|
I'm building my second QMX, now for 20 m - 10 m.
Today I was winding the toroids for the LP filters. I was curious about the real inductance. Using two vectorial analyzers I checked the values of the coils. For each I've found values with a huge difference to the values provided by the assembly manual. Here are two examples:
L511 640 nH, 0.80 ?H, 855 nH @ 14 MHz and L513 270 ?H, 0.43 nH, 445 ?H @ 28 MHz (first value from assembly manual and calculated for the toroid, second with one analyzer and third with the second analyzer).
With these the shape of attenuation should go to lower frequencies, also the notch made by L513 and C522.
I've simulated the LP for both versions with LTspice and got the desired and the shifted graph.
I've build the LPF with the toroids on a tiny rf board and got these expected graphs.
Reducing the windings the graph was more like the assumed for the values from the manual.
What were the main results by the LP with the higher conductivitys?
1. Additional attenuation of 1,6 dB for 28 MHz (69 % of Power out)
2. less attenuation for the harmonics, but maybe enough (?)
Do you have some information about these topic?
73, Ludwig
|
According to the T30-17 data sheet tolerance for AL is +/- 5%.
From my L measurement the tolerance seems to be at least +25 %.
Any idea?
Here are some attenuation values (all in dB, LP for 12/10 m):
v1: Simulation with L values from manual
v2: Simulation with found L values for original winding
v3: measured attenuation values for original winding
v4: measured attenuation values for reduced winding
?? f/MHz? 24.9???? 28.5???? 49.8???? 57.0 ??? 74.7???? 85.5
v1??????? 0??????? 0??????? 23?????? 31.1???? 55.1???? 56.3
v2??????? 1??????? 4??????? 37.2???? 49.1???? 51.5???? 50.2
v3??????? 0.2????? 1.8????? 37.6???? 51.7???? 55.9???? 54.9
v4??????? 0??????? 0.2????? 25.4???? 36.2???? 63.0???? 55.5
There seems to be only one critical value for v3: 1.8 dB @ 28.5 MHz. It would reduce output to around 70 % compared to v4.
Especially for the first four frequencies the values from version 1 and 4 and from version 2 and 3 are similar.
So I guess my L measurement is not wrong.
73 Ludwig
|
Two more information:
1. All tests were made with turns spread over the whole toroid. So inductivity is lower than with turns pushed together.
2. Ref. to manual page 40 L508 should get only 11 turns and not 13. Calculate with 13 turns L should be 270 nH like shown in table (286 nH). I measured 270 nH when made with only 11 turns! So this points also to a higher AL than from the data sheet.
73 Ludwig
|
Hi Ludwig
FYI the toroids in QMX are purchased directly from micrometals in California, US (), in large quantities (10,000 or 20,000 etc). I have no reason to suspect they aren't the very best quality and with correctly specified AL values. If anyone feels the AL values are wrong then I'd be more inclined to suspect the experimental setup and/or measurement accuracy or some similar factor.
But it's tempting to overthink the inductance situation.?
Simulation is a good guide. Or calculation based on theory. In the real world there are frequently many other factors not accounted for theoretically. Results usually differ between practical measurement and theoretical prediction, sometimes only by a little, sometimes by a lot. Bear in mind measurement is subject to errors also.?
In all my work, if I use simulation or calculation, I consider that a first guide only. Then build and test. Then tweak things, see if improvement can be found. If the practical number of inductor turns differs from the theory, it doesn't upset me.?
The number of turns in the manuals are what we have found, experimentally, works best (for example with LPFs, power output and PA efficiency, as well as harmonic levels). Even so, there will always be some variation from builder to builder.?
The manual number of turns are my recommendation; most people can count, but most people don't have an inductance meter (let alone an accurate one at such small inductance values used in HF LPFs).
Having said all that... If you are confident of the accuracy of your inductance measurement for such low values, and if you wish to build it according to theory, I don't see any problem with that... And even better if you report back on your results, it will be interesting.?
