X100 backengineered here

Re: Few Questions if may

keypunch said:
Hi Petter,

1) There was mention of a V2 and V3 in your postings. Is it correct to assume the circuirts and PCB you posed are for V1?
2) It seems you have made some refinements after posting the circuit and PCB re resistor values and added a Vbe. Any chance of posting or eMailing me those updates?
3) I like to see how you implemented the 10W Class A/100W AB and 50W Class A/100W Class AB so you can change via the external case switch you installed. Pic not needed just what points you tapped to, the values you used to switch between or add in parallel or series to enable the bias value change and method of value change as in did you just switch in different value or switched in a series or parallel value?
4) Is there a formula to determine different Class A/Class AB threshold points? If so what or where is this formula?
5) There seemed to be some reference to using output drivers in quadruplets? Is this a fixed requirement of this design that one has to use or can one choose to use 2 or one set of output devices per side of X? I realize this implies less power, that is ok, I am well aware that may be case.
6) Indication is this design scales well. What PSU or componet value calculations does one use to scale the design? For your implementation I cannot recall the Secondary Vac you used and resulting VRail of your supply design. I have a number of PSU based calculations, but seems the PSU calculations vary widely due to topogogy of the amp. Some topologies need certain key component values recalculated for different PSU rail voltages.
7) Any general thoughts on use of PSU chokes in terms or pros and cons that may be specific to this amplifier design?



1. Yes
2. I decided that the project was dead since there was limited interest and contribution from all parties dwindled as people became more interested in the rather vibrant newer Pass project (AlephX). I guess I will stick with that decision. Bear in mind that the original drawing (change the values of the components) with a Vbe multipler is likely close to the original. If you want to increase gain, reduce or even short out the "X" resistor.
3. The level of power you burn off at idle (which defines the class A drive level) is only decided by how much current passes in the output stage. This in turn is defined by the source resistors, the gate voltage of the devices (4-5V typically). So the method of changing this is to set up a relay which causes the Vbe (or other method used to set bias) to shift to the new desired voltage. Since constant current passes in the "Vbe leg", you could just as well use a fixed resistor as shown in drawing one. It is probably better to use a Vbe multiplier for reasons of stability, but the principle holds. If you use the drawing on page one, you can have a resistor of 140 Ohms with a relay switch between the top and bottom of both resistors and you get a resultant 70 Ohm total resistance. This yields half the voltage across the resulting compound resistor. You probably don't want to do something that extreme, but consider what happens if you just short with a relay and zero ohms - you have turned off the output stage completely. This is the principle for changing the level of "class A drive"
4. This is a general question and I really must refer you to the textbooks.
5. The way I understand, you are asking about number of output devices, and how you change the number of them. In the smallest configuration, you use 4 devices, one in each quadrant of the drawing. If you want to scale the current capability etc. you would add 4 more, then 4 etc.
7. Chokes would likely work well in ClassA designs. Bear in mind that the usage of chokes (or the "mode of the choke" if you will) depends on how you use it. Search for "Duncan PSU Designer" and use that to see the difference between a CLC design and an LC design. It will surprise you. The problem with units that draw highly variable amounts of current and an LC design as I see it is that (depending on capacitance at the end) the power supply might not be able to supply the desired current quickly. Still I am a fan of inductors, but I think it is important to think about what you want to achieve before applying them. Suitable goals can be: Lower ripple, more effective use of transfomer (longer charge cycles - i.e an LC design rather than a CLC design), less noise coupled into the system etc. Your mileage will vary. This is also a rather general question, and even though I believe this and many other designs is highly suitable for inductor usage in power supply I cannot answer your specific question with authority.

Also bear in mind that testing the X design is kind of hard. You will probably think that everything is broken even though it might be fully working when you have built it. Check out my tips written elsewhere in this thread about setting up the prototype.

One of my plans for the future is to build the input stage as a complete amplifier which I sense passdiy are either planning, or possibly have even done. This would be an interesting exercise, especially for sensitive speakers - the natural evolution of Zen.

Good luck.

Petter
 
Adding on to question 3 and 4:

Class A can be though of as a standing current in the output stage capable of driving the load fully without shutting off the opposite leg. For a normal unbalanced amplifier with split supplies, this means that if your speakers desire 10 peak Amps at the level you ask of them, the standing current needs to be able to handle that. However since you are modulating both phases, you ned a standing current of minimum approximately 5 amps. For a balanced amplifier, you need half that per phase.


