S-meter calibration

[Daniel Estévez]

Hi,

This is not so much Linrad-related, so excuse me for the off-topic.
Perhaps it could be interesting to other one than myself.

I'm using Linrad as a panadapter for my FT-817ND with a Funcube Dongle
Pro+. Since the S-meter on the FT-817ND is terrible (it is very far from
being 6dB per S-point), I'm thinking of using the S-meter in Linrad.
However, I first need to calibrate the signal level in Linrad (so I
guess this could be of interest to anyone wishing to calibrate their SDR).

I don't have a signal generator to do the calibration, but I have an
oscilloscope. My idea was to make an HF crystal oscillator with some
parts I had around, to measure the output of the oscillator with the
scope and to take it through an attenuator of known value into the radio.

The circuit I have so far is the following:

http://oi67.tinypic.com/28im89d.jpg

The idea with the 1M and two 100R resistors is that they will attenuate
the output a lot and provide a 50R impedance match. If I knew the
attenuation factor they, I could just measure the oscillator at the
emitter and get the output level from that.

However, I'm having some trouble understanding the impedance
matching/attenuation properly. If I replace the 1M resistor by 100K the
signal level in the radio increases by 10dB, and not by 20dB as I think
it should (the voltage divider with 100K gives roughly 10 times more
voltage across the output than with 1M).

I guess I'm trying to do something similar to this:

http://www.elecraft.com/manual/E740084%20XG2%20Manual%20Rev%20E.pdf

but a bit more simple, since I don't need 50uV output as long as I can
measure the output in some way with the scope.

Any comments on how to compute/understand the attenuation factor or how
approach this problem/make this kind of circuit better are welcome.

73,

Dani EA4GPZ.

[Leif Asbrink]

Hello Dani,

You can not make a RF attenuator with a 1M resistor.
Already at 1 MHz a stray capacitance of 0.2 pF would
have an impedance of 1M.

It seems your 1M is more like 200k while the 100k
would be something like 66k. (stray capacitance 200k)

You might use a 10 k resistor to attenuate 20 dB.

I suggest you put 50 ohms from ground to your 6.8 k
emitter resistor. Then you can bring the signal out
and add external attenuation - or you can apply adequate
screening and attenuate in steps of 20 dB with T or PI
attenuators.

In total you will not get very good accuracy because
uncertainties add up in an unknown way.

There is a simpler way. Find out the noise figure at your
FT817 input. If you are on VHF you might use a preamp
with a known NF.
Table 4 here:
http://www.sm5bsz.com/dynrange/gavelstad/gav.htm
lists the FT817 with NF=5.4 dB.

If you want to calibrate the Linrad according to the IARU
standard S9=-93 dBm (VHF) it is easy.

NF=0 is -174 dBm/Hz.
NF=5.4 dB is -168.6 dBm/Hz.

Set your baseband filter to 1.0 kHz.

The noise floor is then at -138.6 dBm.

That is 45.6 dB below -93 dBm. That is 7.6 S units.

You should calibrate youe S-meter to show 1.4 S unit
for the noise floor in 1 kHz. That is 2.4 dB above S1.
In SSB bandwidth, 2.3 kHz, 3.6 dB more bandwidth, the
S-meter should show S2.

If you have a LNA in your system, you probably know the
NF to within 0.2 dB so you could then calibrate your
S-meter to within almost 0.2 dB.

Calibration will most probably be different on different
bands.

73

Leif

> --
> You received this message because you are subscribed to the Google Groups "Linrad" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to linrad+un...@googlegroups.com.
> To post to this group, send email to lin...@googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.

[Daniel Estévez]

El 19/01/16 a las 04:13, Leif Asbrink escribió:

> Hello Dani,
>
> You can not make a RF attenuator with a 1M resistor.
> Already at 1 MHz a stray capacitance of 0.2 pF would
> have an impedance of 1M.

Hi Leif,

Thank you very much for your reply.

I was afraid that stray capacitance would be a problem here, but (stupid
me), I didn't do the calculations and I didn't think it was so significant.

