EMRFD Message Archive 15094

Message Date From Subject
15094 2018-09-26 09:58:02 Ken Chase LO square wave or sine wave
Hi All

I would like opinions/facts regarding using a square wave or sine wave for a LO when using a diode double balanced mixer.

Any disadvantages or advantages in using either?

73

Ken VA3ABN
15095 2018-09-26 13:50:16 kb1gmx Re: LO square wave or sine wave
Sine has a lower harmonic power.  

I found with 5351 source the odd harmonics have a very significant harmonic power 
and that contributes to additional products  at the output of the mixer.

It is said the sine gets "squared" but the diodes and that is true to a point but the 
rise time that results has a lower harmonic power.  Its sine up to diode conduction 
then as the voltage rises the dynamic resistance of the diode conducts harder.  So 
the result is not a hard step function and contained less energy in the harmonics.

Of course if the output is well band passed it makes little difference but even then 
unwanted products are possible.

For some mixers like H-mode and those using digital driven analog switches
the game plays different.   DBMs however being wide band by nature tend to 
process everything provided often to great annoyance.

Allison
15098 2018-09-26 14:26:12 Dana Myers Re: LO square wave or sine wave
15105 2018-09-27 08:42:31 Ken Chase Re: LO square wave or sine wave
Dana and Allison

My understanding is that with square wave, the fast rise and fall of the edges turns on the diodes faster and cleaner. With a sine wave, what happens with a slower rise and fall of the edges? Is there distortion produced?

73

Ken VA3ABN

15106 2018-09-27 09:15:04 Dana Myers Re: LO square wave or sine wave
15107 2018-09-27 09:17:12 Steve Dick Re: LO square wave or sine wave
One additional comment. A perfectly symmetrical square wave’s spectrum consists of a fundamental and odd harmonics.  An asymmetrical square wave can be thought of as a pulse train and its spectrum contains both odd and even harmonics.  So when a square wave is used, it is important that it be symmetrical.  One way to achieve this is to use it to clock a flip flop, presumably with close to identical rise and fall times.  Of course, this divides the input signal by two, often not practical when you are at the frequency limit of the driving source. One way to get better performance out of an asymmetrical square wave is to feed it through a lowpass filter and an attenuator, which provides a better wideband match to the mixer. The lowpass filter will greatly attenuate both the even as well as odd harmonics, making it approach the shape of a sinewave, which is perfectly symmetrical.
-Steve K1RF
 
15108 2018-09-27 10:03:28 Bill Carver Re: LO square wave or sine wave
I looked at the Marki Microwave paper.
My measurements on +7 and +23 diode mixers showed conversion loss is slightly LOWER with square wave LO (mixer driven by paralleled sections of 74AC540), not higher.  Third order intercept was much higher.
Perhaps mixers optimized for 2-4 GHz produce different results than a lower frequency mixer.
W7AAZ

15109 2018-09-27 10:26:48 Chris Trask Re: LO square wave or sine wave
In general, a square wave will give you better IMD perfoamance as there is less distortion when the diodes are in full conduction. Slow rise and fall times will degrade IMD performance.

But there is another factor that gives very little attention, which is the spectral folding created by the square wave LO. Basically, the square wave LO is composed of a large number of odd harmonics, and these also convert input RF signals down to the IF. The wideband noise seen by the RF input, as well as the noise created by the bulk resistance of the diodes, will be converted down to the IF. This results in a 3dB increase in noise figure (NF), even though your conversion loss may be the ideal 3dB.

The ideal mixer would be a noise- and distortion-free analogue multiplier driven by a clean sine wave. You can get those by way of Santa Claus and/or the Easter Bunny.

>
> It is said the sine gets "squared" but the diodes and that is true to a
> point but the rise time that results has a lower harmonic power. Its
> sine up to diode conduction then as the voltage rises the dynamic
> resistance of the diode conducts harder. So the result is not a hard
> step function and contained less energy in the harmonics.
>
>
> The slower transition of the diode switch in this case creates a longer
> period of time in which lower levels of RF energy can modulate the diode.
>
> For example, from https://www.ece.ucsb.edu/~long/ece145b/Mixer1.pdf :
>
> When the diodes are conducting with LO current >> RF current, the mixer
> should behave linearly. At large RF signal powers, the RF voltage modulates
> the diode conduction, so lots of distortion will result in this situation..
> The diodes are also sensitive to RF modulation when they are biased close
> to their threshold current/voltage. For both reasons, we prefer high LO
> drive with a fast transition (high slew rate - a square wave LO is better
> than sine wave) between on and off. The IMD performance is very poor with
> small LO power.
>

Chris Trask
N7ZWY / WDX3HLB
Senior Member IEEE
http://www.home.earthlink.net/~christrask/
15110 2018-09-27 10:36:05 Dana Myers Re: LO square wave or sine wave
15111 2018-09-27 11:50:54 peter_dl8ov Re: LO square wave or sine wave
Four or five years ago I spoke at length with the engineers at Minicircuits regarding this very subject. Their recommendations were as follows.

1) Most double-balanced mixers are designed to see 50 ohms resistive on all three ports, the rest are 75 ohm devices which are tools of the devil.

2) All mixers are designed to be fed with a symmetrical sinewave signal at the specified level, this MUST be free of DC bias.

3) All mixers are designed to be fed with a signal at the RF port which is a maximum of 6dB below the level of the LO, using -8dB is a safer limit and will give lower IMD in transmitter circuits. This signal MUST also be free of DC bias.

4) Pay attention to the maximum current requirement of the IF port which should always be used as an output.

Since then I have followed these four rules and never had a problem.

