Cost + timefactor + cosmetics
Hi guys,
Thought I'd add my two cents.
First of all, I am single. That means my amps can look terrible even though I prefer them to be beautiful (High performance, yes but beautiful would be a stretch). Making a pretty box is a technique I have not mastered thus far. Anyone with a girlfriend or wife is going to have considerable trouble bringing ugly and potentially lethal equipment into the living-room.
Secondly, electronic design which I am more familiar with seems to be a craft which is more or less mutually exclusive with boxuilding. This is the fun part for me.
Third, cost of power supplies, box, connectors is likely to be high to very high. I recommend you buy a completed box. Building your own functional unit is a problem.
Fourth, risk. Will it actually work or will it blow up your speakers? Will you burn your face with acid since you hooked up the power supply wrong? Will your cat die etc.
Fifth, market. Selling high end audio is not easy.
Sixth, plumbing. Most of the work involved with making an amp is plumbing or infrastructure and "buildability". I don't find it much fun to work on infrastructure, and I won't even begin to comment on buildability. If you don't believe this claim, look at a preamp service manual and count pages devoted to this.
Finally, time. If I were to spend the time doing this (and other projects) at work, I would simply have to get a raise.
So, it is unlikely to very unlikely that a potential customer of say an X600 would even attempt to build something like this himself. If someone goes after this in a commercial way and leave themselves open to patent litigation, let them have it. If patents are a problem, go after something more "normal".
Hi guys,
Thought I'd add my two cents.
First of all, I am single. That means my amps can look terrible even though I prefer them to be beautiful (High performance, yes but beautiful would be a stretch). Making a pretty box is a technique I have not mastered thus far. Anyone with a girlfriend or wife is going to have considerable trouble bringing ugly and potentially lethal equipment into the living-room.
Secondly, electronic design which I am more familiar with seems to be a craft which is more or less mutually exclusive with boxuilding. This is the fun part for me.
Third, cost of power supplies, box, connectors is likely to be high to very high. I recommend you buy a completed box. Building your own functional unit is a problem.
Fourth, risk. Will it actually work or will it blow up your speakers? Will you burn your face with acid since you hooked up the power supply wrong? Will your cat die etc.
Fifth, market. Selling high end audio is not easy.
Sixth, plumbing. Most of the work involved with making an amp is plumbing or infrastructure and "buildability". I don't find it much fun to work on infrastructure, and I won't even begin to comment on buildability. If you don't believe this claim, look at a preamp service manual and count pages devoted to this.
Finally, time. If I were to spend the time doing this (and other projects) at work, I would simply have to get a raise.
So, it is unlikely to very unlikely that a potential customer of say an X600 would even attempt to build something like this himself. If someone goes after this in a commercial way and leave themselves open to patent litigation, let them have it. If patents are a problem, go after something more "normal".
Aleph X
You're right, the Aleph is mostly an output stage
design, and the X is mostly a front end design. It
is a reasonably trivial task to mate the two, and
I have several examples, some of which will surface
at the appropriate time.
The initial effort with the X circuit was partly to
demonstrate the effectiveness of the simple X circuit
to perform well at very high power levels. The other
high power products use really complex circuits, for
example I counted what appeared to be nine gain stages
is the big L product.....
Anyways, keep it up and have fun.
You're right, the Aleph is mostly an output stage
design, and the X is mostly a front end design. It
is a reasonably trivial task to mate the two, and
I have several examples, some of which will surface
at the appropriate time.
The initial effort with the X circuit was partly to
demonstrate the effectiveness of the simple X circuit
to perform well at very high power levels. The other
high power products use really complex circuits, for
example I counted what appeared to be nine gain stages
is the big L product.....
Anyways, keep it up and have fun.
I don't think that Pass has anything to worry about from the activities on this board. Even if the folks here manage to completely reverse engineer the X series, the most they can legally do with that information is make a prototype for themselves. Building copies for sale, or even to give to friends, would be a violation of Nelson's patent and would open them up to a lawsuit.
When you patent an invention you disclose details of its design and construction in exchange for exclusive rights to manufacture and sell products based on it for 17 years. The laws are designed to protect the inventor's investment, while allowing others to understand and experiment with the invention. There is even a possibility that someone experimenting with the X series may reach an insight that leads to another different, better way of constructing an amp. Far from being a violation of Pass's rights, this discussion is EXACTLY what the patent system is designed to encourage.
