L-Adapter

diyAudio Chief Moderator
Joined 2002
Paid Member
- We many times wonder what if that SMPS brick that feeds the Raspberry Pi or the mini PC or the MiniDSP or the Squeezebox or the DAC or the desktop Class-D amp etc. was a linear PSU? How such wonder boxes would perform without switching noise polluting their rail and most crucially their many times interconnected ground?

- What if we had a simple and strong linear PSU instead of fixed SMPS brick adapters that it could cover the voltage range for such applications by only setting it up with a jumper?

- Enter the L-Adapter. A versatile Sziklai pair stabilizer / capacitance multiplier based on LED voltage reference.

- Its output voltage is how many LED + trimmer minus one Vbe. Roughly 1.5V to 20V range. Move the jumper, trim Vout, ready. Connect the load.

- Is it any good? Yes its good. Not noisy at all and stable. For 1.5V input ripple it produces 1.5mV thick DC line. For 3 Ampere load it measures 0.02Ω output impedance. Which is flat and extended in frequency. When you pulse it the recovery is clean of ringing. Because there is no feedback between the output and the voltage reference. We want a general purpose PSU staying insensitive to random gear loading peculiarities.

- The recipe isn't anything new but the details are well researched. Low noise unity gain reference, but adjustable too. Which LED bar with which CCS experimentally chosen between many styles for very good Vref stability, what pair of transistors, the layout. Various ways to sink it, accepts quality TO-220 bridge diodes, two reservoir capacitors, fused like a Π filter. The board is 136mm x 63mm.

- What about its output current ability? Well, it uses a 15A audio amp grade TO-3P pass transistor. But that alone says nothing much. Its also the transformer the diodes the reservoirs the sinking the load's average consumption. Say up to 7.5A average can be catered for.

- For any light or heavy current application it takes that the chosen Tx the bridge diodes and the reservoir caps won't lose the plot for a target Vout. That's about rectification and filtering basics. It does not like less than 2.5V input-output voltage difference. That's the DC difference between C4 and the output. Can probe that between the fuse and V+ out. Or across D11. Although it keeps working on smaller differences it gets progressively goofy. If you see the LEDs dimming a bit its tell tale you crossed the raw DC section's losses good limit.

- Here is a schematic with typical reservoir caps values and some pictures. That soldering iron pulled 55W peak from the mains through the PSU to boot and idled at 12W. The scope pic displays Vin ripple on C4 vs Vout status captured at a point when the iron was still pulling hard to heat up started from room temperature cold state.

The Fluke reads C4's raw DC level in another picture. Started at 27V idle with worst loss of 4V during the soldering iron's boot cycle (trafo, diodes, rippleV). At 23V raw DC for 18V output to the iron, that trafo and reservoir caps passed the 2.5V Vin-Vout criterion by double margin. Other type & quality trafo or diodes could lose more or less steam of course. Higher value reservoir caps would achieve less ripple voltage but would also make the diodes work harder. Since the worst raw DC level sufficed, better not increase the caps value in this case.

The example has 4xMUR860 & 2x4700uF/35V B41231 EPCOS/TDK. Also a 38mm tall Q2's sink. Which sufficed due to the irregular current pull of the micro controlled soldering iron. Can it do an RPi3? Yes I tested it with Wi-Fi and streaming vids on the Raspbian OS. Can it do a 12V Windows 10 Cherry Trail mini PC? Yes I tested it. Watched a whole movie stored in a mechanical 2.5 inch USB Hard Disk attached to it. The PC was at the same time charging an OnePlus X phone from a spare USB output so I pushed it further. I now used lower voltage and smaller size transformers than that R-Core. EI or toroidal of average quality. They and the sinks sufficed again because of the irregular current pull of computers with idling gap periods. 45C on the 38mm 35C on the 25mm ones for the diodes. Minimal RPi use shouldn't need diode sinks at all.

- In the final black & gold boards there are actual Q1 Q2 designations. In the first photo of the green prototype you may spot Q3 Q4 instead. They are just a relic from an earlier schematic with extra parts due to various discrete CCS tests until ending up using IC1. Forget about those Q3 Q4 prototype marks.

