REFL2 Prototype rev A.1
REFL2 pcb shown plugged into an evaluation Micro908 system
REFL-2 is a new and improved reflectometer subsystem for the Micro908, providing:
Auto Level Control for the DDS-60 card
Matched & Compensated RF Detectors
Math Co-Processor for advanced function computations
Single card REFL2 replaces existing on-board reflectometer
Easy upgrade …plugs into DDS-60 socket and user adds five wires
Existing DDS-60 Daughtercard plugs into the REFL2 card
Provides improved accuracy, stability and advanced measurement ability
REFL-2 Design Overview
- In the original design we should have nulled the op-amp offset voltage. Failure to do so caused error in detecting low RF voltages. Grebenkemper points out the fact that with an input voltage of 30 mV the detected voltags is only about 450 uV. To minimize error the offset should be a small fraction (perhaps 10% of this). With the LMC6484 we have an offset of at least several mV. Note that the "zero signal" output of the ADC in the 908 is about 04H which is roughly 80 mV, a combination of the offset, errors accumulated in the compensation circuit and ADC error. The MFJ analyzer uses TI TLC274 op-amps which have a spec'd nominal offset well under 1 mV. They do have some error caused by this but apparently chose to live with it since it is fairly small and too expensive for them to correct. Presumably their calibration adjustments (including setting a resistor in the compensation circuit) lessen the effect. Our current op-amps are probably fine.
- Gebenkemper recommends use of matched detector diodes. This will have most effect with low RF inputs where our excessive offset apparently masks any performance improvement.
- Gain errors in our detectors become worse as RF voltage varies, explaining to some extent our increase in errors as output falls off with increasing frequency. A good stable ALC circuit will help minimize these effects.
- Our circuit topography incorporates offset nulling, using a -5V regulated voltage along with one potentiometer and a couple of resistors per detector.
- For best results we do need a compensated Vf detector but it does not need offset nulling since it will always see a relatively high RF input. Surendipitously it also serves as a temp-compensated ALC detector. Naturally we also need detectors for Va, Vf and Vr. The diode detector configuration is symmetrical.
- We are using AvagoTech (nee HP) HSMS282X Schottkly diodes. They have a lower forward drop than the 1N5711 devices, offer half the inherent capacitance and the HSMS2825 is a matched diode pair in an SMT package.
- There will still be some detector error (perhaps 5% max) as can be seen in Grebenkemper's article. That will limit ultimate instrument accuracy thought it points out the fact that going beyond an 8-bit ADC may give no discernable improvment.
- We are using the plug-in card concept again with improved RF layout to lessen any strays present on the current mother board. We'll also use a board-mount BNC connector in the final production run that has less of an impedance bump that the present one.
- ALC ... Automatic Level Control circuitry is used to produce a constant level reference signal for the Reflectometer. The DDS-60 card experiences a roll off in signal level as frequency increases, as predicted by sin(x)/x sampling theory. Further, the response of the video amplifiers on the DDS card also contribute to this signal level roll off . The ALC circuit samples the DDS output signal, and feeds the rectified/filtered signal to an op amp comparator IC. The other input of the comparator is fed by a trimpot whose setting establishes the reference voltage that we ultimately want the DDS output signal to be. The comparator drives an FET transistor that acts as a variable resistor when connected to the Rset pin back on the DDS chip. The value of Rset resistor R10 on the DDS card determines how much signal is produced by the DDS chip, thus completing the closed loop for establishing a constant DDS output level. Ideally, the DDS output level will be set to about 2V pp going into the Wheatstone Bridge.
- Wheatstone Bridge ... We decided to stay with the original basic measurement engine of the Wheatstone bridge. It is computationally straightforward, we can use a great deal of the current computation algorithms, and the technique does not infringe on various other methods in current literature. We felt that if we could eliminate the detrimental effects incurred in the RF signal detection process for Va, Vz and Vr measurement components, we would see significant improvement in measurement accuracy, precision and stability.
- Math Coprocessor ... The Micromega uM-FPU_v3 integrated circuit is a very powerful math coprocessor that is a computational subsystem all unto itself. The main Micro908 processor controls the math chip and transfers data to/from it via the two-wire I2C bus. The math chip uses numerous 32-bit internal registers, RAM, Flash and EEPROM memory to perform integer or floating point computations (both common/low-level ones as well as complex scientific functions) and a lookup table. The coprocessor in the math chip will allow separate algorithmic computations to be performed in parallel with, or instead of, the main Micro908 processor. Performance improvements may be able to be achieved that will offset the extra transfer time for data across the I2C bus. Experimentation and analysis is needed to verify the performance impact/benefits of this math chip subsystem.
An RS-232 level translator chip (MAX233) is also included on the REFL board in order to provide connection of a dumb terminal for debug purposes with the math subsystem. This will be an invaluable tool during the development phase. This RS-232 debugging capability will not be provided on production boards – the MAX233 chip will merely be omitted.
- PC Board Layout ... Special care was taken to orient the Bridge and measurement components very close to the edge of the REFL board and right up against the RF output jack of the instrument. The two pads carrying the output signal may be soldered directly to the back of the BNC connector, thus offering the most ideal short-length connection to the instrument output. (Aside: Probably need to determine a better connection method to the BNC that will allow plug-removal of the REFL board.) Understanding the electrically-sensitive nature of the measurement subsystem, special consideration was also given to the ground plane layout. A single ground “feed point” is located near the center of the REFL board, which connects to four separate ground plane areas: (1) DDS and ALC; (2) Bridge, Log Amps, Video Amps; (3) ADC; and (4) math.
- Power Supplies ... Two TO92-packaged 3-terminal regulators provide 5V power for the REFL2 components. One voltage regulator powers the 5V rail and the other regulator powers the negative 5 volt rail (op amps). The overall current draw of the REFL card is approximately 10 ma.
Adding REFL-2 to the Micro908
Cut the signal and ground
traces at the base of the BNC connector. [Insert photo.]
Insert the REFL-2 board in
place of the DDS card, directly in the 8-pin socket provided for the DDS card
on the Micro908 motherboard.
Insert the DDS card in the
new 8-pin on-board socket. (Prototype boards have the DDS card oriented
perpendicular to the REFL card, while production boards will have the DDS card
oriented parallel to the main board in order to meet the vertical clearance
requirements with the completely enclosed Micro908.)
Solder the two pads at the
far-right end of the REFL-2 board directly to the back of the BNC connector.
Solder seven control wires from the Micro908
motherboard to the supplied 9-pin pinheader socket.
Load the appropriate new
AA908 software that supports the REFL-2 board.
1. Zeroing the op-amp offsets with no DC input.
2. Low frequency (1 MHz) calibration much the same as the current calibration scheme works with open and short circuits, and perhaps load resistors to normalize Va, Vz; and a cal step to accurately set SWR since it is an important parameter in R and X calculations.
3. Re-calibration at several points throughout the operatiing frequency range.
[To be provided]
In case of Troubles
[To be provided.]
[To be provided.]
Back to the Micro908 Kit page
Material and concepts presented on the AmQRP website is Copyright 2005 by the American QRP Club, Inc.
These pages are designed and maintained by George Heron, N2APB (n2apb_at_amqrp.org)
Page last updated: Nov 2, 2007