Double Balanced Mixers

 

Comparative Measurements on Double Balanced Mixers

By Nicolae Fieraru

Document last updated: 28-Dec-2012

The following mixers have been checked and compared:

  1. ADE-1, mixer placed onto a PCB
  2.  Home-made mixer with 1N4148 diodes (PCB + IC socket DIL-8, 2 x ferrite transformers 73uH each coil, 10 turns of trifilar wire)
  3.  Home-made mixer with 1N5711 Schottky diodes (same mixer as above, 1N4148 diodes replaced by 1N5711).
  4.  Home-made mixer with 1N60P Germanium diodes (same mixer as above, 1N4148 diodes replaced by 1N60P).Multimeter measures 0.192V when set on diode.
  5.  Home-made mixer with HSMS-2829 quad ring (same setup as for points 2 and 3. The HSMS-2829 chip had wires soldered to its terminals and the wires were inserted in the IC socket)

RF Frequency = 4.26MHz
LO Frequency = 14.26MHz

RF Level = -20dBm (measured by connecting the oscillator to a 50 ohm Spectrum Analyzer)

RF Level Oscillator (actual value measured by oscilloscope with the oscillator connected at the LO Port of the mixers)

100mVpp for Mixer ADE1
135mVpp for Mixers 1N4148, 1N5711

Conversion Loss (= Level LO Out – Level RF In)

RF = -20dBm ADE-1 1N4148 1N5711 1N60P HSMS-2829
LO Level [dBm] [dB] [dB] [dB] [dB] [dB]
-5 -11.02 -40.20 -11.38 -6.83 -9.14
-4 -9.85 -34.77 -9.77 -6.37 -7.94
-3 -9.01 -28.68 -8.73 -6.04 -7.21
-2 -8.42 -22.93 -8.07 -5.84 -6.70
-1 -7.97 -17.80 -7.60 -5.64 -6.32
0 -7.52 -13.83 -7.23 -5.46 -6.04
1 -7.28 -10.98 -6.97 -5.33 -5.80
2 -7.03 -9.09 -6.72 -5.23 -5.66
3 -6.82 -7.92 -6.57 -5.13 -5.48
4 -6.70 -7.13 -6.39 -5.05 -5.33
5 -6.57 -6.60 -6.29 -5.00 -5.20
6 -6.47 -6.11 -6.19 -4.94 -5.10
7 -6.38 -5.82 -6.09 -4.90 -5.00
8 -6.29 -5.62 -6.01 -4.82 -4.90
9 -6.24 -5.45 -5.93 -4.80 -4.82
10 -6.19 -5.28 -5.86 -4.77 -4.75

RF Rejection at IF Output for various LO Levels:
(formula used: measured RF Level at IF Output -(-20dBm))

RF = -20dBm ADE-1 1N4148 1N5711
LO Level [dBm] RF Level at IF Output [dB] RF Level at IF Output [dB] RF Level at IF Output [dB]
LO off -72.0 -25.1 -25.6
-10 -67.0 -25.9 -26.0
0 -55.7 -26.9 -29.3
7 -52.7 -28.3 -29.7
10 -51.5 -28.2 -29.9

LO Rejection at IF Output for various LO Levels:
(formula used: measured LO Level at IF Output – LO Level)

RF = -20dBm ADE-1 1N4148 1N5711
LO Level [dBm] RF Level at IF Output [dB] RF Level at IF Output [dB] RF Level at IF Output [dB]
-5 -43.8 -63.6 -58.6
-4 -45.6 -64.4 -57.9
-3 -45.5 -66.2 -57.6
-2 -46.7 -71.0 -57.6
-1 -47.8 -78.0 -57.6
0 -49.2 -70.0 -58.0
1 -50.6 -64.5 -58.4
2 -52.3 -61.7 -58.9
3 -54.3 -60.8 -59.2
4 -56.8 -60.4 -59.7
5 -59.8 -59.4 -60.1
6 -63.2 -57.2 -60.7
7 -68.0 -56.3 -61.2
8 -72.2 -55.9 -61.3
9 -68.5 -55.6 -61.1
10 -64.8 -55.8 -60.9

LO Rejection at RF Output, LO = 7dBm, IF Output terminated by 50 ohm:
(formula used: 7dBm + measured level of LO at RF Port)

LO – RF Separation ADE-1 1N4148 1N5711
LO Level [dBm] [dB] [dB] [dB]
-5 -61.7 -62.8 -66.1
0 -59.3 -58.4 -55.8
5 -57.6 -53.6 -51.8
7 -57.2 -54.6 -50.8
10 -56.8 -56.7 -49.8

The conversion loss has been measured by applying RF = -20dBm from an oscillator with an internal impedance of 52ohm.