Peter DL8OV
15112 2018-09-27 12:22:03 kb1gmx Re: LO square wave or sine wave
First I never said much on distortion at all.  Though that is always an issue.  The 
focus is  all of the unwanted additional products and their amplitudes.

Testing has verified the distortion is lower for the fundamental products.

The yabut is in part two.  Due to the presence of the strong odd harmonics you 
get additional products that maybe troublesome.  

Chris Trask:
"But there is another factor that gives very little attention, which is the spectral folding created by the square wave LO." 
That is the evil bunny in the room.  We moved from linear mixing theory to sampling theory.
Not all of the products are where we want them or at the low levels we wish them to be.
and they are not where they are expected.

Best case I know of is the uBitx spurs.  With only 33mhz low pass at he mixer output
and a 45mhz IF as the sources and a LO in the IF+ range we get spurs under 20mhz.  
The best example is a spur at ~17mhz for a LO of 73 mhz...  The spur is about  30db  
down from the desired 28mhz output.  Where does that spurs come from?  
Seems there are several sources including 2IF, maybe 3IF and other sources.  And 
the 3rd harmonic of the LO.  The LO was the very square output of the 5351. 

That is a HF example, at UHF and up frequencies the harmonics and the products have 
greater separation and likely different loss structures as a result.   The ratios of frequencies  
of wide band at 2ghz is not the same as wide band at HF meaning 3-30mhz.  Its hard to 
compare maybe a octave at UHF to 4.5 octaves at HF.

The conclusion is wishing for clean mixer outputs that needn't be filtered is of myth and 
very little substance.  Adding the waveform debate makes it no clearer as in both cases
we get what we what and so much more we didn't want.

Where does this lead...  Right now amateur HF projects are enabled by sources like 
the si5351 and SI570 or DDS.  They offer stability and digital control making for projects 
that relied on analog VFOs or complex PLLs  archaic or at least difficult.  The cost for 
the stability and digital tuning using NCO is that we get a square wave source and we 
get a wider range of  products from mixers that must be managed.   If we use a "nice" 
sine DDS we have SFDR to deal with and likely more.  The first step is to recognize 
the signals and products exist.   Then decide on how to mitigate the problems caused.

To me  DBM is a handy device but also a pre-opened can of worms.  As such it deserves
a far more studied evaluation for the application rather than riding off in all directions at 
once.   


Allison



15113 2018-09-29 17:06:51 Ken Chase Re: LO square wave or sine wave
Haven't been around lately to respond to the discussion.

The response is absolutely appreciated.

The question came up due a discussion in our homebrew group.

73

Ken VA3ABN

15114 2018-09-29 18:29:41 DuWayne Schmidlko... Re: LO square wave or sine wave
More and more of the parts we are now using in our designs were designed
for use in all digital circuitry, or UHF/Microwave frequencies. The
design for these applications has their own set of 'dragons to slay',
which can be very different from what we do in the HF range.
One thing that we can check for early on in the design stage is look at
the mixer products for the IF and LO frequencies we choose. I found a
spread-sheet on Microwaves101.com a couple years ago, and have found it
very useful. It appears to have been written for the Microwave range,
but works well for HF. Just put your IF frequency in the LO field and
your LO range in the RF fields and select a step value that gives good
resolution. Then select your display range and any highlighted range
you want. I have attached a version that has the IF and LO range of the
uBITX, as you can see there are many spurs that pop up under 30MHz.,as
you use the slider to sweep across the LO frequency range.
DuWayne KV4QB


>>That is the evil bunny in the room. We moved from linear mixing
>>theory to sampling theory.
>>Not all of the products are where we want them or at the low levels
we >>wish them to be.
>>and they are not where they are expected.

>>Best case I know of is the uBitx spurs. With only 33mhz low pass at
>>The mixer output
>>and a 45mhz IF as the sources and a LO in the IF+ range we get spurs
>>under 20mhz.
>>The best example is a spur at ~17mhz for a LO of 73 mhz... The spur
>>is about 30db down from the desired 28mhz output. Where does that
>>spurs come from?
>>Seems there are several sources including 2IF, maybe 3IF and other
>>sources. Andthe 3rd harmonic of the LO. The LO was the very square
>>output of the 5351.




[Non-text portions of this message have been removed]
15115 2018-09-30 02:49:33 Alberto I2PHD Re: LO square wave or sine wave
On 2018-09-27 19:26, Chris Trask christrask@earthlink.net [emrfd] wrote:

The ideal mixer would be a noise- and distortion-free analogue multiplier driven by a clean sine wave.   You can get those by way of Santa Claus and/or the Easter Bunny.

Or do it in the digital domain, like in the SDRs... a mixer in the digital domain is just a multiplication, usually between two complex numbers, nothing more, nothing less... and there are NO spurious responses whatsoever....

--
73 Alberto I2PHD
Credo Ut Intelligam


15116 2018-09-30 16:08:39 Ken Chase Re: LO square wave or sine wave
Hi Dana

What is 'At large RF signal powers, the RF voltage modulates the diode conduction' as stated below?

73 Ken

15117 2018-10-05 14:06:55 kb1gmx Re: LO square wave or sine wave
Digital domain has distortion too..

Truncation, Rounding, overflow are just a few of the ills that can best digital.

Allison

15118 2018-10-06 13:10:48 Alberto I2PHD Re: LO square wave or sine wave
On 2018-10-05 23:06, kb1gmx@arrl.net [emrfd] wrote:

Digital domain has distortion too..

Truncation, Rounding, overflow are just a few of the ills that can best digital.


Yes. But those can be predicted and completely avoided using some care.
You can't avoid totally the spurious response of an analog mixer, no matter how much care
you use in designing it...


--
73 Alberto I2PHD
Credo Ut Intelligam