Maker Pro
Maker Pro

in-circuit spectral noise characterization?

B

BW

Jan 1, 1970
0
Hello!

When designing ultra low-noise PSU's and related circuits, what do you
use to characterize them with regards to noise ? One of my component
suppliers for example use an old HP 3571A Dynamic Signal Analyzer,
which is a DC-100 kHz spectrum analyzer, with a 50 ohm coax soldered
directly on top of the measuring-point at the PCB. Supposedly this
allows them to find (spectral) noise on the critical supply-rails on
the order of 10-100 microvolts.

Is this realistic ? I do need to measure noise at an absolute level of
<< 1 mV but it is OK to solder the measurement cable directly to the
design, I don't need to probe around.

Even if that setup works, the problem is that HP discontinued this
device long ago and the Agilent successor the HP35670A is also a very
old technology part. I'm sure there MUST be a modern way of measuring
this, but all spectrum analyzers I find now are geared towards RF
measurements, 3G/GSM/RFID demodulating etc., with frequency responses
of 9 kHz - 3 GHz but I need to go down to almost DC. I guess telecom
is a much bigger market...

Perhaps there are options to buy for oscilloscopes adding a low-signal
FFT mode and I'm sure the sales-guy or FAE would tell me just that,
but input from someone who has actually used it is more helpful of
course :)

I should add that the reason I'm looking for a spectrum-analyzer and
not just a very sensitive pre-amp for an oscilloscope is because my
parts that need power have different sensitivities to noise in
different frequency ranges.

Any recommendations would be received gratefully!

Brgds
/Bjorn
 
E

Eeyore

Jan 1, 1970
0
BW said:
Hello!

When designing ultra low-noise PSU's and related circuits, what do you
use to characterize them with regards to noise ?

With audio equipment I measure the audio band noise. Often anywhere
between 50 and 300uV depending on loading with good PCB layout and common
or garden regulators like the LM317/337 (use the ripple reduction
technique) and 78/79 series.

However it's almost moot since the circuitry isn't very sensitive to PSU
noise anyway if well designed.

Graham
 
J

John Devereux

Jan 1, 1970
0
BW said:
Hello!

When designing ultra low-noise PSU's and related circuits, what do you
use to characterize them with regards to noise ? One of my component
suppliers for example use an old HP 3571A Dynamic Signal Analyzer,
which is a DC-100 kHz spectrum analyzer, with a 50 ohm coax soldered
directly on top of the measuring-point at the PCB. Supposedly this
allows them to find (spectral) noise on the critical supply-rails on
the order of 10-100 microvolts.

Is this realistic ? I do need to measure noise at an absolute level of
<< 1 mV but it is OK to solder the measurement cable directly to the
design, I don't need to probe around.

Even if that setup works, the problem is that HP discontinued this
device long ago and the Agilent successor the HP35670A is also a very
old technology part. I'm sure there MUST be a modern way of measuring
this, but all spectrum analyzers I find now are geared towards RF
measurements, 3G/GSM/RFID demodulating etc., with frequency responses
of 9 kHz - 3 GHz but I need to go down to almost DC. I guess telecom
is a much bigger market...

Perhaps there are options to buy for oscilloscopes adding a low-signal
FFT mode and I'm sure the sales-guy or FAE would tell me just that,
but input from someone who has actually used it is more helpful of
course :)

I should add that the reason I'm looking for a spectrum-analyzer and
not just a very sensitive pre-amp for an oscilloscope is because my
parts that need power have different sensitivities to noise in
different frequency ranges.

Any recommendations would be received gratefully!


I think the key here is that measuring power supply noise is different
from a general purpose noise measurement, in that your souce impedance
should be *extremely* low. So you may be able to improve the noise
measurement by taking advantage of this. You can go all-out for very
low voltage noise, without worrying about current noise (usually you
have to trade-off).

You could make an ultra-low noise preamplifier, perhaps using one of
the very low (voltage) noise opamps such as LT1028 or AD797. And/Or
use a transformer to step up the noise voltage. I seem to recall a
discussion here about using multiple parallel jfets to get really low
voltage noise for power supply noise measurements - was one of the
regulars here (but brain not working today unfortunately).

