where it was I would send it to you. 

My recommendation is to get pretty much anything other than the Hantek.

I meant they both have 2 channels and same price. actually the picoscope was 180JD (254$) and now on offer to be 100 JD (140$). it supports ac and dc coupling 10mv\div sensitivity @ 10mhz but the hantek is 20mhz with dc coupling only.

so you suggest i go with picoscope?

[BrianHG]

but the hantek is 20mhz with dc coupling only.

Get a .1uf cap and place it at the end of the probe.
But I get the drift, why the difficulty in avoiding a single relay which would be needed to switch between AC&DC input.

[gnuarm]

update: I managed to have an offer for Hantek 6022BE very lightly used for 100 JOD (140$). this is the only thing available to me now. it is the same as that 10Mhz picoscope.

the hantek doesn't support ac coupling, i guess it is important to have ac coupling for our use.

which one should I get? since they are the same price.

"Bandwidth" is often the subject of lies. 

The Hantek samples at 48 MSPS, the Picoscope at 100 MSPS.  Which do you think will work better?  What happens when using two channels? 

Read a few reviews.  This is your money.  Spend it wisely based on facts, not other people's opinions.  I have my own biases. 

[VEGETA]

Looks like I will be getting the picoscope. 10mhz will be good enough for now, and it has ac coupling too unlike the hantek. hantek price is too much too.

on a side note, we decided 220uF to be the go-to value since it was shown to be stable with 100uF. However, I ran another test and the circuit got to be stable with about 17uF elec caps only (6.8u + 6.8u + 3.3u) on 5v rail and (6.8u + 6.8u + 6.8u) on 3.3v rail for the 1.6 ohms load. meaning, I could use smaller value like 22uf or 47uf. I still want to get 220uF if I got small enough formfactor though. won't make a final decision until I see the FULL circuit with all components since I am still hoping it will work better.

Then I can test various caps values.

[VEGETA]

I finally got all parts.

Unfortunately, I couldn't solder the switchers properly. I have an air dryer, not a heat gun... despite being very hot, it wasn't enough to melt the solder paste that I got locally (liquid and gel type).

I managed to solder one switcher IC somehow using soldering gun and the result is 7.1v on the 5v output for some reason. I later on lost this IC in solder mania. even though it outputted 7.1v, it still produced that annoying sound which needed an electrolytic cap to disappear.

Maybe tomorrow I could remove the solder mask from switcher pin 9 (thermal pad) to be able to solder it using soldering iron.

the weird thing on this board which had the soldered switcher, is that the 5v regulator outputted 7.1v exactly as the feedback pin on the opamp but when removing the switcher it outputted 5v normally.

I started to get really frustrated on this project after all that time and many hardships + my big fatal mistake on making many boards at once without testing it.

I made another version using the same switching circuit + LC filters but with very simple off-the-shelf linear regulators as seen in attachments. I also layed out the board (same switching parts). but if the switching part is the problem, then maybe I should consider something else??

I will keep you updated on tomorrow's trials but also I would like to hear your opinion on making another version of this project which is simpler and gonna 100% work.

required specs are:

1- straight forward design that is guranteed to work, no homebrew linear stuff. off-the-shelf parts.
2- < 10mV P-P total noise and ripple. Preferably less than that if possible without significantly more costs.
3- can be assembled using cheap PCB assembly services or manually.
4- 50x50mm board size exactly as this one.
5- preferably without heatsink or linear parts if possible.

^ these were clear specs to work with now, and I am fueled by my anger this time to make a better one. one that is clearly better than any other solution on the market.

I am sorry to tire you all with me all that time, but I really learned the very hard lesson from this.

[VEGETA]

I started also a separate one using TPS62913 which is going to be in production soon as TI confirmed, pre-production parts are available and identical as they confirmed. this one delivers in the uV range total noise and ripple as in their datasheet.

I layed the board exactly as their app note and datasheet says, but only one rail for now as seen in picture below.

JLCPCB won't be able to assemble this as they don't have the TI one in their library but PCBWay for sure can, dunno about price for just 5 assembled boards will be.

so now we have this one here to go with or your suggestion.

always thanks for your support and sorry again for dragging you with me this far to get frustration.

[gnuarm]

Hi, good to see you back.  I have to say one thing I learned from the exercise with you previously was that while I'm still not impressed with a capacitor multiplier, it did one thing well.  That is handle higher frequency ripple which an IC based linear regulator can't do because of the limited bandwidth of the amp inside.  That would have been the same problem with your op amp based design since the op amp has limited bandwidth. 

So a linear following the switcher can be effective if it just uses the transistor and not the op amp.  That circuit would remove the ripple very well, but not maintain a stable output from load variations so well.  I don't see that in your requirements, so I don't know if this is an option. 

But it sounds like you want nothing to do with linear stages.  So...

Does TI say how their chip works to reduce the ripple so much?  When did they say they would have samples and when would they have production?  Actually, it looks like they have samples now.  You can buy up to 50.  Why not get a hand full and see how the prototypes work?   The data sheet looks pretty wild.

