| lumanauw |
After reading Leach paper, I get impression of the importance of Re in differential stage. It is to advoid TIM (Differential overload), so Re should be about 100-300ohm attached to source/emitor of differential pair.
But looking at all the designs by Mr.Pass, it never uses Re at all. Starting from A40, ONO, to ZenV7.
Does the issue of TIM is not having any relevance to good sounding power amp? All Mr. Pass designs are good sounding, but not having Re at all. |
|
|
| Christer |
Few amplifiers use as much emitter degeneration as the Leach,
and some use none at all, as you have noticed. While a high
value extends the linear region and improves overload protection,
it also has drawbacks such as lowering the gain, so it is as with
most things, a compromise. The Leach is a quite old design, using
rather slow output devices, for instance. An amplifier with faster
devices and otherwise designed for higher bandwidth may not
need as much overload margin as the Leach, assuming the input
is properly LP filtered. |
|
|
| Nelson Pass |
Re is not the only place where you can degenerate the signal,
although it has appeared in my designs at various times.
More often than not, my circuits are simple and have so low
an open loop already, that they don't need degeneration to
lower the open loop gain. Like anything else, try it both ways
and see what you like. |
|
|
| Bricolo |
Maybe 2 stage designs (Pass) don't have as much gain "margin" so they can't use as much degeneration.
But 2 stages amps have certainly less TIM, due to the simplier circuit path
Am I right? |
|
|
| Nelson Pass |
| quote: | Originally posted by Bricolo
Maybe 2 stage designs (Pass) don't have as much gain "margin" so they can't use as much degeneration.
But 2 stages amps have certainly less TIM, due to the simplier circuit path Am I right? |
That's the idea, anyway. Personally I think the emphasis on
TIM is too high. Real life measured signals in audio material,
vinyl and CD, don't show the kinds of slew rates that would be
postulated to overdrive the input stage (played at reasonable
levels, of course). This stuff has been looked at by myself and
Peter Walker, both published elsewhere, and supposedly for
TIM to be a problem, it requires slew rates to exceed 1/10 the
rating of the amplifier. As a practical matter, this means that
an intelligently designed amp at 100 watts and 15 V/uS shouldn't
have anything to fear.
At the same time, there have been amps rated at 1,000 V/uS
that didn't go anywhere, so even if it is a factor, it clearly isn't
the only one. The same is true of THD. There have been a
number of amps in the past that measured THD in a few parts
per million that were poorly thought of. |
|
|
| djk |
" The Leach is a quite old design"
And the 300 ohm resistors are a very recent addition.
"Personally I think the emphasis on
TIM is too high. "
I'm old. So old I own some of those black things they call 'LPs'. Some of those are even scratched.
On a run-of-the-mill amplifier the scratches can seem to last forever, a broken-glass hall-of-mirrors effect.
On the Leach the scratch is a short 'tic', and its done. On the run-of-the-mill amplifier the scratches are totally mixed in with the music. On the Leach the scratch reproduces up front in a different plane from the rest of the music. You have to hear this to totally understand.
It may not be TIM that makes the run-of-the-mill amplifier sound so poor.
It may not be low TIM that makes the Leach sound so much better.
I hear the newest Leach boards sound much better. Maybe my 25+ year old ones will fail and give me an excuse to find out. |
|
|
| lumanauw |
| quote: | | Like anything else, try it both ways |
I've made a simple experiment, about placing Re or jumper it. Re is 100ohm. The result, with Re the sound is dull, without Re, the sound is more detailed. That is why I ask this question. To me, without Re is better sound. But reading leach paper it gives me a slight fear of TIM.
Thanks Mr. Pass, for the explenation. What is really the sound of amplifier having TIM?
Is this "Re eliminating" can be applied also in VAS? I see in A40, Mr. Pass do not use Re in the VAS transistor. If I want to eliminate Re in VAS transistor, assuming the Vbe is 0.65V, what is the voltage drop in the differential collector resistor? Should it be more than 0.65V or less than 0.65V? |
|
|
| Nelson Pass |
For every diff pair, there seems to be a sweet spot, true of the
Aleph P/ BSOZ and all others. If it sounds better, why worry
about TIM? |
|
|
| GRollins |
One word of caution--if you intend to build a circuit that creates a balanced signal from a single ended input, you will find that putting too much resistance between the Sources/emitters/cathodes will interfere somewhat with the phase splitter aspect of a differential. Otherwise, have at it.
Grey |
|
|
| jam |
How? Please explain that.
Regards,
Jam |
|
|
| lumanauw |
| quote: | | One word of caution--if you intend to build a circuit that creates a balanced signal from a single ended input, you will find that putting too much resistance between the Sources/emitters/cathodes will interfere somewhat with the phase splitter aspect of a differential. Otherwise, have at it. |
What I learn (from you) differential working as unbalance to balance converter MUST muse CCS, not R loading. I've tried this and it is very-very true. CCS makes the signal have the same height.