This current can be determined only by knowing what the load is. Typically one expects a speaker to be resistive at say 8 Ohms. Well that is typically far from the truth so it is impossible to say how much current is required without knowing something about the speaker.

From memory, the Aleph's set up one of the legs as a constant current source (it is a little more complex because the current source did "help" a little by boosting current as required using a positive feedback system from the output). This current then becomes the maximal drive that the amplifier can provide from that phase and thus from both phases if one desires symmetrical drive. In the case of the Aleph, the efficiency is up to double the special case of the constant current source - often termed single ended amplifier (in the tube language). The most important thing is to grasp the concept.

So the formulae you asked for are pretty simple to deduce, but not all that practical unless you are willing to commit some values - such as load impedance. Current in the speaker is a function of voltage applied and impedance. Let's assume 8 Ohms. So 8 Volts applied yields V^2/R "power". In this case 8 "Watts" peak (less RMS). If this is what you want to achieve "power wise", you design at this point. In this case that means 1 A peak though the speaker. I think you get the point. Sounds like you want a whole lot more power, in which case just ramp it up. Remember that in the balanced amp you can halve the current from each phase and still remain with all devices in conduction continuously.

For a fully "loaded" amplifier, you want to set it up so that it can deliver the desired current. If you want class A, then make it so. You know the voltage and hopefully you know the load. Thus you can optimizie + add som extra headroom. So a 100 V split supply Class A unit driving 8 Ohms needs a standing curren of (work it out!)

Good luck!

Petter
 
"Regular" amplifier outputs need all the voltage and half the current.
Balanced amplifier outputs need all the current and half the voltage.
A balanced amplifier's sides only see half the load. There is, for want of a better term, a sort of virtual ground in the middle of the load. Since one side of the amplifier only sees half the load, it needs sufficient current to drive it.
Others may think better at night--speaking only for myself, I find that I think better when I've had sleep. Not that it happens very often.
You can "X" any number of topologies. The basic requirement is that it hinge on a differential. Hence my fiddlings with the Aleph-X.
(Petter--more vibrant thread? Harrumph! I hereby move that this thread be pinned to the top of the Pass Labs forum. I never understood why this thread wasn't as...or more...popular than the Aleph-X thread.)
Hint to see if I can jumpstart Petter's thread: Take John Curl's complementary differential (later called the "Diamond Differential" by a mainstream manufacturer) and use it as the basis for an X amp. For those who aren't familiar with the topology, go to www.marklev.com and look up the circuits John designed for the then-young Mark Levinson company. It seems that John has had bad fortune regarding other people taking his ideas and using them for profit without compensating him for his effort. This has led to him being reluctant to share his ideas. With that in mind, please show some respect when using his topologies. At some remote time in the future, he might be persuaded to share some more schematics if people behave in a civilized manner.

Grey
 
Hey, I second this...

I always thought the regular X-amp was sort of the more elegant concept as compared to single ended class A (balanced or regular). With a variable class A point it can be adapted to all sorts of needs. Not everyone can tolerate 400W+ room heaters ;) As nice as they may sound, there is an impractical side to them...

MBK
 
dveckom said:

Why dont you use feedback from output instead from driver. In that case you could have less distorsion, but amplifier may be less stable.


As I recall, this is covered in the Pass patent.

Feedback around stages (aka Global feedback in this case) is often not desired in high-end audio. One of the reasons I believe such feedback has gotten a bad reputation is that it is often used to correct a poor design by inflating the gain and culling it with said feedback.

Another reason might be that your reactive load might send junk back into the system.

Still, you may take the feedback where you desire. I imagine control in the lower octaves will improve if you do so to the detriment of the highs.

Petter
 
Stepping up

Given the latest endorsements (thanks!), it seems appropriate to step up.

What is the interest on this group in doing a very small power amplifier by using the X input stage configured in a very Zen way to drive the speakers directly?

It is possible to do just about anything, but I imagine a 5-10W into 8 Ohms amplifier could be the net result of such an undertaking. Power dissipation would of course also be Zen'ish, but there would be more transistors to share the power loss.

