Topics

Transmission lines, Input match and Connectors #bestpractice #experiment

Wes Stewart
 

Has anyone measured the input match of the various filter bands?

Wes  N7WS

prog
 

On Wed, Oct 9, 2019 at 10:29 PM, Wes Stewart wrote:
Has anyone measured the input match of the various filter bands?

Wes  N7WS
This discussion could be of interest: https://twitter.com/lambdaprog/status/1183359218323283973

Ron Liekens
 

If interrested there is another nice piece of sftware that has a lot of ponential.
https://zs1sci.com/blog/nanovnasaver/#Edit

There is also a new model NanoVna in the running that will measure higher in frequency and has more dynamic range.

Regards, Ron
ON2RON

prog
 

On Sun, Oct 13, 2019 at 03:24 PM, Ron Liekens wrote:
If interrested there is another nice piece of sftware that has a lot of ponential.
https://zs1sci.com/blog/nanovnasaver/#Edit

There is also a new model NanoVna in the running that will measure higher in frequency and has more dynamic range.

Regards, Ron
ON2RON
Of course. I expect the software section to develop even further.
In any case, make sure:
1) The unit is powered on and streaming
2) The unit is tuned in the pre-selector segment you are assessing

Wes Stewart
 

I asked the question because I had already done it on two units.  I used a DG8SAQ VNWA and to summarize, we agree rather closely. 

Of course, I knew to measure each band with the radio actually tuned to a frequency in that band.  There are some subtleties; AGC, input attenation, preamp state all affect the results to some extent.  I lowered the test signal amplitude as much as possible without affecting dynamic range, to minimize these factors.

I chose to call the bands, A, B, C and D respectively.  An interesting observation is that measuring the frequency range of band "D", 17 to 31 MHz, but tuning the radio to a frequency in Band C, i.e. 16 MHz, gives a much better result.

Wes  N7WS

On Sunday, October 13, 2019, 5:45:21 AM MST, prog <info@...> wrote:


On Wed, Oct 9, 2019 at 10:29 PM, Wes Stewart wrote:
Has anyone measured the input match of the various filter bands?

Wes  N7WS
This discussion could be of interest: https://twitter.com/lambdaprog/status/1183359218323283973

prog
 

On Sun, Oct 13, 2019 at 04:43 PM, Wes Stewart wrote:
I chose to call the bands, A, B, C and D respectively.  An interesting observation is that measuring the frequency range of band "D", 17 to 31 MHz, but tuning the radio to a frequency in Band C, i.e. 16 MHz, gives a much better result.
The return loss of the 10 .. 17 MHz filter is also good for 17 .. 31 MHz because it is a 3rd order elliptic HPF.
That said, the filters were mostly optimized for the best rejection and insertion loss while still being "OK" for the return loss.
The following plot should give you an idea about what has been done.
Sig gen is a R&S SMBV100A followed by an attenuator to give a (mostly) resistive source to the unit.


Wes Stewart
 



On Sunday, October 13, 2019, 8:02:17 AM MST, prog <info@...> wrote:


On Sun, Oct 13, 2019 at 04:43 PM, Wes Stewart wrote:
I chose to call the bands, A, B, C and D respectively.  An interesting observation is that measuring the frequency range of band "D", 17 to 31 MHz, but tuning the radio to a frequency in Band C, i.e. 16 MHz, gives a much better result.
The return loss of the 10 .. 17 MHz filter is also good for 17 .. 31 MHz because it is a 3rd order elliptic HPF.
That said, the filters were mostly optimized for the best rejection and insertion loss while still being "OK" for the return loss.
The following plot should give you an idea about what has been done.
Sig gen is a R&S SMBV100A followed by an attenuator to give a (mostly) resistive source to the unit.

