EMRFD Message Archive 9805

Message Date From Subject
9805 2014-03-06 10:47:02 Ashhar Farhan diode mixer termination and degradation
what happens if the diode mixers are directly connected to the crystal filter? how much is the effective degradation in the IIP3? is there a way to know these figures?

- f
9807 2014-03-06 12:06:44 Sandeep Lohia Re: diode mixer termination and degradation
On 07/03/2014, Ashhar Farhan <farhanbox@gmail.com> wrote:
> what happens if the diode mixers are directly connected to the crystal
> filter? how much is the effective degradation in the IIP3? is there a way
> to know these figures?
> - f

Not having fancy VNA here. but I rememberb,
somewhere I read :

we cannot use something tuned to narrow bandwith @ directly to the
diode ring mixer, they might start whistling...

The diode mixers are inherently broadband and bidirectional in nature.
This is good and bad.

The badness is a little tougher to explain. Imagine that the output of
a hypothetical mixer is being fed to the next stage that is not
properly tuned to the output frequency. In such a case, the output of
the mixer cannot be transferred to the next stage and it remains in
the mixer. Ordinarily, if the mixer was a FET or a bipolar device, it
usually just heats up the output coils. In case of diode ring mixers,
you should remember that these devices are capable of taking input and
outputs from any port (and these inputs and outputs can be from a
large piece of HF spectrum), hence the mixer output at non-IF
frequencies stays back in the mixer and mixes up once more creating a
terrible mess in terms of generating whistles, weird signals and
distorting the original signal by stamping all over it.

A simple LC band pass filter/ or as said above Ladder XTAL that
immediately follows the diode ring
mixer will do a good job only at the frequencies it is tuned to. At
other frequencies, it will offer reactive impedance that can cause the
above mentioned problems. It is requirement that the diode mixer's
inputs and outputs see the required 50 ohms termination at all the
frequencies. In other words, they require proper broadband

A diplexer and a hybrid coupling network is a better


& why not with 2 passive gates, that of any higher Ft switching IC ?
My HC series used previously stops working above 21 Mhz.
why not with something like AD8333A quadrature mixers H-mode ?
9809 2014-03-06 13:47:08 dixonglennb Re: diode mixer termination and degradation
Hi Ashar,

Wes performed this very experiment and documented it in his Solid State Design book.  I consider it one of the classic radio experiments. 
When I get home I'll try to look up and report the exact numbers, but I remember  the crystal termination did degrade IP3, and Wes used the experiment to support the idea that a diplexer termination is necessary.
Interestingly, in an ideal commutation mixer (that is, instant switching, zero impedance paths, etc. etc.) it can be shown that linearity is not termination dependent. ..you can reflect unwanted harmonics back into the mixer from either side without degrading IP3.  Real devices will differ of course, but I tend to think that in this case the actual main source of distortion is the crystal distorting harmonics and reflecting those distorted signals back into the mixer.  You might be able to reflect harmonics back into the mixer, but they better not have any IMD!
Glenn AC7ZN
9811 2014-03-06 16:25:15 w7zoi Re: diode mixer termination and degradation
Hi Glenn, Farhan, and group,

This issue of proper termination of a ring mixer continues to haunt us.   I remember doing the measurements that you mention as reported in SSD, but must admit that I don't recall where in that book it was reported.  I remember that the mixer was a level 7 HP ring that we had as a bench measurement tool.   It was driven with a proper powered  signal source at the LO port.  The RF port was driven with a pair of signal generators combined with a 6 dB hybrid combiner.  The output of the mixer was run into a 10 dB pad and then to a Tek 7L12 or 7L13 spectrum analyzer where we could measure IMD.  After measuring third order IMD, a narrow filter was inserted between the ring mixer and the 10 dB pad.   By "narrow," I mean that the filter bandwidth was similar to the spacing between the driving signal sources driving the hybrid.    I recall that the IF was somewhere around 10 MHz.   The filter that was inserted to upset the purity of the mixer termination was a single pole LC filter.    Alas, I don't remember any of the details other than the critical result:   The mixer third order intercept degraded by 10 dB when the narrow filter was inserted in the system.   I don't even recall if the results were within the bandwidth of the inserted filter.

