EEVblog #1031 - $25 DPS3003 PSU Module Characterisation

[EEVblog]

Characterising the performance and efficiency of the $25 DPS3003 30V 3A lab power supply module from RDtech.
Includes a how-to on measuring DC-DC converter efficiency graphs, with module power loss on the secondary Y axis.

https://www.aliexpress.com/store/product/RD-DPS3003-Constant-Voltage-current-Step-down-Programmable-Power-Supply-module-buck-Voltage-converter-color-LCD/923042_32685187020.html

XL7005 datasheet: http://www.xlsemi.com/datasheet/XL7005A%20datasheet-English.pdf

[w33m4n_]

just a short tutorial for the user interface. after you missed the fast adjust for the unit.
its from julian ilett on a simular converter.

https://youtu.be/mJqPkGe7DzA?t=5m19s

starts at 5 min 19 sec.

[strangersound]

The Signal Path via Hackaday is just in time:

https://hackaday.com/2017/10/10/cleaning-up-a-low-cost-buck-boost-supply/

[ez24]

The Signal Path via Hackaday is just in time:

https://hackaday.com/2017/10/10/cleaning-up-a-low-cost-buck-boost-supply/

This is the first video I have seen that shows a way to reduce ripple from a switcher.  (I asked a year ago)
According to AoE, LC filters do not work, they recommend linear regs.  SP says he did not use one because of variable voltage.  I wonder if a way could be used using LDOs.  Maybe a ganged pot?

Another question

Does the ripple really matter?



Does anyone know of a var switcher using a LDO so there is less voltage drop?

[Brumby]

Another question

Does the ripple really matter?

It depends on the circuit being powered.  With some simple LED lighting it won't matter - but I wouldn't like to run a low level audio amplifier or an ADC directly from one of these.

[EEVblog]

The Signal Path via Hackaday is just in time:

https://hackaday.com/2017/10/10/cleaning-up-a-low-cost-buck-boost-supply/

This is the first video I have seen that shows a way to reduce ripple from a switcher.  (I asked a year ago)

It's a very poor one though. In this case 8W wasted in the opamp (4V drop x 2A), and the nice voltage control on the module is now useless because it does not reflect the output.

According to AoE, LC filters do not work, they recommend linear regs.

If you want very low noise, then sure.
But there is a lot to be said for just adding a simple LC filter, so they do "work".

[KuchateK]

FFcossag experimented with different coils and caps to filter noise from cheap switching PSU. Spoiler (this is very long video): common mode choke with couple ceramic caps on the output reduces noise significantly.

He used 3.9mH 8A rated common mode choke from Epcos and 100nF + 6.8nF caps. Without filter his PSU had 120-500mVpp of noise (tested at various currents). After mod he went down to 20-70mVpp.

[FrankBuss]

This looks really painful to manually write down and enter all these numbers. I know you are not a software person, but I recommend that you install something like the Raspberry Pi logging platform, with which you can connect pretty everything which has USBTMC, GPIB etc. Then you can write a simple Python script to control your power supply and read the measurements of a multimeter, and write it in a CSV file. There are examples for my SPD3303D power supply and for lots of benchtop multimeters for copy-and-paste. Might need some time before you got it all working, but then you save time whenever you want to do it again, and you wouldn't even think "no, that's too much work, I don't want to write all the numbers down manually again", which might result in more interesting characteristics diagrams on your Youtube channel

[kalel]

This looks really painful to manually write down and enter all these numbers. I know you are not a software person, but I recommend that you install something like the Raspberry Pi logging platform, with which you can connect pretty everything which has USBTMC, GPIB etc. Then you can write a simple Python script to control your power supply and read the measurements of a multimeter, and write it in a CSV file. There are examples for my SPD3303D power supply and for lots of benchtop multimeters for copy-and-paste. Might need some time before you got it all working, but then you save time whenever you want to do it again, and you wouldn't even think "no, that's too much work, I don't want to write all the numbers down manually again", which might result in more interesting characteristics diagrams on your Youtube channel

One alternative might be to enter the numbers directly into a table instead of writing them down. Not comparative to the automatic method, of course.

That made me think of a simple question, would it be possible to use text recognition software(or custom software for those with the know how) and a webcam attached on a stand to record data (any data) from a cheap multimeter that doesn't have an external interface? Speed shouldn't be an issue with cheap multimeters, as usually they have 2-3 updates per second. A cheap webcam should be able to get at least a few FPS more than that even in bad lighting conditions and worst case scenarios, and manual focus should allow to get reasonably sharp numbers. A USB microscope and stand could probably be used if it can "see" the entire number.

