Re: Design for a small tube/valve mixer

John Byrns wrote:
In article <fnn65r$1uvc$1@xxxxxxxxxxxxxxxxx>,
Ian Thompson-Bell <ruffrecords@xxxxxxxxxxx> wrote:

Quite right, an equivalent noise resistance is an approximation, however careful placement of zeros in the open loop response will reduce 1/f noise so that the approximation is quite good.

I don't understand how "careful placement of zeros in the open loop response will reduce 1/f noise", at least in the audible range, although I can see that it could mitigate some effects resulting from out of band noise, can you explain?



As you know, in tubes, 1/f noise starts in the KHz region and just keeps going up right down to dc - a real PITA for the RIAA designer. This is less of a problem in a mic pre with it essentially flat frequency response but it is still an issue. There's not a lot you can do in band but ensuring the bass drops off in a controlled manner below the nominal lowest desired frequency will help. 1/f noise has constant power per decade so there's the same power between 1KHz and 100Hz, as between 100Hz and 10Hz. Not much you can do in that region. But there's the same power between 10Hz and 1Hz and between 1Hz and 0.1Hz. This can be a serious issue in transistor designs where a common configuration is a dc coupled triple. This clearly has gain down to zero Hz and this gain is at least unity.

In tubes I have yet to see the equivalent of a dc coupled triple so there is usually at least one zero in the open loop which ensures the gain drops to zero at dc. Unfortunately when feedback is employed is can make matters worse. It tends to extend frequency response anyway but also, where the feedback factor is altered to provide variable closed loop gain, it is all to easy to get a significant gain increase hump at or below 1Hz, thus undoing all the potential good work of the open loop zeros - one reason for my comment about 'careful placement of zeros in the open loop'.

Simple tube triples (two CC + CF) often use a series capacitor in the feedback loop to block dc from the first stage cathode. This zero unfortunately reduces feedback a low/very low frequencies and hence *increases* closed loop gain at low or very low frequencies. By contrast, the zero formed by the coupling capacitor between the two CC stages *reduces* open loop gain and hence closed loop gain at the same low/very low frequencies. Often in simple designs this zero is set at a very low frequency in order to preserve open loop gain at low frequencies.

Perhaps one of the best tube triple designs I have seen is the Pultec MB-1 which someone posted here recently. It uses a 0.05uF coupling capacitor between the first two stages and the second stage grid resistor is 330K giving a -3dB point af about 10Hz. The CF is dc coupled to the cathode of the first stage so there is no zero in the feedback path to cause the closed loop gain to rise at low frequencies. Gain can be altered by a resistor in series with a capacitor placed across the cathode resistor so the feedback also is reduced at low frequencies. The overall result is a flat response across the wanted band with a well controlled falling off below it.

Cheers

Ian
.

Relevant Pages

  • Re: Design for a small tube/valve mixer
    ... capacitor in the feedback loop to block dc from the first ... low/very low frequencies and hence *increases* closed loop ... gain at low or very low frequencies. ... By contrast, the zero ...
    (rec.audio.tubes)
  • Re: Design for a small tube/valve mixer
    ... capacitor in the feedback loop to block dc from the first ... low/very low frequencies and hence *increases* closed loop ... gain at low or very low frequencies. ... By contrast, the zero ...
    (rec.audio.tubes)
  • Re: PC Speaker Tube Amp
    ... the supply is adding "zero" behaviour at the low frequencies, ... larger A (open loop gain) is making things worse. ... entire global loop, so it would be a "dominant" zero. ...
    (rec.audio.tubes)
  • Re: [PATCH] Avoid buffer overflows in get_user_pages()
    ... In particular, "len" is a signed int, and it is only checked at the ... So, if it is passed in as zero, the loop ... I think that, if get_user_pageshas been asked to grab zero pages, ... Which is a bug, and you want to catch it. ...
    (Linux-Kernel)
  • Re: Comment on trim string function please
    ... Call strlenonce before the loop and ... store the result in a variable, and use that variable to control ... existing zero, which is harmless. ... going back and inspecting the moved characters a second time? ...
    (comp.lang.c)