晶闸管电源

Texas Instruments IncorporatedAmplifiers: Op Amps

noise (see Figure 4). For comparison, the THS4012, with a

——

respectable voltage noise of 7.5 nV/√Hzand both current

——

noises of 1 pA/√Hz, is also shown in Figure 4.

Note that the output noise of the THS4012 is the sameas when using the THS3112 with Z = 475 . Again, theseresponses are just like those of a VFB amplifier in the tradi-tionalconfiguration, showing that the basic functionality issound—thereare no differences between a VFB amplifierand this configuration. Figure 4 shows that although usingZ = 1 k produces a very stable amplifier, the output

——

noise is 20 nV/√Hzhigher than that of the THS4012.

Keep in mind that the THS3112 has very low overallnoise but that many other CFB amplifiers will probablyproduce much higher noise. The only way to get aroundthis is if the unity-gain stability of the amplifier requires avery small resistor of, say, only 500 or less. But what ifthere was another way to make the CFB amplifier stableandhave low noise at the same time?

Fundamentally speaking, the circuit needs high impedancewithin the feedback path only at the amplifier’s bandwidthlimit. At frequencies below this point, it really does notmatter what the impedance is, and the amplifier will work

fine. The issues stated previously are alsominimized, resulting in an even bettersystem than one using pure resistors.The first solution that comes to mind isto use an inductor. Inductors have lowimpedance at low frequencies and highimpedance at high frequencies—exactlywhat is desired; but their relatively largesize and high cost are generally consideredprohibitive. An alternative componentthat minimizes these disadvantages andstill functions the same is the ferrite chip.

Testing with ferrite chips used for Z

Ferrite chips have been available for severalyears, are relatively low-cost, and areavailable in verysmall sizes—0402 andlarger. Although several manufacturersproduce ferrite chips, testing was donewith what was availablein the test lab—ferrite chips from Murata’s BLM series.Examining the impedance characteristicsof these ferrites revealed several possiblecomponents that could be utilized.

The first factor in determining the propercomponent was the ferrite’s impedance atthe amplifier’s bandwidth limit. For theTHS3112, this implied an impedance of at least 600 at about 150 MHz to meetstability. This can vary, as the first testresults showed (see Figure 3).

Additionally, the Q of the ferrite chipsvaries from grade to grade. Some have alow Q with a fairly smooth rise to the resonance point that then subsides due toinherent properties and parasitics, whileother chips have a relatively high Q with asharp rise and fall in impedance associatedwith them. Although either style may

meet the impedance requirements, testingwas required to see if this Q had an effecton the circuit. Again, the best way to showthe results was to graph the frequencyresponse of the system, as shown inFigure 5. The responses below 10 MHzwere all identical to the original configu-ration. This figure concentrates on thestability portion of the responses above 10 MHz. For comparison purposes, the681- , pure-resistance response is shown.