I want to measure the noise of a voltage regulator, probably an LM2941 making 5 volts out. So I need a low noise high gain preamp for a scope.
Looks like this would work:
formatting link
Any other suggestions?
I could make something, but it's marginal on time/cost and this looks pretty cool.
Maybe I can bypass the upper feedback resistor to reduce noise, and massively bypass the output to ground, if all that is atable.
I want 5 volts at a few hundred mA and wideband noise well below 100 uV RMS. I could make my own regulator if I had to, but LM2941 is easy. This is going to power a bunch of cmos logic, about 20 ns worth total, and Vcc noise becomes jitter.
I have an old TM500 thing in the dungeon, but I doubt it still works. I guess I'll dust it off and see.
I'd like the noise to be low at low frequencies, down to 1 Hz or less, so an LC isn't practical. And a load of a few hundred mA makes an RC impractical too.
Am 11.09.22 um 00:55 schrieb snipped-for-privacy@highlandsniptechnology.com:
I did not measure the LM2941, but the Lm2940 is about the worst choice possible. It would require real work to be better than a LT3042. Its main drawback is the small load current because of the tiny package.
<
formatting link
>
0 dB is 1nV/rtHz.
In the LT3042 data sheet is a a circuit with an external NPN to take the heat. It measures about the same as the 3042 alone. The noise peak climbs to 100 KHz but does not grow in size.
Adding more output C increases peaking. Cset may be increased at the cost of larger startup time. Cset=100u can make sense and lowers the 1/f corner to a few Hz. It is not really 1/f but
1/f**2 or 3. More than 100u seems not to help.
<
formatting link
>
Some empty boardlets are left over. 1.5A@5V from a 6.6V raw source is ok. The transistor costs a diode drop vs. bare 3042.
The worse-than-1/f did haunt the preamp I used in 2016 when the spectra were taken. It was a much too small input cap. That must be much larger than needed for f-3dB to be able to short the thermal bias-r noise of the amplifier through the low impedance DUT. The DUT _must_ be low impedance or the 220pV/rtHz of the preamp is not useful. (20*ADA4898 in parallel summed up) A 4700uF wet slug tantalum is about right, but that opens another can of worms.
Even board layout counts. Above 500 KHz, voltage noise of the amplifier did increase. on flickr, there is a picture showing the routing with a felt pen. I suspected the tantalum, but it wasn't the culprit as shown by a mesh of wires.
But I forgot about the shorted tantalum when I tried to measure the noise of some Li ion batteries next week. That ended the life of the preamp in a cruel way. Post mortem Pic is on Flickr, more to the right.
I have a new amplifier with 16 * CPH3910 FETs, only 320nV/rtHz instead of 220. But much less noise current, which helps for cross correlation with 2 amplifiers. The Agilent 89441A can do that. 2 b published in october.
I think I'll make my own LDO with an opamp and an n-fet. By the time I have a dpak fet and some giant caps, the opamp and a few passives add not much more board area.
If I use a good 3 volt reference (which I'll have) I expect much less low frequency noise than using the 1.25v bandgap and feedback point of a regulator.
None of that stuff will be on the same die as the big hot pass transistor.
I'll prowl the dungeon and see if I have a AM503 amplifier, and see if it still works, to measure the LDO noise. Or maybe just trust Spice.
One trick I sometimes use is to let the regulator worry about the DC and the load regulation, and then apply an AC-coupled op amp loop to null out the noise.
That way, the AC-coupling goes on the high-Z end (the op amp input), so you can have any corner frequency you like. (If it's very very slow, you'll want to restrict the output swing of the amp to prevent the rail going out of spec.)
It may not be too critical - almost any low noise, battery operated AC-coupled amplifier will give a good indication of relative noise levels, in an iterative process.
You'll want to see a good margin on your actual requirements, then recheck in the physical application.
I used to use a simple BPO two-transistor version of the 'liniac', with increased decoupling capacitors, running off a PP9 battery. It used to take some 10s of seconds to DC-stabilize after turn-on. Switched gains of x10, x100.
Am 11.09.22 um 05:30 schrieb snipped-for-privacy@highlandsniptechnology.com:
LT3042 uses a precision current source to develop the output voltage over a resistor. That is easily filtered.
They even have a proposal of using a post-filtered LT6555 reference with the *1 output stage of the LT3042.
Like their external cheap D44H10 NPN
You can trust spice, but you cannot trust the models. Few have 1/f noise, or even correct data. Many transistor models have been made with the Orcad parts program. You recognize them since all have the same RBB.
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.