More Math for the "SETUP"

[Fred Abse]

On Wed, 05 Feb 2014 13:37:53 -0800, Fred Abse wrote:

> So, what's the problem? I see three possibilities:

Further thought - excessive oscillator harmonics on the Boonton (tired
tubes, etc).

Maybe on your sig gen,too.

That's where a vector voltmeter scores, it's frequency selective.


--
"Design is the reverse of analysis"
(R.D. Middlebrook)

 

[amdx]

On 2/5/2014 3:37 PM, Fred Abse wrote:

On Mon, 03 Feb 2014 10:53:50 -0600, amdx wrote:

If you have any ideas where the errors are creeping in,
other than scope magnitude calibration, scope timebase calibration,
errors in reading the scope, Rsense tolerance and parasitics, and not
knowing the real characteristics of my DUT coils.

Let us look at it from a different angle. What *should* your inductor look
like on your test setup, assuming that your Boonton measurements are
accurate.

Firstly, we'll allow for the probe capacitance. You say that is 14pF.

14pF in parallel with 47.5 ohms , at 3.85MHz looks like 47.49ohms, in
series with 54.1 nF. (Series to parallel impedance conversion). It's
always a good idea to start with everything series, it simplifies
calculation.

That's an resistance of 47.49 ohms, in series with a capacitive
reactance of 0.7641 ohms.

That's an impedance of 47.49 - j0.7641.

We'll call that Zs

Your 5 uH inductor has an inductive reactance of 1330.4645ohms, in series
with a resistance of 1330.4645/250 = 5.2319 ohms.

Just a small correction, 55uh inductor.-

That's an impedance of 5.2319 + j1330.4645.

We'll call that Zl

The whole setup is a potential divider, whose ratio is Zs/(Zl+Zs)

With 10V input, the voltage at the probe is 10Zs/(Zl+Zs).

That is 10 * (47.49 -j0.7641)/(5.2319 +j1330.4645 +47.49 -j0.7641 ) volts

Or 10 * (47.49 -j0.7641)/(52.8119 +j1329.7004) volts

Which comes to 10 * (0.00084 +j0.03568) volts

Or .0084 +j0.3568 volts, which equals 0.3569volts, angle 88.6473 degrees.

That's what you *should* measure at the sensing resistor.

E.&O.E.

Thanks for all that, I'll print it out for further study.
I need to review my numbers but at first look I followed it
up to division of the Imaginary numbers, but I just have to do it
to see how that works.
The only thing I would want different is I start knowing the
measurement at the sensing resistor. I want to work backwards to get ZL.

I did a spice simulation as well (You can have it if you like). It agrees
with the above hand calculation to within 3 decimal places. Using both
parallel probe capacitance, and the series conversion.

So, what's the problem? I see three possibilities:

Scope calibration, it's possible that the two channels you are using,
don't match. Swapping them should show that.

I have checked and swapped the probe/channels, the eye can see a
difference when overlaid on each other.

Have you checked both

channels against a known voltage? DC will do. Don't trust the scope's own
calibrator.

I have not checked against a known voltage, this has it's own
problem, as you switch attenuation, I'd need several known voltage
sources. But where I checked it, the channels read the same.

Reading error, How did you determine that the delay, or phase you're
measuring was accurate, IOW, how did you establish zero crossings?

I use peak and also center the skirts around a centerline, ie.
adjust vert position so two division down from peak the the waveform
crosses equal distant from a center line.
But, I admit this is an easy area for error to show.

Boonton Q meter inaccurate. That's what I'd go for, first. Have you
verified its calibration?

No, I haven't.
In fact while back I saw a method to check 3db points on the Q meter
to test accuracy of the unit.

Don't trust 50-year-old standard inductors,
But it's so pretty! I wonder if there is any helium left inside.



> the proper way to do it involves only measuring voltages. See the manual.

Thanks, Mikek

 

[amdx]

On 2/5/2014 4:13 PM, Fred Abse wrote:

On Wed, 05 Feb 2014 13:37:53 -0800, Fred Abse wrote:

So, what's the problem? I see three possibilities:

Further thought - excessive oscillator harmonics on the Boonton (tired
tubes, etc).

Maybe on your sig gen,too.

That's where a vector voltmeter scores, it's frequency selective.

Ya, where I used to work we used 7 element Chebychev low pass
filters to make sure our harmonics were way down when doing these
measurements.

Mikek

 

[amdx]

On 2/5/2014 3:40 PM, Fred Abse wrote:

On Mon, 03 Feb 2014 07:40:28 -0600, amdx wrote:

On 2/3/2014 3:13 AM, Fred Abse wrote:
On Sun, 02 Feb 2014 20:19:32 -0600, amdx wrote:

You should try that calculation again 1281/33 = 38.8, but we know the
33 ohms is bogus.

Just spotted that, my mistake, big Sunday lunch

33.8 is in the ballpark for the inductor you describe, at your test
frequency.

We can disagree on that. As measured on my Q meter the Q at 3.85MHz
is 196.


That's a feasible value.




The inductor is one I put together, 7/16" polypropylene tube wrapped
with 90 turns of 660/46 litz wire, I have an 8" ferrite rod that I slide
in or out to adjust inductance. It is variable from 8uH to 247uH. The Q
at 472kHz is 550!

What spacing and length. I'd like to try to FEMM it.

