Variable Gain Amplifier For AD9850 DDS
I had been mucking around with the AD9850 DDS modules that are available cheaply on eBay etc
and ended up making several hacked-up RF amplifiers in an attempt to get their very low output up to a useable level, but could never get the level right at the desired frequencies, and even after extensive reading of the datasheet, I couldn’t figure out the output impedance, or what the expected output level was, nor why the output became so low at the higher frequencies.
Anyway, I finally decided that it was time to bite the bullet, make some measurements, and come up with an amplifier that would provide the basis for a worthwhile HF signal generator. My requirements being variable gain up to at least 7dBm output up to 30MHz and with 50Ω impedence, this being suitable for driving level-7 diode balanced-mixers.
I’m quite happy with the end result and present it here.
1. AD9850 Module
The circuit is pretty much straight from the datasheet and shows a 70MHz low pass filter with 200Ω impedance on the output. The 200Ω resistor (R5) strikes me as strange and was perhaps shown on the datasheet only to represent the output load, and was not intened to be actually used, but nonetheless it’s there on the module.
I connected a 500Ω pot across the output and made measurements of power into various load impedances. Maximum power was with 100Ω which suggests that this is the actual output impedance. I mucked around with a 2:1 impedance ratio transformer to bring it down to 50Ω but concluded that the improvement really wasn’t worth the effort.
1.2. Output Power
Using the oscilloscope, I measured the module’s output level into a 200Ω resistor as the load. You can see that it declines rapidly with increasing frequency- so much so, that it seems that something is not right.
It was mentioned somewhere that perhaps some of the modules have the wrong value components in the LPF. As the graph shows, the cutoff frequency seems to be closer to 8MHz than to the 70MHz required.
The modules I tested were 18 months or so old, so perhaps this problem has been fixed. I’ll recheck when I get my hands on a newer one.
rudiswiki9 has done some good work on his website here on replacing the onboard filter with external ones, he also notes that the RSET resistor (R6 on the circuit) can be decreased in value to give an output power of closer to 0dBm.
In my case, as I expect to be using these modules quite frequently, I don’t want to be having to modify the modules and building external filters and so decided that it would be better leave the module unmodified and to work with what I’ve got.
2. Hycas Amplifier
This is a hycas amplifier due to it being a hybrid of a FET amplifier connected to a BJT amplifier, the FET being in common-drain configuration and connected in cascode to the input of the BJT which is in common-base configuration. Thanks to ka7exm for his article explaining these amps, and to Vasily for his extensive practical applications on his now defunct PopQRP site.
The advantages of the cascode arrangement are that it gives wide bandwidth due to minimisation of the Miller Effect (w2aew has a great YouTube video explaing this here ), and secondly that the gain can be easily varied by the voltage at the base of the BJT. Furthermore, by using the hybrid, the input impedance can be very high, meaning little load on the preceding stage. All of these things make the hycas ideal for this application.
RV1 varies the gain and the Schottky diode (D1) provides supply polarity protection; even with polarised power connections and multiple checks, I still need it! Cheap protection.
The 470Ω resistor (R4) sets the output impedance, the 9:1 impedance ratio of the transformer lowers this to close to 50Ω and the 3dB pad further ensures that the output is a solid 50Ω source.
R2 sets the input impedance, 10k seemed like a reasonable value to provide high input impedance and yet to provide enough current to not be effected by the gate capacitance of Q1. It worked so I didn’t experiment with other values, I did, however, try placing 100Ω (and 200Ω) resistor across the input to provide an impedance match for the DDS module, but the effect on the output waveform was undetectable (on my admittedly rudimentary measuring equipment) and so it was omitted in the final prototype.
The protoype was made on a piece of double sided PCB, using, appropiately enough, a hybrid Dremel-etching, Manhattan and Ugly construction method. Single sided board was used for the front panel and for the rest of the case.
To ensure a good connection between the top and bottom ground planes, plenty of wire vias were used, especially near to component leads connected to the top ground plane. Also, the the front and rear panels were soldered to top and bottom ground planes.
The panel-mount SMA connectors are handy in this type of constuction as they can be soldered top and bottom, giving them quite some strength. Once the box is finished I glue around the edges to provide for some extra stress relief.
The blue 10-turn pot was used to find the value for R1 setting the maximum gain.
I measured the output power and gain at various frequencies, but noted that at the lower frequencies and with high gain, the waveform would become quite distorted. By using the scope’s FFT function as a poor-man’s spectrum analyser, I was able to make two measurements: clean power and maximum power. It’s a little hard to see, but the purple traces in the images below show the difference in the harmonics between the two.
As the graph below shows, output is at 8dBm or greater up to 20MHz and gain is relatively flat to 30MHz. I suspect that the dip in gain at the lower frequencies may be due to the drive level from the DDS being too high, and hence causing gain compression. Also, the use of the binocular core in the transformer instead of a torroid is maybe suboptimal; I’ll experiment further next time I build one of these amp.
Output Return Loss measured as very good up to 20MHz, and acceptable up to 30MHz,
meaning that this amp provides a solid 50Ω source impedance to the device that it is driving.
All in all, I’m very happy with this amp and can see myself using the same design quite extensively in future.
3. Power Amplifier
For frequencies above 20MHz a further amplifier stage is required. As shown below, I decided on a bog-standard BJF feedback amplifier, as popularised and described by the design in EMRFD.
The prototype was built using a similiar technique as for the hycas amp, with the board being the same size so it could sit nicely under the hycas.
As this amp can output 70mW or more at the lower frequencies, I used a 2N2222A in a metal can and superglued to it a hat made from a piece of double sided PCB to act as a makeshift heatsink.
The blue pot was for experimenting with the value of R6, in an attempt to set the maximum gain, but in the end I settled on the value shown in the original design.
The long component leads and big spacing between the transistor and the transformer probably didn’t help with the high frequency response.
Unlike the hycas, the gain drops off significantly with increasing frequency.
But nonetheless, it gives a relatively clean output of 8.5dBm at 30MHz and maximum of 18.8dBm (75mW) at 6.5MHz, the input and output return loss are reasonable and the amp should be more than adequate for driving level-7 diode mixers at frequencies up to 30MHz.
All things considered, and pending development of decent software for the DDS controller, these amps and the resulting HF signal generator will be a very worthwhile addition to the workbench.
It was interesting to learn about the hycas amplifier and how it works to reduce the Miller Effect to give wide bandwith gain, and especially interesting to see this confirmed in the measurements.
- A cheap and tiny joystick is a very poor way to control a signal generator, especially when the range is 0-30MHz
- Red low power power-indicator LEDs are much easier to see than are green ones.
- My PCB box building technique continues to improve, but gaps remain, although they’re becoming less noticeable.