Maker Pro
Maker Pro

7805 failure mode

J

John Larkin

Jan 1, 1970
0
AFAIK they have no reverse discharge diode protection. Might have been
killed on the first turn-off. Anyhow, I have pretty much completely
adopted the LM317 for almost everything positive. It can take at least
modest spikes in reverse.

Check out LM1117. It's almost indestructable.

John
 
J

Joerg

Jan 1, 1970
0
John said:
Check out LM1117. It's almost indestructable.

Quote "The ESR of the output capacitor should range between 0.3? - 22?."

Nah, thanks, I think I'll pass.
 
S

Spehro Pefhany

Jan 1, 1970
0
Check out LM1117. It's almost indestructable.

John

Not much input voltage capability to handle transients- 20V, and (like
all LDOs) it's picky about output caps. I still like the 78M05 for
applications where it fills the bill. I've never seen a non-mechanical
failure.


Best regards,
Spehro Pefhany
 
J

John Larkin

Jan 1, 1970
0
Not much input voltage capability to handle transients- 20V, and (like
all LDOs) it's picky about output caps. I still like the 78M05 for
applications where it fills the bill. I've never seen a non-mechanical
failure.

I tested some National 1117's as power amps, with an opamp driving the
adj pin, and specifically did really abusive voltage diffs between the
adjust pin and others, and couldn't blow it up. It's not a real LDO,
as it has an emitter-follower output so it has an inherently low
output impedance. They seem happy with ceramic output caps. The
National datasheet says that ESR should be "below 0.5 ohms."

An LM317 won't quite do a 5.0 to 3.3 volt regulator, but an LM1117
will.

To get the 1.2 and 2.5 volts for a Spartan FPGA, you can ground the
adj pin on the lower regulator, and bootstrap the adj of the upper one
to the 1.25 output of the lower one, and get 1.25 and 2.5 with no
resistors!

John
 
J

Joerg

Jan 1, 1970
0
John said:
I tested some National 1117's as power amps, with an opamp driving the
adj pin, and specifically did really abusive voltage diffs between the
adjust pin and others, and couldn't blow it up. It's not a real LDO,
as it has an emitter-follower output so it has an inherently low
output impedance. They seem happy with ceramic output caps. The
National datasheet says that ESR should be "below 0.5 ohms."

Nope. Sez 0.3 to 22 Ohms on page 9, paragraph 1.3 Whenever I read
something like that goose bumps appear.

At least they openly stated it. For a LM29xx series device there was no
mention of some of the pathologies and it became my first black eye from
LDOs. Didn't like it when Zin went above about 100 ohms, phsssst ...
poof. The answer was pretty much "oops, sorry".
 
J

Joerg

Jan 1, 1970
0
Spehro said:
Well *now* they state it (current (April 2006) datasheet date, and
June 2004 and October 2002 datsheet) <looking...>

Aha! But *not* in the February 2000 datasheet!

Obviously they found a problem; IIRC in 2000 a 10uF ceramic cap was
pretty much unheard of.

Often a company modifies a datasheet after a customer found a problem.
About half way through university I believed datasheets were perfect.
Until I found a hardcore bug in one, then another, and another. This
includes some top notch manufacturers. Typical scenario, call into app
engineering, a group gathers at the other end, more engineers get pulled
in, the noise increases, the crackle of lots of paper is heard, then
suddenly one hears a muffled "Oh s..t!"

WRT low dropout regulators that whole category is on my banned list no
matter from which mfg. I might make an exception if I really, really
have to. Like if the earth would stop turning if I didn't, or so.

[...]
 
J

Jim Thompson

Jan 1, 1970
0
On Tue, 24 Jul 2007 21:47:32 GMT, Joerg

[snip]
Often a company modifies a datasheet after a customer found a problem.
About half way through university I believed datasheets were perfect.
Until I found a hardcore bug in one, then another, and another. This
includes some top notch manufacturers. Typical scenario, call into app
engineering, a group gathers at the other end, more engineers get pulled
in, the noise increases, the crackle of lots of paper is heard, then
suddenly one hears a muffled "Oh s..t!"

WRT low dropout regulators that whole category is on my banned list no
matter from which mfg. I might make an exception if I really, really
have to. Like if the earth would stop turning if I didn't, or so.

[...]

Crap like the current crop of LDO's are the result of customer demand
(*)... minimum number of pins and minimum external components.

With an extra pin, compensating an LDO is trivial.

