How to objectively qualify SSB transmission quality?


Geert Jan de Groot
 

(Before creating this posting, I carefully checked the guidelines and thing this falls under "QRP techniques". If not, I hope the moderators will mercifully point me to the error of my ways..)

There is a lot of QRP construction happening in this community, perhaps I can ask:

When qualifying a transmitter doing FM, there are a few tricks:
  • Check the audio passband: 300-3000 Hz
  • Check the clipping behavior with too much modulation signal, check for audio harmonics (another passband check)
  • Use a square wave as modulation, so one can check frequency and phase relations (after all, a square wave is nothing but 1*f + 1/3*3f + 1/5*5f)
Using this, I've made several radios that had "good modulation". In FM.

However, I am struggling with SSB transmitters. In my experiments, like in many QRP rigs, I use a "crystal ladder" (crystals in series, capacitors to ground). They beat the price, size and weight of a KVG XF-9B filter, but the passband is not flat, and I think it is one of the reasons the comments I get on my SSB modulation are not cheerful.
Like with FM sketched above, I would like to objectively qualify so I can measure and optimize instead of asking "how do I sound now?"
But the square wave trick obviously won't work.

How do "other people do this"?

73,

Geert Jan PE1HZG


Curt
 

Geert

Relative to the filter, measure its response somehow. If you have a nanoVNA that will work. But since you have a rig, use audio software like spectogram to see the pass and just with received noise as the input. Of course you want a flat response from 300 to 2000 Hz or so. A cw filter like shape leads to distortion. Examine the filters designed for bitx, k2 and ubitx for examples. The transmit audio chain is also important. 

Curt


Lee
 

Sweeping the filter will give you an idea of the filter characteristics and its quality; Curt definitely has one piece of the puzzle.

But - if you want to test the whole transmit chain (at least until the drive circuit), a two-tone test is an excellent way to visually determine SSB quality. You will need an oscilloscope for this. Here are two URLs describing the test.

https://www.w8ji.com/transmitters.htm

https://vk6ysf.com/ssb_2tone_test.htm

I've little doubt that there is an app that will get you two tone output from your cell phone or computer.

To do two-tone tests of the output, you will need a dummy load and a loop of wire around it to act as a pickup.

If you have access to a good spectrum analyzer, you can obtain even clear data. See the ARRL Lab Test Manual:

https://www.arrl.org/files/file/Technology/tis/info/pdf/Procedure%20Manual%202010%20with%20page%20breaks.pdf

Lee KX4TT

On Thursday, 26 August 2021, 05:50:43 pm GMT-4, Curt via groups.io <wb8yyy@...> wrote:


Geert

Relative to the filter, measure its response somehow. If you have a nanoVNA that will work. But since you have a rig, use audio software like spectogram to see the pass and just with received noise as the input. Of course you want a flat response from 300 to 2000 Hz or so. A cw filter like shape leads to distortion. Examine the filters designed for bitx, k2 and ubitx for examples. The transmit audio chain is also important. 

Curt


Shirley Dulcey KE1L
 

Frequency response of a filter won't tell you the entire story, though the worst offenders are well in the past (and good riddance). One reason that many SSB radios from the 60s and 70s sounded bad was because of time domain distortions in the filters, something that won't show up in any test with static signals like a two tone test. (They will show up if you test the phase response of the filter, another way of looking at the same phenomenon.) Another reason was intermodulation distortion; some types of that DO show in the two tone test, but there is also transient intermodulation distortion (TIM) that is only apparent with changing signals. Those flaws made radios sound bad in ways that people mostly didn't understand at the time and didn't show up in the tests in common use, but when high speed data modes came along they were exposed.

Many crystal lattice filters, the kind used on older radios that used crystals at different frequencies in specific relationships, were rather bad in the time domain. Ladder filters, the kind we see in most current designs, are better. DSP filters can combine excellent skirts and excellent time domain response, which is why a well implemented DSP can sound very good, rivaling the sound of a classic analog direct conversion receiver while also offering better filtering and all the modern operating conveniences. Those same lattice filters also turned out to limit the dynamic range of the receiver, though that wasn't important until we started approaching 100 dB DR in the rest of the circuit.

Not all of the classic radios sounded bad. The mechanical filters that Collins used and the LC filters in the classic Drake 4 line were both good, as was the crystal filter in most Swan radios. (The latter DID suffer from wider skirts than most competitors.) And they all had excellent tube audio stages with low distortion, better than most of the early solid state designs.

On Thu, Aug 26, 2021 at 11:10 PM Lee via groups.io <kx4tt=yahoo.com@groups.io> wrote:
Sweeping the filter will give you an idea of the filter characteristics and its quality; Curt definitely has one piece of the puzzle.

But - if you want to test the whole transmit chain (at least until the drive circuit), a two-tone test is an excellent way to visually determine SSB quality. You will need an oscilloscope for this. Here are two URLs describing the test.

https://www.w8ji.com/transmitters.htm

https://vk6ysf.com/ssb_2tone_test.htm

I've little doubt that there is an app that will get you two tone output from your cell phone or computer.

To do two-tone tests of the output, you will need a dummy load and a loop of wire around it to act as a pickup.

If you have access to a good spectrum analyzer, you can obtain even clear data. See the ARRL Lab Test Manual:

https://www.arrl.org/files/file/Technology/tis/info/pdf/Procedure%20Manual%202010%20with%20page%20breaks.pdf

Lee KX4TT
On Thursday, 26 August 2021, 05:50:43 pm GMT-4, Curt via groups.io <wb8yyy=yahoo.com@groups.io> wrote:


Geert

Relative to the filter, measure its response somehow. If you have a nanoVNA that will work. But since you have a rig, use audio software like spectogram to see the pass and just with received noise as the input. Of course you want a flat response from 300 to 2000 Hz or so. A cw filter like shape leads to distortion. Examine the filters designed for bitx, k2 and ubitx for examples. The transmit audio chain is also important. 

Curt