The voltage at the LO Input of the mixers was measured with an oscilloscope, probe set on 10:1 to reduce overloading caused by probe capacitance.

When the level of –20dBm was applied to home made mixers with diodes 1N4148 and 1N5711, the actual voltage measured at the RF input of the mixer was 135mVpp (-13.1dBm) due to impedance mismatch.

When the level of –20dBm was applied to ADE-1 mixer, the actual voltage measured at the RF input of the mixer was 100mVpp (-16.0dBm) due also to lower impedance mismatch.

However, the conversion loss was calculated based on the theoretical set level of –20dBm.

It appears the mixer built with 1N5711 shows very similar conversion loss as ADE-1 mixer, despite the home made mixer having an input impedances significantly higher than ADE-1 mixer.

It is interesting to note that for LO levels above 5dBm, the diodes 1N4148 appears to have slightly better conversion loss than the diodes 1N5711.
This is probably due to lower Ron for 1N4148. The best conversion loss was measured with a HSMS-2829 quad ring diode made by Agilent.

Conversion Loss
Conversion Loss
LO Rejection at IF Port
LO Rejection at IF Port
LO - RF Separation (IF terminated by 50 ohm load)
LO – RF Separation (IF terminated by 50 ohm load)

The equipment used:

Rohde & Schwarz Spectrum Analyzer + Tracking Generator  FSAS (100Hz – 1.8GHz)

Rohde & Schwarz Signal Generator SMIQ03B (300kHz – 3.3GHz)

Tektronix Oscilloscope 2236 with frequency counter, 100MHz

Home made LC oscillator, used to provide 4.26MHz, –20dBm, output impedance 52ohm.

All equipment is uncalibrated, but believed to have reasonable accuracy (±0.8dB absolute accuracy error, ±0.3dB indication linearity error for spectrum analyser and signal generator).

Double Balanced Mixer was made using ferrite baluns from Jaycar Australia. I used 10 turns of trifilar windings 0.25mm enamelled copper, which provided an inductance of approx 73uH for each winding.

The next images show the spectrum of DBM with ADE-1 and 1N4148 mixers. There is a table with values and description of each peak.
There is a different value for the marker in the image compared with the table, because I performed multiple sets of measurements at different times.

Marker Id

IF Out [MHz]

ADE-1
[dBm]

1N4148
[dBm]

1N5711
[dBm]

Comment

1

4.26

-69.00

-47.65

-47.62

RF leaking into IF

2

5.74

-68.93

-67.61

-68.32

LO – 2 * RF

3

10.00

-26.47

-25.76

-25.99

LO – RF

4

14.26

-62.91

-39.42

-35.69

LO leaking into IF

5

18.52

-26.50

-24.79

-25.58

LO + RF

6

22.78

-69.26

-72.86

-73.07

2 * RF + LO

7

24.26

-87.54

-52.17

-52.30

2 * LO – RF

8

27.03

-82.54

-81.40

-83.68

3 * RF + LO

9

28.54

-53.44

-73.40

-52.07

2 * LO, second harm. of LO

 

Note that during multiple measurements after reinserting the two sets of diodes (1N4148 and 1N5711), Marker 4 showed levels of -40dBm… – 60dBm for a LO = 7dBm when testing 1N4148 (LO rejection 47…67dB).

I was able to prove by swapping the diodes in various positions that the values I wrote down are in fact correct. People should be aware that when diodes are not matched very well, significant differences in LO rejection can appear when swapping the diodes.

Double Balanced Mixer with 1N4148 - IF Output
Double Balanced Mixer with 1N4148 – IF Output

Next image shows the same screen, for mixer with Germanium diodes 1N60P.

The semiconductor in these diodes is germanium, but they have a metallic point touching the surface of germanium, therefore the junction is metal-germanium, which makes them Schottky diodes.

I sorted them to indicate same forward voltage when measured on diode test on my Fluke multimeter (0.192V).

I notice that the LO level (marker 4) is pretty high, at -45dBm. Probably this happens because the LO Level of 7dBm is too much for the germanium diodes.