[EEVblog]

This looks really painful to manually write down and enter all these numbers.

Not really, it's very quick when you get in the swing of it, maybe 5 seconds per data point. Or just a couple of minutes per curve.
Data entry is pretty quick, again a few minutes per curve.
So I'd wager that I can fully charcterise a converter by hand in less time than it takes to even figure out an automated solution, let alone implement it.
Yes, if you were going to do this over and over again with different converters then it might be worth it, but otherwise manual is juts fine.

[manu]

Andreas Spiess does a low cost SMU based on this device (which has a software control and bluetooth option) and a low cost electronic load: https://youtu.be/QxR-_ZnREQQ?

[hli]

Just stumbled upon this article: https://www.edn.com/design/power-management/4458920/Designing-second-stage-output-filters-for-switching-power-supplies-
It also has some calculations for RC and LC filters, and explains what the pitfalls are (at least some of them).

[RGB255_0_0]

The Signal Path via Hackaday is just in time:

https://hackaday.com/2017/10/10/cleaning-up-a-low-cost-buck-boost-supply/

Someone mentions using a Pi filter instead. What's that?

[tszaboo]

This looks really painful to manually write down and enter all these numbers.

Not really, I used to do these kind of measurements a lot. You just enter a different state of mind, and do it. I don't bother bringing up a script, unless I need say 300+ points. Or the measurements are for production, and it will be done by someone else.

[Brumby]

Someone mentions using a Pi filter instead. What's that?

Google is your friend.

It is simply two capacitors and an inductor placed in the shape of the Greek letter pi



R_L is simply the load.

[FrankBuss]

This looks really painful to manually write down and enter all these numbers.

Not really, it's very quick when you get in the swing of it, maybe 5 seconds per data point. Or just a couple of minutes per curve.
Data entry is pretty quick, again a few minutes per curve.
So I'd wager that I can fully charcterise a converter by hand in less time than it takes to even figure out an automated solution, let alone implement it.
Yes, if you were going to do this over and over again with different converters then it might be worth it, but otherwise manual is juts fine.

Right, for one measurement it is faster to do it manually, but you did such measurements already a few times, and I'm sure it won't be the last time, so it might be worth to invest once the time to automate it. I guess would be piece of cake for David2, who is more into software, and maybe even worth a EEVacademy video, for anyone who wants to know how easy it is to use Python to automate things. Could be even combined with the WiringPi library to control GPIOs for switching external muxes etc.

[metrologist]

 

[Cerebus]

So what is confusing you?

[metrologist]

I didn't mean it literally. Maybe just another filter to try.

[maelli]

Losses @ 1.5V are higher:
I guess this is because the more-lossy-than-the-FET freewheeling diode (Schottky or not) is most of the time conducting at low voltage.
Verly short duty cycle on the FET, long duty cycle on the diode.

[Zad]

Does anyone know of a var switcher using a LDO so there is less voltage drop?

Linear published an interesting design in their in-house magazine "LT Journal of Analog Innovation" July 2014

http://cds.linear.com/docs/en/lt-journal/LTJournal-V24N2-2014-07.pdf

(True 0-24V, 3A)
It uses an LT8612 buck reg followed by a pair of LT3081 linears (drop circa 1.25V) configured so they both track the demanded voltage.

[Kleinstein]

Having a linear regulator following the SMPS part is a good idea to reduce ripple. It would be even better with a LC filter in between.
 Just using a filter is tricky, as the filter will also add some voltage drop from residual resistance and slow down the regulators response. So regulation would suffer from a significant filter. A small filter to get rid of some higher frequency part can be still a good idea.

The combined SMPS + linear regulator should however ideally be controlled by the linear part first and than set the SMPS to keep an more or less constant drop oupout. Just adding a linear regulator for a constant drop is kind of ugly, as the output of the SMPS is not as exact as the ser point.

[Brumby]

All we need is to have a remote sense input for the regulator's output, so it can monitor the voltage on the other side of any filtering.  Default would be to have a link connecting this directly to the output from the regulator.

[KE5FX]

It's not that LC filters don't work -- they obviously do -- but the usual practice of grabbing any old inductor and any old capacitor you have lying around is unlikely to help.  If you can't verify the effect of an LC filter, don't add it.  In the frequency domain, they don't reduce broadband noise as much as they tend to herd it into a small portion of the spectrum near their corner frequency,  They can be effective at suppressing ripple at a specific frequency, but be sure to do the math or at least the measurement to verify that you've improved matters.

