Spikes on bench-top power supplies

[Joerg]

Hello Folks,

Finally found out what killed my prototypes: Bench-top power supplies,
and not just one but several brands. Jeff L had hinted that in another
thread. The first one was a luxury edition (Lambda) yet it could not
stomach 30dB load changes. Voltage dipped, then regulated back and
overshot ... POOF.

So be careful.

John, if you are reading this, IIRC you mentioned these:
http://www.lascarelectronics.com/PR...04e3ece25-842A2914-15C5-5FA1-624EDD23F58C010D

Did you try load changes on them?

 

[Frank Bemelman]

Hello Folks,

Finally found out what killed my prototypes: Bench-top power supplies,
and not just one but several brands. Jeff L had hinted that in another
thread. The first one was a luxury edition (Lambda) yet it could not
stomach 30dB load changes. Voltage dipped, then regulated back and
overshot ... POOF.

What currents? Long during overshoot?

So be careful.

It has never bitten me. Sometimes, on dry winterdays, I had
times that everything latched up under my nose, fiddling
with a stubborn prototype. I'm more worried about static
discharges.

Damn, you made me even more paranoid than I already am

 

[Ancient_Hacker]

Joerg wrote:
Voltage dipped, then regulated back and

overshot ... POOF.

That's why one should always follow Bob Pease's advice-- check every
power supply under pulsed load. Get your heftiest MOSFET, hook it up
to a pulse generator, put a big resistor on the drain and hook it up to
the PS. It's just *amazing* the things you'll see
on your scope. If it's your PS design, you'll likely need to add a
few RC rolloff networks to improve the transient response. Think of
the PS as a big DC amplifier that needs taming.

Another source of glitches-- line glitches upsetting the regulator:

The worst example was, hmmm, may have been a Lamba PS. 5 volts at 60
amps, with a SCR overvoltage clamp set to 5.5 volts. Turning on a
vacuum cleaner in the next room would trip the crowbar every time!

 

[Joerg]

Hello Frank,

What currents? Long during overshoot?

It already happenened going from 10uA to 0.3mA but this design can also
jump from 10uA to 10mA. The regulator chip on the circuit went with a
plume of smoke and a stench. No problem with a battery.

I don't remember the timing but one supply was pumping a spike of less
than a msec. Enough to drive stuff into latch-up or other destructive
modes. Another supply actually began to oscillate.

Damn, you made me even more paranoid than I already am

Sorry, didn't want to cause that reaction...

 

[Joerg]

Ancient_Hacker wrote:

Voltage dipped, then regulated back and



That's why one should always follow Bob Pease's advice-- check every
power supply under pulsed load. Get your heftiest MOSFET, hook it up
to a pulse generator, put a big resistor on the drain and hook it up to
the PS. It's just *amazing* the things you'll see
on your scope. If it's your PS design, you'll likely need to add a
few RC rolloff networks to improve the transient response. Think of
the PS as a big DC amplifier that needs taming.

We just figured a way to do it with batteries until we can find a decent
power supply. If we can find one, that is. This takes those risks out of
the equation.

Another source of glitches-- line glitches upsetting the regulator:

The worst example was, hmmm, may have been a Lamba PS. 5 volts at 60
amps, with a SCR overvoltage clamp set to 5.5 volts. Turning on a
vacuum cleaner in the next room would trip the crowbar every time!

Now that is a true sign of a poor design. Did they ever go through
susceptibility tests or did they somehow sneak by?

 

[John Larkin]

Hello Folks,

Finally found out what killed my prototypes: Bench-top power supplies,
and not just one but several brands. Jeff L had hinted that in another
thread. The first one was a luxury edition (Lambda) yet it could not
stomach 30dB load changes. Voltage dipped, then regulated back and
overshot ... POOF.

I *hate* it when that happens.

So be careful.

John, if you are reading this, IIRC you mentioned these:
http://www.lascarelectronics.com/PR...04e3ece25-842A2914-15C5-5FA1-624EDD23F58C010D

Did you try load changes on them?

No, but I'll give it a check. It's a switcher, so I wouldn't expect it
to be very good. It does kick out typical switcher noise.

John

 

[Joerg]

Hello John,

I *hate* it when that happens.

What really surprised me was that while the "high tech" supplies
misbehaved I had tested all these prototypes with one of my old Russky
power supplies. No problems whatsoever and I even disconnected and
re-connected a whole unit on occasions without turning down the supply.

<boast_mode>
And I might add that none of my own power supply designs ever did that,
including the switchers.

No, but I'll give it a check. It's a switcher, so I wouldn't expect it
to be very good. It does kick out typical switcher noise.

That wouldn't be a problem for one of my clients as long as the feedback
loop in there is properly designed.

A lot of people might wonder why a supply should have such a wide mains
range. It can be very practical for folks who travel a lot or must debug
foreign designs. Once had I almost fried the Philips FET probe. It got
really hot but miraculously survived.

 

[Joel Kolstad]

And I might add that none of my own power supply designs ever did that,
including the switchers.

Yeah, but I bet you "cheated" and actually *tested* the things before you
declared them as ready for prime time!

Perhaps hardware design is heading the direction of software design -- little
initial testing, figuring that faults can always be fixed after the product is
released.

 

[Spehro Pefhany]

Ancient_Hacker wrote:



We just figured a way to do it with batteries until we can find a decent
power supply. If we can find one, that is. This takes those risks out of
the equation.



