What limits noise in voltage regulators?

Gold plating is too expensive for simple prototypes. I don't want to scare prospective users away.

I tried to get on Dropbox but for some reason it would not accept my Windows XP. Just about everyone else has no problem. I'm going to try getting a web site on GoDaddy to upload files instead of using Google Drive.

Whoops, there goes our average. ;) Dead bug all the way for me. If it's something I'm likely to use again, I build it into the lid of a cast-aluminum stomp box and put a folded-up paper schematic inside.

Live-bug weakens the leads so badly that you can't reef on them the way you can when they're intact. A DIP8 gives you eight nice sturdy standoffs for free, as long as you don't weaken them like that.

Of course it's increasingly a moot point--the sorts of parts that need protos don't come in DIP anymore, so it's breakout boards stuck to Cu-clad with foam double-sticky tape. Those are all live-bug.

My dead-bug things are mostly testers or other small instruments these days.

Manhattan is _way_ too slow by comparison with dead-bug, and all those FR4-copper capacitors are unnecessary when you have the aforementioned free standoffs.

Perf board is also very slow, but in addition it's super flaky--almost as bad as a white proto board.

Cheers

Phil Hobbs

I was not even thinking of you.

That has nothing to do with ASCII art. An ASCII circuit proposes specific components arranged in a specific configuration. You have no way to tell if it is accurate, or if it will even work. You need to simulate it or make a breadboard. Simulation is faster and gives much more information.

So you agree. ASCII circuits are very limited.

I don't see your point. It takes skill and experience to work in electronics. We have tools such as calculators, finite element analysis programs, simple ohms law for basic calculations, Matlab, google for searching for different approaches, various programming languages such as Python, C, Basic, and many others, and simulators.

But you have to have the skills, experience, and creativity to know how to fix problems, and how to use simulation to show a solution works.

As far as using ASCII art, you showed the general configuraton of your ripple filter in ASCII, then followed up with the LTspice ASC file for the actual simulation.

My point is to simply skip the ASCII art, which does little, and go directly to LTspice.

Joe Gwinn wrote in news: snipped-for-privacy@4ax.com:

snip Sorry for that, Joe.

My answer... Among other things... PARD.

Every year or two, I have one of our PCB suppliers make me a couple of square feet of ENIG plated FR4. I then shear that up as needed. It's about $100 a square foot, which is trivial when making small prototypes.

Which of the other regulars started out as a SPICE skeptic? I can't think of any. JL started simulating circuits on a PDP-11 or a Marchand calculator or an abacus or something.

Cheers

Phil Hobbs

I can stand behind someone and show them how to draw and run LT Spice circuits in ten minutes. Some advanced things are tough, but basic circuits, using the standard parts, are simple to enter and run.

They have to understand a little electronics first, of course.

The one about 1/4 of the way down, with the thre SMBs and the two blue inductors, is one of mine. Uncredited!

I use 2-56 screws and nuts, coming up from below. They are easy to clip onto, bring grounds up to topside copper, and look nice.

They make nice power connectors too.

John Larkin wrote in news: snipped-for-privacy@4ax.com:

Baloney. A Gold plated (ENIG) 4 layer circuit proto even on a one week turn is pretty cheap. The competition in that 'proto' realm is high, and the going price is pretty low.

John Larkin wrote in news: snipped-for-privacy@4ax.com:

So do COPPER STUDS with a knurl (or not) near the head to bite into the PTH with. Depends on the Amperage one supposes.

Oh yeah... That's right.. you think I do not know any electronics.

I guess you think I think PTH means Putz Top Homo.

It doesn't beat a ten-minute turn though, like JL gets with his Dremel.

My personal preference is Liquid Tin instead of Krylon. YMMV.

Clifford Heath wrote in news:lm5TH.89291$ snipped-for-privacy@fx47.iad:

Yes, but those houses I speak of will send you TEN of them exact copies. And THAT makes for far better proto dev than chipping away at stones trying to make copies of the same statue.

I have done both.

There are plenty of gantry mounts for dremel and x y CNC setups out there to control those far more precisely than Jerky Johnny can.

I use various sizes of diamond concentric drills to create circular islands on the board rather than Manhattan construction.

I cannot visualize your circuit. Can you please zip the ASC and PLT files and upload to

It will take a few seconds to register, but then you will be set for life.

If not, can you post the zip file to your web site?

Thanks

Sure thing.

It'll do a good job of switcher-taming with 330 nF 0805 ($0.015@1ku) and

Cheers

Phil Hobbs

Thanks. That's a lot of work!

But it still has zero db attenuation at low frequencies. My goal is to get the best attenuation as possible at low frequencies, then good attenuation through the higher frequencies.

This is for driving a 10 MHz vcxo for use in a GPSDO (gps disciplined oscillator.) The vco uses varactors to control the frequency, and noise and ripple will result in FM modulation of the signal. This is one of the reasons why GPSDOs are so noisy.

Going to a TL431 regulation improves the low frequency attenuation. This uses the same input and output voltages and supplies the same current as before. The TL431 gives better attenuation, regulated output voltage, and easy adjustment to different output voltages as needed.

The TL431 model is a transistor-level circuit by Helmut, and it matches most of the TL431 device parameters. It is far better than the other models that are out there.

The AC gain peaks at around 0.5 Hz, and it is all downhill from there. The output is 80 dB down at 10 Hz, -160 dB at 20 KHz, and it stays there through the higher frequencies.

The constant current source is not much better, so this gets rid of the

The flicker noise from the bandgap reference might be an issue, but it is going to be a problem to measure. You might be in a better position with your vast horde of HP boatanchors.

Unfortunately, sabercat has introduced a 30 second delay in the download, but this is still better than the old cut and paste.

You can download it at

I've occasionally split the first base resistor and hung a shunt regulator on it. You have to use split feedback, though--resistor to the cmult output and capacitor to the cathode of the TL431. That makes a sneak path that trashes the ultimate rejection, so it's mostly for light-duty use.

The TL431 is extraordinarily noisy, as well--iirc around 200 nV 1-Hz noise in the flatband and bad 1/f. An LT1021-7 with a chopamp providing dc feedback would be one good choice. Run the chopamp off the reference so its kickout doesn't screw up the rejection.

I've used the LT1021-7 in geophysical instruments, and its 1/f noise was by far the best available at the time (2013) in normal sorts of parts. DIY kits such as the LTZ1000 are probably better.

Cheers

Phil Hobbs

It depends on the vendor. The TI LT431 shows 220nv/hz at 10 Hz and 2.5V, or

The lt1021-5 shows about 100nv/hz at 10Hz and 5V output, which is much better. However the output voltage is fixed, and it requires a chopamp at the output:

The Motorola TL431 shows about 60nv/hz at 2.5V, which is 120nv/hz for 5V output. This puts it in the same ballpark as the lt1021:

The NXP doesn't even specify noise:

For my money, I'd go with the Motorola TL431. It has a parabolic temperature coefficient that is flat from 25C to 50C, which is roughly the same category as the LTZ1000. The Motorola TL431 is $0.45 at digi-key, so it is hard to beat for performance and noise.

The sabercat delay is actually 15 seconds. I moved the mouse and had to start over.