| dimitri |
And my wish is let us follow the topic (that words in black bold in the upper part of the screen, just below diyaudio.com) :)
I understand that you all have that much to type here, but please don’t allow Nonsense Expands To Fill The Bandwidth Allotted To It |
|
|
| Steve Eddy |
| quote: | Originally posted by hitsware
I=V/Z is a linear relationship BUT
the Z of a driver (or any reactance)
is (by definition) not. Something to
do with 'rate of change' |
So what if the Z isn't linear versus frequency? That's not what the author said. What the author said was that current and voltage will only have a linear relationship if the circuit is purely resistive.
Here, read it again:
By Ohm's law, the current in a speaker voice coil is the amplifier output voltage divided by the speaker impedance. Therefore, the current can only be linearly related to the voltage if the speaker presents a purely resistive load to the amplifier.
I'm sorry, but Ohm's Law is a linear equation whether you're using resistance or reactance or impedance and therefore there will be a linear relationship between current and voltage even in a reactive circuit.
I = V/R, V = I x R, R = E/I
I = V/X, V = I x X, X = V/I
I = V/Z, V = I x Z, Z = V/I
se |
|
|
| Steve Eddy |
| quote: | Originally posted by fdegrove
Seems to me some people have difficulty in differentiating DC from AC behaviour...
Back to the kindergarten and Thevenin for some of us I reckon?? |
Eh? Since when does Ohm's Law become non-linear under AC conditions? The relationship between current and voltage is linear whether DC or AC.
Here's your kindergarten lesson for today:
Ohm's Law for resistance:
I = E/R, E = I x R, R = E/I
Ohm's Law for reactance:
I = E/X, E = I x X, X = E/I
Ohm's Law for impedance:
I = E/Z, E = I x Z, Z = E/I
Since all of these equations are linear, how does voltage somehow come to have a non-linear relationship to current under AC conditions?
se |
|
|
| hitsware |
Lets start another thread about it and I'll respond best I can.
My fault this one digressed enough to make the tone of some of the responses get sort of aggressive.
(or I may be flattering myself :) )
This forum (new to me ) is a little differant than the FR forum in that more digression is wellcomed over there.
I'd like to see this thread explore active differential
(NOT .....
Counting xfmrs .....
Simply to make it comfortable and stimulating for guru types.
(which i think you and i can agree on ..... we ain't no JC, or NP
or even fred))
front ends to the max.
Off to Raleys ........... mike |
|
|
| dimitri |
>Lets start another thread
... thanks God (and Fred :devilr: ) .... |
|
|
| lumanauw |
Maybe this is not about the main topic. But since there are many expert here, I would like to ask another question.
Once I read an article by Japanese about current power amp. He said that what drives a speaker really is current, not voltage.
But all this time we define audio power amp as a device to amplify voltage. That is like there is sinusoidal 1Vpp input, the power amp is amplifying the voltage to certain value, like 30Vpp.
If the speaker is purely resistive, we can get clean sinusoidal current from sinusoidal voltage, since the load is pure resistive. But since the speaker is Impedance (zr+zl+zc), not pure resistive, the current shape certainly be different from the voltage shape, cause of Zspeaker depends on frequency.
The idea is this. What happens if we make an audio power amp, that sense voltage shape input (like sinusoidal), but gives output of current, with that particular input shape.
Is this what we know as "current feedback" (usually using opamp like the ssm- Analog Device current feedback power amp), or is it something else? |
|
|
| Steve Eddy |
| quote: | Originally posted by lumanauw
Maybe this is not about the main topic. But since there are many expert here, I would like to ask another question.