I don't think we have to worry about Nelson. Both the X series and the Aleph are protected by patents, and anyone who copies them commercially will likely get sued.
While it's certainly polite to ask permission to use someone else's patent for experimentation, it's neither necessary nor should it be expected. In my industry (computers), we routinely read our competitors patents, both to understand their products better and to get ideas to improve our own. It's not only allowed, it's expected.
-Jon
[Edited by Jon T. on 03-30-2001 at 02:53 PM]
When you patent an invention you disclose details of its design and construction in exchange for exclusive rights to manufacture and sell products based on it for 17 years. The laws are designed to protect the inventor's investment, while allowing others to understand and experiment with the invention. There is even a possibility that someone experimenting with the X series may reach an insight that leads to another different, better way of constructing an amp. Far from being a violation of Pass's rights, this discussion is EXACTLY what the patent system is designed to encourage.
I don't think we have to worry about Nelson. Both the X series and the Aleph are protected by patents, and anyone who copies them commercially will likely get sued.
While it's certainly polite to ask permission to use someone else's patent for experimentation, it's neither necessary nor should it be expected. In my industry (computers), we routinely read our competitors patents, both to understand their products better and to get ideas to improve our own. It's not only allowed, it's expected.
-Jon
[Edited by Jon T. on 03-30-2001 at 02:53 PM]
No kidding...you know what they say about sausage...ya don't wanna know what's in it.
However, I'm glad that Nelson stopped by and that we have (I'm assuming--possibly assuming too much) implicit permission to tinker with the X stuff, although for the time being I've got my hands full with an Aleph 2 and a crossover and some thoughts I'd like to follow up that might lead to a preamp...etc. etc. etc.
Not to mention the possibility that I raised above about a hybrid Aleph/X. I had something in mind about a month ago on a 'wouldn't it be nice if' basis, but I was driving at the time and couldn't scribble out a schematic. Now I'd like to see if I could recreate that train of thought. I was hoping that if I prodded petter et. al. that someone would say something that would help me retrace what I had in mind. I'm going to have to sit down with the Aleph schematics, and what's in this thread, and Nelson's patents, and a mug of tea and see what I can come up with.
Grey
However, I'm glad that Nelson stopped by and that we have (I'm assuming--possibly assuming too much) implicit permission to tinker with the X stuff, although for the time being I've got my hands full with an Aleph 2 and a crossover and some thoughts I'd like to follow up that might lead to a preamp...etc. etc. etc.
Not to mention the possibility that I raised above about a hybrid Aleph/X. I had something in mind about a month ago on a 'wouldn't it be nice if' basis, but I was driving at the time and couldn't scribble out a schematic. Now I'd like to see if I could recreate that train of thought. I was hoping that if I prodded petter et. al. that someone would say something that would help me retrace what I had in mind. I'm going to have to sit down with the Aleph schematics, and what's in this thread, and Nelson's patents, and a mug of tea and see what I can come up with.
Grey
First, in regard to a question back on page 4 as to why bother putting an Aleph rear on an X front: Jake brought up the efficiency point, with which I agree. There is also another aspect. In Nelson's patent, he states that the Aleph rear end reduces distortion. I don't remember that it's explicitly stated, but I'm assuming that it's due to the feedback loop coming back in from the output to the current source.
N.B.: I'd like to clear up a possibly erroneous impression that some may have gotten from The Big Thread (aka Opti-MOS). I'm not *against* low distortion. All for it, in fact. It's the mega-quantities of NFB to get those low distortion numbers that bothers me.
Anyway, I see this as a nested loop topology--one node at the back end of the differential, the other purely local at the power stage. With luck (fat chance!), I'll have time to review the relevant material tomorrow and propose a circuit in a somewhat more detailed & coherent manner than I have so far.
For the record, I feel better having Nelson drop in from time to time so that he can voice an objection to proceedings here if he feels he needs to. So far, judging from the cascode hint and the generally mild tone, I'm content to plug along.
Grey
N.B.: I'd like to clear up a possibly erroneous impression that some may have gotten from The Big Thread (aka Opti-MOS). I'm not *against* low distortion. All for it, in fact. It's the mega-quantities of NFB to get those low distortion numbers that bothers me.
