Downscaling resistor values?

[Mick_F]

The feedback loop and corresponding resistor for a LM 38** based amplifier, lets call them Rf and Ri, are usually chosen in the range of Rf~20k and Ri~1k. This is so in the National data sheets, in the original Gaincard and Peter Daniels design (there 22k and 680R) and almost everybody elses too. I was wondering if it is possible to use smaller resistor values, chosen at equal ratio, i.e. yielding the same gain. CarlosFM has done this in his LM3886 design, where he uses Rf=2.2k and a corresponding Ri value. The obvious advantage of this is the lower noise of the lower resistivity.

My questions now are: Are there any benefits or drawbacks in terms of sound? Is there any difference? And are there any other drawbacks encountered when the values are lower? What about using much smaller ones, e.g. 220R and 6R8?

Questions, questions, questions....

Mick

[wintermute]

Hi Mick,

If you go too low I'm guessing you will have excessive current flow in the feedback path....

take for instance 20V on the output (not sure if that is realistic for a GC, haven't done any measurements yet, but should equate to 50W into 8 Ohms) the nfb circuit with 220 and 6.8 ohms would have close to 90ma flowing to 0V through the NFB resistors. that equates to about 1.8W of dissipation!!! someone correct me if I have completely botched the calculations here!!

Tony.

[owen]

Hi.

Yes the current dissipation in the feedback loop will increase, so the dissipation of the resistors will also have to increase. However, of note is that the LM38NN chips *like* a current driven input - hence the almost universal recommendation of a buffer (not that it sounds bad without one), and if you drop the key resistor values (Ri and Rf), the you will need a buffer to drive the lower input impedance....

Ok?

Owen

[wintermute]

Hi Mick,

something I just posted in another thread alerted me to another thing to be aware of if reducing the value of ri (if using a feeback cap Ci)

If you use say a 47 uF feedback cap and a value of 6.8 for ri then the low freq rolloff corner freq would be 497 Hz! Not particularly desirable so when lowering the value of ri you need to use the formula in the datasheet to make sure you don't affect your low freq response too much.

Tony.

[Mick_F]

Yes you are right. Of course you have to recalculate the positions of all the poles introduced.

Mick

[peranders]

Originally posted by Mick_F
...The obvious advantage of this is the lower noise of the lower resistivity.

This isn't obvious at all. You have to know the noise characteristics before you can draw any conlusions.

The more obvious thing is that lower value will reduce influence of stray caps with mean avoiding one source of unstability. Remember also that choice of resistor values has also to to with offset voltage and also effects coming from heating.

You can check my Gainclone designs to see which values I have chosen. I have lower than normal but not particulary low.