I'm sure that other folks must have done measurements that are similar, but I don't recall having seen reports.   Clearly, it is time for the experiment to be repeated.   A bit better reporting would also be in line.  (Do any of you have any references on this matter?)

There are a few interesting details that must be included in our thinking regarding such matters.   The first is the nature of the diplexer that we might use to clean up the termination.   The usual diplexer is a network of inductors, capacitors, and resistors.   The network has the property that the input impedance is constant with frequency.  The diplexer is a selective network that serves to terminate the driving source for frequencies well away from the center frequency.   But it does NOT provide a good termination at all frequencies.   The diplexer only takes care of the frequencies well away from the design center.

Consider an example.   Say we want to use a diode ring mixer to drive a 9 MHz crystal filter with a 1 kHz bandwidth.   The crystal filter is already matched to 50 Ohms at 9 MHz.  So the circuit following the mixer would consist of 
1.  A 9 MHz LC diplexer.  bandpass-bandstop type, EMRFD page 3.38.
2.  The crystal filter.
3.  A 50 Ohm wideband load,   This might be the input to an IF amp.

We would like the input impedance to be constant at 50 Ohms for all frequencies, DC to light (or so.)   It won't be.   Instead, what we will see is a good match exactly at 9 MHz, but with major lumps in the impedance just above and just below 9 MHz.     The crystal filter itself reflects its output termination through to the input within the filter passband, but not in the filter stopband.   The diplexer won't fix these problems because the usual LC diplexer circuit elements are not nearly selective enough to fix the close-in (frequency) sins of the crystal filter.    There is a solution, but it is not trivial.   See the discussion on EMRFD p 6.50, etc.

Another vital detail is to include diplexers or other termination purity enhancements at the mixer RF port if the ultimate performance is sought.    See Stephensen, QEX, May/June 2001.  EMRFD CD item 58.

A ploy I've used for a long time is to follow a mixer with a wideband amplifier with good IMD properties.   This then drives an attenuator and then the crystal filter.   The pad and the wideband amplifier together provide the wideband mixer termination.   The IMD properties of the amplifier are, however, a vital part of the system.   Some folks who have used this topology have enhanced it with diplexers.  VE7CA has done some good things here.

Our more elaborate transceivers tend to use high level mixers with FET switches in a variety of forms.  Some folks report casually that these mixers are less fussy about termination than are the diode rings.   That's all they say.   Numbers would be more useful than blanket statements.  

A recently introduced "appliance" has brought some of these details to the forefront.   This QRP transceiver uses a level 17 ring mixer that is followed directly by a crystal filter.   The performance is good, but it is not clear if it is close to optimum.    Did they do measurements, or did they merely have the good sense to ignore the lore that evolved from early measurements, reported or otherwise?   Clearly, they did overall response measurements, for the final result is a good one.   

The answers to these questions lie in the details.   Some of the details are easily simulated with the modern computer tools we ALL have available to us.   But even when simulations are possible, the ultimate and only real answer will come from measurements.  But lore from a third of a century ago is not good enough (even if we could remember the details!)  If we use lore to assist our thinking, we should at least use modern, up-to-date experiments as a basis for that lore.   

73, Wes

9851 2014-03-21 22:25:38 Ashhar Farhan Re: diode mixer termination and degradation
I took sometime to finally assemble two free running oscillators at 14.000 and 14.0025 and a 6db hybrid combiner and a low pass filter in a chocolate tin box. The output of individual tones is 70 mv on a 50 ohms. When both of them switch on, they give 150 mv peak on the 50 ohms load.

Hence, I have calculated the output of each tone to be -13dbm. 

I used my Minima transceiver to do some testing. First, the KISS mixer was replaced by the ADE-1 mixer. Then, I fed the tone signal generator through a step-attenuator to my transceiver.  This is essentially a low pass filter in the front-end feeding the ADE-1 mixer that is directly connected to the crystal filter.

I turned to the 14.049 to find the IMD signal. and connected the scope probes to the audio jack. I tuned the signal until it was peaked, then used the volume control to reduce the signals on the scope to a 100mV peak. Now, I tune back to the 14.026 MHz signal and kept throwing the switches on the attenuator. The results are as follows:

Crystal filter connected to the output of the ADE-1 on pin 3, oscillator is a square wave from Si570 fed through a -6db pad.

