Arghhh : Awefully bad noise figures in voltage DACs...

[Fred Bartoli]

More on my low power, low noise PSUs...

I have 2 kind of them :
- one which has to have less than 220nV RMS noise, measured through a 40Hz
HPF
- the other, 10µV RMS through a 1Hz HPF.

The system designer choosed a nice octuple voltage DAC.
Having no informations about noise in the DS, I made a noise model with what
I thought really bad figures, having low power bipolar process in mind :
100nV/root(Hz) and 30Hz 1/f corner, and also asked the manufacturer infos
about noise.

The figures just arrived. Would you believe measured figures (i.e. probably
typical) like 150nV/root(Hz), OK I was not so far, but an ignominous 7 to
10kHz 1/f corner (no mistype). Thanks CMOS...

Unbelievable, they manage to make my hard won noise budget skyrocket up to
2uV (starting from 160nV) and 30uV (starting from 2uV).
And of course I have settling time to achieve too.

Bummer!

Fortunately TI seems to do better with their quad DAC7616.

 

[John Larkin]

More on my low power, low noise PSUs...

I have 2 kind of them :
- one which has to have less than 220nV RMS noise, measured through a 40Hz
HPF
- the other, 10µV RMS through a 1Hz HPF.

The system designer choosed a nice octuple voltage DAC.
Having no informations about noise in the DS, I made a noise model with what
I thought really bad figures, having low power bipolar process in mind :
100nV/root(Hz) and 30Hz 1/f corner, and also asked the manufacturer infos
about noise.

The figures just arrived. Would you believe measured figures (i.e. probably
typical) like 150nV/root(Hz), OK I was not so far, but an ignominous 7 to
10kHz 1/f corner (no mistype). Thanks CMOS...

Unbelievable, they manage to make my hard won noise budget skyrocket up to
2uV (starting from 160nV) and 30uV (starting from 2uV).
And of course I have settling time to achieve too.

Bummer!

Fortunately TI seems to do better with their quad DAC7616.

Check for popcorn noise, too. This can be insidious... a DAC can look
great for hours or days, then shift or pulse randomly for a while.
I've seen 15 PPM jumps from a 20-bit DAC. After a while, A*****
D****** said, quote,

"Yes, they're out of spec. No, we won't take them back."


Hey, Jim, what can you tell us about popcorn noise?

John

 

[Jim Thompson]

Check for popcorn noise, too. This can be insidious... a DAC can look
great for hours or days, then shift or pulse randomly for a while.
I've seen 15 PPM jumps from a 20-bit DAC. After a while, A*****
D****** said, quote,

"Yes, they're out of spec. No, we won't take them back."


Hey, Jim, what can you tell us about popcorn noise?

John

Pop! Pop! Pop! Pop! Pop! Pop! Pop! ;-)

Comes from thin-base-region bipolars (super-beta devices).

...Jim Thompson

 

[Fred Bartoli]

"Fred Bartoli"

More on my low power, low noise PSUs...

I have 2 kind of them :
- one which has to have less than 220nV RMS noise, measured through a 40Hz
HPF
- the other, 10µV RMS through a 1Hz HPF.

The system designer choosed a nice octuple voltage DAC.
Having no informations about noise in the DS, I made a noise model with what
I thought really bad figures, having low power bipolar process in mind :
100nV/root(Hz) and 30Hz 1/f corner, and also asked the manufacturer infos
about noise.

The figures just arrived. Would you believe measured figures (i.e. probably
typical) like 150nV/root(Hz), OK I was not so far, but an ignominous 7 to
10kHz 1/f corner (no mistype). Thanks CMOS...

Unbelievable, they manage to make my hard won noise budget skyrocket up to
2uV (starting from 160nV) and 30uV (starting from 2uV).
And of course I have settling time to achieve too.

Bummer!

Fortunately TI seems to do better with their quad DAC7616.

Well, in a hurry I didn't see that the DAC7616 are 1.25V full scale output,
making noise increase accordingly.
Back to square one.

 

[John Larkin]

Well, in a hurry I didn't see that the DAC7616 are 1.25V full scale output,
making noise increase accordingly.
Back to square one.

Oh, as long as I'm raining on your parade, be aware that most serial
DACs have nasty clock feedthrough, so expect spikes and perhaps some
DC offset while you're clocking the data in.

For lowest noise, use an external ref and a DAC that requires an
external opamp. All you want inside the DAC is the switches and the
thinfilm resistors.

John

 

[Joerg]

Hi John,

Oh, as long as I'm raining on your parade, be aware that most serial
DACs have nasty clock feedthrough, so expect spikes and perhaps some
DC offset while you're clocking the data in.

For lowest noise, use an external ref and a DAC that requires an
external opamp. All you want inside the DAC is the switches and the
thinfilm resistors.

Or use a sample and hold. SD5400 are really nice for that purpose. The
S/H needs good RF isolation else the spikes barrel right through. No
autorouter here.

Regards, Joerg

 

[John Larkin]

Pop! Pop! Pop! Pop! Pop! Pop! Pop! ;-)

Comes from thin-base-region bipolars (super-beta devices).

...Jim Thompson

The part I was using was a CMOS DAC.

But 741s were notorious for popcorn noise.

John

 

[John Larkin]

Hi John,

Or use a sample and hold. SD5400 are really nice for that purpose. The
S/H needs good RF isolation else the spikes barrel right through. No
autorouter here.

Regards, Joerg

LT1597 has a built-in s/h deglitcher, probably better than anything
you can make from parts. It works great as long as you don't clock it
too fast. Also external reference, external opamp, nearly ideal for
this app maybe.

John

 

[Joerg]

Hi John,

LT1597 has a built-in s/h deglitcher, probably better than anything
you can make from parts. It works great as long as you don't clock it
too fast. Also external reference, external opamp, nearly ideal for
this app maybe.

Yes, but they want almost $20 for it. Ouch.

Regards, Joerg

 

[Joerg]

Hi John,

The part I was using was a CMOS DAC.

But 741s were notorious for popcorn noise.

The loudest popcorn noise I heard was when my very first board in my
career was stuffed and they reversed all the tantalum caps. There was a
5V 100A supply in the test system....

Regards, Joerg

 

[Keith Williams]

Hi John,

The loudest popcorn noise I heard was when my very first board in my
career was stuffed and they reversed all the tantalum caps. There was a
5V 100A supply in the test system....

I had the exact same experience in the mid '70s. I'd describe it as
shotgun noise though. They went off about 2" from my ear. ...smelled
yummy too.

 

[Ken Smith]

Hi John,

The loudest popcorn noise I heard was when my very first board in my
career was stuffed and they reversed all the tantalum caps. There was a
5V 100A supply in the test system....

I've got you beat:

Imagine a board with about 100 dip packages on it. All the chips were
powered by an LM7805 style regulator running from +28V. There was no fuse
in this +28V wire and the battery could more than easily provided enough
current for what I'm about to discribe.

One second Dino is holding a board with 100 chips on it. The next second,
there are no chips on the board and black marks on the ceiling. Where the
chips were are just lead frames with the ends melted. The board is pitted
and Dino as a very odd look on his face and decides to leave the room.