Don't know about unbalance to balance converter must advoid Re. Could you explain about this? Is this the same case as orignal SOZ (having 100R between sources of differential) VS ZV7 (not having it, using CCS)?
I've always wanted to ask this for a long time. Mr.Pass, why you make the SUSY patent (#5.376.899) as it is? Why using 2 CCS (26) and (27) with (40) bridging the sources? Why not using 1 CCS like ordinary differential instead of this 2 CCS+bridging resistor? There must be something very important that you decided not to use 1 CCS. |
|
|
| Nelson Pass |
If you are using a CCS to bias BSOZ or similar circuits, the
degeneration resistance on the diff pair doesn't degrade the
ability of the circuit to transform balanced to unbalanced or
vice versa. It does affect the gain of the circuit, and more subtley,
the character of the sound.
As to the last part, 2 CCs's instead of one: totally arbitrary.
I felt like it. :cool: |
|
|
| lumanauw |
| quote: | As to the last part, 2 CCs's instead of one: totally arbitrary.
I felt like it. |
Cool answer from the master !. Is this mean 1 CCS can be applied to SUSY circuit? Like the version of x100 backengineered could have 1 CCS only instead of 2CCS+bridging source resistors in the differential stage? |
|
|
| GRollins |
Oh, blast and bother...
I got pulled away from the computer and didn't take time to finish what I was saying properly. The annoying part is that I didn't even leave enough bread crumbs to pick up the thought later.
I remember the idea of fissioning the current source into two, at which point you'd have a node between the two current sources and the two Sources/emitters; essentially a zero ohm connection (no degeneration resistors involved). You can stick a resistor into the link between the halves as Nelson does with the SOZ and also (I believe--don't have the schematic handy) the BOSOZ. As that resistance approaches infinity the non-driven half of the differential will no longer get enough signal to produce the opposite phase of the output. The logical end point of the thing being an open circuit when there's no communication between the two halves, hence no opposing phase at all.
Okay, I'm able to reconstruct that much of my train of thought.
The missing piece of the puzzle was how I was going to relate that to a Leach amp, of all things.
Harrumph. Grumble. Snort.
Oh, well...apologies to anyone I've confused. If it's any consolation, I've gotten myself confused, as Prof. Leach is a pretty unlikely fellow to run parallel current sources, etc. and for the world I can't remember where I was going with that thought.
Grey |
|
|
| GRollins |
Oh, and yes, a SUSY circuit will work just fine with one current source.
Grey |
|
|
| xocoatl |
Is it possible? I don't mean resistive loading. I'm just curious if it would work with JFET dual diff-pair, where one pair is a sink for others current like in Borbely's all fet amps.
http://www.borbelyaudio.com/image310.gif
For sure the cicuit diagram would look cool if it where "quad symmetric". I wonder if there's a way to make upper part of an amp mirror the distortion of the lower, like the left <-> right in normal SuSy. |
|
|
| GRollins |
SUSY without a current source? Yes.
But...
That circuit isn't "X" as it stands. You'd have to duplicate all the stuff to the right of the input differentials over on the left hand side and arrange the feedback properly.
Grey |
|
|
| Nelson Pass |
The X150 through X350 would be said to be "quad-symmetric"
in that sense. They work fine, although cancellations from the
top and bottom halves are "passive", and not from feedback. |
|
|
| GRollins |
Taking them from the realm of "Super Symmetry" to "Hyper Symmetry" I presume?
Grey |
|
|
| lumanauw |
| Quad-symmetric = making susy with complementary differential? The VAS will be push pull for left and right half? |
|
|
| slowhands |
| While I agree that schematics ofr symmetric amps are artistically attractive, there should be technical justification for them, given that they effectively double the parts cost and halve the reliability of a given design. I hate to say it, but a good single ended design seems to work just as well at half the cost. They may be quite boring to look at but meet my design objectives better. |
|
|
| GRollins |
The justification in this case is that it takes a balanced/bridged circuit in order to have an X (aka SUSY) circuit. If you assume that the circuit will have a complementary output--as opposed to an Aleph-style output--then you've already got the P devices in the output and there's very little extra needed in order to make an X front end carrying complementary differentials. Given that it's rarely the front end of an amp that breaks, there's not much of a reliability problem, and you get an extra distortion-reducing mechanism for free. And one that doesn't involve negative feedback, at that.
