DC DC converter

[uneumann]

I recently played with one of those converter modules that are widely selling (got mine on ebay ) that produces up to 390v DC from DC supplies. Here is a link to one example - they seem all to be alike.
https://www.banggood.com/DC-DC-Boost...r_warehouse=CN

I'm feeding it with 24v and getting 350v at 20ma - 7 watts no problem - except for the 30-40 volts of sawtooth ripple.
The ripple persists even with a 20uf filter cap. Does anyone have a positive experience with these modules?
I'm wondering if mine is defective. Of course, I am getting what I paid for, but I was hoping for something under a volt of ripple.

Does anyone know of other modules or simple schematics for a similar function?
I ran into a post by Mike Bailey for a similar supply, but it lacked coil specs.

Thanks for any pointers

[R.G.]

30-40v sawtooth ripple? What frequency is the ripple?

[uneumann]

30-40v sawtooth ripple? What frequency is the ripple?

The freq varies - but it's below 100Hz. I've scoped the input, it's clean (a bench supply).
The ripple level varies as I adjust the output voltage or the input voltage (all tests with a fixed 20K load).
I note the best behavior when I set the output below 220v or the input around 10-12v. Those settings make for low ripple (~200mv) but as I deviate from those settings it suddenly jumps to "high ripple" mode and while the level and freq varies, it stays high - I'm calling 10-50v ripple high.

[The Dude]

It's a switching supply running at about 75Khz per specs, so you shouldn't have much low frequency ripple at the output. Do you have the primary ground completely isolated from the secondary ground?

[uneumann]

It's a switching supply running at about 75Khz per specs, so you shouldn't have much low frequency ripple at the output. Do you have the primary ground completely isolated from the secondary ground?

The grounds are connected on the PCB, so they're not isolated. I can't hear 75KHz, but I can hear buzzing when it goes into ripple mode. The buzz seems to be the ripple freq. It's clearly a sawtooth ripple on the scope. It's not the switching freq. There is no output ripple spec (can't expect much for $5) and it may just not be a problem for intended apps which seem to be "cap charging". I was just hoping to make a HV PS for a tube preamp with this. Likely wishful thinking, but I was hoping that someone had tried these with more success.

[trobbins]

It seems to be a flyback configuration, and they have scratched off the IC labels (not an uncommon practice).

It seems like a switchmode controller regulation issue - perhaps from too noisy a feedback connection on the pcb.

You could try adding a more effective output filter cap - eg. a somewhat smaller electrolytic (4.7uF 400V are quite common in small universal USB supplies) - and connect it directly between the US3M diode pad and the transformer 0V leg (that may be the furthest 0V pad, as the other 0V pad may be for another unused winding). This could be worthwhile, as the 0V return path from the 10uF 400V cap appears to be past the regulator somehow and is a reasonably large loop.

It's difficult to tell how well the input main current path and switching loop is laid out, and whether the regulator track layout is prone to noise (the regulator furthest from the FET will likely just be used for voltage sensing and shutdown).

I have had good results from a 150W dc/dc module aimded for 12V input, and with a few secondary taps. It doesn't use output voltage feedback regulation, but that is fine for valve amps.

[uneumann]

It seems to be a flyback configuration, and they have scratched off the IC labels (not an uncommon practice).

It seems like a switchmode controller regulation issue - perhaps from too noisy a feedback connection on the pcb.

You could try adding a more effective output filter cap - eg. a somewhat smaller electrolytic (4.7uF 400V are quite common in small universal USB supplies) - and connect it directly between the US3M diode pad and the transformer 0V leg (that may be the furthest 0V pad, as the other 0V pad may be for another unused winding). This could be worthwhile, as the 0V return path from the 10uF 400V cap appears to be past the regulator somehow and is a reasonably large loop.

It's difficult to tell how well the input main current path and switching loop is laid out, and whether the regulator track layout is prone to noise (the regulator furthest from the FET will likely just be used for voltage sensing and shutdown).

I have had good results from a 150W dc/dc module aimded for 12V input, and with a few secondary taps. It doesn't use output voltage feedback regulation, but that is fine for valve amps.

Thanks for the suggestions... unfortunately, I'm not versed in these circuits so I'm lost on how to implement any of it. That said, I'd be very open to seeing examples like your 12v supply (if you don't mind sharing it). I'm coming around to the view... "if you want it done a certain way, do it yourself", so it seems I'll need to learn more about these circuits.