- I will be editing & enhancing post #1 in later installments.

- 25/6/2019 Build guide added (includes circuit description & BOM) - attachments rearranged
 

Attachments

  • IMG_20190413_034624.jpg
    IMG_20190413_034624.jpg
    344.6 KB · Views: 8,436
  • IMG_20190414_181710.jpg
    IMG_20190414_181710.jpg
    386 KB · Views: 8,223
  • IMG_20190415_142512.jpg
    IMG_20190415_142512.jpg
    501.7 KB · Views: 7,301
  • IMG_20190413_153127.jpg
    IMG_20190413_153127.jpg
    372.9 KB · Views: 7,044
  • IMG_20190417_155049.jpg
    IMG_20190417_155049.jpg
    321 KB · Views: 7,676
  • L-Adapter Guide.pdf
    292.2 KB · Views: 2,769
  • L-Adapter schematic.png
    L-Adapter schematic.png
    21.3 KB · Views: 5,095
  • Black LA.jpeg
    Black LA.jpeg
    147.7 KB · Views: 4,986
  • L-Adapter_BB.jpg
    L-Adapter_BB.jpg
    555.5 KB · Views: 2,980
  • IMG_20190422_040533.jpg
    IMG_20190422_040533.jpg
    332 KB · Views: 2,761
  • Like
Reactions: 1 user
In a similar vein, you can bias a pair of MOSFETs the Blue LED will get Q2 into its linear region -- this is an adaptation of "Art of Electronics", Volume 3. Figure 9.110. You could use a resistor to bias the LED at the expense of PSRR. If the resistor values are successively doubled, you can get 8-bit resolution.
 

Attachments

  • HV_LED_Switched.png
    HV_LED_Switched.png
    16.2 KB · Views: 2,834
diyAudio Chief Moderator
Joined 2002
Paid Member
Is the Ultrabib still better for the lower current needs?

The UBiB is a successful shunt PSU. Not targeting the same apps. Its mainly meant for internal use. Yes, better in its own envelope. This one isn't a crude external PSU though. It has the DCG3 preamp's DCSTB PSU quality level. It should also give to various gear a bit of extra help beyond the main benefit of skipping switching noise by replacing their SMPS packs.
 
diyAudio Chief Moderator
Joined 2002
Paid Member
It can run anything between 1.5V and 20V for enough Amps of current. Because heaters will be a constant load you can predict your trafo, your ripple loss, and your sinking scheme pretty accurately. 3A at 6.3V is four EL84s plus a couple of drivers, surely doable. Say 8W dissipation for 2.5V across Q4.

You see those side notches. They are to help you mount the diodes and the big transistor on a back surface also. In other words on a potent external sink if needed. Like a solid state amp against a flat sink with its big semis mounted under the PCB edges. He he, now a PSU with a notch and a status bar. Trendy. :D
 
  • Like
Reactions: 1 user
diyAudio Chief Moderator
Joined 2002
Paid Member
That's Tea's area of expertise. Ask him later on. Me I will just send him a finalized PCB design with typical BOM to include in his next GB the soonest I feel I haven't missed the odd mechanical detail to include. Meanwhile I will continue with tests and beta tests acquiring various apps experience and third party feedback from local DIY friends that I will hand most of my few proto green PCBs to. So we can build a better knowledge base on what is best to use it for and what's enough trafo/bridge diodes/reservoir caps/sinks to configure in each case. Because the range of gear it can probably serve replacing their fixed voltage SMPS bricks seems very wide.
 
diyAudio Chief Moderator
Joined 2002
Paid Member
If you will use a higher value VR1 trimmer or a Zener in its place (cathode towards the LEDs), yes. When you don't jumper any side pins all LEDs are full on and the 500R trimmer offers 2.5V extra trim range. Whole Vref is fed with 5mA constant current. Maxing out at 20V output as it is after one Vbe is deducted via an output transistor.