EMRFD Message Archive 14828

Message Date From Subject
14828 2018-05-21 19:25:10 richj_focus mixer if port termination
I see for DC receivers the diode mixer is terminated with 50 ohms via a diplexer that is also terminated into a 50 ohm common base transistor amp.

So why not design the diplexer to terminate into higher impedance so more kinds of amplifier can be used?  That is other low noise amp configureations that have higher input impedance.

What do you think?


14829 2018-05-21 19:33:14 Ashhar Farhan Re: mixer if port termination
The diode mixer before the diplexer needs to see 50 ohms impedance for best performance. That is the design. You could alternatively build a diode mixer using some transformer trickery (read some cool stuff by chris trask about building these) so that the mixer charactersitic impedance is, let's say 200 ohms. You could, then, use a diplexer and an audio amp for 200 ohms.
The diplexer impedance has to be the same impedance as diode mixer. It is not only useless but even performance degrader otherwise.
- f

14830 2018-05-21 21:16:01 Tayloe, Dan (Noki... Re: mixer if port termination

For a DC receiver, a post mixer diplexer seems unnecessarily complicated when other simpler (higher performance?) options exist.


A different (simpler) approach is to terminate the mixer in a capacitor (single ended) or two (double ended) such that the system series R and the detector output C-to-ground form a low pass filter.  Think in terms of a 0.47 uF Mylar or ceramic type X7R detection capacitor.  In this configuration the impedance looks like an open at DC and a low impedance (short) way off frequency.  In this configuration there is no “sum” frequency at all, all the power has been converted into a difference frequency.  The detection capacitor turns this into an integrating type detector (Tayloe detector) that has no need to terminate the sum frequency because there is none.


It does not hurt performance.  In fact the performance tends to improve.  Sensitivity increases (due to lower conversion loss) and additional out-of-band signal rejection is achieved via 1 pole of free R/C low pass filtering.  One pole is not great, but it can be helpful to knock down large signals 10 KHz+ away depending on the R/C frequency cut off selected.  Since the RF input is often not 50 ohms, it is good to under size the detection caps so that the bandwidth is not too tight if the input impedance happens to be quite a bit higher.


Doubly terminated diode mixers work well with the differential input of an op amp, even if one side looks like an open (+ input) and the other side looks like a short (- input).  The inverting input uses the system R as the input gain setting resistor.


If you use an op-amp for the first pre-amplifier, op-amps tend to oscillate given an large input “C”.  A capacitor across the feedback resistor can kill this while providing gain roll off at unwanted higher frequencies.  Something like a 2200 pF across a 2.2K feedback resistor (33 dB gain given a 50 ohm input). 


Depending on how the op-amp is biased, the outputs of the detection capacitors may need to be AC coupled into the inputs of the op-amp.  I would think this would be needed any time a diode mixer is connected to an op-amp unless you happen to have a split voltage supply (like +/- 12v).


For a 50 ohm system (single ended) or 100 ohm (double ended differential), the detection “C” can get kind of large, on the order of 1 uF.  However, even a commonly available 0.1 uF or 0.22 uF detection capacitor buys some protection.


  • Dan, N7VE


14831 2018-05-21 21:26:11 Ashhar Farhan Re: mixer if port termination

The question, I think, was directed towards diode mixers. They are far more sensitive to reflection that passive FET switching mixers. What you are describing too is a diplexer, though very functional while being trivially simple to implement. The sum frequency will still need to see proper termination, won't it? I am not sure I have fullly understood where that energy goes.

- f

14832 2018-05-22 06:52:32 Tayloe, Dan (Noki... Re: mixer if port termination
I have not tried it, but mathematically a diode mixer is similar to a MOSFET type switching mixer in that both types turn on and off and both produce sums and differences.  Adding the large C as a detector cap changes the mode of operation.  The detector cap is taking an RF pulse from the mixer turning on for half a cycle and produces an average detection voltage for that half cycle.  The detector cap is being charged via the series resistance of the system impedance, so a frequency that is far away suffers an R/C roll off.

If the mixer is not driven by a square wave, but rather a sine wave, the mixer will turn on more slowly, so the capacitor is averaging over less than the full half wave.  That should drive the peak detected voltage a bit higher.

All of the energy in each RF pulse is used to drive the charge (i.e. the detected voltage) on the detection cap.  There is no wasted "sum" product in this mode of operation.  The detector tends to "short out" incoming RF signals that are way off frequency, reducing the unwanted junk out of the detector and perhaps helps the mixer to be less sensitive to the impact of other signals on its operation.