[trobbins]

A number of people have had threads on various dc/dc obtained via ebay. I guess a concern is that these are not 'stock' items, so fine for diy but no guarantee on availability or quality.

Some threads:
https://www.diyaudio.com/forums/powe...step-smps.html
https://www.guitargear.net.au/discus...?topic=51615.0
https://www.diyaudio.com/forums/powe...crilege-9.html

[uneumann]

A number of people have had threads on various dc/dc obtained via ebay. I guess a concern is that these are not 'stock' items, so fine for diy but no guarantee on availability or quality.

Some threads:
https://www.diyaudio.com/forums/powe...step-smps.html
https://www.guitargear.net.au/discus...?topic=51615.0
https://www.diyaudio.com/forums/powe...crilege-9.html

Understood - I appreciate the links, a quick browse has already given me some useful ideas... thanks trobbins

[R.G.]

How does the ripple voltage react to loading?

Many recent switching controllers react to too-small loading by going into "hiccup" mode, running a blast of pumping to raise the output voltage, then essentially shutting off for a while until the output goes down a bit again. The repetition frequency depends on a lot of the output loading, leakage, etc. Your description of varying amounts of ripple at different stepup ratios is very reminiscent of this situation. A minimum load might cure it, or a minimum load plus a mild RC filter might fix it.

[uneumann]

How does the ripple voltage react to loading?

Many recent switching controllers react to too-small loading by going into "hiccup" mode, running a blast of pumping to raise the output voltage, then essentially shutting off for a while until the output goes down a bit again. The repetition frequency depends on a lot of the output loading, leakage, etc. Your description of varying amounts of ripple at different stepup ratios is very reminiscent of this situation. A minimum load might cure it, or a minimum load plus a mild RC filter might fix it.

Thanks - something like that seem to be happening. The ripple freq appears proportional to current draw, but variations from 2ma to 20ma don't seem to impact the ripple presence. What does seem to affect the ripple presence is supply voltage. Operating from 10vdc, it's pretty smooth and flawless for a range of output voltages and currents. It's been running for 30 mins now at 10v @780ma input, producing 360v @ 18ma (6.5w). It's producing regulated DC with small switching spikes (.5v @72Khz) that I think I can live with. That's 83% efficient - so it runs cool.

I see two options: 1) try to redesign/fix the module to optimize my use conditions, and the links above could help with that. Or 2) use the module at 10v by adding a step down converter (24 to 10). There seem to be plenty of converters out there so the latter seems my best near term option and gives me time to work out something more permanent.

[SoulFetish]

I was looking into a couple of different boost converters for a recent project. One of my concerns was isolation in the supply (as it would essentially be an inline effects unit). The size and cost are really attractive, but the available products and designs I found were non-isolated boost converters. Then I started some research, looking into designing an isolated flyback SMPS.
Once I started down the Rabbit hole learning about driving the gate capacitance and essentially designing a custom flyback inductor from the ground up, my eyes began to gloss over. I realized that the power supply alone would have taken so much longer than the rest of the project for me.
Ideally, both the low voltage DC supply and the boost converter would have galvanic isolation from the supply for 2 layers of isolation from the mains AC. I know RG designs power supplies for effects, so I'm curious about his thoughts regarding this. Also, Tim, are your supplies isolated architectures?

[trobbins]

For B+ supply, the most recent step-up dc/dc I used has an isolated secondary, although I connected the primary and secondary circuits to the same 0V ground, but....

One issue you may need to manage is ground hum loops - similar to any isolated supply (mains frequency or switchmode) where parasitic capacitive current will flow between primary and secondary sides due to dV/dt. The dc/dc module I used does not include measures to constrain switching waveform voltage induced currents through stray capacitances - so hum will flow between primary and secondary windings and other stray paths on the pcb. I added a local coupling capacitor between input and output grounds, with the aim of keeping that hum loop as local as possible to the dc/dc module pcb, but it will flow externally through whatever loop it can take. A better outcome would be to use winding screens within the transformer (same as in a mains frequency power transformer), but that isn't going to happen in an ebay special.

[R.G.]