     Input signal, each tone is -13dbm
     14.049 MHz with zero attenuation for 100mv on Audio
     14.026 MHz 100mv with 48 db attenuation.
     IIP3 = -13 + (48/2) = 11 dbm

Now, I insert a 10 db pad between the the ADE-1 and the crystal filter.

    14.049 was again adjusted to 100mv peak on audio
    14.026 Mhz needed 52 db attenuation to produce the 100mv signal
    IIP3 = -13 + 52/2 = 13 dbm

I note that there is just a 2db difference! The datasheet for the ADE-1 from their website quotes that their 'excelent' IIP3 is 15dbm. I am almost in that ball-park with a 2db difference with and without a strong termination. So, even going by their own paper, a 4db worsening of the IIP3 may not be a huge impact for most of our work. (I live east of WWV by 8000 km, btw).

- farhan

Any explanations? Should I make the return loss on the mixer stronger? 

9852 2014-03-21 22:27:45 Ashhar Farhan Re: diode mixer termination and degradation
I am heading into a scheduled power cut now. I will test this with the KISS mixer and a few home-made diode mixers as well and report back in the next day or two.

-- f

9853 2014-03-22 15:30:14 w7zoi Re: diode mixer termination and degradation
Hello Farhan, and gang,

Bravo!   You have done a real measurement that takes us beyond mere lip service.    Way to go.

As I think I mentioned, the experiment that I did that was reported in SSD used an LC filter of some sort following the ring mixer.    I saw a 10 dB degradation when the filter was in place versus what we saw with a clean, wideband 50 Ohm termination.    I don't remember the spacing of the two tones that I used, but the spacing could well have been small enough that the IMD products were on the filter skirts just a few dB down from the desired response.   This is much different than what you have done.   You have a crystal filter, probably with a SSB bandwidth.   Then you used tones spaced by 25 kHz.   Hence, your spacing was about 10X the filter bandwidth, so you should be well out of the skirt response.   

This is really excellent work, Farhan!   This is a line in the sand that we can now use to guide us with other experiments.   I'll be really anxious to see the results you get with other mixers.

Again, way to go!

73, Wes
9854 2014-03-23 00:32:05 Sandeep Lohia Re: diode mixer termination and degradation
Farhan, Glenn & Wes

M still a learner but in most MIL commercial-grade HF I found it's passive DDBM
or say Triple-balance mixers which offers provide lower distortion,
better signal-handling capability and higher interport isolation than
basic ADE or SBL single-ring designs, over all greater dynamic range.
disadvantage is cost of four extra diodes, & to match them,
& bit higher drive from LO might not be an issue...

[Non-text portions of this message have been removed]
9855 2014-03-23 20:43:44 Ashhar Farhan Re: diode mixer termination and degradation
wes and the gang,

I have a bit of a bad news. 

First, I'd like to show the the circuit that i am using 

Please note that I am using darlington pairs as the buffers with 100K resistors in series with the VFO output. 

Then, both the oscillators are in the same box. Very little isolation. The box is slightly more than 6 inches across.

Now, for the trouble. I posted about the ADE-1. Then I tried the KISS mixer with almost the same results of 11dbm without a 10db pad and 10 dbm without it. Then, I tried a singly balanced diode mixer, this gave 12dbm without the 10 db pad (directly into the crystal filter) and 15dbm with a 10 pad.

I found this a little suspicious. So, next I decided to confirm the IIP3 were resulting from the mixer and not elsewhere. I tuned back to the IMD tone and flipped the 10 db switch on the attenuator. I noticed the audio dropped to 33% on the 'scope. Corresponding to the 10 db power drop. It should have dropped by 30 db!

Next, I pulled out the retired Ft-817 and tested with that. Even that gave me the same 10dbm result (AGC off, no-preamp).

Now what? I am sure that these IMD products are being created outside the Devices Under Test. I am going to try two different things now:

1. Shield each VFO with copper sheets.
2. Use a common-base (or common-gate) buffer to isolate the VFOs.

Are there any other suggestions?