I by happy to give you all the information, might be a little tough
characterizing 660/46 litz wire.

Sounds like you don't believe 550 at 475kHz.
It is in that ball park. To the error of my Q meter.

Let's say, I'm skeptical.


A ferrite core should bring the Q *down*, significantly.

I agree, ferrites add loss, but there is a trade off between the extra
wire, the additional interwinding capacitance caused by more wire and
the additional losses caused by proximity effect because of more turns
on an air core.
Ben Tongue has a ferrite rod inductor that has a Q over 875 over the
entire AMBCB.

See table 2.
http://www.bentongue.com/xtalset/29MxQFL/29MxQFL.html



Never heard of the guy before.
Ben Tongue was one of the founders of Blonder Tongue
http://www.blondertongue.com/

He states that FEMM doesn't accommodate
> litz wire. My copy certainly does...

I don't know, that paper was written in 2006,
maybe it didn't then.
Oh, it accomidates litz? Ok 660/46 close wound.
I'll send you an email and you can tell me what details
you need. I think the rod is # 61, but I don't know for sure.
According to Ben, some rods of the same material have much
lower losses than others, even out of the same batch.

Thanks, Mikek

 

[Fred Abse]

On Wed, 05 Feb 2014 16:51:49 -0600, amdx wrote:

On 2/5/2014 3:37 PM, Fred Abse wrote:
On Mon, 03 Feb 2014 10:53:50 -0600, amdx wrote:

Your 5 uH inductor has an inductive reactance of 1330.4645ohms, in
series with a resistance of 1330.4645/250 = 5.2319 ohms.

Just a small correction, 55uh inductor.-

Typo, sorry, the math is still correct.

Thanks for all that, I'll print it out for further study.
I need to review my numbers but at first look I followed it
up to division of the Imaginary numbers, but I just have to do it to see
how that works.

Division of complex numbers:

(a+jb)/(c+jd)

Multiply top and bottom by the conjugate of the denominator:

(a+jb)/(c-jd)/(c+jd)(c-jd) = (a+jb)(c-jd)/(c^2 -jcd +jcd -j^2d^2)

= (a+jb)(c-jd)/ (c^2 -j^2d^2)

j^2=-1, so we have (a+jb)(c-jd)/(c^2 + d^2)

Or (ac -jad +jbc -j^2bd) / (c^2 + d^2) = (ac +bd +j(bc - ad)) / (c^2 + d^2)

= ((ac +bd) / (c^2 +d^2)) +j((bc -ad) / c^2 + d^2))

We now have a real fraction, and an imaginary fraction, with only real
numbers in the denominators.

E.&O.E.

The only thing I would want different is I start knowing
the
measurement at the sensing resistor. I want to work backwards to get ZL.

That's left as an exercise

What surprises me is that you didn't ask how to do series-parallel, and
parallel-series impedance conversions. You now have the information to
work it out, see if you can.

If you're going to do much of this sort of thing, I'd advise getting a
calculator that handles complex numbers, it's so much easier.

I did a spice simulation as well (You can have it if you like). It
agrees with the above hand calculation to within 3 decimal places.
Using both parallel probe capacitance, and the series conversion.

So, what's the problem? I see three possibilities:

Scope calibration, it's possible that the two channels you are using,
don't match. Swapping them should show that.

I have checked and swapped the probe/channels, the eye can see a
difference when overlaid on each other.

Aha!

Have you checked both
channels against a known voltage? DC will do. Don't trust the scope's
own calibrator.

I have not checked against a known voltage, this has it's own
problem, as you switch attenuation, I'd need several known voltage
sources. But where I checked it, the channels read the same.

Reading error, How did you determine that the delay, or phase you're
measuring was accurate, IOW, how did you establish zero crossings?

I use peak and also center the skirts around a centerline, ie.
adjust vert position so two division down from peak the the waveform
crosses equal distant from a center line.
But, I admit this is an easy area for error to show.


Boonton Q meter inaccurate. That's what I'd go for, first. Have you
verified its calibration?

No, I haven't.
In fact while back I saw a method to check 3db points on the Q meter to
test accuracy of the unit.

Don't trust 50-year-old standard inductors, But it's so pretty! I wonder
if there is any helium left inside.

Helium diffuses through just about anything.

the proper way to do it involves only measuring voltages. See the
manual.

--
"Design is the reverse of analysis"
(R.D. Middlebrook)

 

[amdx]

On 2/6/2014 1:21 PM, Fred Abse wrote:

On Wed, 05 Feb 2014 16:51:49 -0600, amdx wrote:

On 2/5/2014 3:37 PM, Fred Abse wrote:
On Mon, 03 Feb 2014 10:53:50 -0600, amdx wrote:

Your 5 uH inductor has an inductive reactance of 1330.4645ohms, in
series with a resistance of 1330.4645/250 = 5.2319 ohms.

Just a small correction, 55uh inductor.-

Typo, sorry, the math is still correct.




Thanks for all that, I'll print it out for further study.
I need to review my numbers but at first look I followed it
up to division of the Imaginary numbers, but I just have to do it to see
how that works.