This effect out to be pretty easy to see using a signal generator with enough output, mixer, and a scope.  It is neat to watch what happens both at the operating frequency and as the input is tuned away from the operating frequency.  You can definitely see the R/C roll off.  You also ought to see a higher output for this mode vs. using a more complex duplexer arrangement.

- Dan, N7VE

14833 2018-05-22 07:21:51 richj_focus Re: mixer if port termination
Well i was thinking that the diplexer would have an input  impedance of 50 ohms to satisfy the mixer, and them the diplexer output impedance to be higher to match a non-50 ohm audio amp.

14834 2018-05-22 07:29:20 Ashhar Farhan Re: mixer if port termination

The diode is a non-linear device. Unlike the FET where the channel is largely a linearly resistive path. I say 'largely' only in comparision with the diode. The diode is switched on due to the local oscillator current flowing through it along with the signal current. This complicates the switching when the signal current starts growing large enough to compete with the oscillator current, leading to non-linearity. Passive FET mixers like those in mixers that are popularly known by your name, have no such problem. Due to the bias requirements, diode mixers report much higher losses than the passive mixers. I have never been able to reduce the losses below 7 db.
I am planning to get back to playing with passive fet mixers this week, i will experiment with your suggestion and report the results on the forum. 

- f

14836 2018-05-22 11:28:52 Tayloe, Dan (Noki... Re: mixer if port termination

It does not matter if the diode is not as good of a switch as a FET, the theory of operation does not change.  The diode mixer is delivering RF pulses just like the FET, but maybe not doing as efficient job at turning on or off as cleanly the FET, at least at HF. 

The only real requirement is that the needs RF pulses with a source resistance to work.  I don’t see why this might not work better for a DC receiver than the more normal diplex configuration which throws away half the signal power in an unwanted “sum” product.


  • Dan, N7VE


14837 2018-05-22 12:25:29 Bill Carver Re: mixer if port termination
Unlike MOSFET switches, Diode mixers have LO, RF and IF signals flowing in the same wire/silicon. Applied signals, plus signals reflected back from misterminated ports. A short circuit (bypass cap) on the IF port will reflect LO and RF energy back into the mixer: the usual observation is mistermination produces additional IM products and I don't see how THIS mistermination is somehow magic, producing more signal without any ill effects.

IE, I don't believe in magic.

14838 2018-05-22 14:26:18 Tayloe, Dan (Noki... Re: mixer if port termination

Google “diode mixer schematic” and the first diagram is the common double balanced diode mixer (like a Mini-circuits TUF-1). As I look at that schematic it seems like the operation of the LO is to turn on D1 and D2 for half the cycle which provides a ground path for one of the two “signal windings”.  On one half of the LO cycle D1 and D2 turn on and select one signal winding, and on the other half cycle D4 and D3 turn on, which routes the other half of the “signal” winding to ground as referenced from the “output” signal. 

It appears as though the LO is isolated from the output.  In fact the spec sheet for the TUF-1 shows ~ 60 dB of isolation from LO to the IF output.  Thus a detection cap of the output would not act to “short out” the LO signal.


Thus, the detection cap is not affecting the LO signal.  The LO is using the diodes like switches to alternatively route the RF input to the IF output from one side for ½ cycle, and then the other side of the RF input for half a cycle. 

The detection capacitor (integrating detector) should work fine under these conditions as it is getting the proper RF plus “polarity” every ½ LO RF cycle.  I am not sure, but there might be some slight DC offset from the feedthrough of the LO.  That would not affect anything since this mixer would likely need to be AC coupled to the preamplifier stage unless the preamplifier stage were able to use a split supply so that it could accept the baseband signal that varies either side of ground.

I am headed out of town tomorrow for a few weeks, or I would go “melt solder” and verify this.  Like I said, I don’t think there is anything mystical at work here.  It might indeed work worse.  However just because “we have always done it this way” doesn’t mean that something different might not work better. 

That detector capacity is absorbing and averaging out all the signals in the RF input.  If you think if this detector as a low pass filter with the series R being the system impedance and the shunt C to ground being the detector cap, note that the mixer sits on the “protected” side of the resistor in the R/C low pass filter.  The fact that the detector cap looks like a low impedance “off frequency” compared to the system series resistance means that for the off frequency signal, its signal power is being dropped by the series R such that the power level within the mixer is reduced (I think). 

Thus, this mode of operation may be acting to protect the mixer (not just the audio pre-amplifier) from large off frequency signals. 

Note that this detector loves odd harmonics (even harmonics tend to cancel out in the detector cap averaging process over the ½ LO cycle).  Good RF bandpass filtering is still needed on the front end to get rid of the 3rd harmonic in particular.