73 Hans G0UPL
On Thu, Jan 25, 2024, 3:54?PM < DH8WN@...> wrote: Two more information:
1. All tests were made with turns spread over the whole toroid. So inductivity is lower than with turns pushed together.
2. Ref. to manual page 40 L508 should get only 11 turns and not 13. Calculate with 13 turns L should be 270 nH like shown in table (286 nH). I measured 270 nH when made with only 11 turns! So this points also to a higher AL than from the data sheet.
73 Ludwig
|
|
|
Hi Ludwig,
I got almost the same results by measuring inductance using two methods,VNA and cheap LC meter. Yes, it looks like the tolerance is +20-25% for material #17 and +15-20% for material #6. After some thought, I decided to follow the manual. Soon I will receive a 1N4148 diode for modification and if I manage to solder this SMD microbe without problems, I will power on the transceiver and hope to see real scans of the filters.
I wish you successful completion of the build.
73
Igor, oh2jxa
|
Hi Mushin,
I think it's not the same topic. I'm not about rounding the number of turns.
73 Ludwig
|
Hi Hans,
thank you for this detailed Information. Meanwhile I checked the uncertainty of the measurement. Maybe, the values from the VNWA are a little bit to high, but not so much. I will do some more tests with built LPF, measured with the same VNWA.
73, Ludwig
|
Hi all,
here are my test results and the pleasant final results from the working QMX (20 m - 10 m).
I've made some steps.
1. Checking the uncertainty of measured values for L
I have a set of filter elements to build a high band QDX. I wound some toroids T37-10 following the QDX manual. The measured L were between 10 % and 15 % higher than expected. Maybe I've measured a little bit more than the real value but not 25 % or more like for my T30-17.
2. Checking the validity of values from build LPF
I used a tiny rf test board to build LPF. First I've build the filters from the parts for QDX.
Measurement was done by a VNWA looking at the attenuation at the band frequencies and the first and second harmonic and the match toward the PA.
The measured values were fine. So the setup should be usable.
3. Checking the LPF for QMX
I've build the LPF using the elements for my QMX (again on the tiny test board). The toroids were wound following the manual (L 508 with 11 turns). Measurement was done like in step 2.
LPF 20 m: values usable
LPF 17 m and 15 m: values not good
LPF 12 m and 10 m: values usable
To consider the capacity of the real TX (stray capacity of PCB and capacity from the PA) I've added some 10 pF to the capacitor toward the PA. Here are the results:
LPF 17 m and 15 m: values useable for 17 m and maybe for 15 m
Now I've reduced L 512 also to 11 turns.
LPF 17 m and 15 m: nearly perfect
?
3. Building the QMX following the manual but L 512 with 11 turns, measured output and harmonics
The harmonics were measured using a precise SDR and a spectrum analyzer software. QMX and SDR were coupled with attenuators with an optimized value of attenuation.
The output was 4 … 5 W so ok.
The harmonics suppression was at least 56 dB so more than ok.
?
Finally conclusion: following the manual would be ok, maybe also without the reduction for L 512.
This matches to what Hans wrote in this and in a different thread.
?
73 and wish you success with your QMX
Ludwig
?
|
On Sun, Feb 25, 2024 at 11:05 PM, @Ludwig_DH8WN wrote:
The harmonics suppression was at least 56 dB so more than ok.
One additional Information: at least 56 dB up to 35 MHz at least 59 dB around 42 MHz at least 61 dB above 50 MHz So it's fine according to regulations.
|
Ludwig,?
? ? Am I understanding that you were finding T37-17s to measure 25%+ high when you wound them??
And your further work with T37-10s is validating your measurements?
I'm working on a QDX HB, with the T37-17s. I've been trying to measure the toroids by using a series LC circuit with a capacitor, and using the frequency of resonance to find the inductance. With 12 turns on a T37-17, I was somewhere around 0.35uH. Taking it all the way down to 8 turns got me 0.22, and 10 turns on another got me 0.248. (Target is 0.202 and 0.25).??
On one with more windings, I found I needed 20 turns (to get 0.74uH) instead of the 21 listed to get to the 0.722 target.?
I'm questioning my measurements, especially at low turns.....Modeling the QDX 17/15 LPF in Elsie...I find that one is pretty sensitive to getting the inductance values correct.?
