One aspect of balanced topology is that you can get away with a virtual ground, as significant current will not be flowing in that node.
However, for stability a Zobel network should be used, and this is normally connected to ground at one end. This is not good for keeping the ground 'clean' and avoiding the need for a real ground althogether. Can the Zobel network be connected across the two amps in the balance/bridge, or to a power rail? (+ve for one amp and -ve for the other, to maintain symmetry.)
However, for stability a Zobel network should be used, and this is normally connected to ground at one end. This is not good for keeping the ground 'clean' and avoiding the need for a real ground althogether. Can the Zobel network be connected across the two amps in the balance/bridge, or to a power rail? (+ve for one amp and -ve for the other, to maintain symmetry.)
What's wrong with a "real" ground? You just have to be careful about where the currents flow.
Each amp needs to have it's output damped at high frequencies with respect to gound. This is because the difference signal at the input is measured relative to ground. The supply rails are not ground, nor is the output of either amp.
Your challenge is to create a "virtual ground" through which the zobel currents can flow. You have moved the problem up in frequency where the currents are smaller so that's some benefit.
Each amp needs to have it's output damped at high frequencies with respect to gound. This is because the difference signal at the input is measured relative to ground. The supply rails are not ground, nor is the output of either amp.
Your challenge is to create a "virtual ground" through which the zobel currents can flow. You have moved the problem up in frequency where the currents are smaller so that's some benefit.
traderbam,
I see what you are saying, but what about single-supply operation with AC-coupled inputs and outputs? In that situation the Zobel is connected to the rail as there is no midpoint ground. I'm trying to avoid a real ground as I'm trying to work simply with a single secondary winding. Bridged/balanced topology gives you the opportunity to do this without AC-coupled outputs.
Nelson,
I did actually look at some X-type schematics to see how Pass did it and saw no Zobel. There seems to be some controversy (see chip amps Peter Daniels vs Fred Dieckmann) about whether or not Zobel is needed. I have not tested my amp yet and it may well be stable enough without, I'm just trying to do some preliminary investigation. Note: the amp is my own discrete design.
I see what you are saying, but what about single-supply operation with AC-coupled inputs and outputs? In that situation the Zobel is connected to the rail as there is no midpoint ground. I'm trying to avoid a real ground as I'm trying to work simply with a single secondary winding. Bridged/balanced topology gives you the opportunity to do this without AC-coupled outputs.
Nelson,
I did actually look at some X-type schematics to see how Pass did it and saw no Zobel. There seems to be some controversy (see chip amps Peter Daniels vs Fred Dieckmann) about whether or not Zobel is needed. I have not tested my amp yet and it may well be stable enough without, I'm just trying to do some preliminary investigation. Note: the amp is my own discrete design.
richie00boy said:
The issue comes down to the intrinsic stability of the amp, and
that usually means low or no feedback. The Alephs and X amps
have not needed Zobels because they are very simple gain
stages and don't have a lot of feedback. By contrast, the
chip amps are complex circuits and have relatively huge amounts
of feedback. (I make no reproach here, I'm just relating the
difference.)
Ok, but where is the input gnd connected? To the same rail. In a feedback design the amplifier measures the difference in voltage between the input wire and some reference ground and the output wire and the same reference ground. The zobel is intended to act as an HF damper between output and reference gnd.
Yes. Some designs as you say do not require a zobel to remain stable. In which case some alternative damping path exists or the gain at HF is low enough to avert instability.
The TDA gif has three, a 220n + 1R across the load, a 220n + 1R from the left side 'hot' to V+, and a 220n + 1R from the left side 'hot' to V-
Most all of the amplifiers Crown has sold for the last 30 years or so are bridge designs, the schematic in the link is a partial from a MacroTech 600-1200-2400. It is very similar to the TDA gif, 0.047µF across the load, 0.22µF + 3R from the right side 'hot' to V+, and 0.1µF + 3R from the right side 'hot' to V- (don't let it throw you that the right side 'hot' is the - output of the amplifier).
http://www.crownaudio.com/pdf/amps/MT600-1200 w 3rd gainJ0444-0_A.pdf
Most all of the amplifiers Crown has sold for the last 30 years or so are bridge designs, the schematic in the link is a partial from a MacroTech 600-1200-2400. It is very similar to the TDA gif, 0.047µF across the load, 0.22µF + 3R from the right side 'hot' to V+, and 0.1µF + 3R from the right side 'hot' to V- (don't let it throw you that the right side 'hot' is the - output of the amplifier).
http://www.crownaudio.com/pdf/amps/MT600-1200 w 3rd gainJ0444-0_A.pdf
Don't think so.
The Crown parts values reflect the nature of the parts in the amplifier, the caps going to the rails are different values for a reason. If you look at the emitter resistor values on the driver transistors (for instance) you will see the NPN has 5R6 and the PNP has 22R. Look at the compensation networks on the base of the drivers, very different values here too.
The stability of your design dictates what it will need.
The schematics are one solution for a design with no center tap in the supply.
A different design, a different solution:
http://home.kimo.com.tw/skychutw/
click on 'Ampzilla', and then 'Ampzilla III'
The Crown parts values reflect the nature of the parts in the amplifier, the caps going to the rails are different values for a reason. If you look at the emitter resistor values on the driver transistors (for instance) you will see the NPN has 5R6 and the PNP has 22R. Look at the compensation networks on the base of the drivers, very different values here too.
The stability of your design dictates what it will need.
The schematics are one solution for a design with no center tap in the supply.
A different design, a different solution:
http://home.kimo.com.tw/skychutw/
click on 'Ampzilla', and then 'Ampzilla III'
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.