Linear power supplies

[cnt]

My time machine is broken so I can't go back to 1980 and buy a linear wall wart.  Where you you guys source these for voltage reference projects?  Would a 24v switching supply + an lm317 work well (within a factor of 2 as noisy)?  I remember an eevblog video saying linear regulators don't reject noise that well but the reasons are unclear to me.

[WattsThat]

A time machine wouldn’t help. Older wall worts had transformers with one or two transistor pass regulators. Nothing special there. Don’t kid yourself into thinking they were low noise.

Perhaps you should tell us what you need in terms of voltage, current and noise, perhaps even the intended application.

[Kleinstein]

The main problem with SMPS is the injected common mode signal. So the secondary is not perfectly isolated but there is a capacitive coupling with 2 parts: one is the switching frequency and the 2nd is often half the mains voltage from a capacitor to dampen the higher frequency part.

Linear regulator would not at all help against the common mode part. The differential part can be filtered and an extra regulator can help.

Classic iron transformer based wall warts can sometimes still found with old equipment. With some luck one can sometimes still buy some left overs. Often they are non regulated and relatively low voltage, which can be a slight problem. Here an external regulator can help, usually more a low drop type as the voltage tends to rather low anyway. Old telephone routers a long time still came with classical supplies with a relatively high voltage.

[cnt]

For context: Everything is for fun.  I have no NEED for anything I'm doing. 

I'm kinda hunting for best bang/$.  I'd rather spend 15$ on a power supply if the next 100$ is only going to get me -3db noise.

I need something in the 16-24v range, 20mA.  I don't have a concrete spec for noise and the only way I can measure really low noise is by seeing what the last digit on my DMM does.  I haven't tried, but I don't think my rigol DS1054 is that sensitive (am I wrong?).  This is the background for what I'm doing:


That said, when I read about the lm399 I was both impressed that it regulated temperature so well with a basic circuit and offended that it did so poorly in general (I was once a control theorist).  0.5ppm per degC ambient seems pretty bad when you can buy unheated zeners in the 5-10ppm/C range (1N82x). 

So I'm trying to make my own heated 1N825 with quality temperature control and a low frequency noise floor around the last couple of counts on my 6.5 digit keithley at work.

From my home linear bench supply, and done with jelly bean components (and a 1N825) in a breadboard it's currently stable on the last count of my 5.5 digit meter.
 

[penfold]

[...] I remember an eevblog video saying linear regulators don't reject noise that well but the reasons are unclear to me.

The HF switching ripple is typically way above the control loop bandwidth of most linear regs and therefore the linear reg cannot functionally do anything about reducing it. The parasitic capacitances associated with the pass-transistor also tend to be low impedance at the fundamental (switching) frequency and harmonics thereof which make up the "noise". Which one of the two factors is the biggest problem would depend on the specific device, but each is at play and neither makes it better. I guess both those things came up in the video, may be different wording helps, or not, or makes understanding worse.

Bang-for-buck-wise, I wouldn't recommend the SMPS approach. The reason is that most wall-wart-style ones will probably have quite a high CM noise and be most difficult to tame. An alternative option for SMPS would be to build something up with an open-frame PSU in an enclosure with a little board with filtering, that way you get better control over the CM noise, have an earth connection, and access to the PSU chassis earth if needed - but going to that effort would still get you better more favorable noise from a mains-freq transformer for a similar price.

[cnt]

The HF switching ripple is typically way above the control loop bandwidth of most linear regs and therefore the linear reg cannot functionally do anything about reducing it. The parasitic capacitances associated with the pass-transistor also tend to be low impedance at the fundamental (switching) frequency and harmonics thereof which make up the "noise". Which one of the two factors is the biggest problem would depend on the specific device, but each is at play and neither makes it better. I guess both those things came up in the video, may be different wording helps, or not, or makes understanding worse.

That does help thanks!

Bang-for-buck-wise, I wouldn't recommend the SMPS approach. The reason is that most wall-wart-style ones will probably have quite a high CM noise and be most difficult to tame. An alternative option for SMPS would be to build something up with an open-frame PSU in an enclosure with a little board with filtering, that way you get better control over the CM noise, have an earth connection, and access to the PSU chassis earth if needed - but going to that effort would still get you better more favorable noise from a mains-freq transformer for a similar price.

OK I feel very stupid for asking this but real-life analog anything is not my area, real life power supplies doubly so.  CM noise on a power supply is noise common to +V and the supply "ground" relative to... earth?  Why does that matter for a voltage supply.  If I probe it on my bench meter its earthed and the CM noise is gone(??) and if I use a hand held meter the meter floats with the CM noise ya?

[maxwell3e10]

Jameco has a good selection of transformer wall warts, with and without regulators.

https://www.jameco.com/c/Power-Supplies-Wall-Adapters.html

[penfold]

[...]
OK I feel very stupid for asking this but real-life analog anything is not my area, real life power supplies doubly so.  CM noise on a power supply is noise common to +V and the supply "ground" relative to... earth?  Why does that matter for a voltage supply.  If I probe it on my bench meter its earthed and the CM noise is gone(??) and if I use a hand held meter the meter floats with the CM noise ya?

You're not entirely wrong, as a common-mode voltage, it isn't too big a deal. At the power supply terminals may be the point that you can say "we have a DC level of X, plus a differential mode noise of Y and common mode noise of Z", in a totally symmetrical circuit with very high impedance between each V+ and 0V wrt earth, it will stay quite symmetrical and you could have the same DC+[DM+CM noise] at the load input and the rest of the circuit. The DM noise stays easily filtered and can be quite well filtered still, some may leach through to your outputs, but that's still tolerable or manageable by design - 'not a big deal'/'fun challenge'. The multimeter in reality isn't completely isolated and it will have some capacitance to earth and other "things"; the power supply's wires similarly so and won't have totally symmetrical impedance. (I'm making a bit of a logical jump here between treating CM as voltage and current) So for CM noise, it will find its way through the circuit and to a return path it like the look of, if its entirely symmetrical, the voltages cancel: but in reality, it isn't and they don't.

So I don't mean to say that the CM noise will definitely be an issue, but it has the potential to be an utter nightmare. Good decoupling, filtering, and filtering of outputs will help the CM noise if it infects your DM signal... but the CM noise has a strange ability to uncover any weaknesses there, so it can appear as a problem when you least expect it. When it comes to PSUs, if it's part of a "system" that specific requirements for size and weight, then yeah, it's worth the fuss of SMPS, but for bench test items, I personally go for the less trouble option. Though I'm sure many people will jump in and say "I used This PSU with This circuit and it was totally fine": I'm just skeptical, especially with such a broad range of PSU 'quality levels' and permutations of filters and construction methods.

I'm not sure if I explained that very well, so I might draw a diagram if I get a moment later on. In conclusion, SMPSs are a tameable beast, linear are much more docile, but each one will bite your hand off when it's hungry (analogy broke down there).

[cnt]

Jameco has a good selection of transformer wall warts, with and without regulators.

https://www.jameco.com/c/Power-Supplies-Wall-Adapters.html

Oh! there we go.  an unregulated one is 4$ 
And here I just ordered a 14$ doorbell transformer to use with a rectifier and caps from my parts bin.  I know where to go next time.