Grey |
|
|
| Nelson Pass |
| quote: | Originally posted by lumanauw
Quad-symmetric = making susy with complementary differential? The VAS will be push pull for left and right half? |
Jackpot. :cool: |
|
|
| Fred Dieckmann |
"here's very little extra needed in order to make an X front end carrying complementary differentials." - GR
Perhaps you can draw a schematic on how to do this. I am sure that there would be a great interest in a two jfet pair front end Aleph X and a home-brew version of the jfet input X amps( the second voltage stage should be pretty easy). It would give you opportunity to come up with another X-name and share more of Mr. Pass's design principles with the DIY community. Is anybody else interested in seeing something like this? |
|
|
| lumanauw |
For me it is a big surprise. I always tought that Mr. John Curl and Mr. Erno Borbely are in the field of complementary differential, while Mr. Pass is on the single differential. Just my imagination. It turns out every master can use everything exist, and makes good amp with any circuit.
Just imagining. Maybe the X150-350 looks like A75 mirrored in horizontal direction, using both legs of differential for left and right part of amp (no, no, how the folded cascode goes with this?)
Only X150-350? What about X1000? Still single differential?
| quote: | | Is anybody else interested in seeing something like this? |
I do. |
|
|
| Fred Dieckmann |
| Brwaaaak..... SuperSymmetry!......Brwaaak! |
|
|
| jam |
| Speaking of parrots............... |
|
|
| lumanauw |
| Is this mean for every half (left or right half), X150-350 have 3 stages (Differential(shared together)+push-pull VAS+Curent amplifier), and X1000 only 2 stages (Folded Cascode(shared together)+Current amplifier)? |
|
|
| GRollins |
Although I haven't seen schematics, I would not be surprised if they were also two stage amps.
Grey |
|
|
| moe29 |
| According to the users manuals all the X amps are just two gain stages. |
|
|
| Nelson Pass |
| quote: | Originally posted by GRollins
Although I haven't seen schematics, I would not be surprised if they were also two stage amps. |
Technically, the later circuits would be considered 3 stage, as
the initial stage has a diff pairs driving common source level
shifters. The composite gain has the same open loop as the
original folded cascoded diff pairs, so we let the literature be. |
|
|
| lumanauw |
The initial Xamp and the current Xamp has changed the topology, now without Folded Cascode?
| quote: | Technically, the later circuits would be considered 3 stage, as the initial stage has a diff pairs driving common source level shifters. The composite gain has the same open loop as the
original folded cascoded diff pairs, so we let the literature be |
It seems that it is a 3 stages power amp, complementary differential(shared together), driving pushpull VAS, and Current amp in the final.
| quote: | | common source level shifters |
= VAS? |
|
|
| Nelson Pass |
The X1000 and X600 retain the original topology with the
folded cascode, all other Class AB X amps use the circuit
described above, and the XA amps use a circuit very similar
to Grey's as published in the forum, and have no cascoding. |
|
|
| lumanauw |
| quote: | | all other Class AB X amps use the circuit |
In this forum I read that some expert entering a highend audio show, finds that the best soundsystem is made by X250 (If I dont forget).
Mr. GRollins, would you mind to start a new thread on this amp backengineered? Mr. Pass have give a clue, that it is different than X100 backengineered (it is not using folded cascode).
| quote: | | , and have no cascoding. |
Just experimenting with cascoding yesterday. The results are "THEY ARE GOOD". But only doing it in Differential and VAS (small signal transistors). The sound improves alot.
But have doubt to do this in output stage (Mr. Pass patent place cascode in every stage of power amp), after reading some people's comment. Is it different result if we are doing it in the output stage (with power transistors)? Why? |
|
|
| xocoatl |
| quote: | | VAS (small signal transistors) |
:dodgy:
I read it few times in this forum. It's not true. I was told that small signal is below Ut (thermokinetic potential?!), but VAS swings from one rail to the other.
lumanauw, I don't say it to be annoying, but to be sure that I understand it all clearly. :) |
|
|
| Nelson Pass |
"Common Source level shifters = VAS?"
Lots of things could comprise a VAS, including the folded cascode.
As I said before, topologically you could see the voltage gain
stage of the X150 through 350 as consisting of two stages,
however we set the open loop gain to give the equivalent
of the folded cascode. This saves us the heavy voltage loss
and complexity of the folded cascode and allows a single set
of rail voltages for the front end (passively decoupled) and the
output stage. |
|
|
| lumanauw |
Maybe this is out of theory. But I often see the output stage (current amplifier) as a totem pole (with bias). A totem pole have 2 transistors, NPN in the top and PNP in the bottom. The base tied together, the emitors also tied together. This is usually found in mosfet gate driver (for smps usually). The emitors will follow whatever voltage in the base. The output stage of power amp have similiar topology, but with bias.