Okay.  I obviously have no first-hand knowledge of the internal workings of these devices.  I gleaned what I do know from reading the description on the website and by empirical measurements.  The block diagram shows the input filter as a BPF, so that's what I assumed when I read, "New High Performance Pre-selector for HF with 4 filter banks (corners at ~DC, 5, 10, 17 and 31 MHz)"

I also expected better input matching when I read: "Improved input matching down to virtually DC." Clearly, I should have asked myself, "Improved compared to what?"


prog
 

The BPFs are obtained by combining HPFs and LPFs.

Wes Stewart
 

Okay, understood.  It would be nice to see some plots of this.

Wes  N7WS

On Sunday, October 13, 2019, 11:15:14 AM MST, prog <info@...> wrote:


The BPFs are obtained by combining HPFs and LPFs.
_._,_._,_


Jos Stevens
 

Hello Prog and Wes,

Wes you asked if somebody has measured the input impedance of the Airspy Discovery,  so I tryed to measure this , using my Vector Network Analyser FA-VA3.

Please find attached my SWR graphs, for what they are worth.

The graphs show that 3.5 and 28 MHz responses are a bit schifted to the high side and at  14 an 21 MHz bands the SWR response is remarkable.

Please Prog  you speak aboput 4 filters with corners at ~DC, 5, 10, 17 and 31 MHz, are my graphs as to be expected ?

I measured the measuring voltage to be a few hundred millivolts Vtop, pretty high ofcourse , maybe too ?

I'm worryng about the response at 10 to 26 MHz.

I also mesured the parallel components of the input impedance at some  fixed frequencies :

          50 kHz : 370 Ohms // 4.2 nF

       1   MHz  : 128 Ohms // 3.1 nF

     3.65 MHz : 18 Ohms // 3.2 uH8

      7.06 MHz : 29 Ohms// 915 pF

14.25 MHz : 12.4 Ohms// 295 pF

21.219 MHz : 11.3 Ohms// 7.5 Pf

21.290 MHz : 11.2 Ohms// 7.5 pF

28.500 MHz 11.8 Ohms// 310 nH

Yes I  found the Nanovna and had ordered one,  i'm waiting for it impatiently now, then I can compare the results, and measure the higher VHF responses.

Again Prog are my measurements as te be expexted, or is there something wrong with me ??

Best 73 de

Jos PA3CCE



 



Op 9-10-2019 om 21:54 schreef Wes Stewart via Groups.Io:
Has anyone measured the input match of the various filter bands?

Wes  N7WS

Virusvrij. www.avg.com

prog
 

On Mon, Oct 14, 2019 at 02:07 PM, Jos Stevens wrote:
The graphs show that 3.5 and 28 MHz responses are a bit schifted to the high side and at  14 an 21 MHz bands the SWR response is remarkable.
Regardless of the correctness of the measurements, I want to highlight again that the SWR is much less important in receive than other parameters like the actual MDS and linearity in the desired bands and the rejection in the problematic bands. A reasonable SWR value is sufficient, but good harmonic and sub-harmonic responses are much more important for optimal performance. The latest videos from Leif SM5BSZ demonstrated how this approach compares with other radios with real life signals:

https://youtu.be/yevOmGyRhbk?t=1826
https://youtu.be/uUHr9HilTrQ?t=974

Rewind the videos for the detailed testing methodology.

Jos Stevens
 

Hi Prog and Wes,

I do understand what you mean Prog, no problem.

Please forget the 10 MHz - 26 MHz graph, SDR# was not set correctly, see the attached graph's, this time in csv format if you don't mind.

Regards,

Jos


Op 14-10-2019 om 14:42 schreef prog:
On Mon, Oct 14, 2019 at 02:07 PM, Jos Stevens wrote:
The graphs show that 3.5 and 28 MHz responses are a bit schifted to the high side and at  14 an 21 MHz bands the SWR response is remarkable.
Regardless of the correctness of the measurements, I want to highlight again that the SWR is much less important in receive than other parameters like the actual MDS and linearity in the desired bands and the rejection in the problematic bands. A reasonable SWR value is sufficient, but good harmonic and sub-harmonic responses are much more important for optimal performance. The latest videos from Leif SM5BSZ demonstrated how this approach compares with other radios with real life signals:

https://youtu.be/yevOmGyRhbk?t=1826
https://youtu.be/uUHr9HilTrQ?t=974

Rewind the videos for the detailed testing methodology.

