Are fast transient response regulators worth using?

I've got an analogue circuit with a gain of a few thousand. I note that the 78M05 and 79M05 regulators providing the +/-5V rails have a couple of mV ripple at a few kHz (the input signal being amplified, causing increased drain in IC's down the line) and I'm concerned it will couple into other, more sensitive parts of the circuit. I've tried LM317 / LM337 regs but they don't seem any better, despite a 10uF decoupler on their ref pins, perhaps because their ripple rejection is intended for mains frequencies rather than kHz. I've noticed that there are some expensive regulators touted as "fast transient response" around, do people find these significantly reduce supply ripple, i.e. coupling between IC's in analogue circuits?

Reply to
Nemo
Loading thread data ...

Regulators have an inductive output characteristic, which wants to resonate with the output capacitor. That can cause some nasty noise peaks in the 5-100 kHz range, depending on the size of the output capacitor and the amount of current being drawn.

A low-ohm resistor in series with the output will de-Q the resonance, and can sometimes help a lot.

For sensitive circuitry you're often much better off using a capacitance multiplier. Op amps have practically infinite CMR at low frequencies, where the cap multiplier's vague idea of an output voltage occurs, but are really horrible up in the tens to hundreds of kHz, where a good cap multiplier is like a brick wall.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal
 Click to see the full signature
Reply to
Phil Hobbs

"Fast transient response" is desirable when you want to maintain regulation in the face of drastic changes in current drain. That's not where you're at.

In normal analog circuits, "coupling between stages" via power supply buses in analog circuits is usually not addressed solely through solid state voltage regulation. That's not exactly supply ripple.

Simple RC (or LC, or RLC) decoupling of each stage from the power supply buses can be remarkably effective. It might be nothing more than a 100 ohm series resistor and an electrolytic capacitor bypass.

Most op-amps have substantial power supply rejection ratios already anyway.

It is very possible that your issues have more to do with, say, PCB layout or ground loops than with power supply regulation. The ripple you're seeing on the scope on the power supply, is likely not the cause of your problem, but just a symptom of the problem.

Tim.

Reply to
Tim Shoppa

Nemo, are you sure this ripple ain't coming from somewhere other than the regulators and is just leaking in? Sub-optimal ground structures come to mind.

LDOs can really sing the blues and go into full-blown oscillation up there. Many of those are like the princess on the pea.

I'll second that. Plus cap multipliers are dirt cheap.

--
Regards, Joerg

http://www.analogconsultants.com/
 Click to see the full signature
Reply to
Joerg

Having designed LDO chips in bipolar and MOS, I can say the MOS devices are far superior in terms of noise and stability. A bipolar device near saturation is plain ugly. There are all sorts of gimicks in the error amp to get around how the BJT behaves near saturation, but they are pure kludges. The mos (P-fet pass) is far simpler to design. As far as PSRR, a well designed P-fet pass LDO will settle out to the capacitor divider comprised of the Cds of the pass device and the load bypass cap. You should be able to look at the datasheet typ app performance and see the PSRR versus frequency go flat. If the rejection continuously decreases with increasing frequency, then the error amp is leaking into the pass device, i.e. a less than optimal design.

Reply to
miso

Thank you for the replies.

Some further background: I've already split the circuit into 3 supply zones, each with their own C-multiplier on the + and - rails. These C-multipliers give out about +/-9V so there's plenty of headroom for the regulators, though Miso's comments about DMOS being better than bipolar for LDO's are interesting anyway. The main reason for using +/-5V rails is simply that some of the ICs aren't specified as being able to take the +/-9V coming out of the C-multipliers.

This mid section, with a gain of ~2000 isn't as critical as the front end but I saw the ripple and thought it could be hiding a big problem. Your feedback has helped me calm down a lot. Also, when I did the sums, I realised the ripple simply corresponds to about 1.5 ohms impedance in the supply lines. My next layout will have less impedance in the power tracks (there's already a ground plane), and I will increase decoupling after the regulators - there's only about 30uF there right now because, mainly, of headroom problems but I'll reassess that.

