How do you measure filter offset

[Mike Scott-7]

>I beg to differ. Based on rather careful measurements, my 400 Hz
>8-pole filter needs an offset of .045 to center it with the DSP
>filters.

I don't understand what the filter offset is really doing. My 2.7 KHz filter
for example is -0.87. I assume this means 870 Hz offset from something. What
reference is the offset from? Does this mean that the IF frequency is
actually shifted from some nominal value?

Could someone explain a method of measuring roofing filter offset? I have
spectrum analysis software. Soon [I hope] I will have the second receiver
and I will possibly be setting up a compromise offset for two filters. I
paid a little extra for matching filters but it would be good to do some
measuring.


Mike Scott - AE6WA
Tarzana, CA (DM04 / near LA)
K3-100 #508/ KX1  #1311


_______________________________________________
Elecraft mailing list
Post to: [hidden email]
You must be a subscriber to post to the list.
Subscriber Info (Addr. Change, sub, unsub etc.):
 http://mailman.qth.net/mailman/listinfo/elecraft   

Help: http://mailman.qth.net/subscribers.htm
Elecraft web page: http://www.elecraft.com

[Jim Brown-10]

On Tue, 19 Aug 2008 06:01:35 -0700, Mike Scott wrote:

>Could someone explain a method of measuring roofing filter offset?
>I have spectrum analysis software.

Once you've studied the manual thoroughly and learned to use the
radio, it will become obvious. First, the FFT system must be set for
high resolution scan rates. I use a professional audio FFT (EASERA
SysTun) with band noise as the signal source and lots of averaging. I
measured and stored plots of the DSP filters, then measured and stored
again with the K3 set so that the roofing filter kicks in with the DSP
filters set to be very broad band. Then I overlaid those plots. It's
important to measure at the headphones output -- the Line Out can
overload at relatively modest signal levels, as noted in the manual.

These offsets are nothing more or less than manufacturing tolerances
in the filters. Ain't no big thing, since the K3 can easily be
adjusted to compensate. But it's wrong to deny that they exist.

73,

Jim Brown K9YC



_______________________________________________
Elecraft mailing list
Post to: [hidden email]
You must be a subscriber to post to the list.
Subscriber Info (Addr. Change, sub, unsub etc.):
 http://mailman.qth.net/mailman/listinfo/elecraft   

Help: http://mailman.qth.net/subscribers.htm
Elecraft web page: http://www.elecraft.com

[Ed Muns, W0YK]

> I don't understand what the filter offset is really doing. My
> 2.7 KHz filter for example is -0.87. I assume this means 870
> Hz offset from something. What reference is the offset from?
> Does this mean that the IF frequency is actually shifted from
> some nominal value?

If the filter is not centered on the IF, then the resultant offset can be
entered into the filter configuration in the CONFIG menu so the firmware can
make the correction so that the filter is centered on the IF.

> Could someone explain a method of measuring roofing filter
> offset? I have spectrum analysis software. Soon [I hope] I
> will have the second receiver and I will possibly be setting
> up a compromise offset for two filters. I paid a little extra
> for matching filters but it would be good to do some measuring.

One way is to use an audio spectrum program like Spectrogram which has been
used widely for setting up K2s and by some to look at their K3 IF response.
There are better, more accurate and/or higher resolution equivalents.  K9YC
used a professional program with high quality hardware to produce the
measurements he shared with us.

Another method, more laborious, but fairly accurate is to use the Elecraft
XG1 or XG2 to inject a stable signal at the antenna port and deploy the dbV
function in the K3 to measure the amplitude every 10 Hz or so.  Plot the
data manually or by entering into a spreadsheet that produces a graphic of
the IF response.

73,
Ed - W0YK

_______________________________________________
Elecraft mailing list
Post to: [hidden email]
You must be a subscriber to post to the list.
Subscriber Info (Addr. Change, sub, unsub etc.):
 http://mailman.qth.net/mailman/listinfo/elecraft   

Help: http://mailman.qth.net/subscribers.htm
Elecraft web page: http://www.elecraft.com

[Bill W4ZV]

Ed Muns, W0YK wrote

Another method, more laborious, but fairly accurate is to use the Elecraft
XG1 or XG2 to inject a stable signal at the antenna port and deploy the dbV
function in the K3 to measure the amplitude every 10 Hz or so.  Plot the
data manually or by entering into a spreadsheet that produces a graphic of
the IF response.

You only need the two -6 dB points, which can be measured to a resolution of 1 Hz and 0.1 dB using the internal AFV/dBV function.  You should set AGC Off and ensure WIDTH (i.e. the DSP BW) is >> than the filter you are measuring...otherwise you're measuring the combination of XFIL plus DSP.  I do this by disabling all filters except the one under test and setting WIDTH >> XFIL BW.  Once you have the two -6 dB points, then you can figure how much offset to enter so these two points are equidistant about the center.

