The Mixer-Amplifier Board takes RF signals from a computer graphics card and upconverts, filters, and amplifies it to the frequency and power range needed by the AOM.

For example, A computer typically outputs signals between 0 to 200 MHz between 0 to -10 dBm. The LiNbO3 AOMs developed by our lab typically operate between 350 to 600 MHz and +25 to +30 dBm (depending upon the color and device design).

Originally, all of the components for three channels (for red, green, and blue) were included on a single large board, like this:

I (Drew) cannot state sufficiently how frustrating it was to work with that kind of layout, because if one or more stages hadn't been designed properly out of ignorance or mistake, and the board failed to function properly, or needed to be updated, the entire board was wasted. Fab houses almost always produce multiple copies of the same board per order, which means those other boards are also wasted. This is expensive both in terms of money, man hours, and fabrication/shipping time. It was also a nightmare to track down what stage had been designed or assembled incorrectly (ie shorts, opens, wrong part…). I've spent several years of headbanging on this board to get it to work well, which has only happened after analyzing and scrutinizing every stage, datasheet, schematic, and square millimeter of layout involved in this board. Many times, not just with this project or this lab or this university or with electronics, too much is attempted by an employee or a manager all at once, and when things go wrong, the incorrect effects will be nearly indiscernibly convoluted with each other, and much time and effort has to be wasted to disassemble the system and then find and fix the problem.

(This happens all the time with coding, where “big-bang programming” describes the scenario where a lot of code is written at once, none of it is tested in stages, and several confusing errors result that make it difficult to track down the mistakes.)

To address much of these headaches, the Mixer-Amplifier Board was broken up into single-stage boards. Each of these boards, or segments, has one purpose, and is meant to be as simple as possible, with minimal added features. Segmenting the Mixer-Amplifier board also allows for iterative improvements or fixes without disposing other working stages. This, along with some other developments and lessons learned, has enabled us to make the Mixer-Amplifier board work exactly how we want it to, with minimal compression or harmonic artifacts at the target frequency and power range.

The goal of these segments is to make them easy to assemble, debug, and replace, while also being flexible enough to be used in a variety of different configurations.

These segments are designed to be either modular (by soldering SMA cables on the ends), or monolithic (by directly soldering the boards to each other, forgoing the SMA connectors). Almost all of them have a width of 0.76“ inches, which is the height of a typical TO-220 heatsink pad. This width was chosen so that TO-220 heatsink pads could be used with the HELA-10 and PHA-101 amplifiers (which require heatsinks) without forming any gaps, instead of using messy thermal paste. These pads are already cut to size with holes, thus saving us some assembly effort.

Most segment designs have a 0.01” gap between the copper planes and the top/bottom edges of the board to separate the grounds between parallel channels. This is to help prevent cross-talk between channels and current loops from forming. This gap is not on the left/right edges so that boards can be soldered together left-to-right.

The below image shows examples of successive stages soldered in series on the left, and bolted in parallel on the right.

An example of the TO-220 thermal pad usage is shown below:

Microwaves101.com is a good RF reference site.

A brief list of such:

• RF - “radio frequency”; electrical signals that propagate. When used to describe a circuit board, RF properties have to be taken into account.
• IF - intermediate frequency; typically the lower frequency going into / coming out of a frequency mixer / upconverter / downconverter
• dB - relative gain
• dBm - absolute power level; relative to 1mW
• SMA - the type of connector we use for almost all RF devices; 50-Ohm system
• SMA-RP - be careful not to accidentally buy parts with these connectors; we almost never need them
• Attenuator - decreases the signal strength by a fixed amount, usually labeled in dB
• Termination - the real impedance (resistance) with which a transmission line is terminated
• Impedance Matching - device or process of matching the unequal characteristic impedance of two different systems
• P1dB Compression Point - effectively the maximum output power of an amplifier
• Spectrum Analyzer - shows the frequencies and amplitides in a signal, but not phases or time domain
• Network Analyzer - shows a variety of system properties, such as complex impedance, passive/active frequency response, reflection, etc…
• Signal Generator - generates specific signals based upon input parameters

DAC Frequency Response Equalizing Techniques

• P1dB Compression Point
  
• P3dB Intercept

Difference between 3rd Order Intercept and P1dB

Understanding Intermodulation Distortion Measurements

Short-Term:

• 75-50 ohm impedance matching
  • boards fab'd, not yet assembled (Wed 15 Feb 2017)
  • 
• Software Equalization
  • code working, process not documented (Wed 15 Feb 2017)
  • 

Long-Term:

• FPGA solution to:
  • Digital-Only Graphics Cards
  • blanking
• reintegrate MixAmp segments into monolithic board?