> I suggest you put 50 ohms from ground to your 6.8 k
> emitter resistor. Then you can bring the signal out
> and add external attenuation - or you can apply adequate
> screening and attenuate in steps of 20 dB with T or PI
> attenuators.

I will try that. In my first tests, putting 50 ohms from the emitter to
ground (through an AC coupling cap, of course) loaded the oscillator and
reduced the oscillation amplitude by about a factor of 10. This may
actually be good, since it will reduce the amount of attenuation I have
to apply later.

> In total you will not get very good accuracy because
> uncertainties add up in an unknown way.

Yes. I'm aware that you can't get precise results with this method. If I
get something like 3dB accuracy I will be happy, because it will be much
better than the FT-817 S-meter.

> There is a simpler way. Find out the noise figure at your
> FT817 input. If you are on VHF you might use a preamp
> with a known NF.
> Table 4 here:
> http://www.sm5bsz.com/dynrange/gavelstad/gav.htm
> lists the FT817 with NF=5.4 dB.

I know of that procedure. The problem is that I'm not so sure on the NF
of my receive chain. I want to do this at HF, not at VHF, so the FT817
NF might be different. Also, I'm tapping the IF just after the first
mixer, before the crystal filter. I use a G4HUP PAT board to tap, and
then this goes directly to the Funcube Dongle. Perhaps I should take
into account the NF's of the G4HUP board and Funcube Dongle, and gains
of these stages if I wanted to get an approximated value for my NF (I'm
aware it may happen that the NF of some of these elements doesn't matter
due to high gain before them).

73,

Dani EA4GPZ.

[Leif Asbrink]

Hello Dani,

> > I suggest you put 50 ohms from ground to your 6.8 k
> > emitter resistor. Then you can bring the signal out
> > and add external attenuation - or you can apply adequate
> > screening and attenuate in steps of 20 dB with T or PI
> > attenuators.
>
> I will try that. In my first tests, putting 50 ohms from the emitter to
> ground (through an AC coupling cap, of course) loaded the oscillator and
> reduced the oscillation amplitude by about a factor of 10. This may
> actually be good, since it will reduce the amount of attenuation I have
> to apply later.

This was not what I meant. I suggested you put 50 ohm in series
with the 6.8 k resistor to make the signal level across
the 50 ohm resistor small. (I had expected the oscillator
to not oscillate at all when loaded by 50 ohms - but of course
if you select a reasonably small capacitor it would work fine.

> > In total you will not get very good accuracy because
> > uncertainties add up in an unknown way.
>
> Yes. I'm aware that you can't get precise results with this method. If I
> get something like 3dB accuracy I will be happy, because it will be much
> better than the FT-817 S-meter.
>
> > There is a simpler way. Find out the noise figure at your
> > FT817 input. If you are on VHF you might use a preamp
> > with a known NF.
> > Table 4 here:
> > http://www.sm5bsz.com/dynrange/gavelstad/gav.htm
> > lists the FT817 with NF=5.4 dB.
>
> I know of that procedure. The problem is that I'm not so sure on the NF
> of my receive chain. I want to do this at HF, not at VHF, so the FT817
> NF might be different. Also, I'm tapping the IF just after the first
> mixer, before the crystal filter. I use a G4HUP PAT board to tap, and
> then this goes directly to the Funcube Dongle. Perhaps I should take
> into account the NF's of the G4HUP board and Funcube Dongle, and gains
> of these stages if I wanted to get an approximated value for my NF (I'm
> aware it may happen that the NF of some of these elements doesn't matter
> due to high gain before them).

If the SDR noise floor drops by more than 10 dB when you switch off
the FT817 you will have the NF of the FT817 within about 1 dB.
If the noise flor drops by about 17 dB you will have the FT817 NF
within 0.1 dB. (provided the the G4HUP PAT is powered all the time.)