Anyway, I see this as a nested loop topology--one node at the back end of the differential, the other purely local at the power stage. With luck (fat chance!), I'll have time to review the relevant material tomorrow and propose a circuit in a somewhat more detailed & coherent manner than I have so far.
For the record, I feel better having Nelson drop in from time to time so that he can voice an objection to proceedings here if he feels he needs to. So far, judging from the cascode hint and the generally mild tone, I'm content to plug along.
Grey
You can certainly use the X front-end to drive two Aleph output legs. Another thing that I had planned previously was to actively drive upper and lower legs of both.
If you look at the X schematic, you basically have inverting and non-inverting output at two voltage levels. Let's think about the right side for example. What you want to do is drive the bottom section (normal Aleph drive if you wish) with a normal signal, then you wan to drive the upper leg with an inversion of that. This inversion is available on the left side.
Now "all you have to do" to make this work is ensure voltages are set up right. It shouldn't be that hard. You can probably take the bias portion on the bottom leg and reflect it up to the top for auto bias.
I have elsewhere on this forum provided formulae for how to set up the correct current/voltage levels for an Aleph based on speaker impedance and required power level. Please bear in mind that the voltage requirement in each leg will be halved for a given amount of power.
If you look at the X schematic, you basically have inverting and non-inverting output at two voltage levels. Let's think about the right side for example. What you want to do is drive the bottom section (normal Aleph drive if you wish) with a normal signal, then you wan to drive the upper leg with an inversion of that. This inversion is available on the left side.
Now "all you have to do" to make this work is ensure voltages are set up right. It shouldn't be that hard. You can probably take the bias portion on the bottom leg and reflect it up to the top for auto bias.
I have elsewhere on this forum provided formulae for how to set up the correct current/voltage levels for an Aleph based on speaker impedance and required power level. Please bear in mind that the voltage requirement in each leg will be halved for a given amount of power.
What I ment was drive RHS bottom from below R7, drive RHS top from above R37. For fully balanced bridged operation, do the same treatment for LHS.
Of course, you could use a different scheme to shift voltages than simple resistors, because you will need a lot more voltage differential when you are driving N-channels devices on the bottom and p-channel devices on the top.
The good thing is that you don't need all that much gain.
Of course, you could use a different scheme to shift voltages than simple resistors, because you will need a lot more voltage differential when you are driving N-channels devices on the bottom and p-channel devices on the top.
The good thing is that you don't need all that much gain.
Looking over both circuits again, I'm not sure that the X and the Aleph would go together all that well. The X derives much of its benefit from its fully symmetric, balanced nature, while the Aleph is interesting because it is essentially a single-ended topology with the efficiency of push/pull.
You could use an X style topology to replace the input differential in the Aleph, but I think you'd lose the distortion cancelling benefits of the X. Or you could use two bridged aleph output stages in the X, but that seems like overkill when you've already got balanced drive signals and can run a conventional push/pull output stage.
On a slightly different note, has anyone noticed the remarkable similarity of the Aleph to the SRPP circuit used in many tube driver stages and preamps? The Aleph circuit has the current sense resistor between the load and the transistor output, while SRPP puts it between the two transistors (or tubes). Other than that they seem pretty much the same.
After seeing the Aleph, I've been thinking a lot about using an SRPP output stage in a SS amp.
-Jon
You could use an X style topology to replace the input differential in the Aleph, but I think you'd lose the distortion cancelling benefits of the X. Or you could use two bridged aleph output stages in the X, but that seems like overkill when you've already got balanced drive signals and can run a conventional push/pull output stage.
On a slightly different note, has anyone noticed the remarkable similarity of the Aleph to the SRPP circuit used in many tube driver stages and preamps? The Aleph circuit has the current sense resistor between the load and the transistor output, while SRPP puts it between the two transistors (or tubes). Other than that they seem pretty much the same.
After seeing the Aleph, I've been thinking a lot about using an SRPP output stage in a SS amp.
-Jon
Jon T,
The most attractive feature of the Aleph output stage for me is that it uses N channel devices only, which tend to be better, cheaper, and more abundant than their P-channel counterparts. The Aleph patent speaks to the use of the upper device as a reactive current source (which as you note is similar to the tube SRPP circuit), which makes the overall circuit simple in that the OS can be driven from one side of a differential pair.
