1

I have load that requires a constant DC voltage of \$4.85 V \pm 10 mV\$ and can draw up to \$3A\$.

I have tried to power my load with a \$12V\$ power-supply regulated with a LM350T (+ heat sink), however, I can't dissipate the heat generated by the regulator. My other power-supply provides a voltage of \$5.3 V \pm 200 mV\$, but I don't know which circuit could convert this source voltage to the load requirement.

How can I solve this problem without getting a new power-supply?

Edit

I have not stated my requirement properly. What I need is an average input voltage of \$4.85 V \pm 10 mV\$ with the actual range of \$\approx [ 4.5 ; 5 ] V \$. The load is a \$64 \times 64\$ LED-Matrix powered by \$3.3 V\$ data signal from an ESP32. The whole circuit draws around \$2.2 A\$ when all LEDs are fully powered.

I can dissipate around \$2 W\$.

Edit

The specification of matrix can be found by searching for joyit elv 64x64 matrix (or directly https://files2.elv.com/public/25/2511/251186/Internet/251186_data.pdf and https://files2.elv.com/public/25/2511/251186/Internet/251186_um_eng.pdf).

The datasheet is very vague and basically states:

  • Supply voltage \$5 V\$
  • Max power consumption \$40 W\$
  • Protocol HUB75E (\$\frac{1}{32}\$ scan)

I'm using this library to drive it with a NodeMCU ESP32-WROOM-32: https://github.com/mrfaptastic/ESP32-HUB75-MatrixPanel-DMA

My pin assignment:

Name Pin Name Pin
R1 G25 R2 G14
G1 G26 G2 G12
B1 G27 B2 G13
A G23 B G22
C G5 D G17
E G32 LAT G4
OE G15 CLK G16

Possible Reason

Thanks to the comments I have observed the following:

  1. When I supply \$4.85 V\$ to both the matrix and the MCU everything works.
  2. When I power the MCU from a different source (e.g. USB \$5 V\$) the matrix allows a supply range of \$\approx [3.85; 4.85] V\$ before dimming evenly.

For everyone with the same problem

So I assume that because I have \$3.3 V\$ logic level and not \$5 V\$ the matrix needs less input voltage. But when I provide less than \$4.85 V\$ the voltage regulator on the MCU cannot provide \$3.3 V\$ for the ESP32. I have always assumed the matrix was the problem.

terrakuh
  • 11
  • 2
  • How much heat can you dissipate? – Andrew Morton Apr 02 '23 at 18:06
  • You'll need a switching power supply (not linear). Try Texas Instruments' Webench Power Designer online tool. Put in your requirements and it will spit out multiple solution options. Also consider if you really need 10mV accuracy - that's an extremely tight spec, especially over a wide current draw range. – td127 Apr 02 '23 at 18:06
  • 2
    What kind of load does have requirements like that? – Lars Hankeln Apr 02 '23 at 18:08
  • 2
    Regulators have output impedance, so it is not usually possible to make a very low noise power supply for a load that draws variable current. However +/-10mV should be okay. But you need to explain what the load is and what the current looks like, especially how it varies, because the regulator transient response and output impedance will be a major part of the error budget. – bobflux Apr 02 '23 at 18:42
  • @td127 you can do this with a linear regulator, though you'd need a very low dropout one, which most likely means some control loop idiosyncrasies. – Hearth Apr 02 '23 at 19:00
  • Please make title and body of your post agree on input supply. The WEBENCH® designs seem neither carved out for low ripple, nor fully specified for total noise. Too bad including a, say, PI filter in the feedback loop gives rise to stability concerns. In your specification, try to specify *accuracy* and "each" aspect of *stability* (mV/°, %ripple, mV/h, mV/A, …) separately. Consider buck to about 6.4 V & linear to 4.85 V. – greybeard Apr 03 '23 at 04:24
  • (4S NiMH, if ±10 mV was noise rather than accuracy.) – greybeard Apr 03 '23 at 04:50
  • Thanks for the answers and for the WEBENCH® tip. I have updated the post. I'm fairly new to electronics so I will try my best to fully understand your response @greybeard. – terrakuh Apr 03 '23 at 16:35
  • Why would you need regulated voltage for LEDs? – bobflux Apr 03 '23 at 18:43
  • More and more pixels stop working if I supply less or more than 4.85 V. – terrakuh Apr 03 '23 at 19:07
  • Can you put a schematic of the matrix? – bobflux Apr 03 '23 at 19:47
  • 1
    An LED matrix(-module - MAX7221/7219?) shouldn't be that picky - please provide (a pointer to) specifications. – greybeard Apr 04 '23 at 07:32
  • 1
    @terrakuh If it only works at 4.85 V then there is probably something else wrong. Are all the wires for the power connections of sufficient size and connected properly? Is there anything you can add to "pixels stop working" - such as some of them flicker or are at reduced brightness? Have you tried adding a capacitor of maybe 10 μF very close to the MCU across its power pins? – Andrew Morton Apr 04 '23 at 15:11
  • I have added 330 Ohms resistors on each data line near the input of the matrix to protect against the over- and undershoots I was getting (maybe completely unnecessary with 3.3 V LL for 5 V?). The data lines are about 20 cm longs and all cable sizes should be more than enough. I have tried also a cheap LM2596 circuit board with 12 V input and it seems to working apart from a row at one end. – terrakuh Apr 04 '23 at 16:01
  • Thanks everyone for helping! I have learned a lot and found a promising reason. I will have to redesign the circuit a bit. – terrakuh Apr 04 '23 at 16:24

1 Answers1

0

You can use a LDO to regulate 4.85 V from 5.3 V. 2 Amps with 20 mV variation corresponds to 10 mΩ supply impedance.

Either look for an integrated LDO with appropriate specs or:

  • Pick a P-MOSFET that can dissipate up to 1 W continuously
  • Regulate its gate voltage based on output feedback. This can either be done with a single PNP transistor, or more mundanely, with an opamp.
  • If you specify up to which frequency you need such low impedance, one could suggest more specific solutions.
tobalt
  • 18,646
  • 16
  • 73