Open loop gain, phase margin etc

[ceharden]

I've been searching but there doesn't seem to be any definitive answers or reference to some of the basic issues of amplifier design/optimisation. I'm hoping by posting some of my investigations we can create a reference for other people looking for the same information.

I've gone back to basics with my amplifier design. I have a prototype running but decided before I transfer it to PCB to effectively start again and go through the whole design progress. While my current design works (amp listening to the one prototype channel at the moment) I think there is improvement to be made, especially regarding stability. Doesn't like 100kHz square wave test signals for some reason! Going back to the simulations suggest that the phase margin is fairly non-existent which would explain why it was a PITA to compensate.

So, I'm just looking at the LTP and VAS at the moment. I'm playing with this circuit and a bode plot is shown here

My observations so far are that I've got a fairly decent amount of open-loop gain, even with the quite large emitter degeneration resistors in the LTP. As expected, these resistors improve the phase margin significantly. What surprised me however was that adding degeneration to the VAS actually degraded the phase margin.

My main questions are, how much open loop gain do I need? I intend the final amplifier to have a closed loop gain of either 26dB or 32dB (20x or 40x). How much phase margin is acceptable, good etc? The circuit shown has about 60deg of phase margin.

I'm currently using the crude large inductor and capactitor method of calculating open-loop gain. Is this accurate enough for what I'm doing?

I look forward to a good discussion about this!

Chris

[h_a]

My main questions are, how much open loop gain do I need?

How many negative feedback do you want to apply?

That is, I'm sorry, at the heart of designing an amp. Effectively there are 2 different companies here in this forum: low OLG, low NFB, rather high THD or the same in high (with THD low ).

Plenty of room in between.

Have fun, Hannes

[ceharden]

Thanks for the reply. I am aware that there is some difference of opinion on the subject but I'm trying to put some figures to the descriptions.

What is regarded as low NFB? 10dB? 30dB? What is high NFB? 40dB? 60dB?

My application is for things like studio monitors so I'm looking for accuracy and neutral sound. I would imagine that would put me into the high NFB camp. From my simulations so far, I would expect that I'm likely to get somewhere in the region of 50dB NFB (80dB OLG - 30dB CLG). Is this reasonable?

[h_a]

so I'm looking for accuracy and neutral sound.

Don't we all

Your NFB number might be ok, depending on exact topology and all other design choices.

Why don't you save yourself all the trouble and build a proven low distortion amp like one of Douglas Self's amps? Search for 'blameless' - or better buy his book

I wouldn't trust distortion figures from simulation in general - but that depends very much on your own SPICE-knowledge and experience with device models. You can easily end up with an amp that measures far worse than simulated.

Have fun, Hannes

[ceharden]

There's no challenge in building somebody elses design for me. I look at them for inspiration but I won't copy wholesale. Part of the whole point of this for me is to learn how to design them and I find the best way is to do it! I design my own speakers, so I want an amplifier I've designed myself to go with them!

[h_a]

Yeah, I perfectly understand

learn how to design them

That's the hard part if you want to do a high NFB-design properly. I really recommend to get Self's book on Audio amp design, you will learn a lot on the sources of distortion. Indispensable in my opinion for serious low-THD designs.

For a low THD-amp, it's essential to verify performance. Get/setup a decent soundcard and a low THD-oscillator for measurement. Be prepared to build a number of prototypes that probably will disappoint you in comparison to simulation.

Hey, but it's a lot of fun!

Have fun, Hannes

PS: by the way, have a look at Grey's GR-25 amp; folded cascode topology with pretty little distortion and low feedback - see it's possible It just depends on what is your target.

[Mr Evil]

30dB of feedback at 20kHz is typical. Much more than that and stability becomes troublesome.

90 degrees of phase margin is ideal. 60 degrees should be fine. 30 degrees is low.


Even an amplifier with plenty of phase margin can have stability problems due to its inductive output impedance, so don't omit a Zobel network on the output. Additionally, an output inductor helps with stability into capacitive loads.

[Nrik]

Originally posted by h_a PS: by the way, have a look at Grey's GR-25 amp; folded cascode topology with pretty little distortion and low feedback - see it's possible It just depends on what is your target. [/B]

Isn't Greys GR-25 with NO feedback?

[ceharden]

Ok, thanks for the info. I'm heading along the right lines it seems.

Here's schematic of what I think is a fairly realistic input/VAS stage for my amplifier:

http://www.chaudio.co.uk/AmpDesign/TestAmp2Sch1.pdf

Spec's:

LTP bias current: 2.2mA/leg
VAS bias current: 8mA

Maximum OLG = 103dB
Open Loop phase margin = 60deg

Closed Loop (gain = 20x/26dB) phase margin = 50deg
Response -3dB @ 750KHz

Now obviously the frequency response extends somewhat higher than needed and would probably oscillate freely if built as such. What -3dB frequency is reasonable? I'm thinking 50kHz maybe? Should I control that using more VAS miller compensation or a small cap across the feedback resistor seems to do the trick?

[blueskynis]

There are many ways to compensate an amplifier. I will mention just briefly some of them.

The most common one and very easy to implement is one pole compensation, also called dominant pole compensation, just like you are using. The drawback of this type of compensation is that it sets a pole in transfer function at very low frequency (in range 10Hz to several kilohertz) where the OLG begins to continuously drop 20dB/dec which gives you very low OLG at higher frequencies. Capacitor used is in range of few pF to 100pF. You can also achieve this kind of compensation by tying capacitor between VAS and ground. D.Self calls this a crude way of compensation.

Second also common method of compensating an amplifier seen here is by loading the VAS with a resistor (10 kilohm's and more) to the ground. I have seen some professional designs using this method.

Compensation using zero and pole, known as LAG compensation, can give you more OLG bandwidth and increase OLG at higher frequencies. This allows to get some nice THD numbers, but unfortunately this type of compensation degrades dynamic performances of an amplifier. This type of compensation is achieved by putting a resistor ( few hundred ohms to kilohm) in series with a capacitor (100pF to several nF) to ground or +/-Vcc depending on topology.

Very often you will see a small capacitor (few pF) in parallel with upper feedback resistor. This is called LEAD compensation and it is often used in combination with some other compensation to improve phase margin. It also improves amplifier response times. If you put to big capacitor here, then you can degrade phase margin so use this carefully.

Of course, you can combine these compensations to achieve what you want.

I'm not an native English speaker, so...sorry for any mistakes.