Presumably, the interest in coupling this output stage to the X front end is to be able to actively drive the upper and lower portions of the output stage from the two out-of-phase outputs from the X front end. If successful, this would overcome a couple of performance limitations in the Aleph output stage without adding a third gain stage in series with the signal. There are at least a couple of ways to do this without upsetting the symmetry of the X front end too much, and while achieving symmetrical behavior in the output stage in the process.
It is also possible to modify the Aleph circuit so that the upper device is actively driven from the previously unused differential-pair output, i.e. as a conventional "totem pole" series stage. I have modeled such a circuit but have not yet built it, and have some residual concerns which make me think, for the moment, that the X front end would be the better path to this goal.
The most attractive feature of the Aleph output stage for me is that it uses N channel devices only, which tend to be better, cheaper, and more abundant than their P-channel counterparts. The Aleph patent speaks to the use of the upper device as a reactive current source (which as you note is similar to the tube SRPP circuit), which makes the overall circuit simple in that the OS can be driven from one side of a differential pair.
Presumably, the interest in coupling this output stage to the X front end is to be able to actively drive the upper and lower portions of the output stage from the two out-of-phase outputs from the X front end. If successful, this would overcome a couple of performance limitations in the Aleph output stage without adding a third gain stage in series with the signal. There are at least a couple of ways to do this without upsetting the symmetry of the X front end too much, and while achieving symmetrical behavior in the output stage in the process.
It is also possible to modify the Aleph circuit so that the upper device is actively driven from the previously unused differential-pair output, i.e. as a conventional "totem pole" series stage. I have modeled such a circuit but have not yet built it, and have some residual concerns which make me think, for the moment, that the X front end would be the better path to this goal.
OK, I took a closer look at X patent and I see what you're doing. The power version (fig. 3) looks a lot like two bridged amps, and you could put an aleph output stage in place of more conventional push/pull one.
I think, though, that the output from the driver stages, points 96 & 97 in figure 3 of the patent, are already in paraphase. So what might be interesting would be to cross connect the two output stages.
In such a schem, the upper output devices would be driven as normal, but the bottom devices would be driven by the opposite half of the circuit. I don't know if this would impact the distortion cancelling effects, though.
I'll have to try this one out in simulation...
-Jon
I think, though, that the output from the driver stages, points 96 & 97 in figure 3 of the patent, are already in paraphase. So what might be interesting would be to cross connect the two output stages.
In such a schem, the upper output devices would be driven as normal, but the bottom devices would be driven by the opposite half of the circuit. I don't know if this would impact the distortion cancelling effects, though.
I'll have to try this one out in simulation...
-Jon
Consider the schematic Petter has posted for us.A +/- output supply connected with the + on the drain of Q2 and Q11 and the - on the drain of Q4 and Q12.Parallel connected transistors omitted for clarity.Now connect two floating supplies as follows.The + of supply #1 to the drain of Q2 and the - to the source of Q11.The + of supply #2 to the drain of Q 11 and the - to the source of Q2.Q4 and Q12 are deleted.A pair of 500 ohm resistors added.One from the source of Q11 to ground and the other from the source of Q2 to ground.This circuit is patented otherwise Nelson Pass could have used it.Each half of the bridge acts like a current source for the other.Only one polarity output device needed.A 50W amp could be made from one pair of IRFP22N50A ,a 277W device available from DigiKey for $2.45 .Since no devices are paralleled no matching should be needed.If you have trouble visualizing this just draw it out on a sheet of paper.Prior art shows both tube (Atamasphere) and BJT (Sumo) versions of this idea.
jnb mentioned totem pole output ....