BTW, a serendipitous  discovery. Using the SMD components in VFO is a great thing. The oscillators by far the most stable and cleanest I have ever used. Highly recommended. It drifts less than 10 Hz over an hour!

- farhan

9856 2014-03-23 21:43:11 Tayloe, Dan (NSN ... Re: diode mixer termination and degradation

I suggest you place the two VFOs in different boxes, and place attenuators between the VFO buffers and the combiner.


I notice that at low attenuation levels that my 8640B outputs “talk” to each other.  Thus at high level output, I am measuring the IM of the 8640Bs and not the circuit.


I think there are cap trimmers that can be placed into the circuit to null the output of one VFO from the output of the other.  That would also help.


-          Dan, N7VE


9857 2014-03-23 21:45:55 Ashhar Farhan Re: diode mixer termination and degradation
Dan, thanks, I will progressively apply these band-aids and try it all out.

- f

9858 2014-03-23 22:14:49 John Marshall Re: diode mixer termination and degradation
I think you're on the right track, Farhan. Shielding first, then take a look at the power supply decoupling. Your oscillators are pretty well isolated but not the buffers. Common base buffers should give better isolation but first have a look at the ones you're using. Is there really no bias on the first transistors' bases? Finally, have you verified that your combiner is giving good isolation as well as combining the signals equally? Easy to test by disconnecting one buffer and checking the level of the other oscillator there.

Best luck,

John, KU4AF
Pittsboro, NC

9859 2014-03-23 22:17:33 ashhar_farhan Re: diode mixer termination and degradation
John, that biasing is a mistake in the circuit diagram. If you look closely at the picture, the 100K resistors are present. I have added chokes on the power line with little joy. Now, on to making common-base buffer amps next.

- f
Sent from BlackBerry®
9860 2014-03-24 01:09:15 Ashhar Farhan Re: diode mixer termination and degradation
I have updated the circuit with the original buffers. but i am have changed the buffer to use a common-base buffer. i shall shortly post the results.

- f

9861 2014-03-24 02:08:14 Sandeep Lohia Re: diode mixer termination and degradation

1) Great! I cannot see matched 4148 near 8T Biflar...

2) 2 X 78L09 for isolation

3) Also trying to understand restrictive bridge ( 100 x 2 ) seems it's
for impedance matching! ( proper termination )

4) Collectors of T1 + T2 are sorted, no resistance in between...

5) if with digital, oscillator must be spurs & harmo free...
those unfiltered from LPF might feedback, & both final BJT might work
as mixing even & odd harmos???

6) *Signal output from Emitter of T2 can be available on collector of
T4, ( A+B & A-B )
but bypass-grounding 100 nf on both final BJT, is it that for???

Trying to understand, a small note will be helpful though...

On 24/03/2014, Ashhar Farhan <farhanbox@gmail.com> wrote:
> I have updated the circuit with the original buffers. but i am have changed
> the buffer to use a common-base buffer. i shall shortly post the results.
> - f
9869 2014-03-26 10:11:53 Ashhar Farhan Re: diode mixer termination and degradation
I will take a deep breath and post some confusing results.

First, I noticed that the IIP3 figures were almost the same from mixer to mixer. This lead me to believe the distortion was not in the mixer but elsewhere. After trying a number of things, It struck me that it was the buffer amplifiers of the VFOs used that was creating the distortion. The buffer amplifier were simply a darlington pair. The signal of the other VFO could be travelling back through the hybrid coupler, The darlington pair exposes a forward biased diode of the emitter-base junction to the signal coming in from the other VFO, causing IIP3!

1. Lesson learnt: taking emitter output for driving a measurement setup is a bad idea. Esp when some signals may flow back.

This was changed to a regular feedback amplifier with a 6db pad in the output. See the final circuit at http://www.hfsignals.org/index.php/Measuring_IMD. Each VFO's output as measured at the output port was -15dbm on a 50 ohms load. The two VFOs were set at 14.020 KHz and 14.000 KHz.

2. The test setup consisted of the above circuit fed to the receiver through a homemade step attenuator that allowed us 80 db of total attenuation in 1db steps. The audio output was monitored on a PC based audio oscilloscope as well as a Tektronix 465. Both gave the same results.