Virusvrij. www.avg.com

Wes Stewart
 

On Mon, Oct 14, 2019 at 05:42 AM, prog wrote:
On Mon, Oct 14, 2019 at 02:07 PM, Jos Stevens wrote:
The graphs show that 3.5 and 28 MHz responses are a bit schifted to the high side and at  14 an 21 MHz bands the SWR response is remarkable.
Regardless of the correctness of the measurements, I want to highlight again that the SWR is much less important in receive than other parameters like the actual MDS and linearity in the desired bands and the rejection in the problematic bands. A reasonable SWR value is sufficient, but good harmonic and sub-harmonic responses are much more important for optimal performance. The latest videos from Leif SM5BSZ demonstrated how this approach compares with other radios with real life signals:


Maybe you've said elsewhere that SWR doesn't matter, but I've not seen it in this thread.  Nevertheless, I generally agree, until you get to VHF where transmission line loss increases due to SWR can affect system noise figure.  To be fair, the HF+ Discovery input Z in the upper two bands is pretty good and LNAs designed for NF are often have pretty lousy matching.

I've attached plots of both SWR and Smith format of the six bands as I've measured them.  In all cases a fully calibrated DG8SAQ VNWA was used with post processing in ZPlots to get the graphs.

prog
 

On Wed, Oct 16, 2019 at 05:49 PM, Wes Stewart wrote:
To be fair, the HF+ Discovery input Z in the upper two bands is pretty good and LNAs designed for NF are often have pretty lousy matching.
It's easier to get good matching over half an octave than two or three octaves.

jdow
 

Well, for what it is worth I will reiterate what Youssef said. In general the input match on the receiver does not matter.

In specific, however, there is an issue with filters attached to the input of the receiver. Their passband shape will be distorted. If this is a problem for you then put up a little more wire and add an attenuator to the receiver front end. The attenuation won't matter and the attenuator will make the filter happy.

For what it is worth for most circuit configurations (Leif appears to have exception) an input match gives you an absolute minimum noise figure of 3 dB. However, if you mismatch the inputs creatively you can get a noise figure well below 3 dB. All devices have an input noise current in shunt with the input and an input noise voltage in series with the input. The optimum match is (obviously) En/In, which isn't a neat 1:1 SWR at any impedance. You can, however, do startling things with extreme mismatches so that En or In are immaterial. You do not get the optimum possible noise figure. But, you do get a very very good one.

All that said, what matters in the end is the signal to noise power ratio at the demodulator. And at HF the single largest source of noise is external to your system. This is why you can put in a 30 dB attenuator from a 160 meter dipole and still copy all the weak signals you can find. That is why I declare if input match is critical to a filter you have designed, then put in a 10 dB pad to make it happy. I learned this the expensive way as a teenager. I was on 15 meters at the time in about 1960. I was displeased with SNR on received signals. DowKey made a super low noise preamp. I scraped up the money (never smoking helped me save the money) and bought it. If anything the results were worse due to increased IMD. In college some extra-curricular reading led me to an excellent book, long out of print, on noise figure. And the old Terman handbook/textbook I found used filled in the galactic and other noise sources picture for HF through 2 meters where it becomes nearly negligible. Noise figure cannot be reduced from what your antenna presents your radio. It's something about thermodynamics, the "you can't win" law. It was some years later that I learned about IMD, the other slab of bread surrounding the signal sandwich. Gain can help front end noise mask noise from succeeding stages; gain also puts you closer to overload. (And the really stinky part of this is that AGC systems may reduce signal level; but, most ways you can do it also reduce the large signal handling ability.)

So, Wes, at the risk of being overly blunt, for the most part your excellent efforts were a misinformed waste of time. Sorry for hammering the so many nails in this coffin; but, this ex-cathedral pontificating finally got to me.