Division of complex numbers:

(a+jb)/(c+jd)

Multiply top and bottom by the conjugate of the denominator:

(a+jb)/(c-jd)/(c+jd)(c-jd) = (a+jb)(c-jd)/(c^2 -jcd +jcd -j^2d^2)

= (a+jb)(c-jd)/ (c^2 -j^2d^2)

j^2=-1, so we have (a+jb)(c-jd)/(c^2 + d^2)

Or (ac -jad +jbc -j^2bd) / (c^2 + d^2) = (ac +bd +j(bc - ad)) / (c^2 + d^2)

= ((ac +bd) / (c^2 +d^2)) +j((bc -ad) / c^2 + d^2))

We now have a real fraction, and an imaginary fraction, with only real
numbers in the denominators.

E.&O.E.

The only thing I would want different is I start knowing
the
measurement at the sensing resistor. I want to work backwards to get ZL.

That's left as an exercise

What surprises me is that you didn't ask how to do series-parallel, and
parallel-series impedance conversions. You now have the information to
work it out, see if you can.

If you're going to do much of this sort of thing, I'd advise getting a
calculator that handles complex numbers, it's so much easier.




I did a spice simulation as well (You can have it if you like). It
agrees with the above hand calculation to within 3 decimal places.
Using both parallel probe capacitance, and the series conversion.

So, what's the problem? I see three possibilities:

Scope calibration, it's possible that the two channels you are using,
don't match. Swapping them should show that.

I have checked and swapped the probe/channels, the eye can see a
difference when overlaid on each other.

Aha!

Oops, the eye "can't"
see a difference when overlaid on each other.

I have a new set of probes, I did have a set that measure with
a 6% error between the two.

Have you checked both
channels against a known voltage? DC will do. Don't trust the scope's
own calibrator.

I have not checked against a known voltage, this has it's own
problem, as you switch attenuation, I'd need several known voltage
sources. But where I checked it, the channels read the same.

Reading error, How did you determine that the delay, or phase you're
measuring was accurate, IOW, how did you establish zero crossings?

I use peak and also center the skirts around a centerline, ie.
adjust vert position so two division down from peak the the waveform
crosses equal distant from a center line.
But, I admit this is an easy area for error to show.


Boonton Q meter inaccurate. That's what I'd go for, first. Have you
verified its calibration?

No, I haven't.
In fact while back I saw a method to check 3db points on the Q meter to
test accuracy of the unit.

Don't trust 50-year-old standard inductors, But it's so pretty! I wonder
if there is any helium left inside.

Helium diffuses through just about anything.

Ya, that's I mentioned it.

the proper way to do it involves only measuring voltages. See the
manual.

 

[Fred Abse]

On Wed, 05 Feb 2014 17:07:57 -0600, amdx wrote:

I don't know, that paper was written in 2006,
maybe it didn't then.
Oh, it accomidates litz?

Yes it does, and always has, AFAIK.

There are no pre-written litz models, you have to make your own, the
"modify material" dialog has the necessary functionality. I guess the guy
was in too much of a hurry to get started, rather than play around with
FEMM, and find things out. I always play around with new software, to find
out what it will, and wont, do. Found the litz within about half an hour,
when I started.


Ok 660/46 close wound. I'll send you an email

and you can tell me what details you need. I think the rod is # 61, but I
don't know for sure. According to Ben, some rods of the same material have
much lower losses than others, even out of the same batch.

What I need is winding length, thickness, and diameter. I'm not going to
draw it a turn at a time, that's unnecessary and silly.

What diameter and length is the rod?

I usually make a properly dimensioned drawing with a CAD utility, then
import it as a DXF into FEMM. FEMM's drafting is horribly unintuitive, but
it imports DXFs properly.

I've got Fair-Rite 61 magnetic properties, from their site. No PDFs, just
a web page. Silly people. I had to print it to postscript, and make a PDF
from that. Engineers like to refer to stuff offline.


--
"Design is the reverse of analysis"
(R.D. Middlebrook)

 

[amdx]

On 2/7/2014 1:24 PM, Fred Abse wrote:

On Wed, 05 Feb 2014 17:07:57 -0600, amdx wrote:

I don't know, that paper was written in 2006,
maybe it didn't then.
Oh, it accomidates litz?

Yes it does, and always has, AFAIK.

There are no pre-written litz models, you have to make your own, the
"modify material" dialog has the necessary functionality. I guess the guy
was in too much of a hurry to get started, rather than play around with
FEMM, and find things out.

I don't know about to much of a hurry, he wrote 29 original articles
about crystal radio Some pretty heavily on experimental data.

http://www.bentongue.com/

I always play around with new software, to find

out what it will, and wont, do. Found the litz within about half an hour,
when I started.


Ok 660/46 close wound. I'll send you an email
and you can tell me what details you need. I think the rod is # 61, but I
don't know for sure. According to Ben, some rods of the same material have
much lower losses than others, even out of the same batch.

What I need is winding length, thickness, and diameter. I'm not going to
draw it a turn at a time, that's unnecessary and silly.

What diameter and length is the rod?

I usually make a properly dimensioned drawing with a CAD utility, then
import it as a DXF into FEMM. FEMM's drafting is horribly unintuitive, but
it imports DXFs properly.

I've got Fair-Rite 61 magnetic properties, from their site. No PDFs, just
a web page. Silly people. I had to print it to postscript, and make a PDF
from that. Engineers like to refer to stuff offline.

