Topics

Harmonic performance - SSB vs CW

Warren Allgyer <allgyer@...>
 

I am noting a significant difference in the harmonic suppression in SSB mode versus CW mode. 

The two signals are generated in a very different way. SSB is generated at 45 MHz and mixed down to the final transmit frequency by combining with the first oscillator in the balanced first mixer to make the SSB transmission. In CW the 45 MHz path is disabled and the first oscillator is simply moved to the transmit frequency. The first mixer is unbalanced with a DC bias to allow the first oscillator signal to leak through.

It appears the act of unbalancing this mixer produces harmonics, particularly high order odd harmonics that are clearly in excess of the -43 dBc requirement.

The screenshot shows  the harmonics of an 80 meter transmission at 10 watts. The purple trace is the CW and the blue trace is SSB. Signal source for SSB is a approximately -40 dBV 1 KHz tone with the level adjusted to match the CW carrier level. The SSB signal harmonic suppression is just adequate (green display line) however the CW signal is clearly non-compliant.

I am not sure what to do about it yet but wanted to raise the point. I have done a search on this board since I am new and I do not find this has been reported before..... perhaps I missed it?

WA8TOD

Jerry Gaffke
 

That's an interesting finding, and seems quite likely correct.
Thanks for looking into that.

One possible solution would be to unbalance either the modulator (as done on the Bitx40 by Allard's mods) 
or the second mixer at D3,D4.  In either case, those harmonics should be stripped out by the 45mhz filter
(also the 12mhz filter, in the case of unbalancing the modulator).  Would require minor hacks to the
firmware to set up the si5351 correctly for CW transmissions after this mod.

Jerry, KE7ER


On Thu, Aug 2, 2018 at 09:26 AM, Warren Allgyer wrote:

I am noting a significant difference in the harmonic suppression in SSB mode versus CW mode. 

The two signals are generated in a very different way. SSB is generated at 45 MHz and mixed down to the final transmit frequency by combining with the first oscillator in the balanced first mixer to make the SSB transmission. In CW the 45 MHz path is disabled and the first oscillator is simply moved to the transmit frequency. The first mixer is unbalanced with a DC bias to allow the first oscillator signal to leak through.

It appears the act of unbalancing this mixer produces harmonics, particularly high order odd harmonics that are clearly in excess of the -43 dBc requirement.

The screenshot shows  the harmonics of an 80 meter transmission at 10 watts. The purple trace is the CW and the blue trace is SSB. Signal source for SSB is a approximately -40 dBV 1 KHz tone with the level adjusted to match the CW carrier level. The SSB signal harmonic suppression is just adequate (green display line) however the CW signal is clearly non-compliant.

I am not sure what to do about it yet but wanted to raise the point. I have done a search on this board since I am new and I do not find this has been reported before..... perhaps I missed it?

WA8TOD

ajparent1/KB1GMX
 

Warren,

During my sessions testing the amp it was noticed but the 20Mhz and up spur was higher 
on the priority list.

Another thing I ran into with a friend was defective low pass filters.  The toroids had different 
numbers of turns from other units.  

First try lowering the drive (RV1) and look again.  If it drops its the amp flat topping (overdrive)
(can be any stage in the amp up to including the finals.).  That can be voltage starving (less than 
12V at the board) or other problem unnoticed.

It is possible that the first amp Q90 is being over driven as well and reducing RV1 will not help then.

Generally the CW drive was found to be higher than clean levels of SSB drive so it could 
be amp overdrive.  Lowering the DC bias to the mixer during CW TX can help that.  The
Mixer (DBM) is being used as a switch but they can also be used as a variable attenuator
as well.  R114 is the point of adjustment.  Non linear action by the diodes is not an issue
as the driving waveform is square to start with which is why the odd harmonics are stronger
on the display.  Add to that the push pull amps (driver and final) are balanced so the second
and forth are better suppressed but 3 and fifth tend to be the push pull amps weak spot.

Also sweep the 3-5mhz filter to insure its working as expected.  Easy to do with power off
and tracking generator as its then isolated.  Also make sure the correct filter is being
selected, TX-A/B/C should all be selected (relay active).

If the TX amp is at or near compression try turning up the DC power a Volt or so (caution TA2822!).
If the TA2822 is on separate regulated power its safe to go far higher than 13V.

Test first as there are many possible solutions depending on the problem.

Allison

Warren Allgyer <allgyer@...>
 

Allison
I will try lowering the drive to see if it corrects the problem but I am skeptical. I first noted the problem when probing TP1 with the PA unpowered. The discrepancy is already visible at that point. It is pretty clear the harmonics are being generated at the unbalanced mixer.