@ Soulfetish:
Good thoughts. Doing a good design of a high frequency switching converter does require a lot of picky considerations. Especially where to get small size, you run your switching frequencies up into RF. Energy radiates everywhere it possibly can. Return currents follow paths determined by field interactions, not necessarily where traces go. As Tim notes, isolation becomes both a critical issue, and much more difficult to do.

You have to question what you're isolating, why you're isolating it, and whether the isolation or lack thereof is doing you any good or harm. You want isolation from the AC mains for two reasons - one is hum, the other is shock hazard. You only need the shock hazard isolation from the mains once.

Isolating for EMI related noise gets more involved. You may or may not need galvanic isolation for that. As Tim notes, isolating switching related EMI requires thinking about the current loops where the EMI current moves, and also the capacitive paths that voltage EMI can follow. The idea in keeping current interference out of audio is to keep the current loops where the noise current flows to be as small and local as possible, and to keep the local current paths as low impedance (and that means as low inductance!!) as possible. Getting the currents to follow the paths you want requires thinking in terms of transmission lines set up as planes, with components bridging the boundaries between planes and all of the local loop underlain with a ground plane. Done well, the current will preferentially follow the smallest loops available to it. Minimizing EMI related noise is a big layout related issue because the interconnection conductors ARE circuit components and that can't be ignored.

Tim is also correct - for lowest noise, you need custom magnetic designs, including interior screens. You can sometimes screen out capacitive noise by providing a screen/plane to shunt the capacitive current back to a place it would rather be instead of your output voltage.

There are some subtle issues. A converter working from rectified AC mains will always have a modulation on the emitted EMI that's 1x or 2x mains frequency as the modulation scheme corrects for the ripple in the rectified mains DC. This can be small, but it's always there as a signal riding on the "carrier waves" of the harmonics of the switching frequencies. Your circuit has to avoid letting the pulse-repetition-rate hum get into the audio path where it can be detected by the semiconductors and downconverted back to audio as hum.

One approach is to make your switching converters into screened "black boxes". An isolation boundary where the converter receives its input DC and does all the switching and conversion locally, including post-filtering for EMI components, and presents an isolated DC output as a "differential signal" floating free of the local AC safety ground and free of any other local grounds can sometimes work. This is the instance where you want secondary isolation boundaries. It works best if you can actually cage up the converter in its own screened box, the box acting to screen off any capacitive or EM radiated harmonics. A lot of RF circuitry before truly tiny SMD practice did this, in the form of PCBs with vertical walls of grounded metal to fence off the various sections.

The old saw about knowledge being power is true here. We all have to do our designing from available components, whatever those are. If you can get a chunk of DC-DC conversion that does what you need, you can think in terms of feeding it DC that contains filters to prevent the switching EMI from getting back out into whatever else shares the incoming DC, encasing it in a good-enough screening box to intercept any capacitive or radiated EMI it generates, then filtering the output locally and at the point of use to remove switching artifacts, and considering the output to be a "differential signal" of some DC value, forcing any EMI riding on it to follow only the wires you use to feed your load. This helps to ensure that the sneaky paths of EMI through other grounds can't happen - as well as letting you make the output wires be an absymally poor transmission line for the EMI.

Good on you for doing the thinking. You're getting into the next level.

[tedmich]

There are some subtle issues. A converter working from rectified AC mains will always have a modulation on the emitted EMI that's 1x or 2x mains frequency as the modulation scheme corrects for the ripple in the rectified mains DC. This can be small, but it's always there as a signal riding on the "carrier waves" of the harmonics of the switching frequencies. Your circuit has to avoid letting the pulse-repetition-rate hum get into the audio path where it can be detected by the semiconductors and downconverted back to audio as hum.

RG can correct me but I think this is what DCDC converter maker Vicor does with its FARM module, its a line level AC input FWB front end with associated filtering, unlike their unfiltered AIM modules (370VDC output at 250W about the size of an Altoid box!)
They also sell a ripple attenuation module (RAM) for use after a standard FWB front end-> DCDC converter combo see
http://www.vicorpower.com/documents/.../ds_vi-ram.pdf
these can be had on ebay for $20, see
https://www.ebay.com/itm/Vicor-Rippl...4AAOSwuxFY0Gm1
but they're only good up to 50VDC

I've run a couple 4 x 24V window fans for over 10 years using their old open frontends

with 24VDC DCDC modules (speed variable by +/- trimming ) not too noisy that I can tell.