So, if the output only follows what is in the base, consequently the base driver must have already full swing. This is done by VAS, where the VAS is done by taking output from the collector (this can give full swing, doesn't depend on the VAS transistor's base voltage, like emitor follower, where Vbe is always 0.6V). So, it is right, that in VAS all the voltage swing has happened, just don't have enough current to drive speaker. This is done by final stage (amplifying the current).
xocoatl,
What I mean by small signal is the size of the transistor, not the function of transistor. :)
*Mr. Pass, you haven't read the emails at nelson@passlab.com?* |
|
|
| LBHajdu |
I’m very interested on how these two very different input topologies sound, when compared to each other. I’m also interested as to why a different topology was used for the lower wattage / less costly versions. My guesses are:
a) people wanted singal ended inputs
b) the transformer needs less taps or you need less transformers |
|
|
| Nelson Pass |
| quote: | Originally posted by lumanauw
Maybe this is out of theory. But I often see the output stage (current amplifier) as a totem pole (with bias). A totem pole have 2 transistors, NPN in the top and PNP in the bottom. The base tied together, the emitors also tied together. This is usually found in mosfet gate driver (for smps usually). The emitors will follow whatever voltage in the base. The output stage of power amp have similiar topology, but with bias.
So, if the output only follows what is in the base, consequently the base driver must have already full swing. This is done by VAS, where the VAS is done by taking output from the collector (this can give full swing, doesn't depend on the VAS transistor's base voltage, like emitor follower, where Vbe is always 0.6V). So, it is right, that in VAS all the voltage swing has happened, just don't have enough current to drive speaker. This is done by final stage (amplifying the current).
*Mr. Pass, you haven't read the emails at nelson@passlab.com?* |
(It's nelson@passlabs.com ), you left off the "s"
Your analysis is correct, the X series has complementary followers
as you describe, and the full voltage swing comes off a Drain of
a transistor (which you can call a VAS), but this Drain can be from
a device operated either as Common Source or Common Gate.
In the X600 and X1000 it is Common Gate, in the other X amps
(non XA, which do not have followers) it is Common Source. |
|
|
| Nelson Pass |
| quote: | Originally posted by LBHajdu
I’m very interested on how these two very different input topologies sound, when compared to each other.
I’m also interested as to why a different topology was used for the lower wattage / less costly versions. My guesses are:
a) people wanted singal ended inputs
b) the transformer needs less taps or you need less transformers |
True on both a) and b)
As to the difference in sound, we have tested these changes
(there is yet a 3rd, 4th and 5th front end you haven't heard
about and which are not in production) on unknowing subjects,
and nobody noticed a difference. The later circuits we deem
more appropriate for amplifiers where we have chosen to
include the output stage in the loop, although it's perfectly
suitable for the case where it is not.
The design of the newer front end also does not lend itself to the
higher voltages of the X600 and X1000, and since they were
already so well loved and reviewed, we opted not to alter them
at all. |
|
|
| PRR |
> slight fear of TIM.
Oh, don't be afraid of TIM. A little slew-limiting sometimes "improves" sound. And most low-feedback designs, and most designs with gain compensated for that gain, are unlikely to ever hit bad TIM.
> What is really the sound of amplifier having TIM?
Try it. Build this:
The LM301 is externally compensated. I put it in a circuit with gain of 3 and a 3:1 loss pad in front for overall unity gain.
For LM301 at gain of 3, the suggested compensation cap is about 10pFd, the slew rate is about 2V/µSec.
Put this between your preamp and your power amp. The peak signal level there is around 2V. Jung's early listening tests suggest that it works out that if your slew rate in V/µSec is greater than your peak signal voltage, audio passes cleanly. Common sense suggests that, on most real music, much lower slew rates won't nick the audio.
So with just the 10pFd the slew rate is adequate to avoid TIM on audio signals. (If your source has strong supersonic signals, like a leaky FM tuner or a moving-coil cartridge with a phono preamp that goes flat above 50KHz, fix it.) The LM301 is not a fabulous chip, but within its limits it doesn't suck so bad. And I'm trying to demonstrate gross TIM, not subtle effects.
Now use the DIP switch to switch-in additional compensation capacitance. 300pFd should be noticable on strong treble. 3,000pFd could turn everything to mush, treble splattering all over the audio band. |
|
|
| lumanauw |
Hi, PRR,
Thanks for the cct. I will try it, to hear what is TIM looks like.
| quote: | | Oh, don't be afraid of TIM. A little slew-limiting sometimes "improves" sound |
This is the pleasant answer. |
|
|
| lumanauw |
Hi, Mr. Pass,
| quote: | and the full voltage swing comes off a Drain of a transistor (which you can call a VAS), but this Drain can be from
a device operated either as Common Source or Common Gate.
|
The drain output comes from common base=folded cascode front end, like the x100backengineered.------X600 and x1000 uses folded cascod+susy patent
The drain output comes from common source=ordinary 3 stages amp? (differential, VAS, current output)? I cannot think of other VAS configuration that the signal comes to the base, emitor tied to the rail, besides ordinary/old fashioned design.-----smaller X amp uses conventional 3stages amp+susy patent?
| quote: | | The later circuits we deem more appropriate for amplifiers where we have chosen to include the output stage in the loop, although it's perfectly suitable for the case where it is not. |
This is also important hint, wheter the outputs are in the feedback loop or not. |
|
|
| Nelson Pass |
I see that you will hound me until I deliver more hints.