It would probably be useful to use (sacrilege) active current sources to minimize power dissipation in these. By setting their response to 1Hz or so it might not be so bad?

I have been thinking about this for a couple of years but never actually did anything about it. Could be fun.

I must admit that I haven't been following the PassDIY projects lately. Please advice me if I have missed out and this is simply duplication of what has already been don by The Master himself.

Petter
 
Nelson Pass said:
It's been done, and it works fine. My favorite version is
10 watts with a direct coupled differential pair of power
Mosfets biased by a CCS and whose Drains are power
resistors to the positive rail. I bias the gain stage with
about 4 amps.


Would there not be benefit from building replacing drain resistors with current sources a la the X input stage?
 
Sounds as though Nelson is describing an X'ed SOZ.
Current sources/active loads would help somewhat with the efficiency and increase the gain, but you're left with a tendency for the relative DC offset to wander. Not that the problem can't be solved...
Got some ideas I'll contribute if I can find time to scratch.
Oh, and think in terms of headphone amps, while you're at it.

Grey

P.S.: 'Bout time this thread got nailed up top.
 
In the context of bi- or tri-amping, it would make sense to have several small size (heat sinks! transformers!) a / ab amps such as the X-amp, that run low class a wattage for all normal listening, but have high peak capability for dynamics.

Example:

For say 87 dB / 1m efficiency drivers, room loss of -10 dB at typical listening distance, +6 dB gain from stereo operation, bi-amping with assumed equal average power sharing between lows and mid+highs, +3 dB, so one watt gets you 87-10+9=86 dB average SPL. The typical average classical concert level is supposedly about 80 dB SPL, so 1/4 Watt gets you there. Assuming 20 dB peaks, 25 Watt peak is needed. Movies are mixed to an average 83 dB SPL and 95 dB peak (explosions), so 1/2 Watt average and 8 W peak are needed.

Assuming you'd sometimes like to run it at 92 dB average in-room with +20 dB peaks, you'd like to see 4 W average (class a operation) and 400 W peak, from each amplifier. Alright, most speaker drivers won't take this.

So, assuming only 200 W peak: 24 V rails on an x-amp, balanced into 8 Ohms, using Mosfets with 5 V drop, you'd get those 200 Watts peak with no clipping.

Running it biased for say 5 W class a into 8 ohms would get you a still sane power dissipation, and a very high peak capability. Typical classical music at realistic levels would likely always run in class a.

BTW of course it would be nice to run just several low wattish-class A amps in such a setup. But the problem is that while average power needs are very low, the potential peak per amp even in bi- / triamping setups is still the maximum , i.e. 0 dBFS, so to avoid clipping on peaks you still want those hundred of watts peak capability even though you never exceed 1 W average... That's basically the case for class a/ab, and the x-amp.

200 W into 8 Ohms is just 5 Amperes. Even a single Power Mosfet-Quadruplet (1N and 1 P per balanced branch) could do the trick on an X-amp. The heat sink could be dimensioned quite smallish, for say 20 W typical average dissipation.

MBK
 
This is close to my thinking except that I don't think it is necessary for my needs to have more than about 10-100W max per amplifier depending on how they are crossed over and the needs.

The problem with having higher voltages is that they lead to the standing current having less "effect" in terms of "Class A" drive. Also, if I paraphrase (change) what NP often indicates - having high current in the FET's appears effective.

So, the lower the voltages, the more current you can pass (no pun intended) for a given thermal power budget.

So it remains to be seen whether an Zenish input stage configured as output stage, or a normal amp with few output devices, each carrying high current is "optimal", and for what application they would be "more or less optimal".

Again, your thinking is spot on IMHO

Petter
 
Ex-Moderator
Joined 2002
That sounds interesting MBK, and certainly would bring X-ing to the masses. My long term plan was to use my AX boards as midrange, with possibly a chip amp on the top, and something meaty of around 100W on the bottom, but if there was a scalable A/B X design I would have to rethink my options!
 
I guess what is needed is a scalable X design optimized for 10-100W power output, and capable of handling single ended inputs effectively.

Since I am using switched mode power supplies, I can vary the voltages as well ....

And it would of course be very useful to have the various current sources driven with a drive circuit which means we loose less power there while removing need for users to adjust. It is a complicated project to troubleshoot as I found out first hand.

Petter