It may be Tuesday before I can get you any info. I'm out of town.
Mikek

 

[Fred Abse]

On Sun, 02 Feb 2014 20:19:32 -0600, amdx wrote:

The inductor is one I put together, 7/16" polypropylene tube wrapped
with 90 turns of 660/46 litz wire, I have an 8" ferrite rod that I slide
in or out to adjust inductance. It is variable from 8uH to 247uH. The Q at
472kHz is 550!

I did a quick FEMM on it. Without knowing its exact dimensions, I guessed
the OD of the wire at 1.5mm, since my tables don't go up to 660/46. I
extrapolated it from 270/46.

90 turns, close wound, is 135mm. Internal 7/16" diameter = 11.113mm.

FEMM gives me 8.08 microhenries, and a Q of 253, with air core, at 472kHz.
8.075 microhenries, Q=138 at 3.85MHz.

Inductance is close. Q not so close. That reinforces my suspicion that
your Boonton is out of cal. It's easy enough to check, it's really only
two voltmeters.

Getting FEMM-acceptable B/H data for Fair-Rite 61 was a a pain. They use
old-fashioned gauss and oersteds. FEMM wants tesla/amps per meter. I had
to read points off the plot, convert, and import it.

Using #61 data, both linear, and nonlinear B-H gets nowhere near your
247uH, higher than that, in fact. That's with a 5/16" diameter rod.

--
"Design is the reverse of analysis"
(R.D. Middlebrook)

 

[Fred Abse]

On Fri, 07 Feb 2014 20:31:55 -0600, amdx wrote:

I don't know about to much of a hurry, he wrote 29 original articles
about crystal radio Some pretty heavily on experimental data.

That explains why I'd never heard of him. I stopped playing with crystal
radios when I got my first tubes, at the age of about 12, nearly 60 years
ago. To me, that's like being an expert in buggy whips ;-)

His company appears to be into consumer/entertainment stuff, which
explains why I'd never heard of them, either.

--
"Design is the reverse of analysis"
(R.D. Middlebrook)

 

[amdx]

On 2/10/2014 2:50 PM, Fred Abse wrote:

On Sun, 02 Feb 2014 20:19:32 -0600, amdx wrote:

The inductor is one I put together, 7/16" polypropylene tube wrapped
with 90 turns of 660/46 litz wire, I have an 8" ferrite rod that I slide
in or out to adjust inductance. It is variable from 8uH to 247uH. The Q at
472kHz is 550!

I did a quick FEMM on it. Without knowing its exact dimensions, I guessed
the OD of the wire at 1.5mm, since my tables don't go up to 660/46. I
extrapolated it from 270/46.

90 turns, close wound, is 135mm. Internal 7/16" diameter = 11.113mm.

FEMM gives me 8.08 microhenries, and a Q of 253, with air core, at 472kHz.
8.075 microhenries, Q=138 at 3.85MHz.

Inductance is close. Q not so close. That reinforces my suspicion that
your Boonton is out of cal. It's easy enough to check, it's really only
two voltmeters.

Getting FEMM-acceptable B/H data for Fair-Rite 61 was a a pain. They use
old-fashioned gauss and oersteds. FEMM wants tesla/amps per meter. I had
to read points off the plot, convert, and import it.

Using #61 data, both linear, and nonlinear B-H gets nowhere near your
247uH, higher than that, in fact. That's with a 5/16" diameter rod.

I'll get some time tonight to get you some numbers.
I think I said my inductor measures 8uh to 247uh so your 8.0xxuh is
close my measurement.
I don't understand what you got for a maximum inductance, higher than
247uh?
And I think my ferrite rod is 5/16".
Mikek

 

[Fred Abse]

On Mon, 10 Feb 2014 15:48:37 -0600, amdx wrote:

I don't understand what you got for a maximum inductance, higher than
247uh?

Yes.

--
"Design is the reverse of analysis"
(R.D. Middlebrook)

 

[amdx]

On 2/10/2014 4:35 PM, Fred Abse wrote:

On Mon, 10 Feb 2014 15:48:37 -0600, amdx wrote:

I don't understand what you got for a maximum inductance, higher than
247uh?

Yes.

Ok, here's the data on my ferrite rod inductor.

Ferrite rod---8.160" x 0.337"
Coil length---6.125"
Coil OD---0.563"
Coil ID---0.4375"
Wire diameter---0.0625"
91 turns

The 660/46 litz is wound over a polypropylene
tube with an OD of 0.04375"

The inductance measured 250uh at 475kHz with the rod centered
in the winding.
I measured the Q with an air cap and my scope using the three db
method. I made 3 measurements at 475 kHz, I got 350, 361 and 385.
I also used the 3db method on my Boonton and got the same 550 I
get when using the normal method.

See Calibration on this page. I don't see where this could be wrong.

http://www.qsl.net/k5bcq/qmeter/qmeter.html

I am using a frequency counter on the Boonton , not the frequency dial
readings.

I wonder if my variable cap has an extra 0.7 ohms of loss over the cap
in the Boonton.

Mikek

 

[Fred Abse]

On Mon, 10 Feb 2014 22:02:08 -0600, amdx wrote:

Ok, here's the data on my ferrite rod inductor.

Ferrite rod---8.160" x 0.337"
Coil length---6.125"
Coil OD---0.563"
Coil ID---0.4375"
Wire diameter---0.0625"
91 turns

The 660/46 litz is wound over a polypropylene
tube with an OD of 0.04375"

That's pretty much what I'd guessed.

