开云体育

Updated multi band screen layout.


With bands with span less then 1/3 of the total span the info will start to overlap, not sure how to solve this without too much code.
Increasing the space below the chart to stack the info conflicts with the waterfall.
Feedback?
--

Perhaps only the start frequency and the division (/) is enough información for a band

On Mon, Feb 27, 2023 at 06:02 PM, Erik Kaashoek wrote:

With bands with span less then 1/3 of the total span the info will start to overlap

In this case it is possible to replace the overlapped full band info strings with abridged ones (or may be simply?just use?band numbers (i.e. B#1, B#2, ...).?
And leave the detailed info available at the sidebar menu.

With 5 bands this is just possible


I see there is still an error with the band split (blue line), it walks away.
The grid is the same for all bands with the current space allocation (based on span per band)
With less than 1/5 of the pixels per band some info has to be removed.

Just to show the effect of multi band on the dynamic range (20dB extra)? see this, similar measurement time


and with 10kHz RBW (same as multi band) you get 16 seconds scan time


--

I think displaying the bands one at a time, with a selectable delay between each scan - say 0 to one second - would be best. You'd still be able to simultaneously monitor up to 8 bands without touching anything, and no data would need to be lost. There could also be an 'infinite' delay selection where each band would remain until the screen was touched, which would advance to the next band.

On Mon, Feb 27, 2023 at 07:02 AM, Erik Kaashoek wrote:

Updated multi band screen layout.


With bands with span less then 1/3 of the total span the info will start to overlap, not sure how to solve this without too much code.
Increasing the space below the chart to stack the info conflicts with the waterfall.
Feedback?

Erik,
? On older spectrum analyzers without markers the "freq/" annotation is a necessity.? With the Ultra's ability to place up to 8 markers, the "freq/" annotation could be removed to save space as the markers are more accurate than estimating frequency using the division scaling. Maybe an additional mode that cycles full screen through the different bands with "touch to continue" could be added as suggested by those who prefer that display format.

Herb

I think the markers are the more important features when using this mode.

Would it then be possible to add some additional measurement functions such as harmonic levels relative to carrier and also 2nd & 3rd order IMD.

Ideally with the option to specify an off-set to the level of the two test signal carrier levels, to compensate for any notch filters that may be used ahead of the analyser to prevent overload during such tests.

The ability to see small concurrent slices of narrow frequency bands, makes it possible to use small bandwidths in each slice, which is not possible with the wide sweep that would normally be required to encompass all of the carriers and their complex harmonics and IMD components.

IP2 & IP3 could then be calculated and read directly.

Regards,

Martin

Maximum is 8 markers

Hi Erik,

Normally you would manually just decide upon which combination of signals you wished to use, based on the usual formula, but I guess you could set them up automatically (or choose which sets to use) as long as you know the frequencies of the test signals in use.

So for an IMD 2 & 3 test you would probably only need four or six markers in total (including the two test signals) but would maybe need to be able to select which sets of IMD products you wish to use for the measurement. The usual constraints are the operating frequency range of the item under test, and which set of 2nd and 3rd IMD products that are produced are the highest in level. You would normally do a quick check of the various sets and then just use the worst case values.?

Let me think about this and I'll try and give an example. I think there may be some manufacturers guidance notes (possibly Keysight) somewhere on-line too.

Regards,

Martin

Hi Erik,

Although Keysight have other application notes (a lot are password protected), the one in the following link is probably more applicable to the sorts of testing TinySA users would be likely to try.

For practical tests with typical hobby grade equipment, you would typically measure 2nd order products at f2 – f1 and f1 + f2 and the 3rd order products at 2f1 + f2 and 2f2 + f1

This is because the other set of 2nd order at f2 - f1 is often too low in frequency, and the in-band 3rd order products at 2f1 – f2 and 2f2 – f1 are too close to the two test carriers, and it is difficult to notch out he tones without also notching out the much weaker 3rd order products.

Some more notes.

The big problem with these sorts of tests is determining if the IMD you are measuring is being generated in the Device Under Test (DUT) or the test instrument. A quick way to validate this is to attenuate the signal going into the Spectrum Analyser from the DUT. I normally use a 3dB attenuator for this purpose.

