Maker Pro
Maker Pro

Signalhound USB spectrum analyzer

M

Martin Riddle

Jan 1, 1970
0
Nico Coesel said:
I'm looking for a spectrum analyzer for EMC analyses. I've come across
this spectrum analyzer:
http://www.signalhound.eu/

Would this be a good buy or not?

Typical resolution BW is 300khz for EMC (or was). As long as your just
pre scanning and not measuring precise power levels, then 250khz should
be ok. You should use the rBW that CISPR specs out.

You might want to look to see if there is any automated scanning
capabilities. So you can scan from 10hz to 3ghz in one shot.
We had the old tek analyzer that did that, which was nice. With 300khz
BW the scan took a few minutes.

Cheers
 
M

mike

Jan 1, 1970
0
Nico said:
I'm looking for a spectrum analyzer for EMC analyses. I've come across
this spectrum analyzer:
http://www.signalhound.eu/

Would this be a good buy or not?
I'm a big fan of "fix it in software."
BUT
EMC is tricky business.
You need to be very aware of the testing methods
and equipment used for the qualification tests.
If you use different methodology/equipment, you need to
UNDERSTAND the differences/limits in that technology.

You can't just design in more margin just in case their
reading might be higher than yours...and still meet
your cost/schedule goals.

Failing the first qualification tests on your prototype
can be EXPENSIVE.
Failing the qualification tests on the first unit of
that 10,000 unit production build can be CATASTROPHIC.

Look at risk/reward and do the math.
If this is a toy for your bench, it can be very helpful
and save you trips across campus to your test facility.
If this is your ONLY test method, you've got some math to do.

Saving a few bux on the front end will get you a smile
from your boss.
Shutting down sales will get you a frown from personnel
at your exit interview. "Somebody on the web told me it would
work," will be of little help.
 
N

Nico Coesel

Jan 1, 1970
0
Too expensive. Look on eBay for some old AVCOM units.
Of course, you don't supply enough information for us to really help
you...

What puzzles me most is that people pay hundreds of Dollars for broken
20+ years old equipment and over $2000 for a 15 years old HP analyzer
without any warranties.

All this while you can buy a new analog Atten Spectrum analyzer (with
TG) for about $700 and a couple of hundreds $ more buys you the USB
analyzer I mentioned. So either people are paying way too much for old
gear or the new gear I mentioned has serious flaws.
 
N

Nico Coesel

Jan 1, 1970
0
mike said:
I'm a big fan of "fix it in software."
BUT
EMC is tricky business.
You need to be very aware of the testing methods
and equipment used for the qualification tests.
If you use different methodology/equipment, you need to
UNDERSTAND the differences/limits in that technology.

You can't just design in more margin just in case their
reading might be higher than yours...and still meet
your cost/schedule goals.

Failing the first qualification tests on your prototype
can be EXPENSIVE.
Failing the qualification tests on the first unit of
that 10,000 unit production build can be CATASTROPHIC.

Look at risk/reward and do the math.
If this is a toy for your bench, it can be very helpful
and save you trips across campus to your test facility.

I have been out of campus for a while :)
If this is your ONLY test method, you've got some math to do.

My prefered route is to find & fix potential problems, take a
pre-compliance test, fix things and then go for the final test. And
fix things. The 'problem' is that I'm on my own now so I need to
invest in a spectrum analyzer. I don't want to depend on
lending/renting one. Since you can spend money only once I'd like to
spend it as wisely as possible.
 
D

Do I really need to say?

Jan 1, 1970
0
Of course, some of us actually need good performance. My HP 8568B has
-110 dBc phase noise at 1 kHz offset--good luck getting that with a USB
gizmo.

I am quite certain that the USB part is merely the interface to the
software applet that allows you to display the data that the actually
device gathers. So, it comes down to that device, and the 'gizmo' is
likely quite a lot smaller and quite a lot less refined than your HP.
The HP has decades of knowledge behind it. The gizmo is a stab at that
same capability and feature set.

But it isn't due to the fact that it uses a computer and USB to look at
the data. It relates to how the data is gathered by the 'gizmo' itself.
There is a chance it is better, but not likely.
But do keep thinking that way...that's what keeps boat anchors so
delightfully cheap!