I think you can look up the FT817 NF at ARRL Lab or sherweng.com

You might buy one of these:
http://www.minicircuits.com/products/amplifiers_smt_gpw.shtml

That would give you an accurately known NF which you can use to
calibrate your signal generator. Some are rated DC to 1 GHz.
GALI-74+ for example. The problem is to find them in a small
quantity. The NF is 2.4 dB from DC to 50 MHz according to this:
http://images10.newegg.com/UploadFilesForNewegg/itemintelligence/3M/GALI_741445209604523.pdf

73

Leif

[Daniel Estévez]

El 19/01/16 a las 20:51, Leif Asbrink escribió:
> Hello Dani,

>> I will try that. In my first tests, putting 50 ohms from the emitter to
>> ground (through an AC coupling cap, of course) loaded the oscillator and
>> reduced the oscillation amplitude by about a factor of 10. This may
>> actually be good, since it will reduce the amount of attenuation I have
>> to apply later.
> This was not what I meant. I suggested you put 50 ohm in series
> with the 6.8 k resistor to make the signal level across
> the 50 ohm resistor small. (I had expected the oscillator
> to not oscillate at all when loaded by 50 ohms - but of course
> if you select a reasonably small capacitor it would work fine.

Hi Leif,

Thanks for clarifying that point. I will also try to do that.

> If the SDR noise floor drops by more than 10 dB when you switch off
> the FT817 you will have the NF of the FT817 within about 1 dB.
> If the noise flor drops by about 17 dB you will have the FT817 NF
> within 0.1 dB. (provided the the G4HUP PAT is powered all the time.)

I didn't know that result, so I actually did some calculations to figure
out where it comes from.

I'll perform that test and write again with the results.

> I think you can look up the FT817 NF at ARRL Lab or sherweng.com

Thanks. I found the data for the FT817 at the ARRL Lab. They quote
-123dBm noise floor in 500Hz at 3.53MHz (with Preamp off), which gives a
noise figure of 23dB.

Interestingly, they give a -131.1dBm noise floor in 500Hz at 144.02MHz,
giving a noise figure of 16.1dB. This is in contrast with other sources
that give about 5dB.

I think I'm doing the calculations right. To get the noise figure from
the noise floor I have to subtract -141dBm (the Johnson-Nyquist noise in
500Hz) from the noise floor, right?

> You might buy one of these:
> http://www.minicircuits.com/products/amplifiers_smt_gpw.shtml
>
> That would give you an accurately known NF which you can use to
> calibrate your signal generator. Some are rated DC to 1 GHz.
> GALI-74+ for example. The problem is to find them in a small
> quantity. The NF is 2.4 dB from DC to 50 MHz according to this:
> http://images10.newegg.com/UploadFilesForNewegg/itemintelligence/3M/GALI_741445209604523.pdf

Just to make sure I understand how to use one of these: I see two ways
of using it.

1. Connect its input to a 50 ohm load. Then the noise power density at
the output is

G+NF-174 dBm/Hz,

where G is the gain of the monolithic amplifier and NF its noise figure.
Then I can use this value to calibrate the amplitude of any SDR.

2. Connect some signal generator to its input and an SDR to its output.
Set the amplitude of the SDR to get the noise floor at

NF-174 dBm/Hz.

Then measure the signal in the SDR to get its level. Then the signal
generator can be input directly to the SDR to calibrate its amplitude
and get its NF.

I'll give all this kinds of tests a shots, since I want to compare the
results and I'll probably learn quite a bit in doing this.

73,

Dani.

[Leif Asbrink]

Hi Dani,

> Thanks. I found the data for the FT817 at the ARRL Lab. They quote
> -123dBm noise floor in 500Hz at 3.53MHz (with Preamp off), which gives a
> noise figure of 23dB.

Close, but log 500=27 and room temp is -174 dBm/Hz so a NF of 0 dB
corresponds to -147 dBm. That is 24 dB from -123 dBm so NF=24 dB.

> Interestingly, they give a -131.1dBm noise floor in 500Hz at 144.02MHz,
> giving a noise figure of 16.1dB. This is in contrast with other sources
> that give about 5dB.

147-131.1=15.9 ;-)

Seems measurements differ in that an RF amplifier is on or off.

> I think I'm doing the calculations right. To get the noise figure from
> the noise floor I have to subtract -141dBm (the Johnson-Nyquist noise in
> 500Hz) from the noise floor, right?

?????

