Re: Re: Re: "Circumcising" the F3
Oh yes, I forgot to mention the PSU, but it's important that it's a CLC filter, as the PSU noise rejection of this topology is not very high. If only using a C filter, you can hear the 100Hz when using sensitive speakers.
I found that 5mH and 2*68.000uF keeps the noise down.
Magura
apassgear said:
I run something very similar to this with input buffer and a CLC PSU and results are just as good, though different lineamp, in this case toobs.
Oh yes, I forgot to mention the PSU, but it's important that it's a CLC filter, as the PSU noise rejection of this topology is not very high. If only using a C filter, you can hear the 100Hz when using sensitive speakers.
I found that 5mH and 2*68.000uF keeps the noise down.
Magura
Re: Re: Re: Re: "Circumcising" the F3
In my case I use two secondaries, one each channel, with 15mf + 10mH + 20mf with no apreciable noise on a FE164
Magura said:
Oh yes, I forgot to mention the PSU, but it's important that it's a CLC filter, as the PSU noise rejection of this topology is not very high. If only using a C filter, you can hear the 100Hz when using sensitive speakers.
I found that 5mH and 2*68.000uF keeps the noise down.
Magura
In my case I use two secondaries, one each channel, with 15mf + 10mH + 20mf with no apreciable noise on a FE164
Re: Re: "Circumcising" the F3
The current source has a flat response with regard to frequency.
Disabling the AC current gain in the current source does result
in a different spectrum because the circuit cancels some second
harmonic. At the same time though, it dramatically reduces the
total distortion by effectively doubling the load impedance.
So you have a choice - Lots of second harmonic with lower power,
or much lower distortion. Me, I was pleased to see THD figures
as low as .003% at 1W, 1KHZ. You would not see anything like
that with the constant current source.
Magura said:
The current source has a flat response with regard to frequency.
Disabling the AC current gain in the current source does result
in a different spectrum because the circuit cancels some second
harmonic. At the same time though, it dramatically reduces the
total distortion by effectively doubling the load impedance.
So you have a choice - Lots of second harmonic with lower power,
or much lower distortion. Me, I was pleased to see THD figures
as low as .003% at 1W, 1KHZ. You would not see anything like
that with the constant current source.
Re: Re: Re: "Circumcising" the F3
Hmm, we have made a few blind tests of this, and the conclusion was that people preferred the CCS ov the active curret source. Yes distortion is sure higher with the CCS, this is also the result I have come to, so what could the reason for the preferrence of the CCS be then?
I wonder if this is one of those cases where it's the distortion people like?
Magura
EDIT:
Untill I have found the graphs that showed the non liniarity of the active current source....let's just say that it's my memory playing tricks on me
Nelson Pass said:
The current source has a flat response with regard to frequency.
Disabling the AC current gain in the current source does result
in a different spectrum because the circuit cancels some second
harmonic. At the same time though, it dramatically reduces the
total distortion by effectively doubling the load impedance.
So you have a choice - Lots of second harmonic with lower power,
or much lower distortion. Me, I was pleased to see THD figures
as low as .003% at 1W, 1KHZ. You would not see anything like
that with the constant current source.
Hmm, we have made a few blind tests of this, and the conclusion was that people preferred the CCS ov the active curret source. Yes distortion is sure higher with the CCS, this is also the result I have come to, so what could the reason for the preferrence of the CCS be then?
I wonder if this is one of those cases where it's the distortion people like?
Magura
EDIT:
Untill I have found the graphs that showed the non liniarity of the active current source....let's just say that it's my memory playing tricks on me
Once upon a time I made lengthy posts and was ridiculed for it.
My available time became shorter and my posts shortened as a consequence.
Now, half the time, people think I'm mad at them because I've gotten so terse.
I posted the other day that I thought that someone might have misunderstood something that Nelson wrote...and at the time, that was all I could write before I turned into a pumpkin (no, not a preamp). But it's been bothering me ever since, because that's not the way I like to leave things. However, I don't have sufficient time at the moment to write a small novel like I used to in the days of yore. So, I'm going to try to do a quick review of the Aleph current source, since there seems to be a bit of confusion over exactly what it does and how it does it.
Know ahead of time that I'm already dissatisfied with this post, because I've only got about five or ten minutes.