JNB mentioned something about using the input diff amp to drive both upper and lower output devices in a totem-pole likke configuration. That made me remember the old Siliconix Auto Bias Amplifier from about 1982. JNB, do you remeber this circuit. Is that close to what you're thinking? In the schematic, there's an NPN diff amp with resitive load. Each diff amp collector drives an output MOSFET gate. They do some funky filtering to derive bias control. The particular version I'm looking at is AC coupled to the load. If you're curious, I'm looking at the book: "MOSPOWER Applications" by Siliconix, 1984. Pages 6.97 through 6.94. (As an unrelated aside, this book also has the Cordell MOSFET amp circuit - error correction for the output stage). I'd post a schenmatic but I don't have a scanner :-(
Michael
JNB mentioned something about using the input diff amp to drive both upper and lower output devices in a totem-pole likke configuration. That made me remember the old Siliconix Auto Bias Amplifier from about 1982. JNB, do you remeber this circuit. Is that close to what you're thinking? In the schematic, there's an NPN diff amp with resitive load. Each diff amp collector drives an output MOSFET gate. They do some funky filtering to derive bias control. The particular version I'm looking at is AC coupled to the load. If you're curious, I'm looking at the book: "MOSPOWER Applications" by Siliconix, 1984. Pages 6.97 through 6.94. (As an unrelated aside, this book also has the Cordell MOSFET amp circuit - error correction for the output stage). I'd post a schenmatic but I don't have a scanner :-(
Michael
djk,
I'm also intrigued by the floating bridge (aka "circlotron") output stage, and think that it would work well mated to the X front end. In addition to the changes you suggest above, I would omit R7 and R36 and replace R6 and R37 with Vgs-multiplier type adjustable bias voltage sources (as used in the A75). For a 100W amplifier, Supply #1 and Supply #2 need to be about 50V (36V would work for a 50W amp), but the front end supply can actually be a bit less than 50V (maybe +/- 40V) as it only needs to swing about 25V per side.
BTW, the original US patents on the floating bridge concept issued in the 1950s and have long since expired. The later patents by Ralph Karsten and James Bongiorno resurrect the design with the addition of novel biasing methods appropriate to the gain devices used in each case (triodes and BJTs). As long as you steer clear of the proprietary components of these patents, I would say that you are working in the public domain. Two commercial cases in point are the BAT VK200 and VK500 power amps by Victor Khomenko, and the Lamm power amps by Vladimir Shushurin, both of which I gather use MOSFET "circlotron" type output stages. There is a patent on the Lamm design, but I think it covers the method of physically connecting the output devices to the PC board, not the circuit topology itself.
Michael,
From your verbal description of the Siliconix circuits, I think I have something else in mind -- two things, actually:
One is to use the X front end to drive a totem pole output stage. Here the X front end acts as a more or less conventional differential pair with a folded cascode, where the outputs from each side of the folded cascode are direct-coupled to the upper and lower halves of the totem pole output stage. In this scenario, the signal input is presented between the gate of R20 (re: Petter's diagram above) and ground, while negative feedback is returned to the gate of Q1 only. The noninverting folded-cascode output drives the upper half through a resistor tied between the gates of the upper output devices and the output rail. The inverting folded-cascode output drives the lower half of the output stage through a resistor tied between the gates of the lower output devices and the negative supply rail. To equalize dissipation between the two sides of the folded cascode, a resistor or grounded-gate P-channel MOSFET is added in series with Q19's drain. Bias and balance are set by adjusting the voltage or current sources in the folded cascode, and these must be temperature compensated to provide thermal tracking for the output stage.
The other approach starts with the basic Aleph circuit, and creates a new drive signal for the upper half of the output stage from the previously unused diff-pair output. The most symmetrical way to do this IMO is to add two more current sources to the circuit, each having the same value as that used to bias the input stage. Add one of these current sources between the previously unused drain of the diff pair and the negative supply rail. Add the second between the positive supply rail and the output rail. Next, add a series resistor between the second current source and the output rail to provide the DC bias for the upper output devices in place of the Aleph's NPN control circuit. The resistor just added should have the same value as the resistor being used to generate DC bias for the lower half of the output stage. Finally, add an N-channel MOSFET to the circuit, with its drain connected to the gates of the upper output devices, its source connected to the previously unused drain in the diff pair, and its gate to a voltage source (e.g. zener) referenced to the negative supply rail. These changes create a drive signal for the upper output devices that is equal but opposite in phase to the drive signal presented to the lower output devices. Again, the current sources must be adjustable and temperature compensated to permit bias and balance adjustment and allow for thermal tracking of the output stage.
Having modeled both circuits, I've come away thinking that the first approach may be the better one, as it seems to have less of a problem with phase shift between the paraphase drive signals. But having built neither circuit, I can speak with only limited authority, as I've found that MOSFET models in particular do not always reflect reality in terms of how device capacitances affect phase shifts and overall circuit stability.
My apologies for the length -- hope this helps!