73,  Bill

[Ed Muns, W0YK]

> You only need the two -6 dB points, which can be measured to
> a resolution of
> 1 Hz and 0.1 dB using the internal AFV/dBV function.

This is true if the filter passband is flat, but I've found both ripple and
slope in some INRAD filter passbands, so I always want to see the entire
sweep to make sure I'm finding the true bandwidth.

> You
> should set AGC Off and ensure WIDTH (i.e. the DSP BW) is >>
> than the filter you are measuring...otherwise you're
> measuring the combination of XFIL plus DSP.  I do this by
> disabling all filters except the one under test and setting WIDTH
> >> XFIL BW.  Once you have the two -6 dB points, then you can
> figure how
> much offset to enter so these two points are equidistant
> about the center.

Yes.  Thanks.  These are important points to get valid measurements.  Also
make sure you are within the passband of the K3 audio circuitry.  For
example, it rolls off below 300 Hz.  And, of course, that the RX equalizer
needs to be flat.

73,
Ed - W0YK

_______________________________________________
Elecraft mailing list
Post to: [hidden email]
You must be a subscriber to post to the list.
Subscriber Info (Addr. Change, sub, unsub etc.):
 http://mailman.qth.net/mailman/listinfo/elecraft   

Help: http://mailman.qth.net/subscribers.htm
Elecraft web page: http://www.elecraft.com

[Jim Brown-10]

On Tue, 19 Aug 2008 11:18:54 -0700 (PDT), Bill W4ZV wrote:

>You only need the two -6 dB points

That's looking at the elephant through a pinhole. It can also be
very useful to know what happens much further down the filter slope
-- say at the -20 and -30 dB points.

73,

Jim K9YC



_______________________________________________
Elecraft mailing list
Post to: [hidden email]
You must be a subscriber to post to the list.
Subscriber Info (Addr. Change, sub, unsub etc.):
 http://mailman.qth.net/mailman/listinfo/elecraft   

Help: http://mailman.qth.net/subscribers.htm
Elecraft web page: http://www.elecraft.com

[Bill W4ZV]

Jim Brown-10 wrote

On Tue, 19 Aug 2008 11:18:54 -0700 (PDT), Bill W4ZV wrote:

>You only need the two -6 dB points

That's looking at the elephant through a pinhole. It can also be
very useful to know what happens much further down the filter slope
-- say at the -20 and -30 dB points.

Well why not the -60 to -80 dB points?  :-)  Two points are entirely sufficient to **center** the filter's 6 dB bandwidth, which was the question asked.  Regarding Ed's comment about ripple, we'll never see 6 dB ripple excursions unless there is a defective filter.  The filters I've checked can have a very slightly asymmetrical response (we're talking tenths of a dB) but I set mine so that -6 dB points are equidistant from the center.

73,  Bill

[Jim Brown-10]

On Tue, 19 Aug 2008 13:26:25 -0700 (PDT), Bill W4ZV wrote:

>Well why not the -60 to -80 dB points?  :-)  

Because -35 dB or so is the best I can do. If you lived close to W0YK, you'd
WANT to know what's happening at -60dB!  :)

>Two points are entirely
>sufficient to **center** the filter's 6 dB bandwidth, which was the question
>asked.

I don't think -6dB bandwidth was the question, but rather how to measure the
filter. Whatever -- a single number description is quite limited.

>Regarding Ed's comment about ripple, we'll never see 6 dB ripple
>excursions unless there is a defective filter.

I've not seen enough going on in the passband to even begin to call it
ripple. BUT -- when I'm measuring something, the more detail I can see the
more I know about it. The fact that the filters ARE quite flat in passband is
something that I want to know!  Having read LOTS of product data sheets that
looked like they were drawn by an artist rather than measured, its quite
encouraging to see a product that actually measures like it's data sheet!

73,

Jim Brown K9YC


_______________________________________________
Elecraft mailing list
Post to: [hidden email]
You must be a subscriber to post to the list.
Subscriber Info (Addr. Change, sub, unsub etc.):
 http://mailman.qth.net/mailman/listinfo/elecraft   

Help: http://mailman.qth.net/subscribers.htm
Elecraft web page: http://www.elecraft.com

[Bill W4ZV]

On Tue, 19 Aug 2008 13:26:25 -0700 (PDT), Bill W4ZV wrote:

>Well why not the -60 to -80 dB points?  :-)  

Because -35 dB or so is the best I can do. If you lived close to W0YK, you'd
WANT to know what's happening at -60dB!  :)

        It actually doesn't matter beyond about 30 dB down ina roofing filter...the DSP filters handle things beyond that.