Sorry, but I don't think you thought that one through too well. :)

The HP is far heavier than the gizmo. Hell, the gizmo would probably
float if it were a sealed case. Certainly no anchor.

Anyway, the new Agilent stuff is cool. High function and good looks.
They look nice in the racks next to the other equipment.
 
A

Andrew Holme

Jan 1, 1970
0
Nico Coesel said:
[snip]

What puzzles me most is that people pay hundreds of Dollars for broken
20+ years old equipment and over $2000 for a 15 years old HP analyzer
without any warranties.

All this while you can buy a new analog Atten Spectrum analyzer (with
TG) for about $700 and a couple of hundreds $ more buys you the USB
analyzer I mentioned. So either people are paying way too much for old
gear or the new gear I mentioned has serious flaws.

They don't quote its phase noise performance in the specification; however,
the FAQ says:
<<<
My phase noise is higher with the internal oscillator than with an external
timebase. Why is this?
The internal TCXO has higher phase noise than a precision external
reference. At 1 GHz RF, you typically experience -55 dBc/Hz @ 10 Hz offset
using the internal timebase. This typically improves to -83 dBc/Hz with an
external timebase.
Below this it also says:

<<<
I get some spurs when I am a few KHz away from 1 GHz. Is this normal?
Multiples of 10 MHz have known residual and spurious intermodulation
behavior. Residuals as high as -85 dBm (0 dB attenuator) may be observed.
Spurs as high as -45 dBc may also be observed at certain frequencies when
your RF signal is at a high level, near but not exactly on a multiple of 10
MHz.

If you can, avoid these areas. If not, try adding a 20 dB EXTERNAL
attenuator and set the internal attenuator to 0 dB. If this still does not
reduce the spur to an acceptable level, you may try to use an external
timebase set to a frequency of e.g. 10.02 MHz. This will turn off the
internal TCXO and avoid harmonics at exact 10 MHz multiples.
Sounds like a fractional-N synthesizer with integer-N boundary spurs.

My 20+ year old Marconi 2383 SA has better performance in these areas.
 
J

Joerg

Jan 1, 1970
0
At $700 the best you can usually get with a new unit is a non-PLL
oscillator in there. Which will likely have the stability of Camembert
cheese on a hot summer day. The Signal Hound might have a PLL though.
But, quote "There is also a thermometer for temperature corrections,
allowing accurate amplitude measurements over the entire operating
temperature range". Umm, yeah, well, I have never had or designed RF
gear where the gain would noticeably drift with temperature ...
Of course, some of us actually need good performance. My HP 8568B has
-110 dBc phase noise at 1 kHz offset--good luck getting that with a USB
gizmo.

Yep. I couldn't have said it more succinctly. For example, the
performance of the HP3585 baseband analyzed is IMHO unrivaled to this
day. So is the Tektronix 2465 scope.

But do keep thinking that way...that's what keeps boat anchors so
delightfully cheap!

Right on!
 
M

mike

Jan 1, 1970
0
Nico said:
I have been out of campus for a while :)


My prefered route is to find & fix potential problems, take a
pre-compliance test, fix things and then go for the final test. And
fix things. The 'problem' is that I'm on my own now so I need to
invest in a spectrum analyzer. I don't want to depend on
lending/renting one. Since you can spend money only once I'd like to
spend it as wisely as possible.
My point, exactly. Spending wisely is not always the same as "cheapest
thing I can find." You really need to understand how the thing
works and whether the thing produces equivalent results in the
desired application. You'll find out at the first EMC test, but
you've already spent the $$ by then.

I worry about "artifacts" in a cheap, doitinsoftware solution.
You certainly CAN doitinsoftware, but not cheaply.

The thing that may save your bacon is the small dynamic range you need.
If you set the reference level for the standards limit, you really
only need 20dB of dynamic range. You don't even care about the accuracy
or artifacts of stuff much lower than the limit. All this assumes
that the digitizing is done AFTER a lot of input bandwidth filtering.
Othrwise, you can't even count on small dynamic range.