I'm also intrigued by the floating bridge (aka "circlotron") output stage, and think that it would work well mated to the X front end. In addition to the changes you suggest above, I would omit R7 and R36 and replace R6 and R37 with Vgs-multiplier type adjustable bias voltage sources (as used in the A75). For a 100W amplifier, Supply #1 and Supply #2 need to be about 50V (36V would work for a 50W amp), but the front end supply can actually be a bit less than 50V (maybe +/- 40V) as it only needs to swing about 25V per side.
BTW, the original US patents on the floating bridge concept issued in the 1950s and have long since expired. The later patents by Ralph Karsten and James Bongiorno resurrect the design with the addition of novel biasing methods appropriate to the gain devices used in each case (triodes and BJTs). As long as you steer clear of the proprietary components of these patents, I would say that you are working in the public domain. Two commercial cases in point are the BAT VK200 and VK500 power amps by Victor Khomenko, and the Lamm power amps by Vladimir Shushurin, both of which I gather use MOSFET "circlotron" type output stages. There is a patent on the Lamm design, but I think it covers the method of physically connecting the output devices to the PC board, not the circuit topology itself.
Michael,
From your verbal description of the Siliconix circuits, I think I have something else in mind -- two things, actually:
One is to use the X front end to drive a totem pole output stage. Here the X front end acts as a more or less conventional differential pair with a folded cascode, where the outputs from each side of the folded cascode are direct-coupled to the upper and lower halves of the totem pole output stage. In this scenario, the signal input is presented between the gate of R20 (re: Petter's diagram above) and ground, while negative feedback is returned to the gate of Q1 only. The noninverting folded-cascode output drives the upper half through a resistor tied between the gates of the upper output devices and the output rail. The inverting folded-cascode output drives the lower half of the output stage through a resistor tied between the gates of the lower output devices and the negative supply rail. To equalize dissipation between the two sides of the folded cascode, a resistor or grounded-gate P-channel MOSFET is added in series with Q19's drain. Bias and balance are set by adjusting the voltage or current sources in the folded cascode, and these must be temperature compensated to provide thermal tracking for the output stage.
The other approach starts with the basic Aleph circuit, and creates a new drive signal for the upper half of the output stage from the previously unused diff-pair output. The most symmetrical way to do this IMO is to add two more current sources to the circuit, each having the same value as that used to bias the input stage. Add one of these current sources between the previously unused drain of the diff pair and the negative supply rail. Add the second between the positive supply rail and the output rail. Next, add a series resistor between the second current source and the output rail to provide the DC bias for the upper output devices in place of the Aleph's NPN control circuit. The resistor just added should have the same value as the resistor being used to generate DC bias for the lower half of the output stage. Finally, add an N-channel MOSFET to the circuit, with its drain connected to the gates of the upper output devices, its source connected to the previously unused drain in the diff pair, and its gate to a voltage source (e.g. zener) referenced to the negative supply rail. These changes create a drive signal for the upper output devices that is equal but opposite in phase to the drive signal presented to the lower output devices. Again, the current sources must be adjustable and temperature compensated to permit bias and balance adjustment and allow for thermal tracking of the output stage.
Having modeled both circuits, I've come away thinking that the first approach may be the better one, as it seems to have less of a problem with phase shift between the paraphase drive signals. But having built neither circuit, I can speak with only limited authority, as I've found that MOSFET models in particular do not always reflect reality in terms of how device capacitances affect phase shifts and overall circuit stability.
My apologies for the length -- hope this helps!
Joe-
Wouldn't it be simpler to just put another resister load in the other leg of the diff. input pair and use that signal? It seems to me that using a current source for one but not the other leg of the diff pair would result in a serious inbalance in the drive signals.
Perhaps I'm just not following the desciption of your mods. Would it be possible to post a schematic?
Also, in option 1, you suggested using the paraphase outputs of the X differential input to drive the output stage. Unfortunately, this would lose the X circuit's distortion cancelling attributes, and you'd need to use NFB. For the X circuit, you need to take the output signal between the two halves of the circuit, otherwise it's just a fancy diff. pair. In effect, you need two output stages that are bridged together.
If you're interested in other N-channel only output stages, take a look at http://www.aussieamplifiers.com. He's got a big N-channel only amp with a clever biasing scheme. It's class AB, but should be able to be modified into class A. For a description of the circuit, see "Better Audio from Non-Complements" Wireless World, 12/94 p. 988-992.
I'm also very intrigued by non-complementary output stages, since most of the amps I've liked the sound of use a similar design philosphy.