If the LO level is decreased to -5dBm, marker 4 gets down to -50dBm (same as for LO = 7dBm applied to 1N4148).

If we calculate the drop on two Silicon diodes in series = 1.4Vpp ( = 0.5Vrms = 7dBm) and the drop on two Germanium diodes in series = 0.4Vpp ( = 0.141Vrms = -4dBm),

it becomes clear that this level is sufficient for proper operation of the mixer.

RF Spectrum of Double Balanced Mixer with 1N60P
RF Spectrum of Double Balanced Mixer with 1N60P

Next image shows the same screen, but for ADE-1 mixer. Note the better rejection of  RF leaking into IF (marker 1) and also better rejection of LO leaking into IF (marker 4) compared to images of home made 1N4148 mixer.

RF Spectrum of Double Balanced Mixer with ADE1
RF Spectrum of Double Balanced Mixer with ADE1

I also performed a measurement with the home made mixer, without any diodes inserted.

With LO = 7dBm and RF = -20dBm, I was able to measure at the IF Out:

4.26MHz -> level = -44.27dBm (RF leakage into IF)

14.26MHz -> level = -57.27dBm (LO leakage into IF)

8.54MHz -> level = -79.00dBm (second harmonics of RF)

This means the leakage of RF and LO are not due to imbalance of the diodes, but rather to the length of connection wires or tracks and asymmetry of the balun transformers.

Based on these measurements, I consider 1N4148 are excellent for building RF mixers in HF range when a LO level of 7dBm is available.
It is quite easy to build RF mixers at home that give similar or even slightly better conversion loss as ADE-1 mixers.
For a good LO and RF Rejection at IF port and LO-RF separation, it is important to keep connections short and build the transformers with attention.

In the next image, we show the IF Output due to a significant imbalance of the diodes (3 x 1N4148, 1 x 1N5711). Note Marker 4 (LO leakage) and Marker 9 (2 * LO) are really high.

RF Spectrum of Double (Un)Balanced Mixer with 3 x 1N4148 and 1 x 1N5711
RF Spectrum of Double (Un)Balanced Mixer with 3 x 1N4148 and 1 x 1N5711

In the image below you can see the physical construction of the mixer. Later I discovered that due to the length of the connection wires to the BNC connectors, the performance of the mixer has worsened considerably.

Physical construction of the Double Balanced Mixer
Physical construction of the Double Balanced Mixer
Physical construction of the Double Balanced Mixer
Physical construction of the Double Balanced Mixer (2)

As a generator for RF In, I used a home made oscillator, which can be seen in the photo below.

Variable Frequency RF Oscillator used as LO
Variable Frequency RF Oscillator used as LO

The amplitude is only -20dBm, which I adjusted by varying the power supply voltage. Here you can see the output is clean (second harmonics down 45dB), especially considering there are no filters at the output.

Of course, the low output level makes this job much easier.

RF Spectrum of the Variable RF Oscillator
RF Spectrum of the Variable RF Oscillator

This is the output of the oscillator used as a RF source when it is connected to the RF In of mixer. The first image is with the LO turned off!

Oscillogram of the Variable RF Oscillator
Oscillogram of the Variable RF Oscillator

Next, you can see what happens to the RF input when LO is turned ON. I believe that for a better LO-RF isolation, the effect of LO onto RF is smaller.

Output of Variable RF Oscillator when the LO is turned ON
Output of Variable RF Oscillator when the LO is turned ON

Third order intercept measurements will be performed later.

===================================================

1dB Compression Point

For 1dB compression point, I used a home-made crystal oscillator, 15MHz, 7dBm. The impedance of the oscillator is 34 ohm (I will add a 15 ohm resistor in the future to bring it close to 50 ohm). I adjusted the level for the LO by varying the voltage supplied to the oscillator, until I obtained a conversion loss of 6.2dB on the mixer with 1N4148 diodes.

As RF Source, I used a signal generator SMIQ03B, set on 5MHz. During the measurements I discovered that 15MHz is a multiple of 5MHz so I shifted the frequency to 4.9MHz.

For level measurements, I used the spectrum analyzer FSAS.

Problems I had are related with linearity of the Spectrum Analyzer, linearity of Signal Generator and Level and Output Impedance of Local Oscillator.

The Signal Generator has a step attenuator that changes in increments of 5dB. To generate a level in between the points, an electronic gain control is used, which inherently has lower accuracy.