Do not use thermals on RF components.

Alternate which side components go to ground in filters.

Place vias as close as possible to RF component grounds.

All SMD parts are 0805 size to be the smallest hand-solderable.

Solder mask should be kept to a minimum in RF areas (per observations of satellite dish receiver PCBs).

A small solder mask barrier should be put around components to prevent solder shorts / spillage.

Digital and Analog ground planes should be kept separate from each other. See http://www.analog.com/en/analog-dialogue/articles/staying-well-grounded.html.

Don't make sharp turns when routing RF circuit paths.

Use appropriate trace widths for RF paths: (calculated using Saturn PCB Design ⇒ Conductor Impedance ⇒ Coplanar Waveguide)

RF Segments should be 0.76“ wide to match the height of thermal pads, so that there's no gaps between thermal pads under the PHA-101 / HELA-10 amplifiers when they're mounted in parallel.

Use J502-ND-142-0701-881/886 SMA connector in con-coax library.

Eagle CAD RF Routing Checklist: ( clarify)

• 4 mil width for ground planes
• 8 mil clearance to all signals
• 0 mil distance to dimension (edge)
• use tRestrict to prevent extra ground copper
• keep .01” gap in copper from channel edges
• use t/bStop to erase solder mask
• don't have solder mask over RF traces
• consider RF return path
• consider using polygons for traces instead of thick lines
• place ground vias very close to RF returns of components
• place ground vias 0.15-0.20“ from each other
• don't use thermals
• use reflection corners or curves in traces
• have no traces providing an easy RF return path
• use solder mask only to wall in solder
• check positioning of part names
• make sure that large/non-RF/post-reflow parts have thermals
• use inductors with highest Q factor
• take note of inductors' self-resonant frequency
• alternate sides of grounds of filter components

Use Roger's PCB material? Not necessary under 1 GHz.

Use DigiKey RF VCOs? They're more expensive.

The broadband 75-Ohm-to-50-Ohm impedance matcher was calculated using the Chemandy Matching PI Attenuator Calculator.

For 75 Ohm in to 50 Ohm out impedance matching, use the following values:

A DC-Blocking capacitor was added after the impedance matcher to allow the graphics card to detect a load (by passing a DC current through its Analog Red/Green/Blue signals) but also to block the graphics card's average 0.35V DC component from passing the frequency mixer's carrier tone to the frequency mixer's output. The NVIDIA Quadro FX 5800 is capable of detecting an analog load up to 200 Ohms.

The Voltage-Controlled Oscillator (VCO) is used to create a single frequency by which GPU signals are upconvterted. In other words, the VCO acts as the Local Oscillator (LO) to generate the carrier frequency which is used by the frequency mixer to upconvert GPU signals.

We use a VCO as the LO because the frequency design parameters of our AOMs vary between batches. If these design parameters become mature and fixed, we can used a fixed carrier frequency generator instead.

We use mainly MiniCircuits VCOs (the ROS-535, out of tradition), although there exists the possibility of using crystals / oscillators / programmable oscillators / digital clock sources / resonators as the local oscillator.

All frequency generators have (integer) harmonics. A low-pass or band-pass filter is needed to reduce these harmonics so they themselves don't become additional carrier frequencies in the mixer. One reference project (its design iterations) preferred to use a Bessel bandpass filter for minimal phase distortion “to keep the LO signal as clean as possible”, but it's currently (28feb2017) not understood why the phase of harmonics matters after their amplitudes have been attenuated. Until this is understood, a Butterworth low-pass filter is deemed acceptable. The cutoff frequency will depend upon the frequency range the AOMs operate in.

In the “Local Oscillator CK605 28feb2017” board, CK605 refers to the MiniCircuits package for the ROS-535 and ROS-625 VCOs. The board was designed to accommodate either. R2 is used to reduce the input Vtune voltage (+24V, used by other components) down to levels safe for the VCO. 4.3k can be used for both VCOs listed above, as this will only lower the maximum output frequency a bit. If the ROS-625 is used, 9V could be used for Vcc, as that bus already exists for the PHA-101 amplifier.

