Sunday, 28 April 2019

More IC9700 learning

I spent some time on Saturday talking to local SDR/DSP experts and as a result have a much better idea of how the IC9700 works.
I am on a steep learning curve!
I hope to publish my thoughts on this in my blog in due course.

Thursday, 25 April 2019

More 9700

No measurements today. I have been reflecting in some of these numbers and measurement techniques.
NPR would be a much better way to test the SDR architecture, but the appropriate filters are not available to me. I used to do NPR measurements on microwave link equipment, so I am familiar with the technique.

I had a first contact using the IC9700 this evening.  Admittedly it was ising the 9700 only on receive on the QO100 satellite. As much as I would have loved to use the transmit side for the uplink, I could not find an 8 pin DIN to connect up the SEND output to drive the transverter and PA, so I had to connect up the FT847 for transmit.
My contact was with G7OCD. Jason was using a Lime  Mini for transmit and receive.

Wednesday, 24 April 2019

IC9700 part 2. Some numbers

I spent all afternoon re-testing and confirming measurements on the receiver sections of the IC9700.
 My measurement of noise figure are based on the Y factor method, using HP noise heads with known and well characterised ENR.
Noise power measurement at the receiver output used Moetronics Spectravue Continuum mode running on a Lenova desktop PC and SDR-IQ.

This Y factor technique is well known, although the use of the SDR-IQ to measure the resulting noise output is possibly a little unusual.

Like all these measurements, there are uncertainties in the measured values due to assumptions, rounding and the tolerance of out-of-cal test gear. The values given here are best efforts to get it right, but no guarantee is given as to accuracy.

Noise figure

Noise figure at 292k ambient.
Preamp on   preamp off
4.2dB          17.3db

Preamp on   Preamp off
4.8dB          15.9dB

Preamp on   Preamp off
4.8dB          8.7dB

A low ENR HP346A was used for the preamp on and an HP346C for the preamp off  (higher ENR required for 144 and 432MHz)
High noise figure is expected with the direct sampling technique, without preamplification, as used in the IC9700 
The 1296MHz down conversion arrangement loses some of the direct sampling advantage but does allow a lower noise figure. The preamp is still worth switching in, however, but does not have such a profound effect as on the lower two bands.
On-air listening tests and signal generator tests seem to confirm that these numbers are not far out. 144MHz, without the preamp switched in, is particularly deaf, but quite acceptable with preamp on.

Spurious free dynamic range
144MHz only so far
Two tone dynamic range
Using two very low phase noise test oscillators
The two oscillators are combined in a high isolation combiner and then well attenuated.
Applied at the receiver input, the levels were adjusted to give a third order product at the receiver noise floor.
The test was done with the preamp on and repeated with it off.
The IP+ was then used to test its effect.

Preamp on
89dB at 20kHz tone spacing. I don’t have any low phase noise oscillators at 2kHz spacing
Frequencies 144.100 and 144.120MHz 

Preamp off
91.7dB at 20kHz tone spacing. Same test frequencies. 

I found that IP+ was only effective with preamp on. The dynamic range was not changed, only shifted upwards. However, the noise floor increased by 3dB. This was as expected.

This form of dynamic range test, with a direct sampling SDR receiver, can give misleading results as an SDR does not follow the usual rules for third order intermodulation. It is possible that the third order product may approach the noise floor level at some lower levels of two tome input, dropping away again as the tone levels are increased. The quoted levels were found without difficulty but I cannot be sure that one or more lower levels may have been missed. IP+ helps to overcome this problem by introducing noise as a ‘dither’ signal. As I don’t have the IC9700 service manual and schematics I cannot be sure this is the technique that ICOM used. The 3dB shift of dynamic range (only a shift of position, not an increase in dynamic range) when IP+ is on suggests that the  -89dB figure is not that far off.

S meter calibration
As you would expect of an SDR based receiver the S meter calibration is fairly good. 
Each S point is 3dB and above S9 each 10dB measures 10dB!
Preamp on 
S9 = -95dBm

Preamp off
S9 = -84dBm
If  an 11dB (approximate) gain preamp is used on 144MHz then  if the noise figure of that preamp was about 0.8dB, the overall preamp on and off noise figures are about right.

I still need to test on 432MHz and 1296MHz

These measurements are my copyright. If you want to quote them, please acknowledge this blog.

I offer no guarantee as to the accuracy of the measurements. They are best effort and may be amended following further testing.

Tuesday, 23 April 2019

IC9700 noise figures part 1

Now I am a little more familiar with the rig and have updated to v1.06 firmware I have at last started to make measurements on this much-anticipated transceiver.

The first measurements made are the receiver noise figure. Unlike preamps and transverters, transceivers don't usually have an RF out from their receiver, so that noise figure/gain analysers like the HP8970 can’t be used to measure sensitivity. The analyser receiver section doesn’t usually go down to audio.