  • Dan, N7VE




14840 2018-05-22 15:17:12 Dana Myers Re: mixer if port termination
14841 2018-05-22 15:22:59 Tayloe, Dan (Noki... Re: mixer if port termination

The RF-IF path is bi-directional. The detection cap charge carries over from RF cycle to RF cycle and is in turn reflected back into the RF input and shows up on the RF input waveform like distortion.


Yep, that is the way it works.


  • Dan, N7VE


14842 2018-05-22 15:35:01 Dana Myers Re: mixer if port termination
14843 2018-05-23 06:41:59 John Levreault Re: mixer if port termination
I don't know what kind of diode is typically used in a DBM, but I'm not sure
that the 1N5818 is the best choice. It's designed for much higher operating
currents and has high capacitance. Even the lowly 1N4148 may be an improvement,
although it will probably need more drive voltage. LTSpice has some low-current
Schottkies like the RB500 series that seem to have much lower capacitance.

John NB1I

14844 2018-05-23 09:31:05 Dana Myers Re: mixer if port termination
14845 2018-05-26 05:44:02 kb1gmx Re: mixer if port termination
RIch, it sounds like your trying to get around a problem that doesn't exist.

Max power transfer occurs when impedance are matched.  zit does not 
make a difference if its 15 ohm or 1500.

So why not design the diplexer to terminate into higher impedance so more kinds of amplifier can be used?  That is other low noise amp configureations that have higher input impedance.

I have no clue what that is supposed to mean.   Low noise amps are that and its not an impedance thing.

14846 2018-05-26 06:08:07 kb1gmx Re: mixer if port termination

Terminating a DBM IF port into a capacitor works to a point and the point where 
it increases spurious mixer products.  The products are the result of reflections from 
the effectively shorted at IF port.  If there were only one signal no problem but its 
rare in the real world for there to be just one.  So we have all the likely input frequencies
and the LO and the signals reflected from the IF port.  This is why its bad.  

Sampling mixers and fet based switching mixers do not suffer the same problem.

A diplexer aka an absorptive filter is both simple and effective.  They are 
simple in that at audio is low loss coupled and at some RF frequency above 
audio the RF is absorbed.  Its a fact of life for DBMs if you want all the parameters 
of performance possible.  DBMs are popular as they appear simple building block
and can be used bidirectional and are well understood.  Yes, I am about to say DBMs 
have feet of clay and are imperfect. They have their liabilities and problems while 
understood are often ignored.  The solution can be treat DBMs as the require or 
use something that fits the case better.

Opamps are not generally lowest noise as the feedback and input resistors are 
contributors in part to the noise of the system.   


14847 2018-05-26 06:20:27 kb1gmx Re: mixer if port termination

The math may work if you use perfect diodes and fets.  For real world devices diodes and fets
are imperfect switches and the differ greatly in the type of imperfection.  Soon as we get to 
second order and higher issues the two devices diverge and other complex notions appear.  
Then the mathematical fun begins.

To see the real world effect you need three signal generators.  The reason for the third is to see
the effects of a unwanted signal.  The difference is we are now looking at second and third order 
effects such as IMD.

14848 2018-05-26 06:30:43 kb1gmx Re: mixer if port termination
At RF getting filters [diplexers are filters] to translate and match can be done but with 
some work to get them right.  At audio its not so easy as the values and Qs attainable
are awkward at best if not huge.

It is easy to build an audio amplifier that matches 50 ohms. Low noise is as 
attainable there as 600 ohms.   The auso stage can also perform impedance 
translation as well at high levels of voltage gain.  An example of this is the 
BJT common base at .52 mA Ie with a collector load of 5000 ohms.  Voltage 
gain is 100, input impedance is 50 ohms from near DC to potentially 
approaching the device alpa cutoff (greater than 200 mhz for a 2n3904).
An alternate way if parallel common emitter bipolars at higher currents.
Both ways have been demonstrated many times and work well.

14849 2018-05-26 06:43:33 kb1gmx Re: mixer if port termination

When you get to fet mixers there are two flavors.  Switching like I7SWG and friends, and the 
sampling mixers such as Dan's.  That also gets into the world of balanced and doubly 
balanced types.  For receivers that can be an insignificant difference for transceivers
its a real big deal.  A good example of a signal balanced mixer that worked well one way
was in Minima, the kiss mixer.