The mixer in the Spectrum Analyzer is affected by its own compression and I noticed that when the mixer IF Output level was getting closer to the Reference Level set on the Spectrum Analyzer, there was some compression taking place.

To prevent this, I changed the Reference Level from 0dB to 10dB, but I should repeat the measurements while documenting where I changed the Reference Level.

The Level of LO was set by adjusting the power supply voltage to a level of 7dBm, as measured by Spectrum Analyzer. However, when connected to the mixers (of unknown impedances), the level was different and this means that the performance of ADE-1, compared with the performance of the other three mixers could be misleading. Also, the 1N5711 appears to have larger losses than 1N4148 and the difference seems unreasonably high.

So, with these comments, consider the values below as temporary, later I will repeat these measurements and I will pay more attention to all the things discussed.

LO = 7dBm (*)
1N4148
1N5711
HSMS-2829
ADE-1
Comments
RF Level [dBm]

[dB]

[dB]

[dB]

[dB]

-30
-36.20
-36.96
-36.16
-37.22
-20
-26.20
-26.93
-26.05
-27.22
-10
-16.19
-16.95
-15.99
-17.21
Used as Reference Level
-9
-15.18
-15.94
-15.03
-16.20
Sig Gen fixed attenuator changed
-8
-14.22
-14.95
-14.06
-15.21
-7
-13.23
-14.01
-13.10
-14.24
-6
-12.23
-13.05
-12.08
-13.30
-5
-11.19
-12.03
-11.04
-12.29
-4
-10.13
-10.97
-9.95
-11.24
Sig Gen fixed attenuator changed
-3
-9.19
-9.87
-8.93
-10.18
-2
-8.23
-8.91
-7.97
-9.16
-1

-7.16

-7.92
-7.03
-8.22
0
-6.19
-6.88
-6.09
-7.33
Changed Ref Level to 10dBm on Sp An (?)
1
-5.28
-5.97
-5.28
-6.62
Sig Gen fixed attenuator changed
2
-4.24
-5.05
-4.46
-5.99
3
-3.19
-4.22
-3.78
-5.53
4
-2.18
-3.45
-3.32
-5.07
5
-1.26
-2.82
-2.84
-4.59
6
-0.43
-2.36
-2.33
-4.08
Sig Gen fixed attenuator changed
7
0.33
-1.90
8
0.81
-1.50
9
1.16
-1.12
10
1.44
-0.79

 

The charts below represent my first measurements, using a LO with 34 ohm impedance and the spectrum analyzer uncalibrated.

1dB Compression Point
1dB Compression Point

After doing these measurements, I found the document http://www.minicircuits.com/app/AN00-010.pdf and it appears that for diode mixers level 7 (which means LO = 7dBm), the 1dB compression is 1dBm.

For ADE-1, my measurements indicate 2.4dBm, which is 1.4dB higher. I suspect this could be caused by the higher level applied at the LO port of the mixer.

I already added a resistor to the LO module to compensate the internal impedance to 50 ohm and I performed the measurements again.

Here are the new results (and I am much happier with them). The 1dB compression point is about 1dB lower than in my previous measurements.

LO = 7dBm (*)
1N4148
1N5711
HSMS-2829
ADE-1
Comments
RF Level [dBm]

[dB]

[dB]

[dB]

[dB]