The input stage of the X150 - 350 is a dual differential JFET
input feeding quad-symmetrical common emitters. These have
been degenerated at the JFET outputs and the emitters until we
get the same transconductance for the whole as the original
folded cascode MOSFET circuit, which had a transconductance
of about .05 Siemens. |
|
|
| lumanauw |
Surprise surprise!!
I always tought passlabs using Mosfets for differential, especially TO-220 size. I see Jfets only in ONO. It is true, every master can make good things with everything.
Jfets have low voltage. Cascode is a big chance here.
Got 2 more question.
1. Full complementary Jfets (K389-J109) can use 2 way of biasing. One is with CCS or resistor load, two is with stacking each other (like Mr.Curl likes to use them). Which is the one in smaller Xamp?
2. Is there any "Magic Resistor" involved?
*Mr. Pass, I forgot the "s" in nelson@passlabs.com, so I resend the email. Any comment?* |
|
|
| Nelson Pass |
| quote: | Originally posted by lumanauw
Jfets have low voltage. Cascode is a big chance here.
1. Full complementary Jfets (K389-J109) can use 2 way of biasing. One is with CCS or resistor load, two is with stacking each other (like Mr.Curl likes to use them). Which is the one in smaller Xamp?
2. Is there any "Magic Resistor" involved? |
The JFETs are cascoded.
The JFETs are biased with resistors.
I need a definition of magic resistor, but the Sources of the
differential JFETs are directly connected. |
|
|
| lumanauw |
| quote: | | I need a definition of magic resistor, but the Sources of the differential JFETs are directly connected. |
The sources of Jfets are directly connected. It means no Re here.
Mr.Pass, what I mean by magic resistor is the definition I borrowed from the alephx thread. The resistor from both the outputs to the junction of emitors of differential (to help DC offset). IF this full complementary design uses such a "Magic Resistor", will there be 4 of them? (while in alephx there is only 2 of them)
| quote: | | The JFETs are biased with resistors. |
Mr. Pass, what is your personal thinking of stacking mosfets (Like Mr. Curl likes to do it)? What is the pro and cons than having CCS or R load? |
|
|
| lumanauw |
Just think of something.
| quote: | | The JFETs are biased with resistors. |
| quote: | | the Sources of the differential JFETs are directly connected |
We are talking about X configuration here. If the Jfets differential having no Re, just tied together, would it stable just to bias it with Resistor? There is no Re, shouldn't it be CCS instead of R loading to get matched operation? (example, if the amp is driven by single ended input, not balanced input). Or I miss something? |
|
|
| Nelson Pass |
You are missing something. The dual pairs of Matched JFETs
are self biasing (in the manner taught by Curl) and are not
biased on their Sources with anything except a single resistor
to set the overall current. The output bias resistor (one for
each JFET drives the Common Source level shifters. |
|
|
| lumanauw |
Thanks Mr.Pass for the answer. It answers 2 things. The Jfets are self biased to each other like Mr.Curl used to.
Also if the configuration is like this, there cannot be any "Magic Resistor" Involved.
In the alephX theread there is problem with DC offset, so it comes "Magic Resistor". Does your small X amp using other kind of DC stabilizer? Servo?
Is the level shifter (VAS) mosfets ? If N and P mosfets are used, the resistor in the leg of differential cannot be the same, because they have slight different Vgs, between N and P type (just like A75, the resistors can be adjusted by VR or have different fix value)?
Mr. Pass, in your personal opinion, is there any merit of this configuration (stacking fets, because they have negative Vgs), instead of biasing them with CCS to rails?
One more question. Why I hardly find your design with Bipolar transistor (in differential or final stage)? |
|
|
| GRollins |
At the rate you're going it would be more efficient to simply ask Nelson for a full schematic of the X amplifier.
I've got two separate ideas in mind as to how to arrange a single pair of 'magic resistors' to do the trick, but until I get some parts in I can't check to see if either works.
Knowing Nelson, it's unlikely that he's using a servo.
It seems rather unlikely that he would use self-biased complementary differentials if he didn't think that the topology had merit. JFETs have less noise than MOSFETs or bipolars and the 2SK389/2SJ109 duals offer a nifty way to get FETs that are fairly well matched and will track together thermally. Biasing with a current source to the rail works--Nelson did so in the A-75, for instance--but this is more elegant.