The frequency response of this board and the ROS-535 VCO is shown below (built and tested Thu 9 Mar 2017):

This splitter is meant to split the carrier signal three ways for the red, green, and blue channels. It's designed for the MiniCircuits SCA-3-11, with a bandwidth of 100 to 940 MHz and an attenuation of about 6dB. An amplifier (such as an attenuated GVA-81) needs to be placed between its input and the carrier generator to maintain the same power level. It has a maximum power input of +27dBm.

This board is designed to connect to Frequency Mixer boards that are parallel to each other, with a width / pitch of 0.76”. These SMA Male-Male Angled Adapters can be used to connect to the Frequency Mixer boards.

An example of its use is the following:

MiniCircuits JCPS-8-10

10dB loss, +27dBm max input power, $40.00

Frequency Mixing, upconversion, or heterodyning is the process of shifting (up) a range of frequencies from [0,B] Hz to [C,C+B] Hz, where C is a carrier frequency. Frequency mixing can be used to shift frequencies up (upconversion) or down (downconversion). For instance, human hearing (audio) ranges from 20-20,000 Hz, but is upconverted with a transmitter to radio frequencies greater than 540kHz (the lower end of AM radio) to easily propagate, and is then downconverted with a receiver back to normal audio ranges, more or less.

Simple upconverters actually generate two sidebands: an upper sideband (“USB”, [C,C+B]), and a lower sideband (“LSB”, [C,C-B]). We remove the lower sideband with either a high-pass filter of cutoff frequency C, or a bandpass filter of cutoff frequencies [C,C+B]. Wikipedia helps to give a description of these sidebands.

Both of the following boards were made for MiniCircuit's ADE-2 frequency mixer, which operates between 5-1000 MHz. It costs $3.15 each.

The following is the best configuration for upconversion:

• RF - input signal to be upconverted
• LO - local oscillator's input carrier signal
• IF - upconverted output signal

There should be no DC bias on any of these signals. Since graphics cards' analog outputs have a 0.35V DC bias, a small series capacitor is used to block this (a pad for this is available on the “decoupler” board below). If this DC bias is not removed, the carrier frequency will be about 12dB stronger than the desired upconverted signal. If it is removed, the carrier frequency will be about 20dB weaker than the desired upconverted signal, which is what we want.

Left SMA port is the input; center SMA port is the carrier; right SMA port is the output.

This board was made to potentially decouple any/all signals. The part values for our application are given below:

Left SMA port is the input; center SMA port is the carrier; right SMA port is the output.

This board was made as a pure breakout for the ADE-2.

On the boards below, shunt components (going to ground) should be placed on alternating sides. IE, capacitor C1 and inductor L1 in a bandpass filter should have their grounds on opposite sides of the signal trace. This is to help prevent the RF return signal of some components from interfering with the grounds of other return components.

The high/low pass boards can serve as either highpass or lowpass filters depending upon which components are installed.

Filters are necessary to remove unwanted signals (harmonics, artifacts, etc…) that would saturate the amplifiers and drop their maximum output power.

A bandbass filter was designed for 350-550MHz, but a limitation of parts constrained this to a simulated passband of 400-600MHz. When this was actually built, it had a passband of 350-525MHz.

* targeting maximum +30 dBm output.

The GVA-84 is a high-gain, medium-power amplifier used to increase the signal levels coming out of the GPU, frequency mixer, and filter. Its job is to get power levels up to those needed by the PHA-101 amplifier.

The PHA-101 is a medium-gain, high-power amplifier used to get signal levels up to the maximum safe input for the HELA-10 without compression (targeting an output of +20dBm but having a P1dB point of +25dBm). Its high power output can allow it to directly drive some of the AOMs we produce. It requires active cooling (a heatsink and air flow). It costs $8.95 each.

Link to development log: sat_22_oct_2016_-_pha-101_amplifier.

We have these in the RF Boards drawer as well:

Both the GVA-## and PHA-101 amplifiers are in SOT-89 packages (MiniCircuits package DF782). The following application circuit, from page 4 of the PHA-101 datasheet works for all of these types of amplifiers:

LED D1 and resistor R4 are used to indicate when power is applied. ATX color code convention should be followed when placing LEDs. The following table shows approximate resistor values to be used in series between Vcc and the LED. The current was subjective to give approximately equal LED brightness.