If all of the signal levels decrease by the value of attenuation, the IMD is being produced in the DUT. If the IMD products decrease? (or increase) by a greater amount than the value of attenuation, then the Spectrum Analyser is contributing to the measured value. This is usually because the two test signals are overloading the Spectrum Analyser, and a much better notch filter is required to further reduce their signal levels.

Another quick rule of thumb is that the IOP3 value of an amplifier is usually about +15dB above the 1dB compression point (or about the saturated output power) This can also be used as a confidence test that the values you have obtained are in the right 'ball park', and you haven't done something stupid in your setup.



I hope this helps to provide further insight.

Regards,

Martin

Martin

On Tue, Feb 28, 2023 at 02:45 PM, Erik Kaashoek wrote:

Do you suggest a extra measurement focussing on broadband mixer measurement?

Yes, and also amplifiers, in fact any signal path that is likely to introduce non-linear distortion.

My main interest is in measuring the IMD performance of amplifiers used in active antennas. Some of these have quite impressive performance figures. For example, one I measured (at the limit of my existing test rig), produced an OIP2 = +59.1dBm and OIP3? = +43.8dBm

Two tone transmitter measurements are generally less taxing of the measurement chain, as the IMD products tend to be quite high level and easily observed.

Measuring devices with much better performance becomes more demanding, and I don't think you could achieve it with carrier spacings in the audio frequency range. The close in phase noise would probably be too high, and it would be difficult to notch out just the test carriers in the RF domain when using such close frequency spacing.

However others may disagree, and I'm always keen to find out new ways to perform such tests.

Regards,

Martin

Phase noise will indeed limit the range you can measure with close in IMD products.
I tried to manually setup the bands and the markers for IP2 and IP3 using the far away IMD products and it is indeed a lot of thinking and clicking and I did not even do the calculation.
But what product to chose for IP2 and what for IP3?
In the current OIP3 setup I display both results so you can check if it is balanced but that may lead to too many bands.
Would only using the X+Y for IP2 (so not the X-Y) and 2X+Y for IP3 (so not X+2Y) be sufficient?
--

Hi Erik,

Yes I think X+Y and either 2X+Y or X+2Y would be ideal.

Maybe allow selection of one or the other of the latter pair, or separate out IP 2 & IP3 measurements ?

Personally I like to see both IP2 & IP3 (either pair) at the same time, as I often use these whist adjusting amplifier bias current in order to find the best compromise between the two.

Regards,

Martin

I forgot to add.

The ability to switch in an attenuator quickly as part of the test, to check for any SA contribution (hopefully none) would ?be very useful and save a lot of menu hopping.

My typical IMD workflow.

Test configuration

Two tone test source feeding stepped attenuator, feeding DUT, feeding fixed value power attenuator, feeding switched notch filter, feeding SA.

Set stepped attenuator to give output of DUT at approx 10 to 20dB below saturated output level

Measure output level with SA via fixed attenuator, with notch filter off.

Setup SA to track test tone peak levels, and add level offset for value of external fixed power attenuator.

Switch in notch filter and note level reduction of test tones, compensate for notch attenuation as additional offset in calculations and displayed values.

As the SA should track the test tones in frequency and level, any drift in test sources should be continuously compensated for.

The test tone frequencies and compensated levels can then be used to accurately and automatically set and track the IMD component measurement markers in a narrow bandwidth, in order to maximise the dynamic range.

The IMD traces will be at a lower level than the test tone level, even though these are notched for measurement purposes, so the SA noise floor with the DUT running may have to be used to determine the optimum reference level for these measurements.

The marker information can then be used to directly calculate the IP2 & IP3 values and display them.

As a confidence check an extra 3dB can be switched in to the SA input and the calculated values should not vary by more than approx +/-1dB from the previous measurement without the extra 3dB.

The stepped attenuator on the DUT input can also be varied to determine the effect on IMD values at different power levels up to the DUT saturation point.

I hope this makes sense.

Regards,

Martin

Should have added as an extra check.

Look at both sets of IP3 frequencies and use pair with worst case values.

I like having multiple bands when scanning upload frequencies on some LTE channels. It would be great to be able to set multiple frequency bands from the Console with the sweep command.