>Two points are entirely
>sufficient to **center** the filter's 6 dB bandwidth, which was the question
>asked.

I don't think -6dB bandwidth was the question, but rather how to measure the
filter. Whatever -- a single number description is quite limited.

        The question was how to center filters...not make detailed plots of them.

>Regarding Ed's comment about ripple, we'll never see 6 dB ripple
>excursions unless there is a defective filter.

I've not seen enough going on in the passband to even begin to call it
ripple. BUT -- when I'm measuring something, the more detail I can see the
more I know about it. The fact that the filters ARE quite flat in passband is
something that I want to know!  Having read LOTS of product data sheets that
looked like they were drawn by an artist rather than measured, its quite
encouraging to see a product that actually measures like it's data sheet!

        If you want full plots of 8-pole filters, go below and click on "See Details" under each filter (I think the 250 is missing but the others are there):

http://www.inrad.net/home.php?cat=140

...and plots of the 5-pole 200 and 500 are shown here:

http://f1.grp.yahoofs.com/v1/YEKrSIMok74B_mjAO8WAb3OqdauIV7jhVEWFK5zAcLSr3JJoFH79RRz_dbxeVUNQicFDDsQu2hGN42lvYFAIIJUmHw/Elecraft%205%20Xtal%20CW-filters%20for%20K3.jpg

73,  Bill

[W8JI]

>        It actually doesn't matter beyond about 30 dB down
> ina roofing
> filter...the DSP filters handle things beyond that.
>
>>Two points are entirely
>>sufficient to **center** the filter's 6 dB bandwidth,
>>which was the
> question
>>asked.
>
> I don't think -6dB bandwidth was the question, but rather
> how to measure the
> filter. Whatever -- a single number description is quite
> limited.

If I wanted to center and match a typical SSB or CW filter
for passband centering,  I'd use something like 5 or 10 dB.
I wouldn't match them at a higher attenuation because it
really doesn't mean anything and can be very misleading and
cause the actual useful centers to be misaligned.

It really should just be at some attenuation that is much
greater than the ripple amount, but not so far down on the
skirt it is meaningless for the actual passband.

73 Tom

_______________________________________________
Elecraft mailing list
Post to: [hidden email]
You must be a subscriber to post to the list.
Subscriber Info (Addr. Change, sub, unsub etc.):
 http://mailman.qth.net/mailman/listinfo/elecraft   

Help: http://mailman.qth.net/subscribers.htm
Elecraft web page: http://www.elecraft.com

[Bill W4ZV]

Tom W8JI wrote

If I wanted to center and match a typical SSB or CW filter
for passband centering,  I'd use something like 5 or 10 dB.
I wouldn't match them at a higher attenuation because it
really doesn't mean anything and can be very misleading and
cause the actual useful centers to be misaligned.

It really should just be at some attenuation that is much
greater than the ripple amount, but not so far down on the
skirt it is meaningless for the actual passband.

There is very little ripple in either the 5-pole or 8-pole filters (as can be seen in the previous plots), so -6 dB is well beyond any ripple effect.

I'll write up a simple procedure later and post it under a more descriptive Subject.  All you need to do this is a stable signal (like an XG1), the K3's internal AFV/dBV meter and the VFO.  It's not necessary to use Spectrogram and can actually be done more accurately using the internal K3's internal capability.

[Gary, W7TEA]

Using this method I tweaked the Config menu filter offset adjustments for
my 500hz and 200hz filters.  It seemed to make a difference  centering
the passband, especially on the narrower filter.  

Although my K3 was factory built, I have found that
the REF CAL freq, S meter and now filter offsets all needed a bit of
tweaking.  

tnx and 73,  Gary W7TEA

Bill W4ZV wrote

Ed Muns, W0YK wrote

Another method, more laborious, but fairly accurate is to use the Elecraft
XG1 or XG2 to inject a stable signal at the antenna port and deploy the dbV
function in the K3 to measure the amplitude every 10 Hz or so.  Plot the
data manually or by entering into a spreadsheet that produces a graphic of
the IF response.

You only need the two -6 dB points, which can be measured to a resolution of 1 Hz and 0.1 dB using the internal AFV/dBV function.  You should set AGC Off and ensure WIDTH (i.e. the DSP BW) is >> than the filter you are measuring...otherwise you're measuring the combination of XFIL plus DSP.  I do this by disabling all filters except the one under test and setting WIDTH >> XFIL BW.  Once you have the two -6 dB points, then you can figure how much offset to enter so these two points are equidistant about the center.

73,  Bill