(*) I run into this from customers all the time, but I tend to be my
usual nasty self and tell them that their request is fine... it's no
skin off my back if it oscillates ;-)

...Jim Thompson
 
V

Vladimir Vassilevsky

Jan 1, 1970
0
Joerg wrote:

About half way through university I believed datasheets were perfect.

What datasheets?

In the old good times, you could have one handbook for *all* electronic
components ever made by all companies. And only a small part of those
components was really available. The handbooks did have few mistakes,
and those mistakes where well known.
Until I found a hardcore bug in one, then another, and another. This
includes some top notch manufacturers. Typical scenario, call into app
engineering, a group gathers at the other end, more engineers get pulled
in, the noise increases, the crackle of lots of paper is heard, then
suddenly one hears a muffled "Oh s..t!"

It is very unpleasant to discover bugs in the top vendor's hardware or
software. But in our days everybody seems to loose any sense of shame.
AD Blackfin errata has a couple of lines from me.
WRT low dropout regulators that whole category is on my banned list no
matter from which mfg. I might make an exception if I really, really
have to. Like if the earth would stop turning if I didn't, or so.

You should forgive them for your own sake :)



Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com
 
J

John Larkin

Jan 1, 1970
0
Well *now* they state it (current (April 2006) datasheet date, and
June 2004 and October 2002 datsheet) <looking...>

Aha! But *not* in the February 2000 datasheet!

Obviously they found a problem; IIRC in 2000 a 10uF ceramic cap was
pretty much unheard of.

http://www.futurlec.com/Datasheet/Linear/LM1117T.pdf


Best regards,
Spehro Pefhany

The older datasheets say 0.5 ohms max, and that larger C values just
improve stability. The new ds says 0.3 to 22 ohms, silly on its face,
but also says that larger caps "merely improve loop stability", which
is silly based on 0.5r min.

Anyhow, they seem happy with 22u ceramics or aluminums. But maybe I
should do a little more stability testing.

Lately we avoid alums because of temperature, and tants because of
explosions.

It's not clear if 317's are stable with ceramics. All the caps on the
datasheet are footnoted as solid tantalums.

John
 
S

Spehro Pefhany

Jan 1, 1970
0
I tested some National 1117's as power amps, with an opamp driving the
adj pin, and specifically did really abusive voltage diffs between the
adjust pin and others, and couldn't blow it up. It's not a real LDO,
as it has an emitter-follower output so it has an inherently low
output impedance. They seem happy with ceramic output caps. The
National datasheet says that ESR should be "below 0.5 ohms."

Current datasheet says it must be within the range of 0.3 ohms to 22
ohms, and at least 10uF (* page 9, 1.3). Presumably (??) it will be
stable with ceramic caps if you have a larger electrolytic in
parallel.
An LM317 won't quite do a 5.0 to 3.3 volt regulator, but an LM1117
will.

Yes. That's where I'd use it.
To get the 1.2 and 2.5 volts for a Spartan FPGA, you can ground the
adj pin on the lower regulator, and bootstrap the adj of the upper one
to the 1.25 output of the lower one, and get 1.25 and 2.5 with no
resistors!

John

That's cool. The 'sequencing' might even be better that way.

*
http://cache.national.com/ds/LM/LM1117.pdf


Best regards,
Spehro Pefhany
 
S

Spehro Pefhany

Jan 1, 1970
0
Nope. Sez 0.3 to 22 Ohms on page 9, paragraph 1.3 Whenever I read
something like that goose bumps appear.

At least they openly stated it.

Well *now* they state it (current (April 2006) datasheet date, and
June 2004 and October 2002 datsheet) <looking...>

Aha! But *not* in the February 2000 datasheet!

Obviously they found a problem; IIRC in 2000 a 10uF ceramic cap was
pretty much unheard of.

http://www.futurlec.com/Datasheet/Linear/LM1117T.pdf


Best regards,
Spehro Pefhany
 
J

Joerg

Jan 1, 1970
0
Vladimir said:
Joerg wrote:




What datasheets?

In the old good times, you could have one handbook for *all* electronic
components ever made by all companies. And only a small part of those
components was really available. The handbooks did have few mistakes,
and those mistakes where well known.

In was in a databook alright but the bug was not well known. At least
not to the manufacturer ....

It is very unpleasant to discover bugs in the top vendor's hardware or
software. But in our days everybody seems to loose any sense of shame.
AD Blackfin errata has a couple of lines from me.

There may soon be others from me but for the ADN8831 TEC controller chip.