* not part of ATX voltages / color code
† orange is the color of +3.3V in the ATX code, but nothing we use uses +3.3V.

Resistor pads R1-R3 (see board layout) are available for adding attenuation to bring down the gain of the amplifier (it's easier to create attenuation than to adjust amplifier gain). When no attenuation is needed, place a 0-Ohm resistor across (series) resistor R2.

Chemandy Matching PI Attenuator Calculator

For a 50-Ohm-50-Ohm impedance PI-configuration attenuator, the following resistances will produce the following dB of attenuation:

The above links go to DigiKey product pages for each of these resistors.

The HELA-10 is a low-gain, high-power amplifier used to attain close to +30dBm (1W) output power. Each chip costs about $25 +/- $5 depending on quantity.

We are using the “HELA-10B” configuration for 50 Ohm operation up to 1 GHz.

White papers / more information:

• https://www.minicircuits.com/app/AN60-009.pdf
• https://www.minicircuits.com/app/AN60-033.pdf
• http://www.faculty.ece.vt.edu/swe/lwa/memo/lwa0203.pdf

It must have the following before being turned on:

• less than +20dBm input power
• air flow from a fan
• termination (an RF load)

The connection sequence is as follows:

1. Connect output load
2. Apply +12V power
3. Apply RF input

We've been able to create our own boards for the HELA-10 that have a good frequency response:

Resistor R2 and LED D2 are for power indication, using the LED/resistor table given above.

Link to development log: mon_21_nov_2016_-_hela-10_amplifier_boards.

The MiniCircuits TB-17 module for the HELA-10 costs about $75:

Typical setup of the Spectrum Analyzer, Signal Generator, and computer are as follows:

Review the Instruments section below beforehand.

power on
start freq - 0 MHz
stop freq - 1000 MHz
if you're using an attenuator (attach if you'll be receiving more than +20 dBm)
amplitude - ref lvl offset - 30 +dB
display - display line - 0 +dBm

if needed
power on
frequency - whatever you need
power - if testing mixers, +7.0 dBm, else, whatever you need

Typically you'll need to use the frequency sweeper program located in settings:
separate
VCO frequency set
unused channel amplitudes set to 0

The DVI breakout board makes available the red, green, and blue analog output signals from the DVI-I port on the graphics card. The DAC follows a typical sin(x)/x (sinc) envelope, as roughly shown below. NVIDIA graphics cards have a maximum clock/sampling rate of 400MHz, which we configure them to run at for the HoloMonitor. However, they don't always have a flat frequency response. At 200MHz, the desired signal and the lowest image combine to produce a ~6dB spike.

(0MHz to 1000MHz span, 100MHz/hdiv; top is +0dBm, bottom is -80dBm, 10dB/vdiv)

In this example, we'll be using a carrier frequency of 350MHz at +7dBm (as recommended by the ADE-2 datasheet).

Without a DC block, the DC component of the graphics card signal is passed through about 12dB stronger than our signal.

With the DC block added, the carrier signal drops to being 20dB less than our signal.

In the output of the frequency mixer, we have DAC images/aliases and the lower sideband (LSB) present. We need to remove these by filtering to prevent them from dropping the P1dB compression point of our amplifiers.

Before we begin to add amplifiers, we want to add protection to the Spectrum Analyzer so we don't burn it out (even though the analyzer is rated for +30dBm max input).

Attach a 30dB attenuator to the input of the Spectrum Analyzer, and update the readings by pressing AMPLITUDE → REF LEVEL OFFSET → enter “30 +dBm”. Draw a horizontal line to indicate 0 dBm by pressing DISPLAY → DISPLAY LINE → enter “0 +dBm”. The noise floor will now be at -37 dBm avg.

Use the GVA-84 amplifier (+5V, 24dB) after the filter.

Use the PHA-101 amplifier (+9V, 15dB) after the GVA-84 amplifier. In this example, the power levels going into the PHA-101 are too high (producing around +24.19 dBm output), which causes harmonics to be significantly amplified (up to 0 dBm). Lowering the power input by a small amount (7dB), farther away from the P1dB point will greatly drop (by about 17dB) the harmonic's strength. Also, the PHA-101 will feed into the HELA-10, so these power levels need to be < +20dBm anyways (the HELA-10's maximum input power).