I'm going to assume that everyone knows how a "normal" current source works. If you're at all hazy, I put some links into a recent post in the Aleph-X thread which will take you to simple explanations of what a current source is. That will save me...you guessed it...time.
Take a look at an Aleph. Any Aleph. They're all pretty much the same. There's a MOSFET on the bottom of the output stage that's hooked up in common Source mode. That is, the signal goes in the Gate and comes out the Drain. This hookup has both current gain and voltage gain. It's unusual for an output to supply voltage gain, but not unheard of. Most outputs provide only current gain because they're followers (common Drain).
Now, the thing about common Source is that it needs a load to function into. Normally, you'd expect to see something like a resistor. It's not too unusual to see a current source. Why use a current source? Well, the voltage gain of a common Source stage is directly dependent on the load impedance. The more impedance, the more gain. So if you want lots of gain, you use lots of impedance. The problem with this is that a really big resistor will give you lots of gain, but the voltage swing is far from optimal. A current source is an attempt to have your cake and eat it too. It provides enormous impedance and yet allows (nearly) full voltage swing. You'll lose a bit to the Vgs (or Vbe, as the case may be). But if you think back to the schematics you've seen where people have used a current source, you might remember seeing that they used a device of the opposite polarity. In this case, seeing as how the gain device is an N-ch MOSFET, you might expect to see a P-ch current source, with its Drain aimed squarely at the N-ch MOSFET.
But that's not what Nelson did.
He turned the current source upside-down, so to speak, so that it's Source is presented to the N-ch MOSFET. So what does that do?
Well, for one thing, your voltage gain drops through the floor. In fact, it's pretty close to unity. That's because the Source has a really low impedance, which is what makes them so attractive in the follower mode. The available current is the same, for the simple reason that anything in series always see the same current at any point along the string. If you set a current source so that 1A comes out of its Drain, then by golly, 1A is present at the Source, too. That's just one of the laws of the universe.
So far, you haven't really gained anything, except perhaps a lower output impedance. At this point, the circuit still looks kinda like a Zen with a front end attached.
But...
If you were to actively drive that upper MOSFET, using the Gate, you could, perhaps, make something of that. So let's run through a few permutations:
--Drive it with higher amplitude and inverted version of the same signal presented to the lower MOSFET. That's a quasi-complementary output. Been there. Done that.
--Drive it with a signal derived from a resistor in series with the output. Hmmm...not such a good idea. Wrong phase. The upper MOSFET actually works against the lower one. This is not good.
--Okay...uh...got it! Invert the phase by sending it through another transistor in common Source (oops! I mean common emitter) mode. This transistor need not be very big because it doesn't have to supply a lot of drive current. The Gates of MOSFETs just don't draw much. In fact, you could probably even get away with a TO-92 case. Hmmm....let's see what's lying around on the bench...oh, look...a ZTX450. Cool.
(Wanna really push the upper MOSFETs? Think in terms of a TO-220. But that's overkill.)
All right, so you've derived a signal from a resistance in series with the output. There are other options, like a voltage derived from a resistor to ground, etc. but I'm already past my time limit now so I'm going to speed past that idea, tempting as it may be. There's an interesting possibility inherent in the series resistor idea, in that it senses the current flow between the amp and the speaker. This allows some Alice In Wonderland things if you want to go there. Suppose the speaker is drawing more current. That means more voltage across the sensing resistor. Ho, hum. But then we look at that little ZTX450. You know what...that booger is capable of gain on its own. What if you were to ask it to put just a little bit of oomph! into driving that MOSFET? (Which after all, functioning as a follower, kinda-sorta, has no voltage gain of its own.) This is an odd concept. The current source actually can shove more current than the gain MOSFET on the bottom. But then your signal starts getting lopsided, which is one reason that it's a good idea to stay near the 50/50 drive ratio, at least assuming that you want to keep distortion down.
Up against the wall. I'll try to take up the narrative again, but I don't know when. I normally try to run through and edit things for clarity. This one's going in rough. It may be a bumpy landing.
Grey
My available time became shorter and my posts shortened as a consequence.
Now, half the time, people think I'm mad at them because I've gotten so terse.
I posted the other day that I thought that someone might have misunderstood something that Nelson wrote...and at the time, that was all I could write before I turned into a pumpkin (no, not a preamp). But it's been bothering me ever since, because that's not the way I like to leave things. However, I don't have sufficient time at the moment to write a small novel like I used to in the days of yore. So, I'm going to try to do a quick review of the Aleph current source, since there seems to be a bit of confusion over exactly what it does and how it does it.