WA8TOD

Warren Allgyer <allgyer@...>
 

The signal at TP1, the output of the first mixer is a pure square wave as it is derived directly from the Si5351 in CW mode. The signal in SSB mode is much closer to a sine wave as it is a mix of the Si5351 square wave and the 45 MHz drive that has been filtered by the roofing filter of Y1 and associated components. 

The only filter that the CW signal encounters before the output filters is the 30 MHz LPF between TP1 and TP2. It has a 2 dB point of about 34 MHz in my case so any harmonics below that are unattenuated until after the final.

The obvious way to fix this is to do the CW generation prior to the roofing filter. What is not obvious to me is the best way to do that. As it stands it is pretty messy.

WA8TOD

ajparent1/KB1GMX
 

Warren,
I it was soemthing easy to try but I'd call it the same as you.  Lots of harmonics
in square waves and little filtering.   

At frequencies below 20MHz the filters at the amp output are the only filter.
Not my favored way, either. 

If there were band pass filters at the amp input this would not be an issue
nor would spurs be an issue.

One way and likely there are many:
An improvement is to move the keying (CW_TX) to the product detector/ TX modulator
and move the BFO to the edge of the pass band and key that.    The easy way is
remove R105  and run a wire from the junction of C1/R104 under the board to
a replacement R105 at the modulator junction C222-1 and the transformer T7.   
Then both the 12 and 45mhz filters can clean it up.  That would be the wire changes.

Doing that requires a code change.  Actually any change that comes immediately
to mind requires that.    The code change would remove the mix of VFO switching
to only the BFO and the signal CW-key would drive the Balanced modulator.  That
works well as It has been done on many other radios.  Code is easier than solder.

Allison

Warren Allgyer <allgyer@...>
 

No one is going to want to hear this but the design of the PA output filter is probably fatally flawed and it is likely due to board layout issues.

I have been puzzled how the high order harmonics of 80 meters, for example, could possibly survive a 7th order filter which should have well over 50 dB of attenuation at the 5th harmonic. So today I set out to test the results with a 9th order filter, figuring more is better. The harmonics remained. But in troubleshooting that installation the source of the issue was revealed.

In the attached screenshot the purple trace is the 80 meter CW output with the harmonics clearly far in excess of the green -43 dBc legal threshold. The blue trace is the output with L20 removed which disables the 80 meter filter and blocks RF output from the finals from reaching the output connector. (Don't worry..... the tests were carried out with the drive reduced so the output was only one half watt. And I have a bag of spare IRF510s on hand......). The harmonics remain! How can that be?

Pic #2 shows the path 80 meter RF follows through all three switching relays in order to reach the filter. It is a very novel and inventive concept to use three relays to switch both ends of 4 filters. Most manufacturers would require 8 separate relays to do this job. And for a very good reason.

In order for a filter to provide the required 50+ dB of attenuation, the filter itself must have 60 dB or more of input-to-output isolation. If it does not then the unwanted products simply bypass the filter and go directly to the output. That appears to be exactly what is happening in the filter complex of this unit.

There is no easy fix for this. No amount of on board filtering and tweaking is going to improve harmonic suppression until the path around the filters is blocked. I suspect that path is via the relays themselves but board layout often is critical as well in such cases. My guess is the only solution would be either extensive external filtering or a carefully laid out daughter card to replace the existing filters and relays.

WA8TOD

Warren Allgyer <allgyer@...>
 

BTW..... in re-reading this, unfortunately after it was posted, I note that my threshold line is set at -53 dBc rather than -43. In your imagination move the green line up one division. It does not change the outcome or the analysis.

WA8TOD

Arv Evans
 

Warren WA8TOD

That is interesting, and a bit disturbing as well.  What might be the effect of adding
a series LC trap to ground at the antenna end of the filter, or maybe at the RF PA
end of the filters? 

In the now ancient BITX20A Dan Tayloe added a cap in parallel with one of the inductors
in the LPF to form a notch at the 3rd harmonic frequency.  This worked without degrading
performance of the LPF.  However, this current problem seems more complex as it
involves leakage around the filter instead of a simple filter performance issue.

Are you continuing to experiment to find a workable solution that might be applied to the
uBITX?

Arv  K7HKL
_._


On Sat, Aug 4, 2018 at 6:43 AM Warren Allgyer <allgyer@...> wrote:

No one is going to want to hear this but the design of the PA output filter is probably fatally flawed and it is likely due to board layout issues.

