Topics

Low pass filters

Richard G4TGJ
 

A number of people have had problems with the power output of their QCX and the solution has often been to remove a turn or two from L1-L3. I have been working on my own transceiver and I have been using the W3NQN CWAZ LPF designs from the GQRP Club datasheet. I've been analysing the 20m LPF with my nanoVNA and have had to remove 2 turns from L2 and L6 and 1 turn from L4 to get a satisfactory passband return loss and insertion loss. That's a change in inductance way beyond the 5% tolerance of AL or the 5% capacitance tolerance. Has anybody else done similar tests? Is there a widespread problem with the toroids? Mine came from GQRP Club Sales so I assume they are Micrometals.
--
73
Richard
G4TGJ

Andy Brilleaux <punkbiscuit@...>
 

Why do you instantly start blaming toroids ?

Have you not considered that the original W3NQN design was never quite right in the first place ?

It's only now that these days as amateurs so we can easily deal perform the measurements we can see this for ourselves.

As you have seen, you have made a small adjustment and the problem is solved.

Andy Brilleaux <punkbiscuit@...>
 

sorry for the dratted autocorrect, but I think you get the message :-)

Richard G4TGJ
 

Sorry, have you woken up after a bad night?

I've used Tonne Software's Elsie to analyse the W3NQN design and it should work well but my measurements do not match it. So I've hardly instantly come to any conclustions. It was a question for the community to give their considered and informed response.

The problem isn't solved because it means designs may not be repeatable.
--
73
Richard
G4TGJ

SkipF, NT1G <skip.flem@...>
 

Another list member (me!) want to know...do you have any
 inductance meter or bridge? I takes a lot of guessing (and doubting) out of the picture.
Skip Flem, NT1G
Boston-ish

Andy Brilleaux <punkbiscuit@...>
 

The problem, as discussed many times on this group is that the transition frequency is a bit too low
for comfort, especially with our little milliwatt rigs.

There's a reason for that, the designer opted actually STATES in his own article that he lowered that cut off freq
so he could use standard value capacitors and also decided to count UP the winding to the nearest whole integer.

Source - W3NQN himself in his own SWM mag article.

It also pays not to use ancient software that relies upon SVC tables in it's calculations.

Gave up that rubbish decades ago.












Arv Evans
 

It is difficult for on-line resonnance calculator programs to know how much distributed capacitance and distributed inductance is present in the actual circuit.  This sometimes causes the calculated values to be incorrect.  

Arv. K7HKL
_-_


On Fri, Dec 13, 2019, 4:00 AM Richard G4TGJ <rpt@...> wrote:
Sorry, have you woken up after a bad night?

I've used Tonne Software's Elsie to analyse the W3NQN design and it should work well but my measurements do not match it. So I've hardly instantly come to any conclustions. It was a question for the community to give their considered and informed response.

The problem isn't solved because it means designs may not be repeatable.
--
73
Richard
G4TGJ

Hans Summers
 

Hi Richard

There is a lot of variation, but I believe more of it is in winding styles than in the toroid characteristics.

The theoretical values for the number of turns are often quoted. But these are calculated assuming somewhat perfect conditions, including equally distributed winding around the circumference of the toroid. Practically we are often told to leave a from start to end, which some say is for the purposes of minimizing capacitative coupling; the recommended gap is sometimes 30-degrees or 10% (36-degrees) for example. Squeezing turns together increases the inductance; leaving a gap is like squeezing turns together. 

Another thing that increases inductance is looser windings. It's very difficult to wind so tightly that the wire touches the toroid surface all the way round. With fine wire such as the AWG #28 (0.3mm) used in QCX, if you try to pull it too tight, it snaps. With thicker wire, it doesn't bend so easily so it's even harder to get it wound tightly. This also increases the inductance. 

I feel that all this variation in winding style is more to blame for inductance variation than the material tolerances themselves. 

I can't remember having seen a real world measured inductance that is LOWER than the predicted inductance from the calculation. The calculated number of turns always seems to result in too high an inductance value. It makes no difference whether the cores are original micrometals or not. 

I think these differences are all down to the gap between theoretical models and real world situations. 

