Modern low noise bench top power supply

You say "low noise" and "SMPS" like that is an
easy thing to come by, industrial or home-
built.

The "Dr. Meter" supplies I have for bench work
have tens of mV of hash on the output when
enabled. Is that low noise? I guess it depends
on what you're looking at (and how closely).

Example of buck converter, adjustable output voltage.
The invert-gate produces snap action On-Off. The NPN biases the PNP in order to allow a low voltage to control a high supply voltage.
It's safer if the supply voltage comes from a step-down transformer.

dick_freebird said:

You say "low noise" and "SMPS" like that is an
easy thing to come by, industrial or home-
built.

Click to expand...

Not an easy task, fully agreed, the linear stage after the SMPS should remove some of the noise, as proposed in Elektor article (that article seem like a great choice for me, limitation i see is current output set to max 1A. I can run a set of calculations/simulations to reach 3A, but would like to avoid this if possible).
SMPS should be easier to dimension with available chps on the market today which run on high frequencies reaching 1MHz -L7986). What i'm looking for is modern linear stage that will have fast response and steady precise regulation of voltage with good thermal stability of the output voltage. What are expected parameters of stability, lets say noise of less than 5mV and fast response to current change of 10-90% with voltage spike/undervoltage with Vpp smaller than 0,1V

Hi,

"low noise SMPS" is more than a schematic.

Even with identical schematic there are many parameters that influnece the ouput noise.
One needs to understand what causes the "noise".
Generally for sure the switching of the transistor ... that causes a current (pulses) to a capacitor.

So you
* need a sufficient big capacitor
* the capacitor needs to be LOW ESR
* you need a LOW impedance connection to the GND plane (several vias, short wide traces)
* the GND plane needs to be solid to be low impedance (no "cut into pieces" copper pour)
* you need to care about the current flow on the GND plane (direction...)
* maybe even need to take care about stray capacitance of the switching node to the GND plane
* you need to take care about the power path:
good: transistor - capacitor - load
bad: capacitor - transistor - load
* you need continous switching (low load may cause some burst_mode, which causes the ripply frequency to get much lower and the ripple amplitude (noise) increases)
Ripple frequency may vary with load, thus it´s difficult to design a good filter.

... there will be even more influence..

The problem I ssee with a bench top SMPS is that you don´t know the load. It may be zero, low or high. It´s not unusual that you need a minimum load current to get low ripple)

Now one may think you just add a big ouput capacitor. But big capacitors often have bad ESR. And the drawback is that they stor a lot of energy.
Example: LEDconnected to a "15mA current limited" supply: then the capacitor may be initially charged to a high voltage...so if you connect the LED it will cause a huge peak current until the C is discharged to the steady state voltage.

Klaus

KlausST said:

low noise SMPS" is more than a schematic.

Click to expand...

Fully agreed, on all points, to go even more deeper:
- PCB copper thickness to be as high as possible
- Keep direct lines between bypass capacitors and component that is "bypassed"
- Not only paralleling capacitors with same values, it is great tactics to parallel capacitors with different capacities to care not only about Rs but also Is
- Caring about how you lay copper tracks to avoid parasitic signal influence from "noisy" to "high impedance control" lines.
- Introduce guard layers/tracks to shield sensitive high impedance control lines
- For high frequency SMPS, poor PCB design might require specific shielding to avoid EMI radiation...

And many, many more thing to care about as already noted.. thus the initial request is for already designed, tested DIY solution in which most of these and many more details are already taken care about. If there's nothing existing, then i'll have to buy one and try to analyze/improve it.

If we skip the SPMS part, can anyone propose a good design for precise modern Linear bench top power supply solution?