3. The FT-817 was used as a receiver. 
3.1 Its AGC was switched off, preamp was switched on.
3.2 the IMD response at 14.040 was tuned in. The AF gain (volume) was cranked up and the RF gain was increased just to the point where the signal appears to be clean on the 'scopes. 
3.3 The receiver was retuned to 14.020 and RF attenuator was adjusted until output showed the same amplitude as the previously noted IMD response. The attenuation was 41 db. 
3.4 IIP3  is calculated as -15dbm (single signal) + 41/2 = -5.5dbm
3.5 Now, the FT-817's preamp was switched off and the the measurements were repeated. The attenuation now required was 56 db. The IIP3, came to -15+56/2 = 13 dbm.
3.6 The attenuator was thrown in and the measurements were repeated. Now, the attenuation required was 58db, The IIP3 had climbed to -15+58/2 =14dbm

Note that the attenuator should have pushed the IIP3 way down, it didn't. On the other hand, shutting down the premap had a dramatic effect. We'll have more to say about it in a bit.

4. ADE-1 mixer
4.1 The Minima's KISS mixer was replaced by  an ADE-1 diode mixer. This circuit has just an LPF in the front-end that feeds to a passive FET mixer that directly drives a crystal filter without any diplexer or post-mix amp. The attenuation needed to match the signal at 14.020 to the IMD response at 14.040 was 50db. Thus, an IIP3 of -15+(50/2) = 10dbm was noted.
4.2 A 10 db attenuation pad was inserted between the mixer and the  crystal filter, Now a 58 db attenuation matched the signal to the IMD response. Thus, the IIP3 was -15dbm + (58/2) = 14 dbm.

Comments: The ADE-1 mixer is sensitive to termination. But just to about 4dbm.

5. A two diode mixer
5.1 A two-diode mixer using a junkbox binocular core was made with 8 turns of trfilar coil and unmatched 1N4148 diodes.The output was taken from the center tap of the transformer and fed directly to the crystal filter. the RF input was at the point between the two diodes. The attenuation used now was 63db. Thus the IIP3 comes to 16.5 dbm.
5.2 10 db attenuator was now added BETWEEN the generator and the receiver. The measurements were repeated. The IIP3 remained exactly the same!
5.2 A 10 db attenuator was inserted between the mixer and the crystal filter. The measurements were repeated. Amazingly, the attenuation used now was 62db. However, this figure is suspect. the signal at 14.040 (IMD response) was quite shaky and even a 3db difference would have been hard to spot. The IIP3, in any case works out to 16 dbm.

Comments: First, this mixer seems quite resilient to termination. Second, the 3db increase in IMD for 1db increase of signal level doesnt seem to hold true for the diode mixers.

I have repeated these measurements a number of times. Can someone evaluate the singly balanced diode mixer to confirm my measurements? I am particularly troubled by the lack of 3db-to-1db ratio of increase in IMD vs signal levels. Because this leads to trashing of using IIP3 measurements as the goodness of front-end blocks.

On the other hand, active mixers like the NE602 display this degradation. But not the diode mixers. The KISS needed more power to show up IMD on the scope. So, I will post those results as soon as I can get an AD8307 in the mix.

I have gone through some results posted by others (PA3AKE). But none of them show comparision of diode mixers when measured at two different input levels. That is, measure the IIP3 at, let's say -15dbm and then repeated it at -25dbm. In both cases, the IIP3 should measure up  to the same number (as per the theory) however, my measurements are finding that attenuation before the mixer has very little impact on the ratios between the IMD response and the actual signal. Probably, an independent confirmation might help.

- f


Next, the  minima transceiver's mixer was swapped with a simple two diode mixer. The diode mixer directly drove a crystal filter with 50 ohms characteristic impedance.

9876 2014-03-27 20:43:10 biastee Re: diode mixer termination and degradation
The hybrid combiner can isolate the two VFOs at their fundamental frequencies because all 4 arms of the bridge are balanced. However the combiner is unbalanced at the 2nd harmonic because the low pass filter does not present a 50 ohm termination outside the passband. Therefore, the 2nd harmonic (2*f1) from one VFO is reflected back to the other VFO buffer and mix with the other fundamental (f2) to produce the third overtone product, (2*f1 + f2). I suspect all this mixing happens in the buffer stage. Replacing the LPF with an attenuator will ensure that the combiner is balanced at both fundamental and its harmonics. The LPF can be moved to the buffer output, before the 6 dB pad.