His older designs are chock-full of bipolars. My Threshold S-500s have quite a few. But bipolars tend to lend themselves to circuits with high gain. The high gain, in turn, needs to be gotten rid of. That implies high rates of negative feedback, which brings along unpleasant baggage as far as sound quality. MOSFETs also tend to be easier to work with, thermally.
Grey |
|
|
| lumanauw |
Grey,
Do you think Mr.Pass will give the sch, since the amp is still in production? I will try anyway.
I want to ask you a question. In full complementary X design, what happens if we put current mirror in top and bottom of both the differential legs? Will the cct work at all? Current mirror acts to balance current in both legs. If we use both legs for audio signal, would current be imbalance in purpose, to drive each side of X amp?
Or if we put these current mirrors, the X cct still works and DC becomes better? |
|
|
| lumanauw |
X amplifier schematic is not available for public, since the amp itself is still in production. This is the answer I got from Mr.Pass.
Thanks anyway, Mr.Pass |
|
|
| slowhands |
| quote: | Originally posted by lumanauw
I've made a simple experiment, about placing Re or jumper it. Re is 100ohm. The result, with Re the sound is dull, without Re, the sound is more detailed. That is why I ask this question. To me, without Re is better sound. But reading leach paper it gives me a slight fear of TIM.
....
What is really the sound of amplifier having TIM?
Is this "Re eliminating" can be applied also in VAS? I see in A40, Mr. Pass do not use Re in the VAS transistor. If I want to eliminate Re in VAS transistor, assuming the Vbe is 0.65V, what is the voltage drop in the differential collector resistor? Should it be more than 0.65V or less than 0.65V? |
OK, TIM and Re and VAS design, a lot of concepts. At the heart is TIM, but I'd like to sort of unravel the tangle of ideas here.
ABOUT Re: You might not understand why designers choose to "add Re" to input pairs. Here is my understanding of the usual reasons.
The main reason to add modest Re's to the input pairs is to match the trasistors, even if they are not matched pairs. Ideally, you would not need to do this, because you would use a matched dual transistor, on the same die and thermally coupled in the same package. But if these are too expensive, or if you need higher voltage rating than available dual transistors (60v max), then you use unmatched parts and depend on a small Re to equalize them. When you see Re in the range of 22-68 ohm, probably the intent is to equalize the pair.
When you go to higher Re, you are intending something else: trading transconductance for linearity of the input pair. Leach makes this point very clearly, as does Self. Using no Re, the linear range is typically +-57 mv. Using a 100 ohm value, the range is +-200mv. Using Leach's 300 ohm Re, the input differential voltage range for linear operation is +-1v, extremely wide.
Why would you want to do this? Because you want to avoid input overload, which sounds bad no matter what label you give it. Basically, it stops the music for a time. For example, after a sharp sound like a loud snare drum, you would not hear vocals for short but audible interval.
Here is how this works: If you drive the input with a rapidly rising signal, there is a delay through the stages of an amplifer to the output. The negative feedback signal from the output will be delayed, causing a large input differential voltage to the input pair for a time -- until the output reacts and the feedback signal returns. This interval is when the input pair can overload (and cease to pass signal because it's maxed out already). You don't want it to ever overload because it blocks the music during that time, until the input stage recovers (either because the input signal falls or the output catches up with the signal).
So now we recognize a problem, how do we solve it? There are many techniques for preventing input overload. One is to "add Re" to the input pair, to get a wider linear range. A second is to put passive filters on the amp input, to limit risetimes. (That is very good practice, because other circuits, such as the VAS, are also likely to overload.) A third is to use "lead compensation" from the VAS to give an earlier negative feedback signal to the input stage.
Leach used all three of these techniques. Pass used the input filter and lead compensation in the A40. Both work well, designer's choice. I can overload either amp with an artificial test signal (not real music), so we're talking about a matter of degree of resilience to overload.
ABOUT TIM: You asked "What is really the sound of amplifier having TIM?" It's a catch-all term for a lot of different distortions related to handling transients in the music. It doesn't have any one sound, because it's a label for a collection of bad effects from poor handling of transients. Some early solid state designs had excellent measured performance, using the steady-state metrics of THD and IM, but sounded bad on transients. Some amps blocked, which merely sounds dreadful, but some burst into oscillation when overloaded (which could be destructive).
To quantify these effects, in the early 70s audio engineers labelled the collection of effects TIM and developed new tests to measure it (Dynamic Intermodulation Modulation "DIM"). DIM is measured using a low frequency, higher amplitude square wave mixed with a high frequency sine wave (specifically, 3.18 kHz square wave, 15 kHz sine wave, mixed 4:1, amplifed and spectrum analyzed, distortion products measured in %). Because designers became aware of TIM and because it could be measured, it could be avoided -- with additional care in design.