If my 8 and 10 turn windings are really closer to the calculated 0.10 and 0.15 that Toroids.info comes to, the input SWR of that LPF gets REALLY bad... But if my measurements are correct, then the filter performance is right where it needs to be.?
I'm considering building a whole LPF on a breadboard and measuring the response to help validate what I've measured (jsut like you did), but just double-checking that you were seeing high AL values with the T37-17s.?
-Nate?
N8BTR
|
Hi Nate
toggle quoted message
Show quoted text
On Mon, Feb 26, 2024 at 07:43 PM, <natereik@...> wrote:
? ? Am I understanding that you were finding T37-17s to measure 25%+ high when you wound them??
And your further work with T37-10s is validating your measurements?
From Material 17 I used the size 30 (T30-17). Validation by using T37-10 showed two things: 1. My setup brought probably to high values for L. 2. Despite the deviance (see No 1) the AL of the T30-17 (QMX) could be probably a little bit to high but probably not +25% or more.
With 12 turns on a T37-17, I was somewhere around 0.35uH. Taking it all the way down to 8 turns got me 0.22 ?H
These values are not consistent. Reducing from 12 turns to 8 should reduce the inductivity by theory from 0.35 ?H to 0.16 ?H (not the measured 0.22 ?H). Did you use the same allocation of turns across the toroid?
I'm considering building a whole LPF on a breadboard
Please be aware of the different stray capacitance and the capacity of the connected QDX circuits. And use a rf suitable board!!! My final result of my whole testing was: Following the manual brings good results or at least a good start point to tweak the filters. 73 Ludwig
|
On Tue, Feb 27, 2024 at 08:20 AM, @Ludwig_DH8WN wrote:
Please be aware of the different stray capacitance and the capacity of the connected QDX circuits. And use a rf suitable board!!!
My final result of my whole testing was: Following the manual brings good results or at least a good start point to tweak the filters.
Thank you for your thoughts and clarifications! The more I play in Elsie with LPF simulations, the more I agree that following the manual will be a good start. It seems in the QDX HB, the 15/17m LPF is the most sensitive to the inductance values. The others are more...robust. Especially as far as reflected power to the PA. And there are more downsides to using less turns on the 15/17 than the recommended number of turns, even if inductance is higher than spec'd/calculated. So I will follow the manual pretty closely, maybe doing 20T instead of 21 on L2, but that's the only change.? -Nate? N8BTR
|
Restarting an older thread....
I'm trying to move the cutoff frequency of my 15/17 LPF filter on my QMX 60-15 about a MHz higher. I tried modeling it in Elsie, but it looks nothing like a LPF.
What should I be using for Zin and Zout ?
Besides the basic filter made up of L510, L513, C516, C519, C525, are there other components I should be including ?
Thanks and 73,
Steve, N2IC
|
After a little experimenting, I'll answer my own question...
With Zin and Zout both set to 21 ohms, the filter response curve perfectly matches what I am seeing on my QMX.
73,
Steve, N2IC
|
Steve,
Agreed, those input/output impedances have bitten me in the behind before; I had an engineering manager who believed that the software model was 'more accurate' than what we built on the bench and if the results were different it was because we were incompetent at actually building a circuit. Thankfully he was not long for the engineering-world and was quickly "promoted" to sales.
Like Hans had indicated, a model is a reasonable starting point but it eventually turns in to an exercise where the designer is hand picking components that are close in value and inserting them in to the circuit to come up with the accurate values.
You should look up the reference to something called a "gimmick capacitor". Sometimes I work on old radios and find one left behind; It is humorous when some other restorer either cuts it out or just solders the two ends together.
--
Tisha, AA4HA
|
Steve, also you missed C522 which is in parallel with L512.?
Ron
On Wed, May 29, 2024 at 09:51 Tisha Hayes, AA4HA via <Tisha.Hayes= [email protected]> wrote: Steve,
Agreed, those input/output impedances have bitten me in the behind before; I had an engineering manager who believed that the software model was 'more accurate' than what we built on the bench and if the results were different it was because we were incompetent at actually building a circuit. Thankfully he was not long for the engineering-world and was quickly "promoted" to sales.
