|
|
explanations, is next to worthless compared to move capacitor's terminating point. Why? How? Remember Whitehead's dictum:
"Seek simplicity and distrust it."
And while we are at it, Henry Adams observation on simplicity:
"The most deceitful mistress that ever betrayed man."
And of course, Einstein's famous remark:
"Make everything as simple as possible, but not simpler."
The "UltraPath" is too simple. In the article on hybrids in this issue there is the case where the power supply noise was reduced by connecting the bypass capacitor to ground and not the power supply rail. "UltraPathing" this circuit would increase the noise at the output. Furthermore, the output stage does not exist in a vacuum; the previous stages bring their own noise problems.
I have helped friends implement the noise canceling tricks from April and May 99 articles. They now call me, bubbling with enthusiasm, asking if I could "Ultra Path" their single-ended amplifiers for them. When told that for over half a decade their amplifiers all ready had a complete and true implementation of the "UltraPath," their ardor was dampened, not eliminated. "But this is something new," they complained. The name is new. The name seduces. I can imagine some enterprising fellow renaming and re-labeling the 6550 pentode, let's say the "Ultra Plate" for example. Then I will get calls asking for help to replace the WE 300Bs with "Ultra Plates." Maybe I am wrong and Proust was right: renaming, recreates.
Surely, our aim is a quiet amplifier. Achieving that aim is possible from two approaches. The first is the way of the brute force: large chokes, large filter capacitors, and voltage regulation. The second is the way of canceling the noise by artful noise injection, balancing, and nulling. This last approach is not one single topology, but rather a family of techniques and stratagems. The problem is that this practice lacks a name. As Nietzsche pointed out "As people are usually constituted, it is the name that first makes a thing generally visible to them." So here is a name: "Audio Aikido." Like Aikido, force is not met with force, but instead the force is used against itself;
|
|
|
|
Subject: misc. notes
Hi JRB (John, I think). I am writing to say I VERY MUCH enjoy your webzine and to offer a few suggestions.
First suggestion: in your latest issue, page 20, the reality check headphone amp, I suggest using either an LM833 or LM837 instead of the LF157. The first is a dual Op-Amp, the second a quad. These are both
intended to be audio amps. In fact, each section of the LM837 is intended to drive a 600 ohm load, and will drive +/- 40 mA. So in parallel one IC could drive 150 ohms to "full output" meaning within a couple volts of the rails. It would probably drive fewer ohms if you did not need as many volts p-p. Or you could parallel several ICs and drive (at least some) headphones directly with the ICs and skip the output transistors. The LM837s are less than $2 each from Digi-Key. The LM833s just over $1.
Second suggestion: in the same headphone amp, I suggest that the resistor connected to the base of the lower transistor have its upper end connected to the + 7.2 V bus.
Third, in a later article in the issue, instead of using the IRF series MOSFETs use Hitachi 2SK1056,7,or 8 for the N channels, and 2SJ160,1, or
2 for the P channels. These are rated at 100 watts each and are *intended* to be used as audio amps. (What a concept!) NOTE: the pin outs are different than the IRFs. The gate is in the same place, but the
source and drain are reversed. I have a set of the N channels in a Nelson Pass Zen type amp and they sound great. (My opinion, of course) Hitachi also makes some 125 watt units but their data sheets and P/Ns are hiding from me at the minute. Hitachi also makes some lower power FETs 2SK213,4,5, and 6 N channels and 2SJ76,7,8, and 9 P channels. I humbly suggest these N channels in place of the BJTs in the headphone amp mentioned above. But I admit I am a "FET-head". (I am off line as I write this and I think you mentioned some other Hitachi FETs which I think are no longer made.) Yeah, I know, TUBE Cad Journal. I'll get out of the sand box.
Fourth suggestion: As you know, the RIAA equalization curve has two parts, one above and one below about 2 KHz. I suggest incorporating the part above 2 KHz into the feedback circuitry in or around the output stage. Thus for frequencies above 2 KHz any harmonics or noise generated get more feedback (and attenuation) than the fundamental. This feedback might be in the form of an inductor in
|
|