I'd consult the vendor and ask them for some testimonials from people
using their product for EMC pretesting.
 
J

Jon Kirwan

Jan 1, 1970
0
<snip>
So is the Tektronix 2465 scope.

I've been considering the idea of buying a Tek 2465B and had
found this link, earlier. Now that you bring this up to my
mind again, I'd like to post the link and some of the text
there:

http://www.tek.com/forum/viewtopic.php?f=5&t=4368

"I bought a 2465B (or so it appears) from a calibration
outfit in CA USA. The unit had fresh calibration.
Attempting to measure a signal in the 400MHz range, I
didn't like what I saw. I opened the case (again ;-)),
and noticed at the A1 PCB part number - 671-0720-08 -a
2445B PCB, not a 2465B PCB.

"I queried the seller, and this is what he claims:

'Tek made 2445B and 2465B with only one internal
difference: a 2445B has a bandwidth choke on the
backside of its main A1 board etched into the
printed circuitry of the bare board. This was
also true of the 2445A 2465A pair, and the
original 2445 2465 pair introduced in 1983.

'For the privilege of not having that bandpass
limiter, Tek charged clients $2000 more, but was
too lazy to make 2 really different scopes. So
since 1985, when I discovered Tek's laziness, I
have earned my living upgrading the 3rd digit
from 4 to 6.

'The clock stays with the A5 board. The 2445B A1
board, once it has its bandwidth choke bypassed
(I cut traces in 4 places on the back and spliced
in 2 short straight wires to carry the signals
from the DL100 delay line to the inputs of U600),
can do 400MHz.

'The only other difference is to ground a contact
(pin 23 J511) that tells the A5 processor board
that it is now has a 6 instead of a 4 (as 3rd
digit) that allows the 5nS/cm fastest sweep speed
to be accessible, instead of just going up to
10nS/cm.

'This is what many of my 2465B clients have, but
so far you are the only one to have noticed... '

"I looked at the circuits, and discovered a 3rd difference
- the 2465B PCB has inductor L403, apparently for
adjusting the HF circuit.

"Leaving aside the apparent legal ramifications of what
he has been doing for 25 years, without disclosing to
his customers, I would like to learn from the technical
people here:

"1. If there are any other technical differences between
the 2445B and 2465B A1 PCB's?
2. If the integrity of the A1 PCB has been compromised by
cutting the PCB traces?
3. If there is any spurious interference of any kind
caused by the new spliced wires?
4. If this modified 2445B A1 PCB is, in fact,
functionally equivalent to a real 671-0722-08 2465B A1
PCB (disregarding Tek's marketing considerations), and
if not, why not?"

Jon
 
J

Joerg

Jan 1, 1970
0
Nico said:
I have been out of campus for a while :)

I have. And man did I feel old there :)

My prefered route is to find & fix potential problems, take a
pre-compliance test, fix things and then go for the final test. And
fix things. The 'problem' is that I'm on my own now so I need to
invest in a spectrum analyzer. I don't want to depend on
lending/renting one. Since you can spend money only once I'd like to
spend it as wisely as possible.

Some advice here: Research any new analyzer well. RBW, sensitivity and
all that are not the whole game. A major factor is intermodulation, IOW
how sturdy the first mixer and preamp are.

In that context this remark gives me pause: Quote "No external power
supply needed - draws power from USB". Now that's only 2.5 watts max,
doesn't instill much confidence. When this little power is available how
can there be a nice fat DBM inside and a frontend with tons of dynamic
range? Maybe they designed in some miracle? Lack of raw dynamic range
can be utterly frustrating when you have to do noise hunts in the
vicinity of powerful RF sources, whether they are outside your house or
part of the DUT functionality (switch mode supplies et cetera).

Then there is the USB connection itself which can inject noise that can
mess with your measurements. But that is usually "debuggable". Make sure
you are allowed to ground the analyzer.

