Tag Archives: Multipliers
In making the datasheets for the first Microlithic frequency doubler (MLD-1640), it occurred to us that not enough has been made about the difference between isolation and suppression.
In mixers and amplifiers, some parameters are expressed relative to the input powers, while some are expressed in terms of the output power, with the conversion loss or gain calibrated out. This includes third order intercept point (IP3), which can be expressed as either input IP3 (IIP3), or output IP3 (OIP3). In general it is better to use OIP3 for mixers, since what really affects the dynamic range of a system is the amplitude difference between the output signal and output spur, expressed in dB relative to the output signal or carrier (dBc). This is illustrated in the table below, where the difference between the T3 and competing mixers is even greater when the superior conversion loss of the T3 is considered.
|T3-05||33 dBm||6.5 dB||26.5 dBm|
|Imitator 1||25 dBm||10.7 dB||14.3 dBm|
|Imitator 2||30 dBm||9 dB||21 dBm|
Note that it is better to use IIP3 in amps, for the opposite reason, namely that you want to give the amp credit for it’s gain. So in parts with gain the appropriate measure is IIP3, while in parts with a loss the appropriate spec is OIP3.
When the same logic is applied to spurious products in mixers and multipliers, the input referred value in dB is called isolation, while the output referred value in dBc is called suppression. Suppression is the preferable number to use, because it expresses the important value to the system. The isolation can always be improved by increasing the conversion loss of the mixer or multiplier, but this is obviously undesirable. There are, however, some issues using suppression.
The first comes with mixers. In all mixers we express the spurious output of the LO in terms of isolation, since it is dependent on the input LO power. Since the LO power does not change the conversion loss referenced to the input, this means that the suppression can vary by several dB with different LO drive levels.
The second complication is that the input signal, converted signal, and spurious tone are all at different frequencies. For example, when using a doubler with an 8-20 GHz input range, the output doubled frequency is 16-40 GHz, and the undesired tripled frequency is 24-60 GHz. This means that the isolation curves look like this:
While the third harmonic suppression looks like this:
The curves look slightly different. The suppression is more stretched out, and distorted by the curve of the conversion loss of the doubler. This is the result of the suppression calculation, which can basically be thought of as a 1:1 mapping of the isolation through the conversion loss.
I’ve posted about this before, but there is a persistent myth in the industry that conversion loss ripple comes from bad mixer VSWR. In this post we will dispel that myth. First we start with a thought experiment: how could we make mixer VSWR create flatness problems? Obviously if the mixer reflects power from it’s RF or IF inputs, then that power could reflect from whatever else is in the source. This would indeed cause ripple, as with any microwave component. In this case, however, the problem is that the source is not well matched. As we will see this is a problem whether or not the mixer is well matched.
Next, we imagine that if the VSWR is bad then internal reflections could cause conversion loss ripple. If this was the case, then it would show up in any sort of testing, in any environment. It would be obvious in the datasheet. This doesn’t happen, though, because there would be no path length delay from the output, which is the second necessary part of a standing wave. There has to be a reflection, and there has to be a line for the standing wave to exist on.
Where does conversion loss ripple come from then? The short explanation is that the mixer is both a load and a source from all three ports. If you look carefully with a spectrum analyzer, you will see various products coming out of each mixer port at all sorts of frequencies. This is the mixers job. At a minimum it will have both the sum and difference frequencies. What happens when the desired difference frequency is well terminated, but the undesired sum frequency is terminated in a reflective load? This is often the case when a low-pass or band-pass filter is used next to a mixer.
Ferenc Marki performed this experiment this morning. He was upconverting 2-8 GHz to 12-18 GHz using a 20 GHz LO. First he terminated all the ports in 50 ohm loads, and then with an 18 GHz low pass filter. Here are the results:
As you can see the wideband 50 ohm system shows a smooth conversion loss, while the filter-terminated system has significant ripple. The frequency of the ripples can be increased by placing a length of cable between the mixer and the filter:
Here is the spectral output of the upconversion, along with the illustration of the problem from Ferenc:
What he is showing is that the high side product is reflected from the low pass filter, which then reconverts to the I port and becomes a new signal to reconvert to the desired product. Since the phase is rotating across the frequency band, it shows up as a ripple and not just a fixed loss.