{^_^} Joanne

On 20191016 06:16:37, Wes Stewart via Groups.Io wrote:
On Mon, Oct 14, 2019 at 05:42 AM, prog wrote:
On Mon, Oct 14, 2019 at 02:07 PM, Jos Stevens wrote:
The graphs show that 3.5 and 28 MHz responses are a bit schifted to the
high side and at  14 an 21 MHz bands the SWR response is remarkable.
Regardless of the correctness of the measurements, I want to highlight
*again* that the SWR is much less important in receive than other parameters
like the actual MDS and linearity in the desired bands *and* the rejection
in the problematic bands. A reasonable SWR value is sufficient, but good
harmonic and sub-harmonic responses are much more important for optimal
performance. The latest videos from Leif SM5BSZ demonstrated how this
approach compares with other radios with real life signals:
Maybe you've said elsewhere that SWR doesn't matter, but I've not seen it in this thread.  Nevertheless, I generally agree, until you get to VHF where transmission line loss increases due to SWR can affect system noise figure.  To be fair, the HF+ Discovery input Z in the upper two bands is pretty good and LNAs designed for NF are often have pretty lousy matching.
I've attached plots of both SWR and Smith format of the six bands as I've measured them.  In all cases a fully calibrated DG8SAQ VNWA was used with post processing in ZPlots to get the graphs.
Band_A Custom.gif
Band_A Smith.gif
Band_F Custom.gif
Band_F Smith.gif
Band_B Custom.gif
Band_B Smith.gif
Band_C Custom.gif
Band_C Smith.gif
Band_D Custom.gif
Band_D Smith.gif
Band_E Custom.gif
Band_E Smith.gif

Wes Stewart
 

On Wed, Oct 16, 2019 at 08:58 AM, prog wrote:
On Wed, Oct 16, 2019 at 05:49 PM, Wes Stewart wrote:
To be fair, the HF+ Discovery input Z in the upper two bands is pretty good and LNAs designed for NF are often have pretty lousy matching.
It's easier to get good matching over half an octave than two or three octaves.
Since it doesn't appear to me that there is enough room in this package for many Ls and Cs how (if you care to say) do you build these elaborate filters?  Are these active inductors?

prog
 

If a perfectly resistive load is required for something else other than reception, just add attenuators. This will cost you the valuable MDS we fought to get you here, but I'd assume at this point you know what you are doing.
Spectrum analyzers do not make for good receivers for the same reasons.

prog
 

On Thu, Oct 17, 2019 at 07:44 AM, Wes Stewart wrote:
Since it doesn't appear to me that there is enough room in this package for many Ls and Cs how (if you care to say) do you build these elaborate filters?  Are these active inductors?
Better filters are always better, of course, but definitely not for the reasons you are stating. I would again oprimize for the MDS and the rejection.
Yes, designing receivers is not as intuitive as it seems.

jdow
 

With antennas suitable for efficient transmitting 10 dB is generally safe through 15 MHz maybe 28 MHz. With poor antennas - well - if you install an attenuator and the noise still rises when you replace a dummy load with the antenna, you're good to go. You won't be able to improve things more than a dB or so. That is the bottom line test for what you have.

{^_-}

On 20191016 22:55:39, prog wrote:
If a perfectly resistive load is required for something else other than reception, just add attenuators. This will cost you the valuable MDS we fought to get you here, but I'd assume at this point you know what you are doing.
Spectrum analyzers do not make for good receivers for the same reasons.

jdow
 

{^_-}

On 20191016 22:59:05, prog wrote:
On Thu, Oct 17, 2019 at 07:44 AM, Wes Stewart wrote:
Since it doesn't appear to me that there is enough room in this package for
many Ls and Cs how (if you care to say) do you build these elaborate
filters?  Are these active inductors?
Better filters are always better, of course, but definitely not for the reasons you are stating. I would again oprimize for the MDS and the rejection.
Yes, designing receivers is not as intuitive as it seems.