Many of the manuals (PDFs) for the equipment we use is in the J-Drive under J:\groups\holography\GPIB & Equipment Documentation.

START FREQ - set the start frequency of the analyzer window (usually 0 GHz) STOP FREQ - set the stop frequency of the analyzer window (usually 1 GHz)

REF LVL OFFSET - use this option to correct the power level readings when an attenuator is attached to the spectrum analyzer input (set it to “+30 dBm” for a 30 dB attenuator)

DISPLAY LINE - use this to draw a solid horizontal line at a given power level (usually 0 dBm)

There are two trace in the HP 8592A: TRACE A, and TRACE B. These are digital storage arrays for either the live input to the trace or maximum value of the trace. These can be used in several useful combinations, such as measuring / comparing frequency responses while showing live input.

These options are accessed by pressing either the TRACE A or TRACE B buttons.

CLEAR WRITE A - show the live input on TRACE A; clears previous maximum holds
MAX HOLD A - show the maximum power input on TRACE A
VIEW A - freeze TRACE A while still displaying it; inputs won't change it in this state
STORE BLANK A - hide TRACE A; this does not clear it

The above four options apply to TRACE B as well.

TRACE B Menu only:

A XCH B - swap TRACE A and TRACE B to switch which trace the marker measures

The marker, or “seal's ball” in some labs, is used to measure the power at a specific frequency.

This control is activated by pressing the MKR button and either rotating the knob or typing in a frequency on the keypad.

This button prevents input to the last selected variable.

Manual: agilent8648operationandserviceguide.pdf

The Agilent 8648D Signal Generator generates RF signals up to 4 GHz at specified frequencies and powers. It's also capable of modulating this signal with AM/FM/PM modulation, although we never use this feature. Its controls are fairly straight-forward.

To toggle output RF power, press the RF ON/OFF button next to the RF output.

It retains settings over being power-cycled. Its default GPIB address is 25. (or, currently as of Wed 15 Feb 2017).

Included in the picture is the wooden box containing the expensive precision calibration SMA terminators.

Manual: hp_8719c_20c_22a_22c_operating_manual.pdf

Programming Guide: hp_8719c_20c_22a_22c_hp-ib_programming.pdf

GPIB address is 16.

Useful GPIB commands:

FORM4	  outputs data in ASCII format
OUTPMEMO  outputs the trace data in memory
OUTPDATA  outputs live trace data, I think...
IDN?      returns HEWLETT PACKARD,8720C,0,1.60

Gavant Industries adapter alt link

Prologix Adapter

• Manual: prologixgpibusbmanual-6.0.pdf
• The one we have is actually Dr. Shultz's

See the Software Overview for more information on this program.

This program is heavily used to generate tones and sweeps as input to the RF components.

GPIB is a very old (1975) and prevalent communications port for controlling scientific equipment. It is very useful for automating test/control equipment. It is essentially a serial line interface to these devices. In our lab, we use GPIB control to automate the translation motor and signal generator in the AOM Automated Characterization setup, and to record exact frequency responses from the spectrum analyzer for characterizing the HoloMonitor RF boards.

We can use either a PCI-GPIB adapter (usually for the Automated Characterization setup) or a USB-GPIB adapter (usually for the HoloMonitor dark room setup). The PCI-GPIB adapter is controllable in LabVIEW and the USB-GPIB adapter is usable with any computer that can communicate via a serial port (ie, TeraTerm on Windows).

The USB-GPIB adapter in use in the HoloMonitor dark room is a Prologix GPIB-USB adapter. Connect it to the computer and equipment before powering the equipment on, and it'll show up as a COM port (not COM1). Default baud rate is 9800. The “Local Echo” option is useful for seeing what you type in your terminal (in TeraTerm, this is under Setup>Terminal…>“Local echo”). A particular GPIB device is addressed by typing “++addr #”, where # is the GPIB address of the device (given below). Commands not preceeded by a “++” are sent to the addressed device. Commands preceeded by “++” go to the Prologix GPIB-USB adapter itself. For instance, after addressing the HP 8592A (address 18), typing “tra?” will return the values within TRACE A.