I'll try a few ohms in the output of the regs, add filtering to each critical IC's power, and relax. I thought I was seeing some subtle and major gotcha, but your comments have helped me get this in perspective.

Cheers!

Reply to
Nemo

I'm a little confused about what the problem is, but if you are trying to reduce power supply noise, then I'd put the cap-multipliers after the voltage regualtors. Use some of the LM317's(337) to make (perhaps) +/- 6 volt supplies, and filter that with the cap- mult's.

Is the 'ripple' synchronous with the signal you are amplifying? Or is it random noise? Or something else?

Stay calm, George H.

Reply to
George Herold

I hadn't considered it that way round. But you're right, the drop across the C-multipliers is a pretty constant volt or so. That's a great idea.

It's most noticeable when a 2.5kHz test oscillator on the board is used. The ripple doesn't matter then, but I was concerned it was symptomatic of possible coupling when real signals go through. I had problems on an earlier iteration of the board where I derived a reference voltage from a supply rail, rather than a dedicated 25 cent reference IC, and it picked up a ~300kHz clock from an IC I've since designed out. In effect I was mixing analog and digital supply rails. So, having seen it once, I am a bit wary of echoes of the signal feeding back to an earlier stage of the circuit and causing a self-reinforcing instability. That's why I split the circuit into 3 sections with individually filtered rails.

Reply to
Nemo

Reply to
garyr

ow

t
e

peaks

nd

nd

e
,

are

Did ya try google?

formatting link

George H.

Reply to
George Herold

As long as you're multiplying caps, start with a nice polymer aluminum, like 180 uF and a few milliohms of ESR. And another on the c-mult output gobbles up residual feedthrough.

John

Reply to
John Larkin

Or else use higher beta transistors that let jack up the resistor and use ceramics. Two pole multipliers are better.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal
 Click to see the full signature
Reply to
Phil Hobbs

Fast transient regulators are used with high-performance MCUs and FPGA systems. As others have mentioned, you won't get any benefit from them in your application.

Leon

Reply to
Leon

Strong word of caution here: Do not use the opamp variant to feed a voltage rail that has bypass capacitances on it. Opamps generally do not like capacitive loads and might sing the blues. Or worst case ... phut ... *POOF*

--
Regards, Joerg

http://www.analogconsultants.com/
 Click to see the full signature
Reply to
Joerg

And while at it, use low noise transistors with a low 1/f knee if that matters. The BCX70K comes to mind.

--
Regards, Joerg

http://www.analogconsultants.com/
 Click to see the full signature
Reply to
Joerg

. How

hat

le

ple

on

or

e peaks

and

and

nce

es,

t are

Hmm, I didn't see the opamp variation... What's the purpose of that? I thought the cap multiplier was a bjt transistor circiut.

George H.

Reply to
George Herold

That's my favorite bipolar transistor, beta graded yet. 2N7002 is my favorite mosfet; it will switch 50 volts in under 1 ns.

John

Reply to
John Larkin

Well, in theory the opamp works. In reality, not so well if there's capacitance on the output.

--
Regards, Joerg

http://www.analogconsultants.com/
 Click to see the full signature
Reply to
Joerg

is. How

ge

that

uple

ng

ouple

/

r on

for

e
o
g

ise peaks

or and

e, and

tance

cies,

but are

ot

I must be slow today. But the opamp circuit makes no sense to me. It's a simple buffered low pass but... "What's the purpose of R2?" It's either working against the opamp output impedance to feed some of the noisy input signal to the output. Or it's postive feedback working against the output impedance of whatever is sourcing the input voltage Vi. Or I'm missing something?

(Let's forget any capacitance on the output for the time being.)

George H.

Reply to
George Herold

The 2N7002's threshold is too high for most of my uses. I've recently been using the Diodes DMG1012T (cheaper and smaller than the IRLML2402). As far as bipolars go, I rarely use anything other than the MMBT3904/6.

Reply to
krw

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.