Also, always check the transient behavior during power on/off with a scope.  You may find you're generating a spike that can hose your downstream devices.

[nuno]

I wonder if these modules/module-based-PSU would pass FCC/CE/... EMC compatibility tests... and if not, what changes would be necessary to make them pass.

[Kleinstein]

Adding a filter inside the regulator loop is limited. Too much filtering would make the control loop unstable. In a lab supply like application there usually is not much phase reserve anyway, at least not with an unfavorable load. The main part of extra filtering would thus be filtering the sharp edges from switching and if present ringing of the switch - the lowest ripple frequency would be largely unaffected. Slightly more filtering might be possible if it is considered in the loop filter of the SMPS, but changing things there would be rather difficult.

To really get rid of the ripple it would take more LC type filtering outside the loop and a linear post regulator to get back full load regulation. At the rather high frequencies the PSRR of the linear regulator is likely limited, but it also helps.

I very much doubt the simple PSUs would pass FCC/CE EMC rules. On the input side one might be lucky and the raw supply (e.g. a notebook supply) has enough filtering.

[hli]

I tried to get a LC filter at the output working, using the link I provided above. This resulted in a 10µH coil, 220µF for the first cap and 22µF for the second (although this one could be smaller from the formulas). It did work quite well to reduce the ripple from the switcher (at 67kHz), but there is an additional 1kHz ripple super-imposed on the output. And that one I could not get reduced to where I wanted it.
The scope signals (using a 47µF coil and two 220µF caps, in addition to a damping RC-combination) look like this (filter input at the bottom, output at the top) are attached. The DPS5005 was set to 5V, with a 700mA load.
It might be that the control loop of the module runs at this frequency.

[technogeeky]

I tried to get a LC filter at the output working, using the link I provided above. This resulted in a 10µH coil, 220µF for the first cap and 22µF for the second (although this one could be smaller from the formulas). It did work quite well to reduce the ripple from the switcher (at 67kHz), but there is an additional 1kHz ripple super-imposed on the output. And that one I could not get reduced to where I wanted it.
The scope signals (using a 47µF coil and two 220µF caps, in addition to a damping RC-combination) look like this (filter input at the bottom, output at the top) are attached. The DPS5005 was set to 5V, with a 700mA load.
It might be that the control loop of the module runs at this frequency.

I haven't actually attempted to repair the noise from these modules yet (I'll be looking into this very soon, I hope). But I wanted to point out a few things:

There are at least two noise problems:

• the noise imposed on the load (which you are combatting)
• the noise sent back toward the input power source

Obviously this first kind of noise has to be dealt with no matter what, if it's important to you to reduce the noise. But if you want to make more sophisticated power supplies using this module (for instance, I am trying to combine two units to have either a pair of isolated supplies, or a single tracking (or not) bipolar supply). In my case, I'm using a big audio transformer to provide the two isolated supplies and this means all of the noise that is sent back from one DPS5005 is picked up by the transformer and sent into the second DPS5005.

I don't have much to say about this yet, other than I presume at the very least the standard double choke setup (2 capacitors, 1 common choke, 2 capacitors, 1 common choke, 2 capacitors) is probably a good starting point here. I plan to put one of these between each power supply and the transformer -- and one between the transformer and the wall input.


As for the second kind of noise (applied to the load)...

1) It might be worth it to see where exactly one could bodge in filtering capacitors and chokes inside the supply. Any filtering that can be done in-situ is far more effective than any filtering done after the fact. I wonder what kind of absurd things are possible, like lifting legs of ICs and putting ferrite beads over them. I know there are critical current loop paths that are fundamental to reducing noise, and I don't know if these critical paths were analyzed in the design of the modules.

2) More importantly: the noise of these supplies changes. The worst noise comes when V_out is at the extremes: near zero, or near V_in. In the next week or so I'll try to get captures demonstrating this. I am building my supplies to maximize the range of the unit, so I'm feeding them with 55-57V input. This is probably a bad idea for lots of reasons, but I have a bunch of these modules so I'm not to afraid to break some eggs to make an omelette.

[alex-sh]

I wonder if anyone manage to get rid of ripples for this PSU? I am feeding it with a quality 24V PSU and yet I do see some ripple