When GPIB instruments are being controlled remotely, the control panels are locked up. They are unlocked by either pressing the LOCAL button on their front panel (for both the HP 8592A and Agilent 8648D) or by sending “++loc” to the Prologix GPIB-USB adapter from the serial terminal.

Plug in the Pentalogix GPIB-USB adapter into the HP 8592A Spectrum Analyzer, and the computer. Turn on the spectrum analyzer. Open up TeraTerm or an equivalent terminal program and connect to the serial port (COM# or /dev/ttyS#) that corresponds to the Pentalogix GPIB-USB adapter (ie, COM9). The baud rate can be left at 9800. Turn on “local echo” to see what you're typing and enter the following:

++addr 18   // tell the Pentalogix GPIB-USB adapter to talk to the spectrum analyzer
            //   the ++ prefix sends commands to the adapter
            //   without this prefix, commands go to the addressed device
id?         // ask the spectrum analyzer to identify itself
            //   if you get "HP8592A", everything's working
fa 0mz      // set the start frequency to 0 MHz
fb 1000mz   // set the stop frequency to 1 GHz
tra?        // get the values of TRACE A in dBm; this will return 401 samples
            //   including 0 MHz and 1000 MHz

Manual: prologixgpibusbmanual-6.0.pdf

++addr 18	address GPIB device 18
++addr		get current address
++loc		enable device front panel
++eos		get/set termination characters (see manual)

These start on page 76 of its programming manual: hp_8592a_programming_manual_hewlettpackard-3707_.pdf

Its default GPIB address is 18. (or, currently as of Wed 15 Feb 2017).
This is viewable by pressing CONFIG Menu Button→ANALYZER ADDRESS Item Button.

FREQUENCY
fa 200mz	set lower frequency bound of analyzer to 200 MHz
fa?		get lower frequency bound
fb 800mz	set upper frequency bound of analyzer to 800 MHz
		get upper frequency bound
cf 500mz	set center frequency to 500 MHz
cf?		get center frequency
sp 500mz	set span to 500 MHz
sp?		get span


AMPLITUDE
roffset 30dm;	set reference line (top line) to be +30 dBm
roffset?	get reference line
dl 0db		draw a horizontal line across 0 dBm


TRACES
tra?		get data for trace A
trb?		get data for trace B
a1 or CLRW TRA	clear/write Trace A (page 4-11)					
a2 or MXMH TRA	maximum hold Trace A (page 4-11)					
a3 or VIEW TRA	view Trace A (page 4-11)					
a4 or BLANK TRA	blank Trace A (page 4-11)


MARKER COMMANDS
mf? or mkf?	get marker frequency
mkf 400mz	set marker frequency to 400MHz; mf 400mz doesn't work
ma? or mka?	get power at marker
mkoff		turn off the marker
any marker command except mkoff turns on the marker


CONTROL
hd		hold; remove text from last command/error
stb?		clear SRQ message
id?		returns HP8592A
rev?		returns 871218  (18 Dec 1987 ?)
ip		resets all settings to default state (p. 4-43)
ek		enable the knob while under remote control
done		returns "1" after all commands have completed


DATA FORMATING
mds b or w	set measurement data size to 8 (b) or 16 (w) bits
mds?		get mds value
tdf p or b	set data formatting to human-readable (p) or binary (b); more options on page 4-79
tdf?		get trace data format
aunits dbm	set amplitude units to dBm (page 4-20)
aunits?		get """


MISCELLANEOUS
write a string for fun:
pu;pa 100,100;pd;text%all your base are belong to us%;pu;hd

em		erase video memory (clears temporary drawings)

Manual: agilent8648operationandserviceguide.pdf

Its default GPIB address is 25. (or, currently as of Wed 15 Feb 2017).
This is viewable by pressing the ADRS button, next to the power button.

POW:AMPL -4.9DBM	set the output power to -4.9 dBm

OUTP:STAT ON		turn the RF output ON

FREQ:CW 110.445MHZ	set the output frequency to 110.445 MHz (max 9 digits, 10Hz resolution)

SYST:LANG?		returns "scpi"

*IDN?			returns "Hewlett-Packard, 8648D, 3847M00796, B.04.09"

if you ever get an error message about "reverse power", dig through the LabVIEW code for the Automated AOM Characterizer