As to mixer losses, you pay for transformers how much is dependant on the transformer but
its there and measurable.  For DBMs its possible to get lower losses at the expense of 
signal handling.  However the loss low and adds little noise. A limited amount of low noise gain
prior can easily offset it when lower MDS is required.  Another item to look at is preselection
before the mixer to tamp down out of band signals.

Its a good design challenge and fun to build.

14850 2018-05-26 06:57:47 kb1gmx Re: mixer if port termination
Your on the mark the 5818 and for that fact the 1n4148 is good but
you can get Schottky diode quads that offer excellent match, tracking 
with temperature (same die) and high speed.  This is what the mixer 
companies use.

Farhan used the dual version of BAT45s as a better device of the 
5818 or 1n4148.

FYI the 4148 is interesting as its not a Schottky diode, its a fast switching 
junction rectifier.  As a result you need a higher LO and it has a higher 1db 
compression point as a result.   For some things that can be helpful.

I added the 4148 as a mixer for an example of best or better and 
how that can be a hard to define.

DBM have a demand for symmetry and balance to perform at 
predictable  levels and across wide bandwidths.  Individual diodes
require careful matching to be a good choice.

14851 2018-05-26 08:04:32 Richard Johnson Re: mixer if port termination

I don't think i am communication well and also i am new to the RD world.

So here goes again.

I read that a pi-network used for attenuation cold be designed for 50 ohm input and a higher impedance output.

If so, then could not a duplexer be designed with a 50 ohm input and a higher impedance for the output.

That is the center of my question.

The reason is, i thought it would open up higher input impedance low noise amplifier designs for the post mixer amp.

Thank you for all the answers.

14852 2018-05-26 09:28:24 Lasse Moell Re: mixer if port termination

Yes that should be doable. I see a diplexer as a lowpass and highpass filter with one common port. And a filter can easily be designed for diferent impedance on in vs output port. Having a large impedance transformation may lead to some odd component values, so you have to watch out when specify and design the diplexer.

/Lasse SM5GLC

26 maj 2018, 17:04 centraleuropeisk sommartid, skrev Richard Johnson richj45@gmail.com [emrfd] :


I don't think i am communication well and also i am new to the RD world.

So here goes again.

I read that a pi-network used for attenuation cold be designed for 50 ohm input and a higher impedance output.

If so, then could not a duplexer be designed with a 50 ohm input and a higher impedance for the output.

That is the center of my question.

The reason is, i thought it would open up higher input impedance low noise amplifier designs for the post mixer amp.

Thank you for all the answers.


14853 2018-05-26 11:31:24 Brooke Clarke Re: mixer if port termination

There are a number of factors.
1. The mixer equation is:  IF = +/-m*RF +/-n*LO
So, at the diodes there are a lot of frequencies involved.

2. The idea of a "balanced' mixer is to reduce some of those frequencies at some of the mixer ports.  Bob Mouw invented
the doubly balanced microwave mixer.
This also means you can make an image enhancement mixer by properly reflecting some mixing products back into the mixer.

3. Some diodes work better than other diodes in mixers.  There were some misleading WJ app notes about this, but the
real answer is space charge current limiting.

Have Fun,

Brooke Clarke, N6GCE

-------- Original Message --------
> Re: mixer if port termination
14854 2018-05-26 11:51:18 jwharding3 Re: mixer if port termination Diplexer Q
Mixer world folks .. Regarding diplexers. The most common I see have a Q=1. Others state a Q=10 is desirable for Thank John
14855 2018-05-28 18:12:11 Bill Carver Re: mixer if port termination Diplexer Q
As long as the i.f. port looks like 50 ohms, from a theoretical standpoint ANY Q in the diplexer produces the same result: constant impedance with frequency.  Unfortunately while the transformer of the gate-source feedback amplifier can produce an input resistance of 50 ohms, it also has a shunt reactance. That can be tuned out at the i.f. frequency, and then it looks like 50 ohms in parallel with a very low Q resonant circuit. But that is not the same thing as 50 ohms at ALL frequencies, so the diplexer does not present a 50 ohm impedance at all frequencies to the mixer. The resonance of the gate-source feedback amplifier has less effect on the mixer as the diplexer Q is increased. 

Assume the gate-source transformer has a gate winding inductance of 15.3 uH and is resonated at 5.2 MHz. With a Q of 1 the diplexer has convenient identical inductor and capacitor values. The extremes of impedance seen by the mixer will be 47.5+j0 at 3.23MHz and 8.5MHz, and worst-case phase angle of 1.8 degrees at 2 and 14 MHz.  With a Q of 2 the mixer will see 50+j0 at 5.2 MHz, 48.75 ohms+j0 at 4.075 MHz and 6.62 MHz, and worst-case phase angles of 0.8 degrees at 3.0 MHz and 9.1 MHz.  A Q=10 diplexer would present 50+j0 to the mixer within a few tenths of an ohm and tenth of a degree for all frequencies.