-40
-45.54
-45.76
-44.82
-46.15
Ref Level = -30 dBm
-35
-40.54
-40.76
-39.88
-41.15
Ref Level = -25 dBm
-30
-35.54
-35.82
-34.88
-36.15
Ref Level = -20 dBm
-25
-30.54
-30.82
-29.90
-31.15
Ref Level = -15 dBm
-20
-25.54
-25.76
-24.85
-26.12
Ref Level = -10 dBm
-15
-20.56
-20.84
-19.90
-21.17
Ref Level = -5 dBm
-10 (Ref)
-15.51
-15.76
-14.82
-16.12
Ref Level = 0 dBm
-9
-14.57
-14.76
-13.82
-15.12
Ref Level = 1dBm
-8
-13.51
-13.75
-12.82
-14.12
Ref Level = 2 dBm
-7
-12.51
-12.74
-11.82
-13.15
Ref Level = 3 dBm
-6
-11.49
-11.74
-10.82
-12.15
Ref Level = 4 dBm
-5
-10.49
-10.74
-9.82
-11.15
Ref Level = 5 dBm
-4
-9.46
-9.74
-8.80
-10.12
Ref Level = 6 dBm
-3
-8.46
-8.74
-7.80
-9.12
Ref Level = 7 dBm
-2
-7.46
-7.74
-6.82
-8.12
Ref Level = 8dBm
-1
-6.49
-6.76
-5.82
-7.17
Ref Level = 9dBm
0
-5.46
-5.74
-4.85
-6.32
Ref Level = 10dBm
1
-4.56
-4.92
-4.10
-5.65
Ref Level = 11dBm
2
-3.61
-3.87
-3.43
-5.11
Ref Level = 12dBm
3
-2.66
-3.15
-2.94
-4.69
Ref Level = 13dBm
4
-1.74
-2.50
-2.55
-4.35
Ref Level = 14dBm
5
-0.87
-1.98
-2.26
-4.09
Ref Level = 15dBm
6
-0.15
-1.57
-2.01
-3.80
Ref Level = 16dBm
7
0.42
-1.23
-1.81
Ref Level = 17dBm
8
0.80
-0.91
Ref Level = 18dBm
9
1.05
Ref Level = 19dBm

 

The following is a chart for Compression Point 1dB with improved accuracy.

1dB Compression Point with Improved Accuracy
1dB Compression Point with Improved Accuracy

The mixer with 1N4148 has a 1dB compression point of about 6.7dBm, while the same mixer with 1N5711 has a 1dB compression point of 4.5dBm.

The improvement of 2.2dB comes because of the larger dropdown voltage on the Silicon P-N junction than for the Schotcky Silicon-Metal junction.

For an even higher 1dB compression, two 1N4148 diodes in series could be used for each branch (a total of 8 diodes).

Of course, the 7dBm LO level will have to be increased accordingly (from 7dBm to 13dBm).

===============================================

Oscillator used as a LO for 1dB compression point measurements

to be continued…

 

10 thoughts on “Double Balanced Mixers

  1. Fascinating. I occasionally do these type of measurements, but not as thouroughly as you did!
    I plan to build a couple of diode-ring DMBs soon, but wondered on their performace w/ audio modulation – for DSB generation.

    1. Hi Rick,
      Congratulations and thank you for the first comment on the website! 🙂
      It would be interesting to find out if these mixers are causing any distortion of the signal. The only way I can think of measuring distortion is to apply a high quality AM signal and then demodulate the signal and measure the distortion. In fact, we have at work a modulation analyzer that does both functions, it can generate 10 MHz, -10 dBm with variable AM and AF and also demodulate the signal and read the distortion, function of AM or AF. It’s a high quality instrument.

  2. We are building a new SDR application and a simple up-converter, easy to be build by Makers.
    During tests with mixer diodes (4148) and NE602 it appears that the NE602 works absolutely better. Maybe there is something wrong in our test components…
    Another possibility is that the dynamics of the receiver we use (RTL 2835 U) , is so limited (8 bit only = 60dB) that the dynamics of the mixer is not important. And probably the gain (17 dB) of the NE602 compared with the loss (6 dB) of the mixer makes a great difference in our circuit.

    What do you think about?

    Can you please write the value of the inductances used in the ADE-1 mixer (this will help our Spice simulations)?

    bye and thanks for your work
    Livio

    1. Hi Livio,
      I haven’t made any measurements yet with NE602 (I think I have some of these chips). Your tests involving a system using both a mixer and a receiver may not provide clear information about the dynamics of the mixers used. The fact NE602 offers a gain could be the reason. You could add a 23 dB attenuator at the output of the NE602 (you could make a PI attenuator using SMD resistors, there are calculators on the net, provided you know the impedances).
      I will measure the ADE-1 inductances and publish it.

    1. Hi David,
      Thanks for notifying me about missing images. I haven’t visited my website in a long time. I’ll try to fix it, if I remember how 🙂

      1. Hi Nicolae,
        Thank you for your answer.
        I hope you can fix this issue, your article is really interesting for people working on experimental diode ring mixer.
        Regards,
        David.

  3. Had to re-build ring mixer in Yaesu FT7B, used old 1n914’s, works as good as original Shottky’s!
    Brian Williamson M6BAQ

Leave a Reply to Nicolae Cancel reply

Your email address will not be published. Required fields are marked *

CapTitle * Time limit is exhausted. Please reload CAPTCHA.