Once I read an article by Japanese about current power amp. He said that what drives a speaker really is current, not voltage. |
Yes, ultimately it's the current flowing through the loudspeaker's voice coil which produces the time-varying magnetic field which interacts with the fixed magnetic field in the gap and causes the cone to move.
| quote: | | But all this time we define audio power amp as a device to amplify voltage. That is like there is sinusoidal 1Vpp input, the power amp is amplifying the voltage to certain value, like 30Vpp. |
Yes. That's because pretty much all loudspeakers are designed to be driven by a voltage source rather than a current source.
| quote: | | If the speaker is purely resistive, we can get clean sinusoidal current from sinusoidal voltage, since the load is pure resistive. But since the speaker is Impedance (zr+zl+zc), not pure resistive, the current shape certainly be different from the voltage shape, cause of Zspeaker depends on frequency. |
No, if you feed a sinusoid into a reactive circuit you'll get a sinusoid output.
| quote: | | The idea is this. What happens if we make an audio power amp, that sense voltage shape input (like sinusoidal), but gives output of current, with that particular input shape. |
You'll get a sinusoid either way.
The problem is that you'll get a frequency response that resembles the speaker's impedance curve.
For example, the image I've attached is of the frequency response and impedance plot of a Fostext FE208E Sigma full range driver.
The plot along the top is the driver's frequency response as driven from a voltage source amplifier and the plot below is its impedance versus frequency.
If you drove the driver from a current source, its response would resemble its impedance plot, and instead of the gradual rolloff of its low frequency response, you'd get a big peak in its low frequency response.
se |
|
|
| lumanauw |
What will it sound?
My idea of current amp is like this. I'll take an example from schematic in post no.81 (by hitsware). Let's assume that this is a working amp (not experimental).
The output is taken from the drain junction of SK and SJ. Push-pull amp will have this kind of output, usually taken in the junction of 0.22ohm/5W resistor (after emitors, if it is EF)
To get current sensor, before going to loud speaker, between the junction of those drains before output to sepaker, we put R, like 1ohm, to detect what is the current delivered to the speaker.
Then, this current figure is compared to the voltage signal in the differential input.
This way we don't need the 10k and 510ohm voltage feedback divider, since we are detecting the output current in the 1ohm resistor drop. That is the data for the differential to measure the difference in signal input and current output. Maybe this is "voltage input - current output" audio power amp. But is this possible? What will it sound?
This idea is coming from the statement of japanese I mentioned before that speakers are driven by current, not voltage. |
|
|
| john curl |
Dimitri, it would seem that this current drive question does belong on a separate thread.
For the record, current drive is not unusual for motor drive applications, and has been used for many decades. However, loudspeakers, under some circumstances, could benefit from current drive, BUT NOT typical speaker systems.
Also, I could care less whether SE is on this website or not. I like this website, because it is fast moving, and many inputs post interesting schematics and other references. I would prefer to keep on subject, if possible, but I'm sure I also have diverted from the original subject on many occasions, over the years. We were having an interesting discourse on differential input stages, but now this is pretty much lost in the noise of other inputs. |
|
|
| hitsware |
>I would prefer to keep on subject, if possible,
ref:post 199
If I understand Hugh right, the single pair configuration has more even harmonics not from the pair itself, but from the single ended (as a rule) next stage. Correct? Isn't the pair itself push pull by definition? .............. mike |
|
|
| lumanauw |
Sorry, my mistake. I will get back into track.
Like one post said, is it true dual differential gives less harmonics, because they cancel each other? What is really the benefit of having less harmonics? Isn't that tubes have warm sound because they have big harmonics?
About the current drive power amp, I make it a new thread, because it is about very different topic in http://www.diyaudio.com/forums/show...&threadid=22959 |
|
|
| fdegrove |
Hi,
| quote: | | Like one post said, is it true dual differential gives less harmonics, because they cancel each other? What is really the benefit of having less harmonics? Isn't that tubes have warm sound because they have big harmonics? |
No, differential would give less distortion with less even harmonics due to distortion cancelation.
Tubes can be made to sound icecold just the same.
The idea of tubes having a warmer ( richer?) sound is just plain BS.
Cheers,;) |
|
|
| dimitri |
Jc wrote: The complementary differential has lower distortion, all else being equal.
This is absolutely right. The differential pair alone is symmetrical, but when the signal is passed on to the VAS stage, it loose symmetry in current mirror, or in resistive load, or simply the VAS stage transfer function is not symmetrical. |
|
|
| hitsware |
>Tubes can be made to sound icecold just the same.
