Could someone explain this Hafler/Strickland output topology to me?

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

Recently I was curious if I could find schematics of the last Haflers (the 9500/9505 specifically). I did, but I don't understand how this output section works - i.e. how the current is supposed to flow through the output drivers relative to the + output without shorting the power supply. Could someone explain it to me?

Thanks,
Paul

http://k-amps.8m.com/cgi-bin/i/PowerAmps/Semicond/hafler_915c.jpg
 
OK. Now I think I see it, but it looks more to me like the caps and power supply are referenced to ground by the outputs, not by the speaker. I.e. when the N-channels are saturated, the + end of the power supply will be near 0V, thus the + output will be about -63V from the - output.

Interesting.

What's the advantage of something like this over a regular common collector / common drain setup?
 
The output stage's configuration isn't really any different.

However the place where the load is, is different.

I am not sure, but you do get some benefit - I think - when driving "wierd loads", and there is no "DC" connection to the load, since you are running through the caps...

The feedback has to come from another place too - loop feedback, since the usual junction is now at ground.

_-_-bear
 
I don't think so. The caps are not in series with the current path here.

If you draw an equivalent I think you'll see that the output is DC coupled.

The difference seems to be that the outputs vary the coupling of the power supply to ground, rather than varying the coupling of the power supply to the load.
 
Originally posted by SQLGuy
What's the advantage of something like this over a regular common collector / common drain setup?

The caps block DC when outputs fail. No need for relays and sensing circuits.

If you plan on doing this, I suggest using Low ESR caps and a healthy margin in voltage rating. I have quite a few Altec and EV amps with their supply caps blown off. No problems with the QSC though.

Mike
 
Michael Chua said:
The caps block DC when outputs fail. No need for relays and sensing circuits.

And I still say, no. If the N-channels were to short, for example, you would have -63VDC on the + output and 0V on the - output. C16 would be in series with the load, but so would the two, forward-biased, rectifiers in the bridge, which are in parallel with C16. F1 would eventually blow, but in the meantime you'd be feeding DC to the speaker.

C16 is a power supply filter cap, not an output decoupling cap.
 
Originally posted by SQLGuy
If the N-channels were to short, for example, you would have -63VDC on the + output and 0V on the - output.

I am not very familiar with this arrangement so readers feel free to correct me.

The +OUTPUT is the midpoint of 2 capacitors in series. Note that it is referenced to the center tap of the transformer and is not connected to Signal Ground or Mains Earth. In other words, its "floating".

In operation, the +OUTPUT swings when current passes through the power mosfets. When the amplifier is in quiescent state or when fuses are blown, +OUTPUT is at 0V.

QSC is similar as wg_ski mentioned.

Mike
 

Attachments

  • qsc.jpg
    qsc.jpg
    87.5 KB · Views: 732
The center tap is connected to the + OUT, so an equivalent circuit for the center tap, one power tap, the two rectifiers and the filter cap would be a battery between C (+ OUT) and +V.

So, you have A/C current between the center tap and the power tap. This will be rectified through the bridge producing DC between +OUT and +V. If the N-Channels are short, the current path is +OUT, through speaker, through shorted N-Channels, through +V and back through the recitifier to the transformer.

Yes, you're right(ish), when the amp is quiescent the +OUT will be at 0V relative to ground because of bias current. When the fuses are blown, the +OUT will be unreferenced from ground because there will be no common current path. When the fuses are not blown and one side outputs are short, there will be high current flow through the speaker.
 
Exactly my point. The caps and the output design do not obviate the need for protection circuitry. This is not a cap coupled amp like a Zen, or a tranformer coupled amp like a McIntosh, where the speakers are protected from DC in case of output failure.

F1 is 10A in a 9505. The rails are listed as 63V (on another schematic I saw). If you have a true 8 Ohm speaker, full rail DC will be about 8A, so the fuse will never blow and you'll instead feed about 500W into the voice coil until something else blows.
 
I do agree with you that an output series cap or a transformer coupled amp like a McIntosh is more obvious in blocking DC in the event of output failure.

However, I don't think that this Hafler arrangement is not equally effective.

I can't make out in the schematic whether the rails are 55V or 63V but whichever one it is, when any N-fet shorts, there will be a full discharge of 20,000uF at 55V or 63V through the mosfets to ground. I'm quite certain a 10A fuse will not survive that. In my view, its actually quite an elegant solution for speaker DC protection. No added cost. Nothing between amp and speakers too.
 
Hi Michael,

The one restriction, though, is that that full discharge will still be through the speaker load, so it's still going to be current-limited by the resistance of the voice coil. If that resistance isn't significantly below 8 Ohms, the fuse will not be seeing enough current to blow it. The 8A continuous DC through the voice coil, though, probably will cook it.

The only safety factor will be that the continued drive signal into the P-Channels, and their feedback loop, would probably short them quickly after the N-Channels short. At that point the current will bypass the speaker load and blow at least one, and ideally both, of the fuses. If only one blows, you'd be back in the original situation, though, with 8A going one way or the other through the voice coil.

Cheers,
Paul
 
I can see two advantages:

* These mosfets have the source connected to case - and with this configuration all FET:s can be mounted on the same heatsink without insulating washers (which would make heat transfer worse). The QSC circuit using bipolars has this advantage too. Many amps get around this by using separate heatsinks instead - but in this case the heatsinks will be floating at high voltage which might not be a good idea.

* The front end of the amp doesn't need high voltage supplies - the output stage only needs +-10V or so on the gates for full output. The drivers can be run at high idle current without high power dissipation in them. Look at those transistor numbers, they are in the small TO-92 case all of them!

The obvious disadvantage is the need for separate power supplies for each channel.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.