The more usual way to measure receiver sensitivity is to measure MDS (minimum discernible signal  or minimum detectable signal). This can be a very misleading name as when measured, as MDS is defined, the signal is usually most definitely still discernible!

The accepted definition for MDS is when an applied signal power equals the receiver noise floor in whatever bandwidth is selected  (usually 2.7kHz for SSB and 500Hz for CW).  There are two problems here.

First, the signal level must be measured for its true noise power. This requires a true RMS voltmeter. This should preferably be an analogue (moving coil) meter rather than a digital meter, since the nature of noise is such that when measured in these relatively small bandwidths the digital display will be flickering all over the place, unless well damped, and this can lead to other errors. You are measuring a sine wave signal against band limited noise. Never an easy thing to do when the signal is close in level to the noise.

The second problem is that the signal generator must be well screened and preferably a very low minimum output level. It is usual to operate the signal generator a little above its minimum level to improve screening, however, in order to minimise poor screening effects. As the required signal levels will most likely be in the range -130 to -140dBm, the generator is often set at about -90dBm and then an additional 40dB of in-line fixed attenuators plus  an 11dB step attenuator used to reach the low level required. The step attenuator is used to exactly set the signal generator signal into the receiver so that the meter shows a rise in level, over no signal, of exactly 3dB, indicating that the signal power is at the same level as the receiver noise floor. The definition of MDS.

This method can be quite accurate and repeatable, but often is not so due to less than perfect screening in the signal generator, uncertainty in the fixed and step attenuator calibration and coax cable leakage. Double screened coaxial leads are most definitely required.

For these reasons I am a little sceptical about some published MDS results and decided I wanted a better way to measure the sensitivity of the IC9700 receivers on each of its three bands.

Since I have an HP346A (5dB ENR) noise head I thought it would be interesting to use the Y factor method of measurement.

The well specified, low level, noise output from an HP346A noise head can be used instead of a signal generator to measure the complete receiver noise figure from RF input to audio output.  

Just one problem. How to measure the audio power and overcome the noise flicker?

My solution is to use my SDR-IQ, as it can be used down to 500Hz, to effectively replace the receiver section of the HP8970 noise figure meter, and then measure both the noise power from the receiver with no input and with the noise head switched on and producing the excess noise. The ratio of these two noise powers is the Y factor and it is easily converted to indicate the noise figure of the receiver.  The switching on and off of the noise head is done manually as described below.
 It is important to ensure that the SDR-IQ range is set a bit bigher than the bandwidth selected in the receiver. I used 4kHz with a  centre frequency of  2kHz 
The SDR-IQ, in Continuum mode, can be set to give a great deal of smoothing and increase the FFT. Since the noise source is manually switched on and off, the amount of time in each state is effectively unlimited, leading to improved resolution of the Y factor.

The SDR-IQ, when running Spectravue in Continuum mode, shows the noise power as a dB difference (Y factor when converted) between noise on and noise off. The HP8970 driving the noise head can be switched between these two states with the 14.1 and 14.2 Special Function and the actual ENR used in the measuring band can also be shown using the 5.x Special Function.
To be continued..

Sunday, 21 April 2019

At last!

I collected my IC9700 from ML&S on Easter Saturday. Braving the A12/M25 Easter traffic I was able to arrive mid morning and pick up the new rig, a dust cover and an SP38 external speaker (to complement my IC7300).
I spent Easter Sunday sorting out my 2m and 70cm antennas. They had been plumbed into linear amplifier and masthead preamp systems. For my evaluation of the IC9700 I wanted to bypass these items. That turned out to be a little less easy than I expected.
So far I have a 9 element 2m yagi and a 23 element 70cm yagi directly connected to the shack cable frame, with the 2x44 element 23cm still to be re-connected after removing (temporarily) the masthead preamp.

Initial impressions are quite favourable. I’ve only used the rig to listen to 2m and 70cm beacons and some data signals, so far. The thing you notice is how ‘clean and clear’ the signals sound compared to several of my other rigs like the FT847 and the K3 with 2m transverter. This is something I also noted with the IC7300 on 6 and 4m.

Once I’ve sorted out the transmit parameters and installed firmware v1.06 i’ll see if I can find a few local stations to test with.

I will be posting some of my measurements on the IC9700 in due course. 
I will be publishing a user review in PW, if all goes well.


Wednesday, 10 April 2019

No IC9700

Sadly, it looks like I missed the first cut for an IC9700.
What is particularly galling is that I chose to place a deposit with one of the large dealers in the belief they would have the early stocks. The irony is that some smaller dealers seem to have received stock at about the sme time, and because they didn’t have so many pre-sale deposits, some had uncommitted stock.  These wre quickly sold. The result is a longer wait to receive one from my chosen dealer.
I had been asked to submit a user review to one of the UK ham radio magazines. That is now delayed.
Packed the test gear away again!