-Jon
Wouldn't it be simpler to just put another resister load in the other leg of the diff. input pair and use that signal? It seems to me that using a current source for one but not the other leg of the diff pair would result in a serious inbalance in the drive signals.
Perhaps I'm just not following the desciption of your mods. Would it be possible to post a schematic?
Also, in option 1, you suggested using the paraphase outputs of the X differential input to drive the output stage. Unfortunately, this would lose the X circuit's distortion cancelling attributes, and you'd need to use NFB. For the X circuit, you need to take the output signal between the two halves of the circuit, otherwise it's just a fancy diff. pair. In effect, you need two output stages that are bridged together.
If you're interested in other N-channel only output stages, take a look at http://www.aussieamplifiers.com. He's got a big N-channel only amp with a clever biasing scheme. It's class AB, but should be able to be modified into class A. For a description of the circuit, see "Better Audio from Non-Complements" Wireless World, 12/94 p. 988-992.
I'm also very intrigued by non-complementary output stages, since most of the amps I've liked the sound of use a similar design philosphy.
-Jon
Jon,
The purpose the current sources is to transfer the signal from the negative supply rail to the output rail without inverting its phase and without passing the signal through additional active stages. In fact, the signal does pass through what amounts to a folded cascode on its way up to the output rail, but IME this can be a pretty transparent experience. In simulations at least, the balance between the drive signals is quite good, as evidenced by relatively little second-order harmonic distortion. It's the added phase shift experienced by the noninverting signal that I think may limit the performance of the circuit. If I ever get a chance to build it, I'll let you know whether it's a practical alternative to the Aleph.
It's true that my proposed modification of the X front end does away with its unique distortion-canceling ability. But this is appropriate to its use as a driver for a single totem-pole output stage. If the goal were to retain the fully balanced nature of the X topology as you suggest, and retain the balanced output stage, I'd be hard pressed to suggest anything better than the complementary-symmetry approach shown in Petter's diagram (also used in the Pass X series, I understand).
One alternative that I might want to try in that case is the later version of the output stage used in the Aleph 0 and 0s amplifiers. In fact, I onced asked Nelson Pass why he didn't use that OS in the X series, and he explained that it really needs to run in class A to work well, which would be prohibitive in a high-power amp.
But this was back when the smallest X amplifier was 350W/ch; now that DIYers are contemplating 100W versions, it might be reasonable to entertain the idea of running the OS in class A. In the case of Petter's circuit, you would need 2.5A per side, and given +/-25V rails, you "only" have to dissipate 250W per channel.
The purpose the current sources is to transfer the signal from the negative supply rail to the output rail without inverting its phase and without passing the signal through additional active stages. In fact, the signal does pass through what amounts to a folded cascode on its way up to the output rail, but IME this can be a pretty transparent experience. In simulations at least, the balance between the drive signals is quite good, as evidenced by relatively little second-order harmonic distortion. It's the added phase shift experienced by the noninverting signal that I think may limit the performance of the circuit. If I ever get a chance to build it, I'll let you know whether it's a practical alternative to the Aleph.
It's true that my proposed modification of the X front end does away with its unique distortion-canceling ability. But this is appropriate to its use as a driver for a single totem-pole output stage. If the goal were to retain the fully balanced nature of the X topology as you suggest, and retain the balanced output stage, I'd be hard pressed to suggest anything better than the complementary-symmetry approach shown in Petter's diagram (also used in the Pass X series, I understand).
One alternative that I might want to try in that case is the later version of the output stage used in the Aleph 0 and 0s amplifiers. In fact, I onced asked Nelson Pass why he didn't use that OS in the X series, and he explained that it really needs to run in class A to work well, which would be prohibitive in a high-power amp.
But this was back when the smallest X amplifier was 350W/ch; now that DIYers are contemplating 100W versions, it might be reasonable to entertain the idea of running the OS in class A. In the case of Petter's circuit, you would need 2.5A per side, and given +/-25V rails, you "only" have to dissipate 250W per channel.
Petter,
If I'm guessing correctly on what you're suggesting, you'll still have a pretty fierce power dissipation (i.e. effiency) to deal with. I had considered something along those lines for a while last fall, then decided, purely as a thought experiment, to go a different route. I'm still mulling the idea over, as it's so whacky on the face of it that I'd like to test it to make sure that it works. I may have more to say on this later.