I have been puzzled how the high order harmonics of 80 meters, for example, could possibly survive a 7th order filter which should have well over 50 dB of attenuation at the 5th harmonic. So today I set out to test the results with a 9th order filter, figuring more is better. The harmonics remained. But in troubleshooting that installation the source of the issue was revealed.

In the attached screenshot the purple trace is the 80 meter CW output with the harmonics clearly far in excess of the green -43 dBc legal threshold. The blue trace is the output with L20 removed which disables the 80 meter filter and blocks RF output from the finals from reaching the output connector. (Don't worry..... the tests were carried out with the drive reduced so the output was only one half watt. And I have a bag of spare IRF510s on hand......). The harmonics remain! How can that be?

Pic #2 shows the path 80 meter RF follows through all three switching relays in order to reach the filter. It is a very novel and inventive concept to use three relays to switch both ends of 4 filters. Most manufacturers would require 8 separate relays to do this job. And for a very good reason.

In order for a filter to provide the required 50+ dB of attenuation, the filter itself must have 60 dB or more of input-to-output isolation. If it does not then the unwanted products simply bypass the filter and go directly to the output. That appears to be exactly what is happening in the filter complex of this unit.

There is no easy fix for this. No amount of on board filtering and tweaking is going to improve harmonic suppression until the path around the filters is blocked. I suspect that path is via the relays themselves but board layout often is critical as well in such cases. My guess is the only solution would be either extensive external filtering or a carefully laid out daughter card to replace the existing filters and relays.

WA8TOD

Warren Allgyer <allgyer@...>
 

I am not sure Arv. The real solution here is like a "fork lift" upgrade, replacing everything after T11 in the transmit chain. If I were doing a redesign I would use the same four filters, add five more relays, and decode the three lines to get four lines each driving the pair of relays bracketing the desired filter.
 
I would also use latching relays and move the TR switching to the PA side of the filter bank to allow the filters to be utilized on receive as well.
 
All relatively easy to do...... just takes time and money. And motivation.......

WA8TOD

ajparent1/KB1GMX
 

Warren,

In my testing I also found the isolation across one relay was barely 45db in one configuration and 50 in another.
Using one relay to switch both ends of a filter will compromise the filter attenuation characteristics.

Most other radios commercial put a spdt relay at each end to avoid common mode and coupling.

You are seeing what I saw....

Allison

ajparent1/KB1GMX
 

Arv,

For bitx20 (or 40) a monoband radio traps and Caur filters are a good solution.

For a multi band radio its a spot solution that is likely to case troubles on other bands.

The fundamental issue was inadequate to non existent filtering for the TX path
before the TX.

Allison

ajparent1/KB1GMX
 

Warren,

Latching relays have a higher cost, I like the idea though.  However there are lower power
relays that would do the job.  Since the radio eats power on TX anyway another 200ma is tolerable.

I would want band pass before the TX chain.  The logic is you get cleaner out if you put cleaner in
and it wold fix many of the spur issues.  However that's a lot of relays or diode switching for 9 bands.

Allison

Warren Allgyer <allgyer@...>
 

Allison

Have you shared your past findings here? I did a search and found nothing but I could have missed it. It would seem to me that if you knew the product was illegal for use on the air you would have a responsibility to share that with the group.

WA8TOD

Jerry Gaffke
 

This is primarily an SSB transceiver, and harmonic suppression seems adequate in
that mode on the stock rig.  Best and easiest fix is probably to either not use CW mode,
or do a minor hack to move the CW mixer injection an earlier mixer so the 45mhz
filter takes out the harmonics as previously stated.
 
The uBitx relay arrangement is not that uncommon, though would be better
to double the relay count and use separate relays for input and output.
Board layout is important.  This guy definitely knows what he is doing:
    https://www.w6pql.com/a_1.5kw_lpf_for_160-6m.htm#v1filter

Another solution might be to get an outboard set of filters such as the WA2EBY fllter board:
    http://www.py2nfe.com.br/Arquivos-pdf/linear-kossor1.pdf

LPF's such as seen on the uBitx are usually adequate for use with class C amps
on CW transmitters (which are dealing with high power square waves).  
However the uBitx gets by with as few LPF's as possible by covering multiple
bands with each filter.  The corner frequency of the LPF may be a bit high to 
fully suppress the harmonics of the lowest band of that set.

Jerry, KE7ER

Warren Allgyer <allgyer@...>
 

My unit puts out illegal harmonics on SSB on all bands below 20 meters.