Ya Mon ! I once had to modify a tube preamp to make it sound like a tube preamp.........mike |
|
|
| AKSA |
Yes, true, the fully complementary input stage should give less distortion because of even order cancellation, particularly at the voltage amplifier, which is then fully complementary as well.
The single diff stage will not, however, have particularly low TOTAL distortion, other things being equal. Its configuration merely gives an inverting node at like potential - the base of the feedback transistor - so that like phase signal from the output can be injected as negative feedback.
Truth is, the diff pair as used in most SS amplifiers is dual input, one inv and the other non-inv, but single-ended output. This is an important distinction.
If it were used as a differential output then its even order distortion would be nulled. Lots of people would be happier....
No, as was suggested, the chief benefit of a dual diff pair input stage is that it permits use of a fully complementary voltage amplifier, which will null its own even order distortion, leaving only very low levels of odd order.
However, and here's the rub, do we want to reduce the even order distortion, while leaving the odd order essentially unaffected? Do we really want to skew the formation of harmonics towards low levels of odd-order, with even-order submerged below the noise floor?
I would suggest we don't. It just ain't musical. Another point which NP made, a bloody good point, is that a single diff stage leads to a single-ended voltage amplifier, which is almost always supplied with current from a current source supported from the rail. This arrangement facilitates marvellous offset control - no small feat.
The feedback node on a SS amplifier is required to undertake two tasks; AC voltage feedback, very important and quite obvious; and DC offset control, holding the speaker output at very close to zero volts. This task is vital; it's the amp equivalent of a regular heartbeat for a mammal. If the output deviates even a couple of volts from zero (and the nature of offset control failures are normally rail excursion catastrophes), then the speaker either overheats, or dies outright, and often takes the output stage with it too.
So, here are two very good arguments for a single differential input stage. Of course, I need not emphasize that this topology halves the component count, and is cheaper, simpler - less to go wrong! (that's the third argument!)
Cheers,
Hugh |
|
|
| jcarr |
A comment on terminology and how it is used by different people to mean different things:
At least in Japan, "dual differential" specifically means a topology in which an input differential pair feeds its output to another differential pair (in series). A topology with two input different pairs - one N and one P - is called (at least in Japan) "complementary differential".
I raise this issue because both of these terms are commonly used in catalogs and brochures published by the big Japanese electronic conglomerates (and other audio manufacturers), and unless you are clear as to who is using these terms and what what they mean by doing so, you may be left with a totally erroneous impression of what the circuit topology is.
Previously alluded to in this post.
http://www.diyaudio.com/forums/show...8889#post258889
hth, jonathan carr |
|
|
| lumanauw |
| So, we've got 2 confincing answer about single differential from AKSA and Mr.Pass. Is there any designer from pro-audio that used to applicate dual differential? What is the main reason to use dual differential? Is it just for easiness, or symmetrical look? Even though it LOOKS symmetrical, I think npn and pnp are not exact opposite in behavior. So if the behavior is not an exact opposite, would it be that the output like "have a shadow" or "blurry"? |
|
|
| AKSA |
Thank you JCarr,
I stand corrected. When I mentioned dual diff input stage I meant fully complementary diff input stage. I see it might have meant one diff pair as VAS following the input diff pair.
There is an engineering elegance in the complementary diff input stage, and there is no doubt it can be made to sound very good. As PRR remarked, a lot of the sonics are in the fine detail; components, dimensioning, layout, etc. But we all have a belief structure, and I lean towards the single diff.
Cheers,
Hugh |
|
|
| jcarr |
Hi Hugh:
>I stand corrected.<
No problem.
>I lean towards the single diff.<
You mean a single non-complementary input differential pair, with the second-stage being?
The terminology used in this thread has stayed arbitrary for so long that I would like to confirm where everyone stands and what everyone means. :)
For instance, I can state that I have previously used what I call a dual differential circuit in a commercial product, but I should clarify that this was _not_ a complementary differential circuit.
regards, jonathan carr |
|
|
| AKSA |
Jonathon,
yes, precisely that, with ....a single ended voltage amplifier, some supplied from bootstrap, and some from a constant current source.