Like Hans had indicated, a model is a reasonable starting point but it eventually turns in to an exercise where the designer is hand picking components that are close in value and inserting them in to the circuit to come up with the accurate values.
You should look up the reference to something called a "gimmick capacitor". Sometimes I work on old radios and find one left behind; It is humorous when some other restorer either cuts it out or just solders the two ends together.
--
Tisha, AA4HA
|
As was mentioned, you need to add the missing capacitor in parallel. 47pf in the QDX HB. IIRC this is all the parts Hans shows in his example of the LPF modeling in the QDX manual. I did my modeling for 50 ohm impedances, but knowing that it's not what happens in reality. But using the "tune part" function in Elsie, I could easily see what the relative effect was of changing inductances or capacitor values.?
I modeled this LPF for the QDX using values that Hans put in the manual for inductance. (Which, is what it was modeled with I believe, probably using the earlier T37-10 toroids.)? ?That one used 290 and 250nh for inductor values for the 15/17m LPF. If I leave the one at 290, and shift the other to less ,like 220nh, my -3db frequency increases by about 1 Mhz.
For this LPF, try reducing the inductance on Part #4 in your model.??
If you mess around with the inductance of part 2, you'll see in the plots it wildly affects the SWR...which is another component of filter design that makes it more tricky! Reducing Part #4 inductance seems to be favorable, SWR wise, and still moves the cutoff frequency up.?
-Nate
N8BTR
?
|
Start with Zout is not 100 ohms its 50, the antenna.
Steve why add 10 ohms at the input?? ?ELSIE can set the source and
load impedance to the desired value.? All the 10 ohms adds is
insertion loss and will distort the filter.
Generally with Elsie if I use real values and Qs I get result remarkably
close to reality.? By remarkably close if I use 10% caps, and measured
inductance I get what was modeled.? When they differ its due to strays
or component tolerance (or the selected filter is intolerant!) .
I find the TUNE function useful. With that I can inject some manual
Monte Carlo to see how it behaves with part variation.?
To move the cutoff up start with the base filter and trim the value
by a percentage of the original base frequency and the target.
That's only a starting point.?
So software simulation is as much how the user drives it can influence
the result.? GIGO!
RE: FERRITE over the years I've used a metric boat load of it for all manner?
of designs.? ?EVERY TIME I had an error that didn't measure there was an
error either not uniformly wound or under count (+1 turn).
FYI:? if you pinch the winding close rather than uniform a variation of
more than 10% is easily had and I've seen as much as 18% measured
from close to as wide as possible.
Tools that measure I have, AADE LCII (very accurate), ACE LCR,
NanoVNA (I have three), HP4191A impedance analyzer, Daton M250?
LCR bridge.? When they all agree, then the error is elsewhere.
Its very easy to create your own errors.
--
Allison
------------------
Post online only,?
direct email will go to a bit bucket.
|
Tisha,
Thats the issue. So many newer hams and electronic hobbyists didn't build the simple kits of the 1960s, etc., and never saw or used a gimmick capacitor. ? One of the 4SQRP Club kits used a gimmick cap recently (ie in the past 10 or 15 years). I just can’t remember which kit. ?Ha! Be the REASON someone smiles today.
Dave K8WPE
toggle quoted message
Show quoted text
On May 29, 2024, at 12:51?PM, Tisha Hayes, AA4HA <Tisha.Hayes@...> wrote:
?Steve,
Agreed, those input/output impedances have bitten me in the behind before; I had an engineering manager who believed that the software model was 'more accurate' than what we built on the bench and if the results were different it was because we were incompetent at actually building a circuit. Thankfully he was not long for the engineering-world and was quickly "promoted" to sales.
Like Hans had indicated, a model is a reasonable starting point but it eventually turns in to an exercise where the designer is hand picking components that are close in value and inserting them in to the circuit to come up with the accurate values.
You should look up the reference to something called a "gimmick capacitor". Sometimes I work on old radios and find one left behind; It is humorous when some other restorer either cuts it out or just solders the two ends together.
--
Tisha, AA4HA
|
Tisha Hayes, AA4HA
Ronald Taylor
David Wilcox K8WPE