In air, I get 8.30257uH, Q=239.6593

With an 8" x 5/16" #61 rod, placed on-axis, and centered symmetrically, I
get 632.399uH, Q=621.164

All at 472kHz.

The inductance measured 250uh at 475kHz with the rod centered
in the winding.

I measured the Q with an air cap and my scope using the three db
method. I made 3 measurements at 475 kHz, I got 350, 361 and 385.
I also used the 3db method on my Boonton and got the same 550 I
get when using the normal method.

*One thing I've noticed: your Boonton and scope readings are in almost
the exact ratio 2/pi (or pi/2). Have you mixed some peak and average
readings?*

See Calibration on this page. I don't see where this could be wrong.

http://www.qsl.net/k5bcq/qmeter/qmeter.html

Use the procedure in the Boonton manual. Pages 17 & 18.

I am using a frequency counter on the Boonton , not the frequency dial
readings.

I wonder if my variable cap has an extra 0.7 ohms of loss over the cap
in the Boonton.

What sort of variable capacitor is it? Lab grade, or commercial? 0.7 ohms
isn't much in a commercial capacitor. My GR1422CD is specced at
dissipation factor <10x10^-6. That's about as good as a variable capacitor
gets. The Marconi TF1275 Q meter had a superb capacitor, machined from the
solid, both stator and vanes. I have a photograph somewhere.

--
"Design is the reverse of analysis"
(R.D. Middlebrook)

 

[Fred Abse]

On Fri, 14 Feb 2014 11:55:09 -0800, Fred Abse wrote:

The Marconi TF1275 Q meter had a superb capacitor, machined from the
solid, both stator and vanes. I have a photograph somewhere.

s/TF1275/TF1245

--
"Design is the reverse of analysis"
(R.D. Middlebrook)

 

[amdx]

On 2/14/2014 1:55 PM, Fred Abse wrote:

On Mon, 10 Feb 2014 22:02:08 -0600, amdx wrote:

Ok, here's the data on my ferrite rod inductor.

Ferrite rod---8.160" x 0.337"
Coil length---6.125"
Coil OD---0.563"
Coil ID---0.4375"
Wire diameter---0.0625"
91 turns

The 660/46 litz is wound over a polypropylene
tube with an OD of 0.04375"

That's pretty much what I'd guessed.

In air, I get 8.30257uH, Q=239.6593

With an 8" x 5/16" #61 rod, placed on-axis, and centered symmetrically, I
get 632.399uH, Q=621.164

All at 472kHz.

633uh, I think there is some info that didn't make it into the
calculation. I'm sure the inductance is not that high.

I don't think this analysis includes a resonating capacitor, is that
correct?
I'd have to pull out someone else's derivation to back into the Q
using a Q of 3000 to 5000 for the capacitor. The Boonton cap may
be higher than that, I don't know.
Including the cap loses may lower the Q to my 550.

The inductance measured 250uh at 475kHz with the rod centered
in the winding.


I measured the Q with an air cap and my scope using the three db
method. I made 3 measurements at 475 kHz, I got 350, 361 and 385.
I also used the 3db method on my Boonton and got the same 550 I
get when using the normal method.


*One thing I've noticed: your Boonton and scope readings are in almost
the exact ratio 2/pi (or pi/2). Have you mixed some peak and average
readings?*

Two completely different measurements. The only common element is the
inductor.

See Calibration on this page. I don't see where this could be wrong.

http://www.qsl.net/k5bcq/qmeter/qmeter.html

Use the procedure in the Boonton manual. Pages 17 & 18.

I'll look to see what that is, but I use the machine as it was
designed to measure Q. Then as a check, because of your doubt about the
550 number, I tried the 3db method on the Boonton and got very good
agreement.

I am using a frequency counter on the Boonton , not the frequency dial
readings.

I wonder if my variable cap has an extra 0.7 ohms of loss over the cap
in the Boonton.

What sort of variable capacitor is it? Lab grade, or commercial? 0.7 ohms
isn't much in a commercial capacitor. My GR1422CD is specced at
dissipation factor <10x10^-6. That's about as good as a variable capacitor
gets. The Marconi TF1275 Q meter had a superb capacitor, machined from the
solid, both stator and vanes. I have a photograph somewhere.

The cap I used is a high quality cap, I have several high end caps,
not sure if any are the holy grail, but they are good. Not the standard
365pf cap found all over.
Thanks, Mikek

 

[Fred Abse]

On Fri, 14 Feb 2014 17:14:29 -0600, amdx wrote:

On 2/14/2014 1:55 PM, Fred Abse wrote:
[quoted text muted]
633uh, I think there is some info that didn't make it into the
calculation. I'm sure the inductance is not that high.

That's with a Fair-Rite #61 model, derived from published data. I suspect
your rod either isn't #61, or has been roughly handled / subjected to
excessive field strength.

Your results imply a ferrite with permeability around 79, for the
dimensions I used. #61 is quoted as 125, although I used proper hysteresis
curves, rather than just a blanket permeability figure. Modeling was done
at 100 milliamps.

I don't think this analysis includes a resonating capacitor, is that
correct?

