@steven: VAS stage in your zero-feedback design?

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hi Steven,

you posted this in the Hawksford thread:
http://www.diyaudio.com/forums/attachment.php?s=&postid=330446

Can you explain why you returned the VAS emitters to the output resistor of the right hand side of the diff pairs? I can see how this actually increases the swing of the VAS by using the gain from both sides. At the same time, I would be worried about the assymetric loading, the left hand sides being even buffered by emitter followers while the right hand sides have to drive an emitter with pretty low impedance.

Regards,

Eric
 
Lender

Hi Capslock,

The input stage does not care much about asymmetric loading. You have the same if you drive the VAS from one half of the differential pair and tie the other collector to the supply. The collectors of the differential pair transistors are current sources after all.
This VAS is stage is also called the Lender circuit, after Rudolf Lender (Motorola) and is used by other amplifier designers too, a.o. Borbely.
This stage has good linearity (local feedback), thermal stability (Vbe cancellation), speed (semi-folded cascode) and predictable bias current. Its performance is comparable with a cascoded VAS, but with simpler biasing (no dc voltage source required).

Steven
 
dissecting Lender circuit

Maybe we should change the thread title to this?

Ok, googling for Lender, circuit, amplifier yielded a couple of hit, but no in-depth analysis.

First of all, there seems to be a kind of degenerated common emitter amplifier with equal emitter and collector resistors used as a level shifter and signal inverter, this is clearly not what we are talking about: http://www.mujweb.cz/www/icansoft/zesik/pg26.htm


Then there seems to be the case where the emitter of the VAS is returned to the inverting side of the diff amp, so that the collector load resistor is the same as the emitter degeneration resistor of the VAS. There are two variations, with and without a follower.

The follower (as in Steven's and Borbely's examples) will do two things:
- change current drive to voltage drive (with the corresponding ramifications)
- compensate the V_BE offset and drift

Let's stick with the buffered version because that saves us having to figure V_BE into the equations:

For lack of part values in Steven's example, I have nicked them from Borbely. He runs the diff amp at 2 mA and has a 2k2 resistor on the left hand side and a 200 R resistor on the right hand side.

The left side will sit at 2200 R * 1 mA = 2.2 V below the rail, the right side would sit at 0.2 V if it weren't for the VAS transistor. This initial disparity makes the VAS conduct until its emitter and hence the right side collector sits at exactly the same voltatage as the left side. In this example, this fixes the bias current at 10 mA.

Now with an input voltage on the diff amp, lets assume the collector currents are shifted by 1%, i.e. we have 1.01 and 0.99 mA and hence 0.02 mA differential current. The VAS current increases until the diff amp collector voltages are again equal. This causes a current of 0.01012 A to flow in the VAS, i.e. 0.012 mA of signal current, i.e. a current transfer ratio of 60%.

The local feedback acts to control the VAS current which will no doubt make for a high-impedance, i.e. linear VAS. The downside is that the compensation capacitor has to be connected to ground in order not to bypass this feedback path, so the output impedance is essentially determined by the cap's impedance.

I can see no fundamental advantage compared to an unbuffered VAS that uses a Miller cap both for compensation and the same amount of local current feedback as in the Lender circuit. And this traditional implementation can be improved by cascoding the VAS to give more loop gain into this feedback loop. In a single diff amp arrangement, VAS bias is set by the collector load (current source), in a complementary arrangement, one could e.g. use bypassed emitter degeneration to stabilize the current.

A slight advantage is there in that the Lender circuit keeps the input transistor collectors at the same voltage, so we will see a common mode capacitance variation that cancels if both inputs are driven from a similar impedance.


Has anybody done simulations of Lender vs. various other diff pair/VAS arrangements?
 
I'm afraid this is a bit of a non-answer, but I tried simulating this VAS structure on my TGM amplifier. I didn't find any benefit from it. Bearing in mind I was looking mainly at the harmonic distortion profile and my simulation abilities were not that great. I also searched the web for further information but found next to nothing.

Hugh might have some wisdom to add regarding this topology.
 
Gareth,

Tried the Lender, simulated it also. No OLG advantage over a single ended drive to a stock common emitter VAS, and no discernible advantage sonically. However, it does combine some elements of CM, tending to force current balance in the LTP by adjusting VAS current. But this may not be very useful if the VAS current is fixed anyways, as it normally is.

Cheers,

Hugh
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.