Anyway, then feed the output into your HP35670A and make measurements
as usual, subtracting the gain of your preamp of course.
 
J

John Devereux

Jan 1, 1970
0
John Devereux said:
I think the key here is that measuring power supply noise is different
from a general purpose noise measurement, in that your souce impedance
should be *extremely* low. So you may be able to improve the noise
measurement by taking advantage of this. You can go all-out for very
low voltage noise, without worrying about current noise (usually you
have to trade-off).

You could make an ultra-low noise preamplifier, perhaps using one of
the very low (voltage) noise opamps such as LT1028 or AD797. And/Or
use a transformer to step up the noise voltage. I seem to recall a
discussion here about using multiple parallel jfets to get really low
voltage noise for power supply noise measurements - was one of the
regulars here (but brain not working today unfortunately).

I found the thread now; it was Fred Bartoli I was thinking of.

<http://groups.google.com/group/sci...._frm/thread/f66601473612d5e1/020b5a3302a728f0>

(This may be overkill for you).
 
Hello!

When designing ultra low-noise PSU's and related circuits, what do you
use to characterize them with regards to noise ? One of my component
suppliers for example use an old HP 3571A Dynamic Signal Analyzer,
which is a DC-100 kHz spectrum analyzer, with a 50 ohm coax soldered
directly on top of the measuring-point at the PCB. Supposedly this
allows them to find (spectral) noise on the critical supply-rails on
the order of 10-100 microvolts.

Is this realistic ? I do need to measure noise at an absolute level of
<< 1 mV but it is OK to solder the measurement cable directly to the
design, I don't need to probe around.

Even if that setup works, the problem is that HP discontinued this
device long ago and the Agilent successor the HP35670A is also a very
old technology part. I'm sure there MUST be a modern way of measuring
this, but all spectrum analyzers I find now are geared towards RF
measurements, 3G/GSM/RFID demodulating etc., with frequency responses
of 9 kHz - 3 GHz but I need to go down to almost DC. I guess telecom
is a much bigger market...

Perhaps there are options to buy for oscilloscopes adding a low-signal
FFT mode and I'm sure the sales-guy or FAE would tell me just that,
but input from someone who has actually used it is more helpful of
course :)

I should add that the reason I'm looking for a spectrum-analyzer and
not just a very sensitive pre-amp for an oscilloscope is because my
parts that need power have different sensitivities to noise in
different frequency ranges.

Any recommendations would be received gratefully!

Brgds
/Bjorn

So what exactly is wrong with an old HP DSA if it still works? Most
DSOs have 8 bit sample heads, which is kind of coarse for noise
measurements.

I never used one of their DSAs, but SRS makes products like the older
HP DSAs.
http://www.thinksrs.com
 
B

BW

Jan 1, 1970
0
So what exactly is wrong with an old HP DSA if it still works? Most
DSOs have 8 bit sample heads, which is kind of coarse for noise
measurements.

I never used one of their DSAs, but SRS makes products like the older
HP DSAs.http://www.thinksrs.com

Hi!

Agilent sells a replacement for the old HP, but that is also a very
old piece
of equipment by now. It probably works fine for this purpose, but the
thing
that bugs me is that there has to be a more modern way of measuring
noise in bias-supplies and power-supplies in noise-critical circuits
than this.
Perhaps there isn't, I don't know. Sometimes an old "work-horse"
sticks
because the market changes so the modern instruments are optimized for
other tasks (i.e. telecom).

Thank you for the link to SRS, I'll check that out as well. Also, yes,
my feeling
as well was that oscilloscopes are not really useable for this.

/Bjorn
 
Tek has some nice FFT spectrum analyzers that do 40 MHz at a time, and
work down to (really) DC.  (The Agilent ones don't.)  I have both HP3562
(true double channel) and HP35660A (fake double channel) DSAs, and am
often frustrated by the 102.4 kHz maximum frequency.

Cheers,

Phil Hobbs

I have a HP 3562 also. While it doesn't do DC, it goes very low in
frequency. I don't have the specs handy, but something like 60uHz
resolution comes to mind. Yeah, the high end limit sucks.
 
Top