It was a finite element magnetic model. No resonance. FEMM Modeling doesn't
work that way, it derives inductance, resistance, and flux, from field
equations.


Like this:
Total current = 0.1 Amps
Voltage Drop = 0.30193+I*187.548 Volts
Flux Linkage = 6.32399e-005-I*9.93076e-008 Webers
Flux/Current = 0.000632399-I*9.93076e-007 Henries
Voltage/Current = 3.0193+I*1875.48 Ohms
Real Power = 0.0150965 Watts
Reactive Power = 9.37741 VAr
Apparent Power = 9.37742 VA

That translates to an inductance of 632.399uH, Q=621

At 10 amps, I get this:
Total current = 10 Amps
Voltage Drop = 5.6904+I*7731.37 Volts
Flux Linkage = 0.00260696-I*1.66867e-06 Webers
Flux/Current = 0.000260696-I*1.66867e-07 Henries
Voltage/Current = 0.56904+I*773.137 Ohms
Real Power = 28.452 Watts
Reactive Power = 38656.8 VAr Apparent Power = 38656.8 VA

Inductance now 261uH, Q=1359.

That (extreme) example shows the difference, measuring at different
currents, with a nonlinear core.

I'd have to pull out someone else's derivation to back into the Q using
a Q of 3000 to 5000 for the capacitor. The Boonton cap may be higher
than that, I don't know.
Including the cap loses may lower the Q to my 550.

The manual for the Marconi TF1245 is on BAMA. Get it. It includes a good
treatment of errors arising from instrument strays in Q meters.

Capacitor losses would not account for the discrepancy.

*One thing I've noticed: your Boonton and scope readings are in almost
the exact ratio 2/pi (or pi/2). Have you mixed some peak and average
? readings?*

Two completely different measurements. The only common element is
the
inductor.

Don't overlook the significance. Investigate. It's too close to be likely
coincidental. pi/2, to within less than 1%.

[quoted text muted]
I'll look to see what that is, but I use the machine as it was
designed to measure Q. Then as a check, because of your doubt about the
550 number, I tried the 3db method on the Boonton and got very good
agreement.

I'd expect that.

The cap I used is a high quality cap, I have several high end caps,
not sure if any are the holy grail, but they are good. Not the standard
365pf cap found all over.

It'd have to be a pretty lousy cap to produce the results given. Look into
the effects of injection impedance, yet again, the Marconi manual covers
this in depth. The impedance of your signal generator will have an effect
on the perceived Q.

Herein lies where I think your discrepancy is: Generator impedance in
series with your tuned circuit.

As an example, a 250uH inductor, having a true Q of 100, resonated at
450kHz, will have an apparent Q, by the "center frequency over bandwidth"
method, of only about half that, with a generator impedance of 8 ohms.
Don't assume that, because your generator is calibrated into a 50 ohm
load, that its looking-backwards output impedance is also 50 ohms, unless
the spec actually says so, specifically.


There's another error that can arise, which only becomes significant at
low Q values (less than about 10), due to measuring the resonant voltage
across the variable capacitor, which all Q meters do. The peak you get
isn't actually resonance, but a bit off, if you fine tune with the
capacitor, rather than by tweaking the frequency. Another reason for using
the "incremental frequency" method.
You don't need worry about that, with the Q you have.


--
"Design is the reverse of analysis"
(R.D. Middlebrook)

 

[amdx]

On 2/15/2014 10:50 AM, Fred Abse wrote:

On Fri, 14 Feb 2014 17:14:29 -0600, amdx wrote:

On 2/14/2014 1:55 PM, Fred Abse wrote:
[quoted text muted]
633uh, I think there is some info that didn't make it into the
calculation. I'm sure the inductance is not that high.

That's with a Fair-Rite #61 model, derived from published data. I suspect
your rod either isn't #61, or has been roughly handled / subjected to
excessive field strength.

All that could be correct, but a circuit Q of 550 is pretty high for
a circuit Q. This would include capacitor losses.

Your results imply a ferrite with permeability around 79, for the
dimensions I used. #61 is quoted as 125, although I used proper hysteresis
curves, rather than just a blanket permeability figure. Modeling was done
at 100 milliamps.

Near as I can figure the Boonton is measuring with about 4 microamps.

I don't think this analysis includes a resonating capacitor, is that
correct?

It was a finite element magnetic model. No resonance. FEMM Modeling doesn't
work that way, it derives inductance, resistance, and flux, from field
equations.


That's what I thought, no resonating capacitor.

Like this:
Total current = 0.1 Amps
Voltage Drop = 0.30193+I*187.548 Volts
Flux Linkage = 6.32399e-005-I*9.93076e-008 Webers
Flux/Current = 0.000632399-I*9.93076e-007 Henries
Voltage/Current = 3.0193+I*1875.48 Ohms
Real Power = 0.0150965 Watts
Reactive Power = 9.37741 VAr
Apparent Power = 9.37742 VA

That translates to an inductance of 632.399uH, Q=621

At 10 amps, I get this:
Total current = 10 Amps
Voltage Drop = 5.6904+I*7731.37 Volts
Flux Linkage = 0.00260696-I*1.66867e-06 Webers
Flux/Current = 0.000260696-I*1.66867e-07 Henries
Voltage/Current = 0.56904+I*773.137 Ohms
Real Power = 28.452 Watts
Reactive Power = 38656.8 VAr Apparent Power = 38656.8 VA

Inductance now 261uH, Q=1359.