WA8TOD

ajparent1/KB1GMX
 

Warren,

Not for harmonics.  The spur issue took precedence.
They were shared in other message threads here.

The spur issue was discovered during power amp improvements.
IT was announced here in the forum and also on the ubitx.net site.

IT was why I suggests keying the modulator as then CW cleans up
leaving only the above 20mhz spurs.  

Basically for any band above 20mhz there are spurs below the cutoff of the 10M filter.
Specific example is at 28mhz the spur is 17mhz (+- offsets).  At 21mhz the spurs
are 24mhz (they move up as you move down).  At 14mhz the 20M filter catches any
spurs but not the 30mhz filter.

That is why I said band pass filter before the transmitter would clean up a lot of issues.
Those being harmonics, spurs and some of the oddball mix products visible.

From my point of view my radio is not going to see an antenna other than as RX.
I am looking at how i can make it suitable for use as transceiver. 

Its up to others on this to evaluate the data made available and use or not use.

Allison

Warren Allgyer <allgyer@...>
 

Correction -- My unit is not illegal on all bands. Spectrum Analyzer output for each band can be found in the Photos section at https://groups.io/g/BITX20/album?id=65861. Here is the summary:

80 meters - SSB: OK   CW: No
60 meters - SSB: No   CW: No
40 meters - SSB: No   CW: No
30 meters - SSB: No   CW: No
20 meters - SSB: OK   CW: No
17 meters - SSB: OK   CW: OK
15 meters - SSB: OK   CW: OK
12 meters - SSB: OK   CW: OK
10 meters - SSB: OK   CW: OK

These measurements are for a sample of one. All measurements were made at 2 watts for consistency and to avoid any overload or clipping issues. This does not mean YOUR unit is illegal! I have no way of knowing how consistent this problem is...... I suspect it is the same with all units but it is your responsibility to operate legally and there is good reason to believe if you are using one of the "No" modes above, you might be missing that bar.

WA8TOD

ajparent1/KB1GMX
 

Jerry,

if you move the keying to the either of the earlier mixers CW becomes a victim of
the Spur issue.

As to the corner frequency the offending harmonics are 2nd, for 3.5 its 7mhz above the 80m filter
and the third which is particularly strong at 10.5mhz well above the 80m filter.  I did sweep mine
and it was not at all bad for a 7 element filter.  However relay blow by does not help.

FYI the EBY filter was revised to improve it but its still a low pass filter.  However if it were between
the first mixer and the power amp it would help greatly up to 20mhz.

When I look back at Warren's SA out put it shows the harmonics I was seeing while testing for improved
flatness whenever I over drove the amp.  Which is why I asked if he tried lowering the drive to see
if the harmonics drop faster than the power.   The unmodified amp has a strong tenancy to run out
of power ( current or voltage limiting) in the predrivier and driver stages even at the low end of the
range.  That sweep goes to 30mhz so there are a lot of higher order harmonics and spur products
visible.   Even when your seeing 10W at 80M  its hitting the wall at 10.5mhz (third harmonic)
as well as the higher harmonics.

A reminder for those working with linear amps the power out is the total of all the amplified signals
including those stopped by filters.  There are a lot of milliwatts in those harmonics even if they never
get past filters they still help heat the drivers and finals.


Allison

iz oos
 

I always use external low pass filters I have done some years ago. They are six according to the frequencies and can be put in series to increase harmonics attenuation. No relays, just plug and play. I know it is a rudimentary solution but I think it solves many issues in homemade stuff.


Il 04/ago/2018 18:51, "Warren Allgyer" <allgyer@...> ha scritto:

Correction -- My unit is not illegal on all bands. Spectrum Analyzer output for each band can be found in the Photos section at https://groups.io/g/BITX20/album?id=65861. Here is the summary:

80 meters - SSB: OK   CW: No
60 meters - SSB: No   CW: No
40 meters - SSB: No   CW: No
30 meters - SSB: No   CW: No
20 meters - SSB: OK   CW: No
17 meters - SSB: OK   CW: OK
15 meters - SSB: OK   CW: OK
12 meters - SSB: OK   CW: OK
10 meters - SSB: OK   CW: OK

These measurements are for a sample of one. All measurements were made at 2 watts for consistency and to avoid any overload or clipping issues. This does not mean YOUR unit is illegal! I have no way of knowing how consistent this problem is...... I suspect it is the same with all units but it is your responsibility to operate legally and there is good reason to believe if you are using one of the "No" modes above, you might be missing that bar.

WA8TOD