> > You might buy one of these:
> > http://www.minicircuits.com/products/amplifiers_smt_gpw.shtml
> >
> > That would give you an accurately known NF which you can use to
> > calibrate your signal generator. Some are rated DC to 1 GHz.
> > GALI-74+ for example. The problem is to find them in a small
> > quantity. The NF is 2.4 dB from DC to 50 MHz according to this:
> > http://images10.newegg.com/UploadFilesForNewegg/itemintelligence/3M/GALI_741445209604523.pdf
>
> Just to make sure I understand how to use one of these: I see two ways
> of using it.
>
> 1. Connect its input to a 50 ohm load. Then the noise power density at
> the output is
>
> G+NF-174 dBm/Hz,
>
> where G is the gain of the monolithic amplifier and NF its noise figure.
> Then I can use this value to calibrate the amplitude of any SDR.

Yes. (But G may depend on impedances so accuracy may be not so good.)

> 2. Connect some signal generator to its input and an SDR to its output.
> Set the amplitude of the SDR to get the noise floor at
>
> NF-174 dBm/Hz.
>
> Then measure the signal in the SDR to get its level. Then the signal
> generator can be input directly to the SDR to calibrate its amplitude
> and get its NF.

Yes. This should give a good accuracy.

> I'll give all this kinds of tests a shots, since I want to compare the
> results and I'll probably learn quite a bit in doing this.

Good luck:-)

73

Leif

[Daniel Estévez]

El 23/01/16 a las 17:23, Leif Asbrink escribió:

> Hi Dani,
>
>> Thanks. I found the data for the FT817 at the ARRL Lab. They quote
>> -123dBm noise floor in 500Hz at 3.53MHz (with Preamp off), which gives a
>> noise figure of 23dB.
> Close, but log 500=27 and room temp is -174 dBm/Hz so a NF of 0 dB
> corresponds to -147 dBm. That is 24 dB from -123 dBm so NF=24 dB.

Ok. That's the same calculation I made. I probably made a typo or other
stupid error.

>> Interestingly, they give a -131.1dBm noise floor in 500Hz at 144.02MHz,
>> giving a noise figure of 16.1dB. This is in contrast with other sources
>> that give about 5dB.
> 147-131.1=15.9 ;-)
>
> Seems measurements differ in that an RF amplifier is on or off.

On the FT-817, on 144MHz the RF preamp is always on, so this shouldn't
be a factor.

>> I think I'm doing the calculations right. To get the noise figure from
>> the noise floor I have to subtract -141dBm (the Johnson-Nyquist noise in
>> 500Hz) from the noise floor, right?
> ?????

Sorry, I took the noise floor in 1kHz (-144dBm). I intended to subtract
3dB from that to get the noise floor in 500Hz, but I ended up adding 3dB
(silly mistake).

>> I'll give all this kinds of tests a shots, since I want to compare the
>> results and I'll probably learn quite a bit in doing this.

I did the noise level measurements switching on and off the FT-817 with
the G4HUP PAT board running. The noise floor changes little, by about 3
or 5dB, so the NF of the whole system is not just the NF of the FT-817.

I have built the circuit of
http://norcalqrp.org/files/NorCal_S9_Assy_V1.pdf
This is supposed to give a S9 (50uV) signal.

For me, it gives S8, according to my FT-817 and also to the calibration
I have done using the NF of the FT-817 (but this may not match well the
NF of the whole panadapter system).

I don't trust this circuit very much because it is only specified at
+/-3dB and because I don't quite understand how the particular value of
-73dBm gets generated.

I see that Q1 is a common emitter amplifier with a voltage gain of 1000
(20K emitter resistor and 20 ohm collector resistor) and there is a 56
ohm shunt, 470 ohm series pi attenuator in parallel with the 20 ohm
collector resistor. This attenuator gives about 26dB attenuation.
Therefore, to have -73dBm at the output, I think I need 2.8Vpp at the
base. In my circuit the oscillation seems much lower, although sticking
a 10x probe in the base might load it too much for this measurement to
be valid.

Anyway, I'm going to follow your suggestion and build a preamp with a
known NF. Minikits sells a nice kit for the GALI-39 which I have already
ordered.

73,

Dani EA4GPZ.