There are two ways to create high frequency (low phase noise) tones: use a high frequency oscillator to directly synthesize (probably optically) or use a crystal to create a low phase noise signal at low frequency, and then use a non-linear element to multiply it up to the desired frequency range. A signal passed into any non-linear element will create all possible harmonics, and can be used as a multiplier. More sophisticated circuitry can be used to improve the three metrics that multipliers are evaluated on: conversion loss, undesired harmonic suppression, phase noise. For amplified multipliers there is the metric of additive thermal noise/noise figure, related to the conversion loss of the multiplier core.
For odd order harmonic generation, Marki has found that limiting amplifiers (such as our line of LO square wave driver amplifiers) offer a superior tradeoff of conversion loss, harmonic suppression, and noise to comparable amplified multipliers. Odd harmonic generation is created by the ‘clipping’ effect in these saturated amplifiers. It is this same effect that causes square wave generation, and produces all odd harmonics within the bandwidth of the amplifier. For this reason Marki recommends saturated square wave amplifiers for all odd harmonic generation requirements.
An ideal square wave amplifier produces only odd harmonics, while even harmonics are suppressed as much as possible (they represent duty cycle distortion in a square wave). Even harmonic generation requires rectification. This can be accomplished using a single diode, but with poor conversion loss and harmonic suppression. The preferred technique for producing even order harmonics is to use a standard mixer diode core with passive circuitry in a special circuit configuration.
Microwave Doubler Operation
The basic microwave doubler circuit is the microwave equivalent of a low frequency full wave rectifier circuit. It is shown below.
In Marki doublers the hybrid junction is typically a microwave balun, implemented in a similar fashion to our mixer baluns. The baluns perform the important function of canceling odd harmonics, such that only even order harmonics pass through. The balance of the balun determines the cancelation of the odd harmonics and thus the harmonic suppression, while the bandwidth and insertion loss largely determine the bandwidth and conversion loss of the balun.
Odd harmonics appear as a differential signal to the balun, and thus are canceled. This is similar to a double balanced mixer, where even order spurious products are canceled. In this case, however, the input signal acts as both the LO and the RF signals, meaning that a certain threshold level must be met in order to turn on the diodes that make up the ring quad for the doubler core. As with a mixer this threshold is determined by the turn on voltage of the diodes that make up the doubler, and are typically specified as Low, Medium, or Intermediate level diodes. Each doubler operates over a limited input power range for low conversion loss and low additive noise performance. All Marki doublers use silicon schottky diodes, which provide excellent additive phase noise and thermal noise properties.
The time domain operation of a doubler is shown below. In the first scope capture a 1.5 GHz input at 16 dBm is shown.
This is input to a D-0204MA, and the output from this is shown below. Note that the scale is different (10 dB of attenuation is removed, the conversion loss of a typical doubler is 10 dB).
As you can see the frequency is doubled. Since the output is AC coupled, the baseline remains 0V, whereas in a true full wave rectifier there would be a DC component. The wave distortion from an ideal full wave rectifier is caused by imperfect cancelation of odd harmonics (causing the asymmetric nature of the output wave) and the finite bandwidth of the circuit. Note that if the doubler is underdriven, not only will the conversion loss suffer, but the suppression of the harmonics will also suffer, particularly the fundamental leakthrough. This will result in a distorted output waveform, as shown below with the doubler driven at only +9 dBm.
The asymettric waveform is again a result of the unsuppressed odd harmonics.
Generally the output of a doubler is used as a local oscillator downstream in the system. To counteract the conversion loss of the doubler and create a suitable local oscillator an amplifier is typically used after the doubler. Marki amplified doublers offer an application matched amplifier combined with our world class doubler core to provide high output powers suitable for driving a high frequency mixer.
