Choosing current for LTP?

Zero Cool

Member
Paid Member
2004-09-20 6:10 am
MN
I just read N.P's excellent article on DIY opamps! Man there is a lot of great info there to digest!!!

I have been reading a lot about Long Tailed Pairs lately and gaining some understanding of there operation. and so now I am curious...and my questions may be too broad to answer i don't know...

But how do you determine what current to run the LTP At? what factors influence this decision?


Zc
 
In general, the bandwidth (fT) will increase as Ic increases. (fT will decrease at relatively high currents, but most likely, you would not use it at such currents.) On the down side, increasing Ic will reduce the differential input impedance and increase the bias currents. So these are the trade-offs and the choice of Ic depends on how important these factors are.
 
and as always there are trade offs. and as usual it sounds very complicated.

Looking at several schematics. it seems 2ma is a common choice?

Yes 2ma is a common choice, as higher the current LTP, higher the NFB, but a very high current in LTP more largest DC in base of TRs (BJTs) which increases the noise.
A current mirror doubles the transconductance in LTP, 2ma with a resistor is same as 1ma with current mirror. In my humble opinion, one LTP degenerated into 100 ohms with BC550/560c, max 3,5ma (~50mV on base of TRs).
 
Hi,
if the source impedance is kept low then you will find that high tail current does not increase stage noise significantly.
Adding an RF filtering cap at the +IN input lowers the impedance seen by the base and again helps keep noise low.
LTP tail currents from 0.3mA to 12mA are commonly used.

If you go for high hFE and medium to low Itail then you get low input offset current.
go very high hFE and high Tail current and the input offset current is usually considered OK.

If the two sides of the LTP are carefully matched then input offset current and Tail current do not degrade the output offset.

If you choose transistors designed for low noise then they include noise graphs. These will show IC values against RS with noise curves superimposed. At low Rs you will find that low noise can be had at quite high Id. Whereas high Rs demands low Id to maintain low noise.
 
Last edited:
Hi Andrew

I put an Fig for explain what I was saying, because my English is not good, also forgot to mention that DC is dependent on the impedance of the amp.

The recommendation to keep =<50mV is in the book of D.Self for avoid noise Jhonson (Cap: Small Signal).
 

Attachments

  • DC-Vb.png
    DC-Vb.png
    21.7 KB · Views: 305

Aldotin

Member
2009-12-31 11:56 pm
Iltp=2×pi×fmax×Cl×Vp×X

Iltp: ltp current (A)
fmax: max. frequency
Cm: load capacitance (~Cmiller)
Vp: max. peak o.p. voltage of the Vas
X: min. 3, preferably 6

For a simple conventional 70W/4r audio amplifier using +/-35V rails and BD139 as Vas transistor:
I=2×3.14×20000Hz×100pF×30V×6=2.2mA (1.1mA/leg)
 
that is just the calculation for slew rate to allow full power 20kHz operation.
Slew rate should probably be 5 to 10times that to minimise avoidable distortion. recently a poster came back and said 20times might be a better slew to aim for.

That just goes to show why there can be such a big variation, it's a balancing act between all the different effects that result from setting the LTP tail current.
 
In topology that uses Miller CDOM (collector-base of the VAS), the current LTP determines the Slew Rate of amplifier as higher the current LTP, more higher speed have the amp.
Amp Blameless-D.Self uses 3.5mA in LTP with 10K input impedance, if you want to increase input impedance or want more current in LTP, should increase the gain of the differential pair transistors, as mentioned for Andrew...
 
Jean Hiraga recommended biassing devices (in general) operating in Class A by looking at the data sheet for beta dependency on Ic which are typically shaped like a 'hill' , i.e. reduced beta at low current and high current. You pick a current near the top of the hill.

Not sure I explained it well but if you read his paper on Le Monstre it has a picture to go with it.

My limited experience says use between 1ma and 2ma per active device. If you are using a CFP LTP then you want more than 4mA feeding the LTP in total.

Most small signal devices I've seen would want a fair bit more than 1mA for optimal biassing according to the Hiraga method - perhaps it's best to conduct an experiment and see if it sounds different at different bias levels.
 
Jean Hiraga recommended biassing devices (in general) operating in Class A by looking at the data sheet for beta dependency on Ic which are typically shaped like a 'hill' , i.e. reduced beta at low current and high current. You pick a current near the top of the hill.

Good point Bigun, but in case of a LTP, the transconductance is not dependent of Hfe (different situation in VAS). Is best to choose the most suitable components to the project, may have problems with current, too high or too low.