L-Adapter

So it ate all the source PSU's scope visible switching noises? What will be its client circuit to power?


With switch mode supplies I always measure the switching frequency and put an LC low pass on the output so the regulator doesn't have to deal with it. In my case the regulator's job mostly is decoupling and providing the supply with a low output impedance over a wide enough frequency band.

The LC filter introduces another pole but none that matters in that context.


I had a DDDAC with Pi input that I got rid off. The L-Adapter was supposed to power the Pi. I did build it anyways out of curiosity and hey, who doesn't like to build a regulator and at least measure what it does ;) Maybe the guy who bought the DDDAC will be interested in the L-Adapter.
 
I don't think there will be many people using it for <500mA duty. I don't want to go into extremes with latest LDOs in parallel powering Rasps and PCstreamers. L Adapter seems to other good performance, reasonable consumption and it's simple enough
Tempted to build my pcb and go for Teabag's if I like it.
 
diyAudio Chief Moderator
Joined 2002
Paid Member
With switch mode supplies I always measure the switching frequency and put an LC low pass on the output so the regulator doesn't have to deal with it. In my case the regulator's job mostly is decoupling and providing the supply with a low output impedance over a wide enough frequency band.

The LC filter introduces another pole but none that matters in that context.


I had a DDDAC with Pi input that I got rid off. The L-Adapter was supposed to power the Pi. I did build it anyways out of curiosity and hey, who doesn't like to build a regulator and at least measure what it does ;) Maybe the guy who bought the DDDAC will be interested in the L-Adapter.

I tend to still discover some switching noise creeping to places through inductive coupling and ground return paths, not completely gone no matter the filtering, especially when dealing with DC to DC converters.
 
Maybe you need to switch to 3d rat's nest build instead of PCBs ;)
Joking aside, I am far from being in a spot where I at least feel like seeing light at the end of the tunnel, especially as different types of SMPSs measure quite different. Also when looking at an FFT I mostly see harmonics of the switching frequency creeping through, often in a frequency range where I am not sure anymore if I am looking at noise from the switcher or bench noise or a combination of all the other the "stuff" going on.
 
Last edited:
diyAudio Chief Moderator
Joined 2002
Paid Member
Imagine the hair pulling if there was SMPS and I had to debug where switching spikes are coupling on my old humble cassette deck's cable runs. I dug it up along with some tapes and restoring it for fun. Its a low end 2 head Lux. Seems rather alright after disassembly, Cramolin (Deoxit), and some joints rework :)
 

Attachments

  • IMG_20190422_215029.jpg
    IMG_20190422_215029.jpg
    589.1 KB · Views: 1,327
Here is my test L-Adapter for filaments..
I use it to feed the filaments from my tubes 6.2V
I made my own test pcb to fit my demands box etc..
Cold start without load 6.175V , with load 10Ω start from 6.100V and stabilize finally at 6.200V-6.205V.
The dimensions of the heatsink are 6X3cm.
So my drift in me is 30mv.
 

Attachments

  • IMG_20190505_123714.jpg
    IMG_20190505_123714.jpg
    705.8 KB · Views: 1,171
Last edited:
All parts now in my bench, one question regarding main filtering caps.
I have two 10mf and two 2,2mf at hand. Considering that my first goal is psu for Rasp-type pc which is better suited ? Output caps are Panasonic FCs 220uf
I think MSR860 can do the job for both (on 4cm heatsinks)

I have thought of using it also as pre-reg for DACs but that's another story

Tomorrow I'll fire it up (hope not literally)
 
Last edited:
diyAudio Chief Moderator
Joined 2002
Paid Member
I think MSR860 can do the job for both (on 4cm heatsinks)

Bridge diodes have high peak to average current ratio spec so they should withstand the inrush of the larger caps.
F1 slow blow fuse chosen at your diodes average rectified forward 8A spec should take the inrush as well, because T (time) type fuses have much higher instant blow current ratio to average current nominal too. If it does blow use the next higher nominal spec one until sufficient. NTC Thermistor anti-inrush disc can be also used in the primary when with big transformer and large caps for high power apps. There are 16A average current spec diodes or more etc.
Heat on bridge diodes has mainly to do with the load's average consumption. Your 38mm sinks should suffice for much more output power applications than any sane RPi scenario.
When you also have the main transistor mounted under the board for external sinking, like insulated to a thick chassis floor or to a sizeable sink, then the output power envelope escalates pretty substantially.
In this case alternative main transistor types can also be included for wider choice and sourcing.
TO-264 package with alike spec to the standard TO-3P NJW0302G (NJW1302G too). Those are pin compatible bigger jobs, some with less RθJC also, usable when not restricted by board level sink mounting space opening anymore. MJL1302AG is the ON Semi analogous type.
I had in hand and tested the compatible Toshiba 2SA1943 though with correct results. Digi-Key, Mouser, TME, they have such.
 
It lives !

On first test on resistive load everything seems OK. I've decided to put only one 10mf cap for Tinker and add the second one for the test on my audio PC.
It will take some days to setup Tinker and do an overall audio test on it. After that I'll do the comparison with my linear psu (LT1083 / picoPSU) on my audio pc. I'll post my impressions then, now I have to crawl back to my bench.

Have I mentioned that the board is a kind offer from Nick ? Same on me.
Well done Salas. Thanks for everything.:wave2:
 

Attachments

  • L_adapter_test.JPG
    L_adapter_test.JPG
    155.2 KB · Views: 1,160
Last edited:
diyAudio Chief Moderator
Joined 2002
Paid Member
Easy to assemble and set? Steady? How many mV you have across R2?

All you check in the end is DC across the D11 1N4007 (an easy place to measure Vin-Vout) will not drop below 2.5V worst case. That check confirms the losses in the chosen components for the rectification & filtering section, including the transformer, are still at bay for the highest current consumption peak of a given computer. Probe across D11 with the voltage meter during boot and on highest CPU usage %. Else, you have to use higher secondary voltage transformer and/or higher mF reservoir capacitor(s) to decidedly distance the Vin-Vout voltages and/or reduce the ripple voltage loss contribution.
 
It was very easy to assemble, I have no special instructions for things to pay more attention, it's pretty straight forward.
I measure 64mV across R2 (1.7 A load)

Transformer's secondary is 8VAC, target voltage is 5VDC
For 330 mA load, 5VDC output, input voltage is ~10 VDC
For 1.7 A load, 5VDC output, input voltage drops to 7.6 VDC
Of course this is transformer dependent, I'm only using 25VA of this 50VA transformer

Even for high current test, I measure (multimeter) below 1mVAC at the output.
As I've stated all test (for now) are done with resistive load, I'll measure again with various CPU loads and usb devices attached when I setup Tinker
 
diyAudio Chief Moderator
Joined 2002
Paid Member
Of course this is transformer dependent, I'm only using 25VA of this 50VA transformer

Even for high current test, I measure (multimeter) below 1mVAC at the output.

Doesn't have another same voltage secondary to parallel? When paralleling secondaries we must make sure they aren't connected backwards of course. Its either they are in parallel and there is voltage or in antiparallel and there isn't any voltage. In this case the transformer heats up fast. Blows its primary fuse or it heads to meltdown sometime soon when there isn't a fuse.

Don't know about your Fluke 179 but my Fluke 87V takes patience to settle down in its final AC mV reading when hunting very small figures. Seems like it performs a slow integration and averaging process. Has 20kHz bandwidth spec and its reliable at least. I had checked it finally shows the right sub 1mV figure with highly divided lab generator signals where I know the input level and the L-Pad's division ratio.