Amplified Quadrupler Operation
Marki quadrupler circuits come in several flavors based on the same underlying principle. The basic principle is just to double an incoming signal twice. To provide high fundamental and undesired (2f, 3f) suppression with high conversion gain across a specified bandwidth an appropriate combination of amplifiers and doubler circuits must be used. Our DAD circuits (DAD-0225, DAD-0405) use and input low level doubler, amplify the output signal at the intermediate frequency, and double the intermediate frequency with a low level, higher frequency output doubler. Our AQA circuits (AQA-1933, AQA-2032, AQA-2040, AQA-2050), however, amplify the incoming signal to a high level, double to an intermediate frequency with a high level, low frequency doubler core, double the intermediate frequency with a low level, high frequency doubler, and amplify the resulting output with a high frequency amplifier. The selection of amplifiers and doublers with appropriate power levels and frequency ranges is the art of amplified quadrupler design, and has been honed by Marki for over 40 years.
As with mixer outputs, filtering of the multiplier output signal is critical. While the circuit can suppress the adjacent harmonics (i.e. a doubler suppresses the f and 3f harmonics while maximizing the 2f harmonic), it is usually powerless to suppress the higher order even harmonics (4f for a doubler). This means that filtering is required for the application the doubler will be used in. Without the use of a tunable filter this will frequently limit the useable bandwidth of a doubler circuit to a single octave. If a doubler can provide an output of 1-4 GHz then the 4F harmonic will start to leak into the band when it is used below 2 GHz. This is why many of our doublers are limited to octave band. Much broader bandwidth multipliers are available, but they may require switched filter banks at the output to maintain harmonic suppression in the application circuit.
The conversion loss of a basic doubler is limited by the fourier expansion of a full wave rectifier. The coefficients for these are given by
where n is all even harmonics. This results in the minimum conversion loss for a doubler of 7.4 dB, and for the fourth harmonic from a doubler of 21.4 dB. This is the minimum theoretical conversion loss for a doubler and suppression of the fourth harmonic.
As mentioned above the suppression of harmonics is determined mainly by the balance of the balun circuits, as well as the balance of the diodes in the ring quad.
Additive phase noise is critical for most multiplier applications. Phase noise is fundamentally degraded in the frequency domain during a multiplication as a result of it being constant in the time domain. For example, a 10 GHz signal with 10 ps of jitter on it will occupy 10% of the period, while if a 20 GHz signal has the same 10 ps of jitter on it, it will occupy 20% of the period. Since frequency multiplication occurs in linear time this means that multiplication will always increase the phase noise by a factor that can be shown to be 20*log10(n) (thanks Chris Irwin!), where n is the number of multiplications (2 for a doubler, 4 for a quadrupler, etc.) and phase noise is expressed in dBc/Hz. Conversely, in a frequency divider the phase noise is reduced by the same factor (less whatever phase noise is added by the divider itself).
As detailed in our Mixer Basics Primer, the isolations of a mixer refer to how well a balanced mixer cancels out each of the arms inputs from the outputs of the other arms. There are three isolations of different importance:
1) LO-RF Isolation: This is most important because the LO tends to be very close to the RF (usually the IF is low frequency). Therefore the LO will be in the band of the following components and contaminate the RF circuitry, whether in an upconversion or a downconversion.
2) LO-IF Isolation: This is also important as it is the worst of the isolations. The LO can contaminate the IF circuitry, especially if it is wideband, or cause conversion loss problems if it is reflected from the IF circuitry.
3) RF-IF Isolation: This is less of a problem because the LO is always the strongest signal in the system.
So how do you determine what the isolations are? For a given frequency conversion it is simple: measure the output of the RF or IF on a spectrum analyzer and measure the leaking component. LO power input divided by output power at the LO frequency at the RF or IF port output (or minus, in dBm units) gives the isolation. This is what matters ultimately for the system.
