Some kinda XSoZ

[Fuling]

I had the great luck to find three HUGE heatsinks on a junkyard a couple of days ago, and due to their size the little DIY-devil on my shoulder immediately started whispering -"Son of Zen, Son of Zeeeen..."

I´ve always wanted to build one of these monsters but I´ve also realized that 4% efficiency is.... low. Very low.
What bothers me the most is that a great part of the total power is wasted in the tail resistors, so replacing them with a CCS that runs at a lower rail voltage seems like a good idea.
I guess this has been done before.

After studying the ZV6 & ZV7 at passdiy.com I got a bit inspired to try the X-feedback. What bothers me here is the low input impedance and the electrolytic coupling caps at the inputs.

Input coupling caps can be omitted if we use a bipolar (+-) PSU and the low impedance can be dealt with using input buffers.
One fine side effect here is that the buffers also acts as level shifters, lowering the gate voltage for the power Mosfets by about 4V which gives us a bit more headroom.

Have a look at my conceptual (and ugly) schematic and feel free to comment. I´m not sure that this circuit would work at all, so please tell me if anything is horribly wrong.

[GRollins]

I did a thread once on the SOZ with current sources. It's not difficult, and yes, it can potentially save a lot of heat/power.
Given that the absolute DC offset isn't a problem (as far as the speaker is concerned) with the SOZ, you don't even need to worry about having the front end current sources adjustable. On the other hand, it would give you the chance to even out small differences in relative DC at the output, so you can do so or not, as you please.

Grey

[Fuling]

Yup, I read that thread yesterday. Since this amp runs of +30/-15V rails the power consumption is down to 75% of what it would be with symmetrical rails. This equals to 90W less on the heatsinks.
My plan with the adjustable front end CCS´s is to set the output DC offset, which shouldn´t be that big anyway if the mosfets are matched.

Grey, what do you say? Shall I build this one?


/Daniel

[Nelson Pass]

In that particular schematic, the Sources of the gain devices
will probably sit at -4.5 to -5 Volts. You can build a decent
little current source using 5 volts, so you could take the negative
rail down to -10 volts and get away with it, particularly if you
are driving it balanced (recommended).

If you want more efficiency, look at the coil loading in ZV7 -
this nearly doubles the efficiency over using constant current
source loads, not to mention resistors, and you would have
a positive supply more down around 18 volts easy enough,
and it would work better.

Or you could stick Aleph current sources instead of the CCS
off the Drains of the Diff pair, and pick up some more %.

The key here will be to adjust the CCS which biases the
diff pair and keep it stable, and to provide some resistive
loading from each output to ground, so that things stay
stable. I usually use about 47 ohms or so, which only loads
the circuit about 9% but does the job nicely.



Also, don't be so worried about input caps - remember the Gate
of a Mosfet is a cap, and not a very linear one at that.

[Fuling]

Thanks for your reply!

the Sources of the gain devices will probably sit at -4,5 to -5 V.

I thought each mosfet would drop more like -4 V, so I calculated for -8 V in total. Well, a few volts here and there

look at the coil loading in ZV7

Not that I really *need* to improve the efficiency, but inductor loads are always interesting. The article mentions coupled or centertapped inductors, is that important for the performance?
I happen to have a bunch of 25mH chokes that maybe could be put to use here. Would 50mH per "leg" be enough?

Or you could stick Aleph current sources

Things are getting out of hand here!!

[Nelson Pass]

Originally posted by Fuling
I thought each mosfet would drop more like -4 V, so I calculated for -8 V in total. Well, a few volts here and there

Not that I really *need* to improve the efficiency, but inductor loads are always interesting. The article mentions coupled or centertapped inductors, is that important for the performance?
I happen to have a bunch of 25mH chokes that maybe could be put to use here. Would 50mH per "leg" be enough?

Sorry, you're right, and you probably ought to assume more
like -10 volts max.

50 mH would make a good mid/tweeter load, but if you want
bottom end, you're in the 1H range. But a transformer with
split secondary can give you that.

[Fuling]

I think +30 and -15 V rails will be fine. 12 and 24VAC transformers are easy to find, which is an important issue since I want to keep things cheap
For once I don´t think heatsink size will be a problem, the ones I brought home are huge!

Well, since I´m into biamping I might build a mid/tweeter amp some day using these chokes. Since my speakers are dipoles they might like a bit of output impedance, so maybe some kind of "choked Zen" with the feedback network replaced by a simple source resistor? Another day, another story...

If we take a few steps back and focus on the schematic in the first post, would it be worth a try?

Med vänliga hälsningar,
Daniel

[Fuling]

Here´s a complete schematic.

[lumanauw]

Hi, Fuling,

Nice design. But there is 2 thing that I wanted to ask you.
1. What makes you think of input buffers (the IRF610 at both inputs, like zenV4 but with N mosfet). Aren't they will add unnecessary distortion to the whole system, if we talked about CCT simplicity? Or are you concerned about input impedance so you put this buffers to make it high impedance input (compared if there is no buffer, the input is as low as 1k)
2. The feedback system in the differential should work without 100uF cap. What makes you put it there?

[Fuling]

Hi

1: Yes, the low input impedance bothers me quite a bit.
IMHO a pair of buffers is easier to deal with than a complete preamp than can drive the "original" Xsoz.

2: Without caps the feedback loop interacts with the bias, which might not be a good thing here. Wait a minute, maybe they can be removed without complications after all. Have to check it out.