Longs story short, with the old HP boat anchors there isn't much
research that needs to be done, other than determining the confidence
level that you will receive a working unit where the first mixer didn't
have a head-on collision with a pulse yet. Newer Agilent gear sometimes
did not make such a great impression on me, I'd prefer the big old stuff
such as HP8566 over those. Of course, the disks in your spine won't ;-)
 
N

Nico Coesel

Jan 1, 1970
0
Joerg said:
I have. And man did I feel old there :)



Some advice here: Research any new analyzer well. RBW, sensitivity and
all that are not the whole game. A major factor is intermodulation, IOW
how sturdy the first mixer and preamp are.

In that context this remark gives me pause: Quote "No external power
supply needed - draws power from USB". Now that's only 2.5 watts max,
doesn't instill much confidence. When this little power is available how
can there be a nice fat DBM inside and a frontend with tons of dynamic
range? Maybe they designed in some miracle? Lack of raw dynamic range
can be utterly frustrating when you have to do noise hunts in the
vicinity of powerful RF sources, whether they are outside your house or
part of the DUT functionality (switch mode supplies et cetera).

Then there is the USB connection itself which can inject noise that can
mess with your measurements. But that is usually "debuggable". Make sure
you are allowed to ground the analyzer.

Longs story short, with the old HP boat anchors there isn't much
research that needs to be done, other than determining the confidence
level that you will receive a working unit where the first mixer didn't
have a head-on collision with a pulse yet. Newer Agilent gear sometimes
did not make such a great impression on me, I'd prefer the big old stuff
such as HP8566 over those. Of course, the disks in your spine won't ;-)

I have been eyeballing the HP8920 series mobile analyzers as well. Not
a great dynamic range but tons of other features. The problem is the
weight (about 32 kilo) which essentially makes it non-portable. I
almost bought one from Germany but I messed up the bidding on Ebay
(the seller must be happy though).
 
T

tm

Jan 1, 1970
0
Joel Koltner said:
The SignalHound's internal oscillator says its phase noise is
typically -55dBc/Hz at 10Hz out... I'd have to respectfully submit that
NO WAY does that drop to -110dBc/Hz by 1kHz as Phil's 8568B does. :)

You need to spend a lot of money to get the performance of a HP8568B or
8566B. It's only downside is the weight but then no one will walk out with
it.

If you get either one, be sure to check the display brightness as a new CRT
is around $500 and a LCD upgrade is $1800.

Get the cables also, they run about $500 a set.




tm
 
J

Joerg

Jan 1, 1970
0
That's like comparing the Greenbay Packers (or the Ajax soccer team in
Nico's case) with the boys from the local highschool :)
You need to spend a lot of money to get the performance of a HP8568B or
8566B. It's only downside is the weight but then no one will walk out
with it.

If you get either one, be sure to check the display brightness as a new CRT
is around $500 and a LCD upgrade is $1800.

Get the cables also, they run about $500 a set.

Even those things would not deter me at all. I mean, if someone offered
you a Rolls-Royce Corniche for $4k and it had a wee dent, would you turn
it down?
 
J

Joerg

Jan 1, 1970
0
Joel said:
Something like an HP 8561E from the '90s has significantly better specs
than a Signal Hound there... and if you really want to give your
chiropractor a lifetime of income, check out the 8568B.

The power utility will also be very happy :)

Or you can get something like an Agilent N9020A for $35k+ depending on
what options you want.

I'd rather get the good old stuff, HP-85xx. IMHO it was (is) more
useful. The topper in my case was a "modern" analyzer targeted at the
EMC market. Cost as much as a Lexus sedan. It oozed tons of radiated
emissions through its display.

[...]
 
M

Martin Riddle

Jan 1, 1970
0
Martin Riddle said:
Typical resolution BW is 300khz for EMC (or was). As long as your
just pre scanning and not measuring precise power levels, then 250khz
should be ok. You should use the rBW that CISPR specs out.

You might want to look to see if there is any automated scanning
capabilities. So you can scan from 10hz to 3ghz in one shot.
We had the old tek analyzer that did that, which was nice. With 300khz
BW the scan took a few minutes.

Cheers

FYI, the Signalhound has a free API that you could use to write an
application to do a EMC sweep from 10hz to 4.4Ghz.
But you need someone to write the application ;(

Cheers
 
Top