Know ahead of time that I'm already dissatisfied with this post, because I've only got about five or ten minutes.
I'm going to assume that everyone knows how a "normal" current source works. If you're at all hazy, I put some links into a recent post in the Aleph-X thread which will take you to simple explanations of what a current source is. That will save me...you guessed it...time.
Take a look at an Aleph. Any Aleph. They're all pretty much the same. There's a MOSFET on the bottom of the output stage that's hooked up in common Source mode. That is, the signal goes in the Gate and comes out the Drain. This hookup has both current gain and voltage gain. It's unusual for an output to supply voltage gain, but not unheard of. Most outputs provide only current gain because they're followers (common Drain).
Now, the thing about common Source is that it needs a load to function into. Normally, you'd expect to see something like a resistor. It's not too unusual to see a current source. Why use a current source? Well, the voltage gain of a common Source stage is directly dependent on the load impedance. The more impedance, the more gain. So if you want lots of gain, you use lots of impedance. The problem with this is that a really big resistor will give you lots of gain, but the voltage swing is far from optimal. A current source is an attempt to have your cake and eat it too. It provides enormous impedance and yet allows (nearly) full voltage swing. You'll lose a bit to the Vgs (or Vbe, as the case may be). But if you think back to the schematics you've seen where people have used a current source, you might remember seeing that they used a device of the opposite polarity. In this case, seeing as how the gain device is an N-ch MOSFET, you might expect to see a P-ch current source, with its Drain aimed squarely at the N-ch MOSFET.
But that's not what Nelson did.
He turned the current source upside-down, so to speak, so that it's Source is presented to the N-ch MOSFET. So what does that do?
Well, for one thing, your voltage gain drops through the floor. In fact, it's pretty close to unity. That's because the Source has a really low impedance, which is what makes them so attractive in the follower mode. The available current is the same, for the simple reason that anything in series always see the same current at any point along the string. If you set a current source so that 1A comes out of its Drain, then by golly, 1A is present at the Source, too. That's just one of the laws of the universe.
So far, you haven't really gained anything, except perhaps a lower output impedance. At this point, the circuit still looks kinda like a Zen with a front end attached.
But...
If you were to actively drive that upper MOSFET, using the Gate, you could, perhaps, make something of that. So let's run through a few permutations:
--Drive it with higher amplitude and inverted version of the same signal presented to the lower MOSFET. That's a quasi-complementary output. Been there. Done that.
--Drive it with a signal derived from a resistor in series with the output. Hmmm...not such a good idea. Wrong phase. The upper MOSFET actually works against the lower one. This is not good.
--Okay...uh...got it! Invert the phase by sending it through another transistor in common Source (oops! I mean common emitter) mode. This transistor need not be very big because it doesn't have to supply a lot of drive current. The Gates of MOSFETs just don't draw much. In fact, you could probably even get away with a TO-92 case. Hmmm....let's see what's lying around on the bench...oh, look...a ZTX450. Cool.
(Wanna really push the upper MOSFETs? Think in terms of a TO-220. But that's overkill.)
All right, so you've derived a signal from a resistance in series with the output. There are other options, like a voltage derived from a resistor to ground, etc. but I'm already past my time limit now so I'm going to speed past that idea, tempting as it may be. There's an interesting possibility inherent in the series resistor idea, in that it senses the current flow between the amp and the speaker. This allows some Alice In Wonderland things if you want to go there. Suppose the speaker is drawing more current. That means more voltage across the sensing resistor. Ho, hum. But then we look at that little ZTX450. You know what...that booger is capable of gain on its own. What if you were to ask it to put just a little bit of oomph! into driving that MOSFET? (Which after all, functioning as a follower, kinda-sorta, has no voltage gain of its own.) This is an odd concept. The current source actually can shove more current than the gain MOSFET on the bottom. But then your signal starts getting lopsided, which is one reason that it's a good idea to stay near the 50/50 drive ratio, at least assuming that you want to keep distortion down.
Up against the wall. I'll try to take up the narrative again, but I don't know when. I normally try to run through and edit things for clarity. This one's going in rough. It may be a bumpy landing.