[VEGETA]

Hi, good to see you back.  I have to say one thing I learned from the exercise with you previously was that while I'm still not impressed with a capacitor multiplier, it did one thing well.  That is handle higher frequency ripple which an IC based linear regulator can't do because of the limited bandwidth of the amp inside.  That would have been the same problem with your op amp based design since the op amp has limited bandwidth. 

So a linear following the switcher can be effective if it just uses the transistor and not the op amp.  That circuit would remove the ripple very well, but not maintain a stable output from load variations so well.  I don't see that in your requirements, so I don't know if this is an option. 

But it sounds like you want nothing to do with linear stages.  So...

Does TI say how their chip works to reduce the ripple so much?  When did they say they would have samples and when would they have production?  Actually, it looks like they have samples now.  You can buy up to 50.  Why not get a hand full and see how the prototypes work?   The data sheet looks pretty wild.

hello and thanks for your follow.

I have a low ripple low noise requirement, a total of <10mv p-p.

What do you think about my 1084-based design? also the TI one?

I messaged TI and they said this IC will be available in late march, which is maximum of 2 weeks. the pre-production ones are usable though, same package and same footprint. so I could use the pre-production one in prototype PCBs for now using PCBway right?

check TI datasheet and app note, here is what is in datasheet:

Low output 1/f noise < 20 μVRMS (100 Hz to 100kHz)
•  Low output voltage ripple < 10 μVRMS after ferritebead
•  High PSRR of > 65 dB (up to 100 kHz)
•  2.2-MHz or 1-MHz fixed frequency peak currentmode control

and see figures of 2nd stage ripple and noise, it is well below 0.5mv total!

here is an evaluation unit: https://www.ti.com/lit/ug/slvubx2a/slvubx2a.pdf?ts=1615930746111&ref_url=https%253A%252F%252Fwww.ti.com%252Fproduct%252FTPS62913

it is a new design of this chip which uses ferrite bead inside compensation loop itself to achieve this. the IC is tiny though but can deliver 3 amps. i guess 3 amps is enough for 3.3v rail and more than enough for 5v rail.

[gnuarm]

I don't know anything about the 1084 design.  The TI device is interesting.  I haven't read enough of it to understand how it works.  If you use any of these I recommend you read the data sheet thoroughly and make your design as close as possible to the reference design they provide.

Also be aware that their testing is done under perfect conditions of constant load, etc.  Load regulation is often as important as removing ripple.  Does the TI data sheet show anything regarding load transient regulation?

[VEGETA]

I put a link for a design of mine using 1084 regulator, check it above.

anyway, the TI part uses 2nd pole\stage internal compensation to achieve such low values of noise and ripple. datasheet says < 25uV of total RMS noise and ripple which is insanely good. assume my layout wasn't perfect and caused this to double, it will be 50uV or let's say it was a bit more worse, 100uV... this is still extremely nice.

I didn't find enough data on load regulation in their datasheet but they mentioned that I should do good layout to prevent bad load regulation which I will anyway.

I am gonna continue with this TI part if you don't have a better alternative. Last night I kept doing the layout of the board but didn't finish it. Once I finish it, I will review the layout with you here.

[VEGETA]

I managed to solder the 5v switcher properly and when mosfet is there it works (+ filter cap to remove sound). However, when I put 1.6 ohms load it drops to 4.750v instead of 5.050v when no load.

I also soldered a 3.3v switcher on another board somehow and it outputted the required 4.6v, but this board got ruined again.

[gnuarm]

What is the input voltage to the 5V switcher when it is under load?  Did you get your scope yet?  You can view the waveforms across the high current components and see what is happening.  A switcher should regulate normally until the transistor is saturated. 

Is this a no name switchers from eBay or Aliexpress?  They are often not nearly up to the specs they claim.

[VEGETA]

What is the input voltage to the 5V switcher when it is under load?  Did you get your scope yet?  You can view the waveforms across the high current components and see what is happening.  A switcher should regulate normally until the transistor is saturated. 

Is this a no name switchers from eBay or Aliexpress?  They are often not nearly up to the specs they claim.

No scope yet.

I tried for hours and hours last night but no use, soldering it is difficult. I succeeded twice but somehow it got ruined again. I will try next week when I return.

I got this AOZ1284 switcher from LCSC.

input voltage to switchers are the 12v source, dunno if it changes with load but it shouldn't.

[gnuarm]

I don't think you realize how little information you provide when you answer a question.  Of course the input to the switcher is a "source".  Is it another power supply? a battery?  Can't you measure the voltage with a meter when the switcher is drooping?

I'm asking the question to try to diagnose what is wrong.  That means details are needed.  Your 1.6 ohm load will draw over 3 amps on a 5V source.  Is the circuit rated for that?  I'm not asking about the data sheet for the chip.  I asking about the design.

It's going to be very hard to diagnose a problem that makes a switcher droop.  First thing to try is other loads.  Find out if it is drooping at lower currents like 2 amps, 2.5 amps, 1 amp...  If it is only drooping above some current the current limit must be kicking in.  If the droop depends on the current over a wide range, then something is wrong that is interfering with the regulation.