You should forgive them for your own sake :)

Oh, I forgave, otherwise I should not go to communion on Sundays. But
that doesn't mean I won't be careful and a bit suspicious in the future.
So I got used to whipping up a switcher whenever there wasn't enough
dropout voltage available.
 
J

Joerg

Jan 1, 1970
0
John said:
The older datasheets say 0.5 ohms max, and that larger C values just
improve stability. The new ds says 0.3 to 22 ohms, silly on its face,
but also says that larger caps "merely improve loop stability", which
is silly based on 0.5r min.

Anyhow, they seem happy with 22u ceramics or aluminums. But maybe I
should do a little more stability testing.

Try that over the full expected temperature range including the low end
(unit unloaded from truck in northern Alaska or Siberia and turned on
right away). There can be lots of surprises.
 
D

DaveM

Jan 1, 1970
0
Winfield said:
Paul said:
Winfield Hill wrote:

[snip]
One possible clue: when the 7805 was first
used in this circuit, a hand-wired prototype,
the ground pin was left floating. But when
the failure mode was observed this pin was
properly grounded.

Some clarification needed: by 'the ground pin' you are
referring to the 7805 common pin, correct?
Yes.

When the pin was properly grounded, did the 7805 resume normal
operation? If so, I wouldn't call what you experienced
a 'failure'. It would be 'unspecified behaviour'.

It operates normally, except for the short turn-on
voltage spike every time the power is applied.

Wonder what happens if you were to drive the common pin with a low impedance
square wave source? Do you see the spikes on the edges of the square wave?
How wide are the spikes?
--
Dave M
MasonDG44 at comcast dot net (Just substitute the appropriate characters in the
address)

"In theory, there isn't any difference between theory and practice. In
practice, there is." - Yogi Berra
 
J

John Larkin

Jan 1, 1970
0
Try that over the full expected temperature range including the low end
(unit unloaded from truck in northern Alaska or Siberia and turned on
right away). There can be lots of surprises.

I tried a National LM1117 with various loads and outputs caps. I'm
injecting a load current step and looking at the transient response.

No cap at all looks fine.

Any tantalum looks good. 10uF looks fine.

It rings with ceramics. Very roughly,

10 uF 50 KHz Q=3

22 uF 40 KHz Q=2

44 uF 20 KHz Q=4

At 22 uF, the esr for critical damping is about 0.2 ohms.


Temperature and load variations don't change things much.

John
 
J

Joerg

Jan 1, 1970
0
John said:
I tried a National LM1117 with various loads and outputs caps. I'm
injecting a load current step and looking at the transient response.

No cap at all looks fine.

Any tantalum looks good. 10uF looks fine.

It rings with ceramics. Very roughly,

10 uF 50 KHz Q=3

22 uF 40 KHz Q=2

44 uF 20 KHz Q=4

At 22 uF, the esr for critical damping is about 0.2 ohms.

Interesting. At least it didn't oscillate. Although ringing can cause
brown-out effects in hotrod 3.3V logic or even fry it.
Temperature and load variations don't change things much.

Did you go below freezing?
 
S

Spehro Pefhany

Jan 1, 1970
0
Interesting. At least it didn't oscillate. Although ringing can cause
brown-out effects in hotrod 3.3V logic or even fry it.

The last LDO I designed in oscillated beautifully if you left out the
10uF e-cap (just with bypass caps). I know because the proto assembly
guy mixed up two regulators, one LDO and one not.
Did you go below freezing?

Best regards,
Spehro Pefhany
 
J

John Larkin

Jan 1, 1970
0
Interesting. At least it didn't oscillate. Although ringing can cause
brown-out effects in hotrod 3.3V logic or even fry it.


Did you go below freezing?

Yup. My freeze spray hits -50C.

John
 
J

John Larkin

Jan 1, 1970
0
The last LDO I designed in oscillated beautifully if you left out the
10uF e-cap (just with bypass caps). I know because the proto assembly
guy mixed up two regulators, one LDO and one not.

Some of the newer ldo's are specifically ceramic-stable.

I don't understand exactly why all regs aren't stable for any
reasonable c or any reasonable c+esr. Must have to do with the
available range of internal compensation parts.

John
 
W

Winfield

Jan 1, 1970
0
Jim said:
With an extra pin, compensating an LDO is trivial.

(*) I run into this from customers all the time, but I tend
to be my usual nasty self and tell them that their request
is fine... it's no skin off my back if it oscillates ;-)

Perhaps you should come up with an effective capacitance
multiplier, etc., and internally compensate your circuits.
 
Top