That (extreme) example shows the difference, measuring at different
currents, with a nonlinear core.

Ran into this in one of my early electronics classes. Each person in
the class measured a 5 milliHenry inductor and got something like
150uhs. Way off. The instructor just ignored it. I dug into it and
learned about the BH curve and the slope. Our measurement current was
way down on the slope.
It was sloppily put together experiment by a grad student. But I
learned something even if no one else in the class cared about the
discrepancy.

I'd have to pull out someone else's derivation to back into the Q using
a Q of 3000 to 5000 for the capacitor. The Boonton cap may be higher
than that, I don't know.
Including the cap loses may lower the Q to my 550.


The manual for the Marconi TF1245 is on BAMA. Get it. It includes a good
treatment of errors arising from instrument strays in Q meters.

So does the Boonton manual.

Capacitor losses would not account for the discrepancy.

I think it would. I'll let you do the math, but if you use a
resonating cap with a Q of 4500, this brings the circuit Q down
to 547. I used your 621uh, even though it is not what my inductor is.
I'll rerun the numbers just to see what happens at 250uh.

*One thing I've noticed: your Boonton and scope readings are in almost
the exact ratio 2/pi (or pi/2). Have you mixed some peak and average
? readings?*

Two completely different measurements. The only common element is
the
inductor.


Don't overlook the significance. Investigate. It's too close to be likely
coincidental. pi/2, to within less than 1%.

I can't use the Boonton and a separate circuit to measure Q on my
bench at the same time. Just a strange coincidence.

I'll look to see what that is, but I use the machine as it was
designed to measure Q. Then as a check, because of your doubt about the
550 number, I tried the 3db method on the Boonton and got very good
agreement.

I'd expect that.


[quoted text muted]


The cap I used is a high quality cap, I have several high end caps,
not sure if any are the holy grail, but they are good. Not the standard
365pf cap found all over.


It'd have to be a pretty lousy cap to produce the results given.

Ya, that could be right. As a thought check, using my numbers,
250uh Q=550. I assumed the Boonton Cap has a Q of 6000(?). That would
mean the inductor has a loss R of 1.225 ohms.
My "pretty lousy" second test cap would need a loss R of 0.815 ohms,
That's only a Q of 915. That would be pretty lousy.
I'll have to run that again. I'll found out which of my caps are
really good!



Look into

the effects of injection impedance, yet again, the Marconi manual covers
this in depth. The impedance of your signal generator will have an effect
on the perceived Q.

Yes lots of corrections, for strays and other losses, but they are all
small and under normal circumstances you don't need to run the calculations.

I don't think we are far of on Q if you put in the tuning capacitor
losses. I have not found a number for that in the Boonton Manual.
I will keep looking.
But then again if I used 621uh, our Q numbers probably won't be in
close agreement.

Herein lies where I think your discrepancy is: Generator impedance in
series with your tuned circuit.

I'm not connecting my frequency generator to the tuned circuit. I'm
driving a separate air core coil with my generator. This is placed as
far as I can get it away from my variable inductor and still induce
enough signal to be able to read the voltage on my scope. it's about
8" away.
I have a better (higher) load than my scope probe, I'll pull it out
and run some tests with all my supposedly "good" caps. Should be
interesting.

Thanks for the feedback, Mikek

PS. see if you can figure anything about our 250uh/621uh discrepancy.
The rod is out of an AM radio, I don't know of any that used other
than #61 material, but?

Oh, I also learned something about the connection points. I'm using
banana pin connection on my coil, but the threaded connectors still need
to be tight or there is a loss. After tightening my Q=550 increased to
Q=610 I'll even try removing the banana plugs and see if the Q goes over
610.

As an example, a 250uH inductor, having a true Q of 100, resonated at
450kHz, will have an apparent Q, by the "center frequency over bandwidth"
method, of only about half that, with a generator impedance of 8 ohms.
Don't assume that, because your generator is calibrated into a 50 ohm
load, that its looking-backwards output impedance is also 50 ohms, unless
the spec actually says so, specifically.


There's another error that can arise, which only becomes significant at
low Q values (less than about 10), due to measuring the resonant voltage
across the variable capacitor, which all Q meters do. The peak you get
isn't actually resonance, but a bit off, if you fine tune with the
capacitor, rather than by tweaking the frequency. Another reason for using
the "incremental frequency" method.
You don't need worry about that, with the Q you have.

 

[amdx]

On 2/15/2014 1:58 PM, amdx wrote:

On 2/15/2014 10:50 AM, Fred Abse wrote:
On Fri, 14 Feb 2014 17:14:29 -0600, amdx wrote:

I went through 7 of what I thought were my “good” caps.
A description of how I setup to measure using the 3db method.
Drawing of layout is here.
http://tinyurl.com/n95t3zl
I have a 4” dia air core coil connected to my frequency generator,
I also had a 50 ohm resistor in parallel with the coil.
This coil was 17” from the inductor under test, on axis.
I then attached a capacitor across the inductor.
This was all sat on a piece of 4” styrofoam and away from lossy items.
I have a high impedance, low capacitance amplifier,
like the second amp design here.
www.crystal-radio.eu/fetamp/enfetamp.htm
I connected that amp across the LC circuit and connected my scope to the
output. I set the Sig Gen to about 475kHz and adjusted the cap for
maximum amplitude on the scope. I recorded the frequency. Of course
there was some back and forth to find the resonant frequency. I then
adjusted the Sig Gen output level so the scope reading was 7 units. Then
I adjusted the frequency up until I had 5 units on the scope, I recorded
the frequency, then lowered the frequency until I had 5 units again, and
recorded the frequency. 7 units x 0.707 = 4.949 or 5 units.
I did three measurements on all seven caps and calculated an average.
Most measurements were close to each other.
Cap 4 had the widest range of measurements 499, 524, and 528.