For making a mixer datasheet, however, things are more complicated. We want to know what the isolation is across the entire operating band of the mixer. The obvious way to do it is to hook it up to a network analyzer and measure the leakage. However, since the mixer is by definition a non-linear element, we have to pay attention to the parameters we choose. The power levels usually used to measure linear devices are much too low to measure a mixer, because the mixer diodes don’t turn on until the rated drive level of the mixer (around +10-+13 dBm for a typical double balanced L diode mixer). Some VNAs cannot produce that much power across the whole band. Fortunately ours generally can. Below are two graphs of the isolations of an M1-0412LA mixer, one with measurement power of -5 dBm, and one with measurement power of more than 10 dBm.
As you can see the isolations can vary dramatically with input power. Increasing the input power by 20 dB in this case can lead to an isolation improvement of 20 dB or better!
The last conundrum is how to measure RF-IF isolation, since the normal operating mode of a mixer is with a strong LO present. The truth of this is that generally the LO and RF ports of a mixer are interchangeable. This is actually a trick that we use sometimes to improve the spurious performance of a mixer (sometimes it works better backwards!). So if you measure the RF-IF isolation the same way as the LO-IF isolation (with a high power RF signal input), then it should replicate closely the actual performance you would see under operating circumstances. In any case, this is the least important isolation for most applications anyways.
In this post we gave a block diagram for how to measure spurs from a microwave mixer. This post is to show some actual data for it. Experimental Setup: I measured an M1-0620NP using two Anritsu synthesized sweepers and an Agilent spectrum analyzer. The power into the mixer was measured using an HP power meter (this is important, as the power will change with the added filters). An IF of 2.2 GHz was input at a power level of -10 dBm, and an LO of 7.4 GHz was input with a power level of 13 dBm. Here are the spurs shown with no additional filtering: Here is the output in the X band (8-12 GHz). You can see the high side output product at 9.6 GHz with a power of -16.52 dBm (6.52 dB conversion loss). There are two spurs, one at the 2*LO – 2*IF at 10.4 GHz with a value of -57.55 dBm (41 dBc) and one at the 1*LO x 2*IF at 11.8 GHz with a value of -68.73 dBm (52.2 dBc). Next we add a filter to the IF. Since this a low frequency we used our DPX-3, which will pass the 2.2 GHz but suppress the 2*IF of 4.4 GHz. Here are the results from that: Fundamental: -15.95 dBm (input power increased .5 dB) 2 LO x 2 IF: -63.8 dBm (improvement of 6 dBc) 1 LO x 2 IF: -62 dBm (improvement of 6 dBc) This is the most important filter, as the IF harmonics tend to be the most prominent. Next we add an LO filter: Fundamental: -15.91 dBm (same) 2 LO x 2 IF: -58.8 dBm (degradation of 5 dBc) 1 LO x 2 IF: -60.4 dBm (degradation of 1.5 dBc) The spurs got worse with an LO filter? This illustrates how touchy spur measurements can be with mixers. The filter at the LO will change the impedance seen by the mixer across many frequencies, which will cause the spur powers to change unpredictably. Ideally we would put an attenuator on each port, which would ensure a good 50 ohm match at all ports. We didn’t do this in this experiment because we didn’t have the LO power to spare. Also things as small as slightly tightening connectors can affect spur measurements. Every little thing counts. Fundamental: -26.37 dBm (power decreased by 10 dB, same as the pad) 2 LO x 2 IF: -70.72 dBm (44.35 dBc, 3 dB better than originally measured, but not the lowest we’ve measured) 1 LO x 2 IF: -71.4 dBm (44.63 dBc, not as good as we originally measured) Here again the spur levels change unpredictably. This is because the ESA front end consists of a mixer as well, and adding the pad changes the way the intermodulation products reflect back and forth in the system. Spur levels are very low; they can be unpredictable in actual systems.
This is our final answer. The good 50 ohm match at the IF caused a significant improvement. Note that the filter may be matched at the IF frequency, but for our purposes it has to be matched at the spurious frequencies, which for our purposes means all frequencies.
With the RF and IF ports matched to 50 ohms and the filters in place for the synthesizers, these are the spurs. Note that these aren’t the spurs that you’ll see in your system, unless your system provides filtering at the synthesizers and a good 50 ohm match at all ports.