As is the way of things, TIM was overemphasized in the 70s, then became passe. But because the most excitement in music is in the transients, it's still important to handle them accurately. You can introduce TIM at any stage of an amplifer, so it's important to keep it in mind throughout the design. |
|
|
| lumanauw |
Hi, Slowhands,
Thanks for the explenation. What makes that I feel that without Re the sound is more detailed?
Hi, Mr.Pass,
if it is possible, I would like to ask a final question.
What is the bias current (mA) in the differential stage? |
|
|
| stefanobilliani |
| quote: | Originally posted by slowhands
For example, after a sharp sound like a loud snare drum, you would not hear vocals for short but audible interval.
But because the most excitement in music is in the transients, it's still important to handle them accurately. |
Sounds a bit of Auto-Digitalism , BTW ;) |
|
|
| Nelson Pass |
| quote: | Originally posted by lumanauw
What is the bias current (mA) in the differential stage? |
I seem to recall about 4 mA per device. |
|
|
| Fred Dieckmann |
"At the rate you're going it would be more efficient to simply ask Nelson for a full schematic of the X amplifier. I've got two separate ideas in mind as to how to arrange a single pair of 'magic resistors' to do the trick, but until I get some parts in I can't check to see if either works. Knowing Nelson, it's unlikely that he's using a servo.
It seems rather unlikely that he would use self-biased complementary differentials if he didn't think that the topology had merit. JFETs have less noise than MOSFETs or bipolars and the 2SK389/2SJ109 duals offer a nifty way to get FETs that are fairly well matched and will track together thermally. Biasing with a current source to the rail works--Nelson did so in the A-75, for instance--but this is more elegant. His older designs are chock-full of bipolars. My Threshold S-500s have quite a few. But bipolars tend to lend themselves to circuits with high gain. The high gain, in turn, needs to be gotten rid of. That implies high rates of negative feedback, which brings along unpleasant baggage as far as sound quality. MOSFETs also tend to be easier to work with, thermally.
Grey"
Uh Grey........ with all the talk about intellectual property and reverse engineering, you really want to come out and say something that boneheaded even in jest. Hell maybe you should just ask someone to take one apart and draw a schematic. At this point even that would not supriise me. I am sure glad you aren't the president of my fan club. I would hate to spend that much time on damage controll. I thought jam sent you some J109 and K389s to play with. Oh well another circuit design with an excuse not to show it.
If elegant means the least number of parts to do something..... yes.
There are some real potential advantages in using two current sources instead of self biasing in terms of PSRR and CMRR both these are most likely not a big effect when cascoding the jfet pairs and in the X amp topology were these errors show up mainly as common mode output noise which is not seen by the speaker. With more typical topologies and especially with a mosfet second gain stage, I would defiinitely go with the two current sources instead of the self biasing route, due to the different turn on voltages and transconductances of P channel and N channel mosfets that would be used in mosfet voltage gain stage. Unlike you, I don't know Mr. Pass well enough to predict how he does this voltage gain stage. I am sure you can probably figure it out with your ever growing insight into his design techniques. It really is uncanny and I stand in awe of your talents in this area.
"But bipolars tend to lend themselves to circuits with high gain. The high gain, in turn, needs to be gotten rid of. That implies high rates of negative feedback, which brings along unpleasant baggage as far as sound quality"
You mean like emiitter followers? Try some emitter degeneration you might be suprised......... |
|
|
| lumanauw |
| quote: | | I seem to recall about 4 mA per device. |
If I'm not mistaken, you are using dual transistor (Like K389 and J109). Is it 4mA per device means each single transistor is 2mA or it is 4mA per single transistor (K389 consist of 2 transistors in each die)?
Mr. Pass, Just experimenting with this K389 and J109, but using ordinary CCS bias (to rails). I intended not to use any C in the feedback-ground, but the DC is big (about 150-300mV). Is putting C in feedback (to ground ) is a must? John Curl doesnt use this C, but he uses Servo. What do you use? |
|
|
| jam |
| Just to keep Grataku happy........................;) |
|
|
| lumanauw |
Hi, Fred,
| quote: | | I would defiinitely go with the two current sources instead of the self biasing route, due to the different turn on voltages and transconductances of P channel and N channel mosfets that would be used in mosfet voltage gain stage. |
That makes sense, if we are using P and N mosfets (with different Vgs), like the picture below? (adjustable CCS)
The selection of VAS device, if we want to consider its gain, Mosfets and Fets have lower gain than bipolars. But bipolars have much more predictable Vbe (about 0.65V always). Using mosfets must have different value CCS, or different resistor from differential legs.
| quote: | | You mean like emiitter followers? Try some emitter degeneration you might be suprised......... |
Can we use bipolars with Re to get about the same gain as mosfet or fet, for VAS device? Will the sound of degenerated bipolar the same as if we use mosfet/fet?