With two identical coils and capacitors the Q=1 diplexer is easiest to make and tune. The Q=2 has a 4:1 radio between inductors and capacitor values, still practical, slightly more critical to tune. At Q=10 coils of 15.3uH and 153 nH are required in the diplexer, and self-resonance in the 15.3 uH would be in the HF or low VHF range and could be seen by the mixer, and tuning it to achieve that would probably require a VNA. But it could be done.

So how big a difference does 0.8 degees in the mixer load make, versus 0.1 degree? I'll let someone with more diode mixer experience comment on the effect of a few degrees from resistive has on intercept.

Bill W7AAZ

14856 2018-05-30 09:01:08 kb1gmx Re: mixer if port termination Diplexer Q

The exact answer is always related to the level of performance desired and the capability of the part.

I was once told you solve 80% of a problem with 20% of the effort, the remaining 
20% will take way more than you think.  Also is what remains a problem that 
needs to be solved?

If port is known to be most sensitive but what the threshold for pain?  if it sees a 
reflection how bad is it and how much is bad?  What was missed is mixer and 
diplexer and the stage after it all count and if the stage after is crap why go to 
the effort.  It also never talked about getting the IF or baseband through the 
diplexer with minimum loss as that loss adds directly to the noise figure.  
Dealing with DBMs sometimes its don't break the fix.  Test it and if things 
go the way desired call it done or adjust as needed.

DBM is a conundrum in 3 ports. Its a simple passive block that offers all sorts of 
handy uses.  It also has all sorts of liabilities and interactions.

In the end maybe a DBM was the wrong part to use for a specific task or its 
being used in the wrong way.  If you use it you have to appreciate the conundrum
and accept the compromise.

Oh, and its only 50 ohms at some specified (or not) drive level.


14857 2018-05-30 15:46:26 Bill Carver Re: mixer if port termination Diplexer Q
A good general response, Allison, to a tough question.

The specific question that was asked used a post-mixer amplifier that has a noise figure well under 2 dB, and as I recall an input intercept around +26 dBm. With very practical and reasonable element Q's the diplexer loss from mixer to post-mixer amplifier is negligible: a tiny fraction of 1 dB. I cranked out the IF port termination for diplexer Q of 1, 2 and 10. So I think the variation in IF port termination is very well defined. Even so, I/we understand there is no exact answer to the question.

Because while the LO can be a pretty good wideband match (as has been discussed recently) obviously the RF port is very unlikely to be a wideband 50 ohms match. And different internal mixer designs might be expected to react differently to the RF port mismatch. And no two mixers are exactly the same, whether it's due to the specific design, small diode mismatches or transformer winding variations. And the mixer could be +7, +10, +13, +17, +20 or +23 and made by a wide range of manufacturers. Those issues are NOT well defined.

Nonetheless, there are generalizations. Ulrich Rohde said diode mixer IF port termination was a key design issue back in the 1970s. This isn't something new. Are there more specific generalizations....a number for i.f. port SWR below which intercept degredation is considered minimal, for instance? Not a specific intercept, just how much degredation. We are homebrewers: we aren't designing to a specification from the Marketing Department, we just want to understand what degredation of the mixer intercept might result from those variations in i.f. port termination.

Regards - Bill

14858 2018-05-30 20:38:40 hyu hyu To terminate or not a CATV splitter
Hello everyone, 

I have two CATV splitters, both with two outputs.

The first splitter outputs are a TV and the other OUT a second splitter (port IN).

The second splitter outputs are the internet cable modem, and the other output not connected.

The question is: Do I have to terminate the unused output port?

Thanks in advance.

PS: I don't know if this is off topic.

Virus-free. www.avg.com
14860 2018-05-31 00:08:14 Andy Re: To terminate or not a CATV splitter
hyu hyu wrote, "The question is: Do I have to terminate the unused output port?"

Well, no, you don't HAVE to.

But should you?  Yes.

Some splitters provide enough isolation (and loss) so that an unloaded output could be forgiven without much signal impairment.  Even so, it is not best to leave unused outputs open.

In the unlikely event that you're talking about active splitters, then the isolation is such that unterminated outputs might not be so bad for the system -- but might be more bad for the active splitter itself.


14863 2018-05-31 20:55:06 hyu hyu Re: To terminate or not a CATV splitter
Thanks Andy for your reply.