73 Hans G0UPL 


On Fri, Dec 13, 2019, 13:08 Richard G4TGJ <rpt@...> wrote:
A number of people have had problems with the power output of their QCX and the solution has often been to remove a turn or two from L1-L3. I have been working on my own transceiver and I have been using the W3NQN CWAZ LPF designs from the GQRP Club datasheet. I've been analysing the 20m LPF with my nanoVNA and have had to remove 2 turns from L2 and L6 and 1 turn from L4 to get a satisfactory passband return loss and insertion loss. That's a change in inductance way beyond the 5% tolerance of AL or the 5% capacitance tolerance. Has anybody else done similar tests? Is there a widespread problem with the toroids? Mine came from GQRP Club Sales so I assume they are Micrometals.
--
73
Richard
G4TGJ

Richard G4TGJ
 

Hi Hans
Thanks for the thorough reply. I tried squeezing the turns but could not get the response to match the expected. I think your point about loose turns is a good one although I did try to make the windings as tight as I could.

So it does sound like their is a toroid issue. Not with the toroids themselves but with the calculations.

I will have to get myself an LCR meter for Christmas.
--
73
Richard
G4TGJ

ON7DQ Luc
 

Richard,

You do know you can measure Inductance with your nanoVNA very well ?

I made a small "component-adapter", and use it very frequently (see picture).
It's a 4 pin header, two pins to ground, two pins to center of the SMA connector.
I also made a simple 4 pin CAL set,first two pins shorted, center pins open, and a 49,9 Ohm resistor between the last two pins.
(50 Ohm CAL shown in the picture)
The two center "hooks" are bent up slightly to measure SMD components.
Put the nanoVNA in single frequency mode on a low frequency (I use 50 kHz, but using your "working frequency", e.g. 7 MHz,  works too).
73,
Luc - ON7DQ

ajparent1/KB1GMX
 

I'm with Hans on this.

Experience is two T37-2 toroids wound with 20 turns had a  15% variation
in measured inductance at 7mhz using a calibrated P4191A.   One tight 
and closely spaced to maybe 65-70% of the space the other loose and 
covering near 95%.  When stripped and measured the loose one also
had a significant amount more wire, inductance free space or on a
former is dependent on the amount of wire used.  As to pulling hard
enough to break #28 you are likely pulling way to hard.  ;-)

Since iron powder is not spec'ed as exact or even 5% seeing variations 
is nominal case.  Most have a Temp coefficient  +35-50ppm!/degree-C.

What does that mean, if shown a picture of how the sample was wound
one might want to consider that a strong suggestion.  The other is if told
there may be tuning likely because its not a wide band design and can't
be with class-E amp.  Do the tuning.

The tolerances of ferrite (actually its iron powder Txx-xx part not FTxx-xx) ,
caps, and winding style can easily fall out of range.  The first two are verifiable
the third tends to map widely.

Add to that the unknowns like small difference in lead lengths and distributed
capacitance all tend to gang up.  Further L4 and C30 are critical  values
for class E performance and they do interact with the low pass filter.

As to Tonne software and experience while I have always seen errors
when I went back and measured and then plugged in the real values it
was always correct.  One of the items you can plug into models is the
real effect of circuit traces!

What are the real values... inductance of the coils at the operating frequency
(It does change with frequency due to distributed capacitance and mu of the iron
powder inductors).  The actual value of the capacitors, and additional capacitance
of the board or wiring.  If all are 5% (in electronics that is pretty good!) consider
the effect on Fc of a low pass filter, that can push the Fc around by 350khz easily
at 7mhz!

Allison
--
Please reply on list so we can share.
Off list email goes to trash, I had to due to scrapers.

Richard G4TGJ
 

Hi Luc
Yes, I had assumed I could but hadn't yet looked into it. I like your arrangement and will do something similar. Thanks.
--
73
Richard
G4TGJ

Andrew Lenton
 

Hi here is a plot using the low pass kit supplied by QRP-LABs, I measured the 7 components all were in 5%, I think the largest influence on performance is the Q of the CAPs and inductors not, necessary their value!