What is the calculation to degenerate gain of bipolar with Re? (To have the gain about the same as if we use mosfet/fet) |
|
|
| fab |
| quote: | Originally posted by slowhands
To quantify these effects, in the early 70s audio engineers labelled the collection of effects TIM and developed new tests to measure it (Dynamic Intermodulation Modulation "DIM"). DIM is measured using a low frequency, higher amplitude square wave mixed with a high frequency sine wave (specifically, 3.18 kHz square wave, 15 kHz sine wave, mixed 4:1, amplifed and spectrum analyzed, distortion products measured in %). |
I am looking for reference documentation on DIM measurement. Do you have something to suggest. Or does it say it all in your description? I have never seen any quantified specification on DIM on commercial audio product.
Thanks |
|
|
| grataku |
Much obliged Jam ;)
By the way, can you send some of them Jfets (the 389s) this way too?
Name your price, you can ask Jocko, I am good for it. Of course that is unless Jocko has them in which case I would buy them from him as I tend to be a loyal customer.
Fred
I can always put an enema to good use. I see your point more than you can imagine ;) |
|
|
| Nelson Pass |
In the 70's (maybe early 80's) Sansui developed an interesting
TIM test, where the amplifier was fed a series of triangle waves
alternating between a sharp rise and slow fall and then a slow
rise and a sharp fall. TIM resulted in a DC shift in the amplifier
forming a square wave with a period corresponding the the
frequency of the alternation of the triangle wave. |
|
|
| fab |
| quote: | Originally posted by Nelson Pass
In the 70's (maybe early 80's) Sansui developed an interesting
TIM test, where the amplifier was fed a series of triangle waves
alternating between a sharp rise and slow fall and then a slow
rise and a sharp fall. TIM resulted in a DC shift in the amplifier
forming a square wave with a period corresponding the the
frequency of the alternation of the triangle wave. |
Thanks Mr Pass! Is this method of measuring DC shift similar (even if the waveform used is different) to this one below I found in E & W.W. of 1981? By the way, in this article they have measured the Threshold 4000 and found it good in that manner compared to 3 other models.
But is there a common standard (other than THD and IMD) that is used to compare amps?
Regarding the degeneration resistors isn't it true that too high a value will change the distortion curve in function of power level;instead of having a sharp increase of distortion only close to the clipping level, the distortion would begin to rise at lower power levels but with a lower slope until clipping? If true, could it be caused by the reduction of overall feedback due to lower gain in input stage? |
|
|
| Nelson Pass |
a) I don't know
b) see a)
:cool: |
|
|
| lumanauw |
Have anyone Experimented with Re in differential?
What happens if we place inductor for Re, instead of resistor? Inductor have raising resistance with frequency.
What happens if we place resistor//with capacitor for Re? Like 1k//with capacitor. Capacitor has less resistance with raising frequency.
Or those experiment is dangerous, because to oscilation possibilities? |
|
|
| fab |
| quote: | Originally posted by lumanauw
Have anyone Experimented with Re in differential?
What happens if we place inductor for Re, instead of resistor? Inductor have raising resistance with frequency.
What happens if we place resistor//with capacitor for Re? Like 1k//with capacitor. Capacitor has less resistance with raising frequency.
Or those experiment is dangerous, because to oscilation possibilities? |
Having 1k//with capacitor: it depends of cource of the capacitor value. I have already used that scheme to polarize a j-fet input stage with a very low cut-off frequency so I would have the stable gain I needed for the whole audio range. Here in a BJT differential pair, there is no need for that unless, if I understand your point is to vary the gain of the input stage against frequency. At low frequency (where C is seen as an open circuit) you would have a gain of about only 1.8 in input stage ! This is too low.
The amount of overall feedback of the amp would increase as frequency increases. I would tend to believe that oscillation problems may be hard to solve. |
|
|
| LBHajdu |
Uli published this circuit in the thread “normal X-Amp”. I’m surprised, it got little attention. I believe it to be very close to the x600 and 1000x. The amp uses one more cascade in the signal path then the x600 and x1000. That is an improvement. The one short full of this amp compared to those smaller X-amps is that it needs a balanced input. It’s not hard for us diyrs to build a balancing preamp for this amp.
Some changes we talked about, that Uli is trying is to remove the 100 ohm magic resistor for even more X effect and removing the 100k ohm resistors on the gates of the Input mosfet. I don’t think you need the RC networks on the outputs of the amp.
If NP or any others know of any other way to get better performance out of the circuit, let me know. |
|
|
|