With Cap 1 average Q = 721

With Cap 2 average Q = 630

With Cap 3 average Q = 701 First 5 @ 475kHz

With Cap 4 average Q = 517

With Cap 5 average Q = 672

With Cap 6 average Q = 689 @ 500kHz

With Cap 7 average Q = 713 @ 600kHz

After removing the banana plugs from my inductor and connecting with
the compression thumb nuts,
The Q measured on the Boonton. Q = 618.
It needs some work, I would think it equal or better my 3db testing. My
first thought is the wiper connection on the capacitor.

So it was a worthwhile exercise, I have marked my caps and know which
are best.

Your thoughts, Mikek

 

[Fred Abse]

On Sat, 15 Feb 2014 18:57:04 -0600, amdx wrote:

On 2/15/2014 1:58 PM, amdx wrote:
On 2/15/2014 10:50 AM, Fred Abse wrote:
On Fri, 14 Feb 2014 17:14:29 -0600, amdx wrote:

I went through 7 of what I thought were my “good” caps.
A description of how I setup to measure using the 3db method.
Drawing of layout is here.
http://tinyurl.com/n95t3zl
I have a 4” dia air core coil connected to my frequency generator,
I also had a 50 ohm resistor in parallel with the coil.
This coil was 17” from the inductor under test, on axis.

17 "somethings". My newsreader doesn't support those silly Windows
characters. Please either use ", and ', or "in", "ft", "mm", "m", etc. I
can't tell whether it was inches or feet ;-)

I then attached a capacitor
across the inductor.
This was all sat on a piece of 4” styrofoam and away from lossy
items.
I have a high impedance, low capacitance amplifier, like the second amp
design here.
www.crystal-radio.eu/fetamp/enfetamp.htm
I connected that amp across the LC circuit and connected my scope to the
output. I set the Sig Gen to about 475kHz and adjusted the cap for
maximum amplitude on the scope. I recorded the frequency. Of course
there was some back and forth to find the resonant frequency.

For best accuracy, fine tune the frequency, not the capacitor. I did a
proof thirty years ago, when I liked calculus more than I do today.

That's why Marconi's Dielectric Test Set, a huge beast, with a 2 foot
scale mirror galvanometer, glass insulators, and micrometer capacitors,
used parallel, not series,resonance. Cost as much as a car did, back then.

I then
adjusted the Sig Gen output level so the scope reading was 7 units. Then
I adjusted the frequency up until I had 5 units on the scope,

5/7 isn't quite -3dB. How much bandwidth error you'll get depends on how
sharp, or flat the peak response is.

I recorded
the frequency, then lowered the frequency until I had 5 units again, and
recorded the frequency. 7 units x 0.707 = 4.949 or 5 units.
I did three measurements on all seven caps and calculated an average.
Most measurements were close to each other.

That's a rather Byzantine way of going about things.

Why not use a parallel connection method on the Boonton? Measure at
several frequencies, and tabulate the results.

Cap 4 had the widest range of measurements 499, 524, and 528.

With Cap 1 average Q = 721

With Cap 2 average Q = 630

With Cap 3 average Q = 701 First 5 @ 475kHz

With Cap 4 average Q = 517

With Cap 5 average Q = 672

With Cap 6 average Q = 689 @ 500kHz

With Cap 7 average Q = 713 @ 600kHz

You don't say what capacitance the capacitors were, or the frequency.

After removing the banana plugs from my inductor and connecting with the
compression thumb nuts,
The Q measured on the Boonton. Q = 618. It needs some work, I would
think it equal or better my 3db testing. My first thought is the wiper
connection on the capacitor.

Lab grade capacitors have multiple wipers, usually bearing on the vane
edges.

So it was a worthwhile exercise, I have marked my caps and know which
are best.

Your thoughts, Mikek

Firstly, without capacitor values, the figures aren't much use. I will say
that they suck as measurement capacitors. Boonton (carefully?) don't state
parameters for their main capacitor. Marconi, however, did. I'll post
their curves to A.B.S.E. Suffice it to say here that, at 500pF, and 1Mhz,
Q is greater than 20000.

Do the capacitor testing on the Boonton, in parallel with the main
capacitor, measure the C, and Q at various settings, and tabulate the
results. I can do curves with gnuplot, from tables, if you don't have
plotting software.

I refer you to my last reply, where I wrote about Boonton test current.
Using 15 milliamps, calculated from Boonton's voltmeter figures, FEMM
gives Q at 620. Close enuf?

Inductance given by FEMM is still high, at 633uH. I still suspect that
your ferrite isn't what you think it is. Might it be #67?

I'm beginning to believe your Boonton ;-)

--
"Design is the reverse of analysis"
(R.D. Middlebrook)