In the meantime, and along the same lines as what you're (I think) suggesting, it might be possible to tuck an Aleph output 'up under' the outer transistor of the cascode (your Q3 & Q19) of an X front end, using the present top of the cascode as the current source (i.e. the top) of the Aleph output. The obvious objection to this is that it would wreak havoc on the careful 50/50 division of current between the two transistors in the cascode, as the output portion of the circuit (Q3 & Q19) would need to run at a pretty high current. The new current ratio would need to be on the order of 90/10. The upper cascode gate would need to be shifted to being driven from the Aleph output current sense array, rather than a fixed V source. Now you're entering the realm of something bizarre as you'd have a pseudo-cascode with both parts driven, which is something I don't recall ever seeing before, and am not sure what to call it. Depending on how you conceptualize things, you could argue that a cascode is one gain stage, and that if you string the other half of an Aleph output in there, then it too is part of the 'cascode' and by the time you're done you have achieved the impossible...a single stage amplifier capable of a decent amount of wattage.
*Don't* beat me up on this. It's just a thought experiment. I run these things through my head while I drive. The upside is that I am able to utilize otherwise empty time. The downside is that I'm not able to draw it out and convince myself that it either will or will not work.
To have the class A output of an Aleph and the distortion cancelling capabilities of an X, all in what is technically (depending on how you look at it, granted) a single stage amp...hmmm.
Comments?
Grey
If I'm guessing correctly on what you're suggesting, you'll still have a pretty fierce power dissipation (i.e. effiency) to deal with. I had considered something along those lines for a while last fall, then decided, purely as a thought experiment, to go a different route. I'm still mulling the idea over, as it's so whacky on the face of it that I'd like to test it to make sure that it works. I may have more to say on this later.
In the meantime, and along the same lines as what you're (I think) suggesting, it might be possible to tuck an Aleph output 'up under' the outer transistor of the cascode (your Q3 & Q19) of an X front end, using the present top of the cascode as the current source (i.e. the top) of the Aleph output. The obvious objection to this is that it would wreak havoc on the careful 50/50 division of current between the two transistors in the cascode, as the output portion of the circuit (Q3 & Q19) would need to run at a pretty high current. The new current ratio would need to be on the order of 90/10. The upper cascode gate would need to be shifted to being driven from the Aleph output current sense array, rather than a fixed V source. Now you're entering the realm of something bizarre as you'd have a pseudo-cascode with both parts driven, which is something I don't recall ever seeing before, and am not sure what to call it. Depending on how you conceptualize things, you could argue that a cascode is one gain stage, and that if you string the other half of an Aleph output in there, then it too is part of the 'cascode' and by the time you're done you have achieved the impossible...a single stage amplifier capable of a decent amount of wattage.
*Don't* beat me up on this. It's just a thought experiment. I run these things through my head while I drive. The upside is that I am able to utilize otherwise empty time. The downside is that I'm not able to draw it out and convince myself that it either will or will not work.
To have the class A output of an Aleph and the distortion cancelling capabilities of an X, all in what is technically (depending on how you look at it, granted) a single stage amp...hmmm.
Comments?
Grey
teflon boards
Hi Petter,
You mentioned you were considering teflon pcb boards above. If you need the teflon board, I found a cheap source through Ebay - US$30 for 12"x18" Rogers Duroid. Guy said he could supply more than two I bought, so he must have more. A steal at that price, I think. Search for "electro mavin" for their website, it has their contact details.
Hi Petter,
You mentioned you were considering teflon pcb boards above. If you need the teflon board, I found a cheap source through Ebay - US$30 for 12"x18" Rogers Duroid. Guy said he could supply more than two I bought, so he must have more. A steal at that price, I think. Search for "electro mavin" for their website, it has their contact details.
Ordered teflon boards as per your suggestions -- thanks! Nice low dielectric constant at 2.2. Dissipation factor at 10GHz 0.0009 or so .... Should be OK for an audio amp 
I am definitely building PCB's for the input stage now.
You can find more info on these products at http://www.rogers-corp.com/mwu/web/sld001.htm
[Edited by Petter on 04-09-2001 at 03:21 AM]
I am definitely building PCB's for the input stage now.
You can find more info on these products at http://www.rogers-corp.com/mwu/web/sld001.htm
[Edited by Petter on 04-09-2001 at 03:21 AM]