YIG Oscillator - Ten Second Explaination

David Straight wrote a good introduction on the basic physics behind the YIG oscillator.

Web blogs in technical communication

Today I read an short article about the utility of blogs in technical communications. I've pasted the article here and the URL of the source. My comments are in blue text.


Link

Weblogs: Fad, or Future?

-- Test & Measurement World, 11/9/2004 10:55:00 AM

Weblogs—or blogs, as the online, Web-based journals are called for short—constitute a true online phenomenon. But what, if anything, do they portend for the future of technical communication?

The Weblog universe, or blogosphere, is exploding. Weblog search service Technorati, for example, now tracks almost 4 million blogs, up from 100,000 just two years ago. But what does this explosive growth mean for the future of technical communication? Can the blogosphere become a valuable conduit for professional test tips and techniques, for instance, or will it remain the domain of amateurs having strong opinions but little useful knowledge?

What information would you like to see a blog provide?

What's your opinion? Have you found any blogs that are helpful in your work? Send your replies—and blog URLs—to rnelson@tmworld.com.

I'm a beginner in blogging and I have not found any technical blogs that taught me something useful in my field (RF/microwave design). My impression of several blogs that claim to be technical were really opinion pieces about technology and those blogs said little about the design process, practical tips, and theory. I created my blog as a way to keep track of little nuggets of knowledge that I have accumulated through experience and studying. If someone else finds my blog edifying, I would be glad.

Digital Down Converters in Receivers

Why the heck do we need a DDC (digital downconverter) in receiver architectures with a DSP backend to do the IF processing?

Based on my textbook understanding of DSP, the analog IF signal should be sampled at 2*fmax, where fmax is the highest frequency signal, to avoid aliasing. Why do receivers need a DDC to reduce the sample rate and do some low pass filtering? Why can't the A/D sample rate be set to the reduced rate and have the DSP chip do all the filtering?

I finally read a clear explaination from R.N. Mutagi's article in RF Design, September 2004.

Analog signal should be oversampled for several reasons.

1. Quantization noise is spread out over wider spectral window (0-0.5*fsample). The total quantization noise power depends on the number of bits in the A/D. Its distribution is uniform across the spectral window. The wider the window (higher fs), the lower the noise level for the same amount of A/D bits.

2. In most receiver architectures, the A/D is preceded by an analog anti-alias filter (AKA prefilter). If we sample at fs=2*fmax, then that analog filter needs to have a steep cutoff at fs. A step cutoff filter is not easy to implement, especially if there are requirements for constant delay (linear phase). By oversampling, the filter's rolloff requirements are relaxed and easier to implement.

The downside of having a high fs is that the DSP may not be able to keep up to do REAL TIME processing. The DSP has to do all its operations on the present signal sample before the next sample arrives. That means all filtering, demodulation or whatever IF processing must be done in time < fs_new="N/M" fs="2fmax">>2*21.4 MHz or fs>>2*1MHz? It's the latter. I'll have a followup post to describe how to apply DSP to bandpass signals.

Noise

Noise has always been a confusing topic. The treatment of noise at the systems level and circuits level seems to be disjointed. The systems level treatment doesn't care about the particular details of the connections between each block. One assumption is that the blocks are all 50 ohms input and output. What happens to the noise when the blocks are not matched to each other? What if the port impedances are complex or pure imaginary?

At the circuits level, I understand the concept of equivalent input voltage and current noise sources. Doing the algebra to get the expression for Vn, In is tough, especially if the frequency response of the circuit is also factored into the analysis. I don't understand how to get the op-amp's equivalent input Vn, In and how those noise sources are affected by the feedback network.

Definition of noise factor

F = total noise power at input of network, including the source / source noise power

If source is a resistor Rs, then F = (4kTRs + Vn^2 + (In*Rs)^2)/ 4kTRs

F = 1 + (Vn^2+(In*Rs)^2)/4kTRs

NF=10 log 10 (F)

perfect noiseless amp
F = 1 , NF=0 dB
noise added by amp is equal to that contributed by Rs
F = 2, NF = 3 dB


Notice In interacts with the source Rs.
Also note that Vn and In are assumed to be uncorrelated. This is a simplifying assumption. The proper, full blown way is to also include a correlation term like (In*Rcorr)^2 in the numerator. If correlation is ignored, then the worst possible NF error is 3 dB. (I think)