I doubt if the problem lies with the darlington topology. Many of the early silicon MMICs, such as Minicircuits MAR series (re-branded from HP/Agilent/Avago's MSA), are based on the darlington topology. These devices were used in many signal sources. If their use have led to artificially poorer IP3 results, a caveat would have been issued.The improvement that you ascribed to the feedback amplifier is more likely contributed by the two 6 dB pads which were added at the same time as the feedback amp. These pads, which increases the isolati
9877 2014-03-28 03:30:37 Ashhar Farhan Re: diode mixer termination and degradation

Thanks for pointing out the LPF issue. I will increase the VFO output and place a 6db pad in-between the LPF and the hybrid mixer.

The darlington pair in itself is not the problem, Using it as emitter follower could be. The forward biased diode at the emitter-base junction can easily act as a non-linear mixer. I injected two crystal oscillators into the emitter of a common collector amplifier and I could get pretty decent sum and difference frequencies from the base!

- farhan

9878 2014-03-28 03:36:18 Ashhar Farhan Re: diode mixer termination and degradation
I finally got around to doing some measurements of the KISS mixer made from two J310s. The IMD response was so weak that emerging it out of the noise was a challenege, much less measuring the amplitude.

First experiment was of connecting the KISS mixer directly to the crystal filterIt took me 70 db of attenuation of -15dbm signals to match up with the IMD response. That puts the IIP3 at -15 +(70/2) = 20dbm.

Next, I put in a 10db pad between the mixer and the crystal filter. The IMD response had dropped into the noise floor. But a very hard attempt at matching wanted signal in the input would yeild similar results,  I presume. I am playing this by the ear.

This makes the mixer performance, only average. The 5db increase over the extremely simple diode mixer with just two 1N4148s may not be worth all the trouble.

Next, I will investigate the original KISS mixer with FSA3157s.

- farhan

9883 2014-03-29 09:55:32 biastee Re: diode mixer termination and degradation
If you have simultaneously changed the buffer topology and added 6 dB pads, how can you be certain that the improvement is due to the former and not the latter? If an emitter follower is more susceptible to mixing, how can it be used in audio amplifier output stage, where IMD is also an important consideration? In diode-based RF circuits, e.g. switches and attenuators, my experience indicates that a forward biased junction is always more linear than a reverse biased one.

> I injected two crystal oscillators into the emitter of a common >collector amplifier and I could get pretty decent sum and difference >frequencies from the base!

Have you tried feeding two signals into the feedback amplifier's collector? I suspect any one of the transistor's three terminals can serve as the mixing input.

73, Chin-leong Lim, 9W2LC.
9884 2014-03-30 14:48:02 Ashhar Farhan Re: diode mixer termination and degradation

I wasn't clear, my bad. 

my analysis is like this: the emitter follower that is feeding a hybrid coupler has the other VFO's signal incident upon it. This is not the case for an audio amplifier or any other application (like a post-mix amplifier) where a relatively strong signal from another source is incident upon the emitter of an emitter follower. I am hypothesizing that the mixing of the two VFOs is happening here. Surely, the hybrid coupler does reduce the leakage of one VFO into the other but it is not absolute. I have measured that to be only about 40 db down. (that is, all ports are terminated in 50 ohms) and one input port's RF energy is measured at the other port.

a reverse biased diode should not exhibit any non-linearity until very near forward bias condition. At least, that's what the diode equation predicts. The non-linearity decreases with increased diode current. Non-linearity is also dependent upon the charge inside the diode's junction. For instance, a forward biased 1N4007 will have greater RF linearity when forward biased than, a 1N4148. Hence, switching needs diodes that do not respond quickly enough to RF current sinusoids but rather let them pass through transparently. 

feeding the two signals into the collector will probably have little effect compared to feeding them to a forward biased emitter.  This is purely a conjecture at the moment. I can do the measurements and report them back to you. 

- f