 

The first plot is the correct values but the Q is low on the 390 pf and 180 PF Caps. Taking a few turns on the coils moves the frequency roll off up a bit, but is hiding the problem. Look at the slop.

 

BR

 

Andrew G8UUG







Bob ZL1RS
 

Andrew, please confirm if the second plot is with 'a few turns' taken off, or (more likely) with higher Q 390pF and 180pF capacitors installed?

Thanks, Bob  ZL1RS


n4qa at_hotmail.com
 

Bob,
 I took Andrew's post to mean:
1. First plot resulted from removing turns and having low-Q caps
and
2. Second plot resulted from no turns removed and having high-Q caps

Hopefully, Andrew will clarify.

72 / 73,
Bill, N4QA

ajparent1/KB1GMX
 

I'd be quick to point out that any variation of right half and left half for a 7 element
low poss will show variation in plots. 

For example did you take 2 turns off L1 only, or L1 and L3?  That is not equivilent
result in all cases.  While the "fix" worked what is it changing and how does it
solve a problem.

Having built and tested  a few of the LPF kits any issues were values as predicted
vs as measured.

--
Please reply on list so we can share.
Off list email goes to trash, I had to due to scrapers.

George Korper
 

Hi, 
Is there anybody out there willing to test switching to a larger wire size on 20 meters? I will do it when my kit arrives but I would really like a collaboration from a trusted source. Verification is good.

I realize all the hypothesis say this is not a factor. In RF circuits real world testing is essential in my opinion.

George

On Fri, Dec 13, 2019, 4:08 AM Richard G4TGJ <rpt@...> wrote:
A number of people have had problems with the power output of their QCX and the solution has often been to remove a turn or two from L1-L3. I have been working on my own transceiver and I have been using the W3NQN CWAZ LPF designs from the GQRP Club datasheet. I've been analysing the 20m LPF with my nanoVNA and have had to remove 2 turns from L2 and L6 and 1 turn from L4 to get a satisfactory passband return loss and insertion loss. That's a change in inductance way beyond the 5% tolerance of AL or the 5% capacitance tolerance. Has anybody else done similar tests? Is there a widespread problem with the toroids? Mine came from GQRP Club Sales so I assume they are Micrometals.
--
73
Richard
G4TGJ

ajparent1/KB1GMX
 

Changing wire size will impact the actual inductance slightly.  
Doing so without measuring and testing is a shot in the dark.

--
Please reply on list so we can share.
Off list email goes to trash, I had to due to scrapers.

Andrew Lenton
 

Hi Just to clarify:

 

1.     Top plot: is correct number of turns, parts straight out of the kit no modifications.

 

2.    Bottom plot: is replacement capacitors for known high Q types. No adjustments to inductors.

 

3.    If anyone is interested I could remove two turns of each coils and make a plot on the bottom plot, I no longer have the poor caps to do the same exercise on the top plot

 

BR

 

Andrew

James Sweeden
 

Well I would never describe myself as a "trusted source" but did use #24 wire when winding L1 - L4 in my QCX 20.  L4 was wound and installed as per the instructions, I did not measure the inductance.  I did measure L1 - L3 however with my MFJ-259B and found all to be higher in value than that given in the manual.  Removing 2 turns from each brought them closer but still a little high.  I built the filter on a scrap of pcb using silver mica caps (I had already install the capacitors supplied with the kit into the radio) and connected it to my old AN/URM-25F RF signal generator.  Comparing the input and output levels while running the frequency up through 15MHz convinced me all was well so L1 - L3 were installed - all with 2 turns removed.  Power out with a 13.8VDC supply and D8 bypassed was 3.4W.  Spreading the turns on L1 and L3 out to cover as much of the core as possible and compressing the turns on L2 brought that up to 4.4W.  I think the MFJ-259B did a fairly good job of measuring the inductance.

Hope you find this of some use.

73,
James/KB7LJP


On Sun, Dec 15, 2019 at 01:24 PM, George Korper wrote:
Is there anybody out there willing to test switching to a larger wire size on 20 meters? I will do it when my kit arrives but I would really like a collaboration from a trusted source. Verification is good.