DIYHiFi.org

For those of you who pay attention, I have stated that designs will come forward as I need them.

Well, I have something that is an outgrowth of a rare consulting gig. (Yeah, bidnis sux.)

So, what I have is a little PCB that supplies either +5 or 3.3 V for digital circuits. It has low broadband and low 1/f noise, so it should be ok for a clock. Doesn't have a lot of expensive or hard to find parts. Since this will be the DIY version, there won't be any friggin' SM parts!!!!

It does use an op-amp for the error amp, and since it is designed to work on +9-12 V supply, you can't stick any ol' op-amp into it.

For those of you with a fondness for nostalgia, I have tested it with a '5534.

So there!

Jocko

Hi Jocko,

Is it a shunt or linear type of regulator?

Linear.............you want a shunt element added to it? More parts, more space.

Jocko

I would have thought that it would be easier to ask questions on the forum, as opposed to my e-mail.............

Oy vey.

OK, here is the rationale:

It had to work on as little as +9 V. (It runs better at +15V, but these guys didn't have that much.) Low broadband noise, say 1-2 nV/rt-Hz. Low 1/f, rising to say..........single digits in the 5 Hz range. Not a whole lot of parts or space. This is what I came up with.

Output z, with light loading, is around 30 milliohms. Goes down a bit with more current or higher rail voltage. Impedance is flat to over 20 kHz, with the -3 dB point between 35-50 kHz, depending on operating conditions.

Yeah, it may cost you 20-25X more to make than a '7805. Try making a low-noise clock with just a '7805. Good luck.

Jocko

20 to 25 times more than a 7805 = 10 to 12 bucks worth of parts.

Just the op amp in one of those Jung jobbies will set me back that much, and you've got comparable specs. So this sounds promising.

How much current will it be able to deliver?

Andrea

Uh........I think that will be limited by device (TO-92) dissipation.

IOW, it should be able to power a digital filter chip.

Jocko

This is not intended to be JSR competitor. It is for digital circuits that need low broadband and 1/f noise.

Basically, clocks, and stuff like that.

Since this was designed to work on +9 V (translation: 9 V battery), PSRR was not a prime design concern. Unlike the JSR. So, it is intended to be used with a decent supply ahead of it.

I'm sure if one were to carefully peruse Walt's site, he in all likelihood points out that there are other solutions for low-voltage supplies on his site. Other than the JSR.

Finding a op-amp that will work with its inputs near the negative rail, and its output near the positive rail (both of which are low voltages) is a bit tricky. Unless you are not concerned with 1/f noise.

Eventually, I will put some performance data for it when using one of those wretched '5534 op-amps.

Actually, I bet it won't be all that bad. So, those of you who are not able to dredge up the one that I used should be able to get adequate performance.


Just a word in general about the designs that I eventually hope to provide...........

Most won't be revolutionary, SOTA, or ne plus ultra. They will be circuits that will be, for the most part, easy to build. Yet should provide fairly good performance, without shelling out $$$$. With the exception of a handful of common-base stages sprinkled in every now and then, they won't be anything really hard to understand.


The only hard part may be understanding why I did it that way.

Jocko

Jocko
Have you looked at any of the newer R to R lower V supply opamps?

Sounds like just what a lot of us need - - something that will strip the noise from a regulated supply for digital circuits. And not cost $60 a shot...

If it can be rigged to do -5V, even better....

LM7332.... rail-to-rail output, beyond r-to-r input, supply voltage from +-1.25V to +-16V. Under $4 in single quantities.

(but dual only, no DIP)

I only investigated singles, and only ones that had low-noise specs. A lot don't spec anything but broadband, and not so good even on that account.

Once I get the proto PCBs done, I will post performance data using a '5534 on my site. You guys should already have tons of surplus ones in your junque boit. If not, they ought to be cheap. And good enough to power a digital filter.

I would use something lower for a clock, though. This cheap little circuit is 2-3 orders of magnitude quieter than a '7805.

Which is what a lot of you are using now for your clocks.............

Jocko

Only if you have a negative supply, that has its own independent ground. I know a lot of big-name audio designers that use that scheme to use LM317s for a negative pre-reg.

Jocko

I'm not

I use something like this (except current feedback pair on output, but since you want smaller out Z..).
Noise: few nV/sqrt(Hz) with 1/f knee depended mostly on RC filter that you can live with.
Out Z - about 0.1 Ohm flat to more than 100kHz @ 20mA (without output cap of course).
Cost...

Specs aren't that good that your circuit, but I hope you like it ;]

Anyone who has worked with me knows that I have pretty much made a career of using that circuit, or something like it. It works really well in 99% of your applications.

The 1% that it doesn't work in is ones where you need to have good 1/f performance.

In theory, that limit is set by the RC network. In practice, there is a source of 1/f after that point. Getting rid of it is very problematic.

So, for a digital filter, where you have tons of crud on the rail, it probably is fine.

The output Z you mention is around the limit for that design. Some circuits need lower. We recently ran across one such circuit, and it was one that surprised us. But there was too much overshoot on the waveform with that circuit. No, bypass changes did not seem to help. That, coupled with 1/f problems, prompted us to use a more conventional scheme.

You don't want to know how much high-$ test equipment we had to use to find out all of this.

Jocko