Thoughts on a Cost-Conscious Bode Plotter

In my quest to design and build high frequency devices, one useful piece of kit would be something that can measure the frequency response of a bit of circuitry.  I think that this is one of the main functions of a device called a Network Analyzer -- but those are quite expensive tools.

It seems interesting to try to build something that performs just this one task, and (through this focus) keep the costs down.

So the idea is:

1. Generate an accurate sine wave in the range of (say) 1 kHz to 100MHz or 200 MHz.  I don't have the chops yet to deal with signals in the GHz range, so I'll start with a more manageable frequency limit.
2. Amplify the signal to a selectable voltage level.  Probably 5V - 10V is sufficient as a maximum, and for a minimum, whatever I can manage (ideally maybe 10mV or so).  For this purpose, it seems like being able to deliver the signal into a 50 ohm load is good enough, so 100-200 mA of current seems like a good goal.
3. Deliver the signal to the Device Under Test.
4. Probe the DUT at a desired point and deliver the voltage signal back to the test unit.
5. Measure the peak and/or RMS of the return signal voltage.
6. Perform this test repeatedly with different frequency values, under processor control.
7. Display the resulting plot

 Each stage has its own challenges and requirements.

Sine Generation

After an extensive datasheet-reading expedition, I settled on the Analog Devices AD9951 400 MSPS Direct Digital Synthesizer chip.  It comes in a hobbyist-friendly LQFP package and can be had for $17 on eBay (I just ordered two).  The on-chip frequency tuning and 14-bit DAC should produce a nicely-accurate wave, and serial control lines should be pretty simple to control from a MCU.

Amplification

It shouldn't be too hard to find an appropriate op amp.  Researching the choices is my next task on this project.

Delivery/Return

A 50 ohm transmission line (using coax cable) should be able to deliver the signal with sufficient fidelity and current.

Measurement

A complete solution would incorporate an ADC with at least 400 MSPS... that approach would allow the measurement of phase information and make measurement of both peak and RMS values pretty easy.  But I'm not really ready to attempt such a high frequency ADC yet (and it would add significantly to the cost).  A future version of this tool might take this path.

For now, it seems to me that it should be enough to do one or more of the following:

• Smooth the signal with a filter and sample the result at a low rate.
• Build some sort of high and low peak detection/hold circuit and sample the output of that.
• Use some sort of op-amp-based integration scheme to sum up the return signal.

Control

A general issue -- most of the projects I'm working on need MCU or CPU support (with varying levels of capability).  I will write a post soon about the choices I have made along these lines.  At any rate, it is not a huge hurdle.

Display

It would be good to be able to save the result for offline analysis if desired (maybe on an SD card or else transmitted over a USB link to a PC).  Besides that, the display aspect is another general issue, as most of the projects I'm interested in require some kind of decent display.  I will be making lots of posts about displays in the coming months, as it is something I spend a fair amount of time on and will continue to do so until I have a display system I am happy with.

More to come!