Grey
Magura said:
I don't believe that is a correct quotation. The Aleph CCS has
a frequency response which is flat beyond 20K unless you slow
it down deliberately. It does, however, contribute its own
distortion, just as any gain device will. The primary distortion is
2nd harmonic, and toward that end the system enjoys some
cancellation of 2nd between the negative and positive halves.
You will find that both the CCS and the Aleph CS have this distortion factor.
In an apples-to-apples comparison found in fig 9 of the ZV9 article,
you will see that the Aleph CS reduces distortion by a factor of
5 to 10 over exactly the same circuit with no AC component to
the CS.
They say that any landing you can walk away from is a successful landing. To the extent that the pilot walked away from my previous post, I guess maybe it qualified as a successful landing.
But no one ever seems to get around to asking the plane's opinion of the matter...
*****
So I'm sitting here with a minute or two, trying to untangle the mess left after the crash, and POW, I get whacked by two or three ideas in quick succession. Here's the most "Aleph" of them:
Step 1) Take an Aleph, set up as per usual.
Step 2) Replace the front end with a complementary differential.
Step 3) Now duplicate the entire output stage in mirror-image, with P-ch devices, feeding to--and from--the same output. The input to the P-ch gain MOSFET (which is up top, since it's a P-ch device) comes from the N-ch half of the input. The take-off points for the P-ch Aleph current source are the same as the ones for the N-ch. There's no trouble in driving an extra Aleph current source because there's plenty of current available at the output.
This leaves you with complementary gain devices in a transconductance hookup. That's not a really excellent output stage by itself because the impedance is way too high, but you've also got a complementary pair of Aleph current sources. If my shoot-from-the-hip scenario is viable, it should offer lowered distortion and give you the option of balancing "overdriven" Aleph current sources against one another so that the output is no longer asymmetrical, i.e. distorted.
An Aleph-X version would be trivial--just hang another output off the other side of the input.
Lemme think about this.
Grey
But no one ever seems to get around to asking the plane's opinion of the matter...
*****
So I'm sitting here with a minute or two, trying to untangle the mess left after the crash, and POW, I get whacked by two or three ideas in quick succession. Here's the most "Aleph" of them:
Step 1) Take an Aleph, set up as per usual.
Step 2) Replace the front end with a complementary differential.
Step 3) Now duplicate the entire output stage in mirror-image, with P-ch devices, feeding to--and from--the same output. The input to the P-ch gain MOSFET (which is up top, since it's a P-ch device) comes from the N-ch half of the input. The take-off points for the P-ch Aleph current source are the same as the ones for the N-ch. There's no trouble in driving an extra Aleph current source because there's plenty of current available at the output.
This leaves you with complementary gain devices in a transconductance hookup. That's not a really excellent output stage by itself because the impedance is way too high, but you've also got a complementary pair of Aleph current sources. If my shoot-from-the-hip scenario is viable, it should offer lowered distortion and give you the option of balancing "overdriven" Aleph current sources against one another so that the output is no longer asymmetrical, i.e. distorted.
An Aleph-X version would be trivial--just hang another output off the other side of the input.
Lemme think about this.
Grey
Interesting...I wonder, would there be sonic advantages to this over the aleph output stage, or an F4 output stage, or an Aleph 0 output stage, or the Aleph X output topology...... or would it end up being roughly equivalent to one of the above, but with extra parts?
Nelson--any thoughts as to how the concept pieces sound??
JJ
well, this thread seems to have gone "phhhhhhht".....
But is it due to lack of interest? I really like my Mini aleph, better than my a-75, and better than my X3 (although I have to say, I think the X3 has best overall clarity and detail, perhaps if it just had more bias....) but I digress, I really like my aleph, and I would really like to do a few incremental improvements on the design. Grey's post was very intriguing and left me pondering, but it seemed to bit too 'complementary' to be called single-ended. At which point--would it still be an aleph?
So I was wondering, did this thread to "phhhhhhht" due to lack of interest, or due to some other reason?
JJ
But is it due to lack of interest? I really like my Mini aleph, better than my a-75, and better than my X3 (although I have to say, I think the X3 has best overall clarity and detail, perhaps if it just had more bias....) but I digress, I really like my aleph, and I would really like to do a few incremental improvements on the design. Grey's post was very intriguing and left me pondering, but it seemed to bit too 'complementary' to be called single-ended. At which point--would it still be an aleph?
So I was wondering, did this thread to "phhhhhhht" due to lack of interest, or due to some other reason?
JJ
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.