Now that is a true sign of a poor design. Did they ever go through
susceptibility tests or did they somehow sneak by?

Susceptibility tests? What applicable US standard for a bench power
supply requires those?


Best regards,
Spehro Pefhany

 

[Joerg]

Hello Joel,

Yeah, but I bet you "cheated" and actually *tested* the things before you
declared them as ready for prime time!

Yes, per your definition I then cheated

Perhaps hardware design is heading the direction of software design -- little
initial testing, figuring that faults can always be fixed after the product is
released.

Sure, but I am a bit disappointed that rather expensive lab power
supplies would exhibit such behavior. I could tolerate a deviation
downwards but when a supply spikes upwards that is bound to cause grief.
Expensive grief at times.

 

[Joerg]

Hello Spehro,

Susceptibility tests? What applicable US standard for a bench power
supply requires those?

Three things would require that:

Liability: If someone hires a lawyer because they found out what fried
that multi-million Dollar linear accelerator this could require the V.P.
of Quality Control to seek a prescription for Prozac.

Reputation: When it is found out that a top of the line piece of
equipment does not behave top of the line chances are pretty slim that
the next pieces of equipment are bought from that manufacturer.

CE: Lab gear is (usually) marketed world-wide. If the thing misbehaves
in terms of EMI, something nasty happens and it turns out that in
hindsight self-cert wasn't such a good idea, oh boy.

 

[Frank Bemelman]

Hello Frank,


It already happenened going from 10uA to 0.3mA but this design can also
jump from 10uA to 10mA. The regulator chip on the circuit went with a
plume of smoke and a stench. No problem with a battery.

I don't remember the timing but one supply was pumping a spike of less
than a msec. Enough to drive stuff into latch-up or other destructive
modes. Another supply actually began to oscillate.


Sorry, didn't want to cause that reaction...

Well, I checked my bench supplies, no problem whatsoever. A tiny
dip at most, and even tinier overshoot. All < 0.1V. Another
30 minutes wasted. Yes, I looked for <1mS events.

 

[Joerg]

Hello Frank,

Well, I checked my bench supplies, no problem whatsoever. A tiny
dip at most, and even tinier overshoot. All < 0.1V. ...

Did you have a capacitive load connected for the test? Almost all
circuits present that type of load because of the bypass caps.

... Another
30 minutes wasted. Yes, I looked for <1mS events.

That's not my philosophy. In the same way that I do not consider the 30
minutes it takes to check the brake pads on a car "wasted".

 

[Frank Bemelman]

Hello Frank,



Did you have a capacitive load connected for the test? Almost all
circuits present that type of load because of the bypass caps.

No, but I would guess there is a small bypass cap on the
output anyway. You're not talking me into another 30 minutes
of my time wasted? Sheesh, would you please provide us with
accurate test models other than simply complaining you see funny
things when current jumps between 10uA and 10mA. Next thing
you're telling us you had 20 miles of cables between target
and supply. Or admit your prototype caught fire because of
some other stupid mistake. At this rate, your credibility drops
below zero. Correct that, *has* dropped below zero.

That's not my philosophy. In the same way that I do not consider the 30
minutes it takes to check the brake pads on a car "wasted".

Well it was your philosphy on power supplies, otherwise you would
not have complained about it.

About brakes; I have that task delegated to the mechanics at the
garage. At least that assures me it is carried out properly.

 

[Joerg]

Hello Frank,

Well it was your philosphy on power supplies, otherwise you would
not have complained about it.

Nope. I spent several hours testing the supplies here in the lab before
anything electronic was connected to one of them. This included testing
with various capacitive loads as it is highly unusual (but not
impossible) that an electronic circuit boards does not contain bypass
caps. These supplies passed. The ones at the client didn't but I have no
control over power supplies that are hundreds of miles away until I get
there

About brakes; I have that task delegated to the mechanics at the
garage. At least that assures me it is carried out properly.

Good. You don't have to spend the 30 minutes for a check-up yourself,
you can certainly pay someone else who is competent to do it. Not doing
that on a lab supply is like playing roulette.

 

[Aristotle Eisenglas]

What really surprised me was that while the "high tech" supplies
misbehaved I had tested all these prototypes with one of my old Russky
power supplies. No problems whatsoever and I even disconnected and
re-connected a whole unit on occasions without turning down the supply.
<boast_mode>
And I might add that none of my own power supply designs ever did that,

There's state and local "government" too, with the same as the
old saying goes "time is money". For sample requests I certainly
understand . But not banquets (banquets are puddings - sweet
courses) Mary Yes. We have a huge retirement community just
south of here and a lot easier to work with that every day, as a
Lineman. Any insight into this problem would be solved, but
coped with..." Shimon Peres Best, James Arthur That's not what I
want, the experiment or about which I have not studied their
history.

I have a 0.1uF cap across the power supply because.... Tell me
about these real estate taxes that apply in the land of the
free, Jim ! Graham Woman on first trip to Europe, seeing a
bidet: "Oh, is that for washing babies in?" Concierge: "No,
madam, that is for washing babies in?" Concierge: "No, madam,
that is for washing babies out." Cheers! Rich hmmm see here
martin Guys thanks again for all your answers I just took that
job.

Except it is in San Diego. The other relevant point which
appears not to be a grandma I've seen that with our chicks
recently, it's amazing to watch! Mary Harvesting - hunting -
isn't the same as 5,000 years ago is niave All things are
probably possible But its not a general given AGREED I have no
idea how to drive the 281's pin.