Q: Many times you see designs with a 4-pole crystal IF filter followed by an IF amp followed by a single crystal filter. I understand the idea that the crystal after the IF amp is to filter out noise introduced by the amp however:
Would it be better to use a wide roofing type filter after the IF amp, or:
Split the 4-pole crystal filter into two 3-pole crystal filters, one before the IF amp and one after it, or:
Would a single tuned circuit after the IF amp be just as good as the single crystal filter?
What are the trade-offs between filtering before and after the IF amp?
A: You are delving into a fairly complex issue: distribution of gain and selectivity in a multistage receiver. A very good reference on this is Wes Hayward W7ZOI's "Introduction to Radio Frequency Design".
But the basic concepts are just that: basic. Reduce the bandwidth and control the signal amplitudes prior to the active stages which have a finite signal-handling capability.
This means that any mixer or amplifier can be driven into its non-linear operating range by high enough amplitude signals. Thus, one must ensure that signal amplitudes are within the linear operating range of the first active device. This is done is two ways:
use filters to reduce the bandwidth of the preceding circuitry feeding the active device to only that required for the desired signal; this keeps unwanted high-amplitude signals from affecting operation..
control signal amplitudes by AGC as needed.
Thus, we commonly see relatively wide r-f filters used ahead of the r-f amplifier and/or first (or only) mixer stage to reduce the likelihood that strong signals can reach the mixer and over-drive it, causing it to produce spurious signals (IMD).
In the early stages of the receiver before the conversion to IF, these filters are commonly called preselectors. They serve the added purpose of minimizing the response of the receiver to image signals which the receiver is prone to receive as a result of the mixer operation. Example: If the signal frequency is 14 MHz, the L.O. frequency is 23 MHz and the IF is 9 MHz, the image frequency is 32 MHz.)
Assuming an LO of 23 MHz, the first-order response frequencies would be LO + IF = 32 MHz and and LO - IF = 14 MHz, the desired response. Thus, 32 MHz is the principal image frequency to be rejected, along with its accompanying noise in a bandwidth equal to the first IF filter bandwidth centered on 32 MHz. The rule is that the image is twice the IF away from the desired signal frequency. Thus the preselector filter(s) must have some minimum required stopband attenuation at 32 MHz to avoid image-response issues.
Once the signal has been converted to IF, gain must be used to make up for the loss in the mixer. So here again are one or more stages vulnerable to being overdriven and creating distortion products. Therefore, their signal levels have to be controlled closely by appropriate AGC circuitry. Usually, a filter whose passband is somewhat wider than the widest main IF filter is placed at the IF output of the first (or only) mixer. This filter is often referred to as a roofing filter.
It is usual practice to set the operating bandwidth of the receiver by one or more filters operating at the IF. Once sufficient amplification has been obtained, the signal is applied to the main IF filter. Again, one must be careful not to drive the filter too hard, if it is a conventional crystal filter, since these too have a signal range beyond which they do not perform as desired.
The filter will introduce loss which must be again be made up for by amplifier stages and again these must have their signal amplitudes controlled to prevent overdriving. Since the signal levels are now rather large compared to those at the mixer input, amplifiers capable of handling large signals must be used. Any amplifier stage will generate noise in addition to increasing signal level. It is the usual practice to follow the IF amplifier(s) with yet another filter to limit the signal- and the noise-bandwidth of the receiver. Note that the filter does nothing to the noise except limit its power through limiting the bandwidth. It does not "filter out the noise" and leave the signal.
This filter usually is not as complex as the main IF filter nearer the mixer output. That is, it may not have as small a shape factor or be as narrow in bandwidth as the main filter. On the other hand, the filter may indeed have a smaller bandwidth in order to further reduce the operating bandwidth of the receiver. For example, the early IF filter may be a 500 Hz filter followed later by a 250 Hz filter for added CW selectivity.
All these remarks concerning filters apply to conventional receivers with crystal filters. In DSP IF receivers, such as the Icom PRO series or the ORION and other recent radios, the stages preceding the analog/digital converter (ADC) stage are operated and controlled so as to ensure that the ADC cannot be driven into saturation (maximum digital count). No particular effort is made to secure a narrow operating bandwidth in the preceding stages since the DSP filtering will accomplish this much better than any feasible crystal filter(s) can.
This is a very complex topic, and I have not done much to reduce the complexity in trying to give you some idea of how to answer your questions. Like most questions in engineering, the answer is "it depends." Many factors bear upon how a receiver is designed, some posed by the user requirements, some by the characteristics of the devices used and some by the laws of Nature.
Finally, for a look at a most elegant receiver design, you might examine the circuitry of the Elecraft K2. Its seemingly oversimplified circuit manages with very few parts and a very simple architecture to provide performance surpassing that of much more complex and expensive radios. You will find there a fairly complex crystal IF filter following the post-mixer amplifier and feeding the IF amplifier which then feeds a very simple crystal filter to the final mixer stage which produces the audio output signal.
Again, Wes Hayward's book, and his new one "Experimental Methods in Radio Frequency Design", are outstanding and I highly recommend their study.
Copyright © 2004, George T. Baker, W5YR
Page created by A. Farson VA7OJ/AB4OJ. Last updated: 11/15/2012