It is always tricky to use words to describe a topology. Different folks, different strokes.....
But nice pictures take longer to put on the forum, dammit......
Hugh |
|
|
| jcarr |
Hi Hugh (again):
From your post:
http://www.diyaudio.com/forums/show...6958#post266958
>One valid, real-world approach is to have extremely fast devices and very short, non-inductive feedback and signal paths which minimize group delay and thus bring down overshoot and distortion to vanishingly low levels.<
I completely agree (some of my previous posts cover the same territory), but will add that the above should be accompanied by a low-pass input filter so that the circuit does not need to deal with HF signals that may possibly give it trouble.
IMHO, unless the designer is willing to implement the aforementioned attention to detail in the actual design (including the physical construction aspects), this thread discussion of what topology is best is a somewhat academic exercise.
regards, jonathan carr |
|
|
| bocka |
| quote: | | Even though it LOOKS symmetrical, I think npn and pnp are not exact opposite in behavior. So if the behavior is not an exact opposite, would it be that the output like "have a shadow" or "blurry"? |
Completely correct. A complementary design reduces even order distortion, not canceling it out. But again, the main mechanism of distortion is the output stage, if this stage produces even order distortion (and it will, because NPN and PNP are not completely complementary) we'll get overall even order distortion, regardless of a complementary input diff stage or not.
The complementary input stage has more sensitivity to the bias influence of VAS than a single LTP with a current source. A big disadvantage. To come out of this problem you need emitter degeneration of the VAS stage. On the other side a complementary design has lower distortion levels as john curl has pointed it out. So we've one advantage and one disadvantage and the freedom to design both topologies right...
BTW complementary designs have only slightly more components (about 10% or so) than asymmetrial designs. |
|
|
| hitsware |
>This may help.
Thanks ! I don't know if it's relevent or not, but something I've noticed is that A seems to have better recovery characteristics than C............. I.E. coming out of clipping there is not as much 'lag' or something (the little step off of the flat place going back down to the wave shape) .......... mike |
|
|
| jcarr |
mcp:
Thanks for the figure. Most convenient.
A Japanese EE would call A "Single Differential", B "Dual Differential", and C "Complementary Differential". Don't know about elsewhere.
regards, jonathan carr |
|
|
| Nelson Pass |
There's quite a few more, so I suggest you with stick with
pictures describing diff pairs after the first two. |
|
|
| hitsware |
>A Japanese EE would call A "Single Differential", B "Dual Differential", and C "Complementary Differential". Don't know about elsewhere.
Makes sense. I tend to call A single and C dual, but your nomenclature is much more succint .......... mike |
|
|
| PRR |
Yes, thanks, MCP.
If we actually keep talking about different diff-input stages, someone should probably paste that chart every few pages, so we all know what hook-up is being discussed.
jcarr> A Japanese EE would call A "Single Differential", B "Dual Differential", and C "Complementary Differential"...
MCP's image with Jcarr's "Japanese EE" terminology, tight layout:
I have a copy on my machine but the URL is ugly. Wrap the UBB URL tags around this tiny-URL: http://tinyurl.com/vuqt |
|
|
| john curl |
Most of the previous comments discuss complementary differential, but don't actually show its true advantages.
The MAIN advantage of comp-diff input is that the SECOND STAGE is symmetrically driven. This reduces the effort of just one device having to drive the output stage, and effectively doubles the available class A current. It also gives a doubling in the open loop gain, without decreasing the phase margin, and third, with bipolar transistors, it can cancel the input bias, so that an input coupling cap is unnecessary. This offsets the extra parts count.
There are other advantages, especially with FET input designs, but this should be enough for this discussion. |
|
|
| PMA |
| quote: | Originally posted by AKSA
Pavel, have you tried the prosaic double emitter follower output stage, particularly Self's Type II, in side by side listening comparisons, against your CFP? I have done tests with others some years back and concluded strongly in favor of the double emitter follower.
You make some challenging comments about RF injection into the amp too, which I'm looking into. Great work, Pavel!
Cheers,
Hugh |
Hugh,
A colleague of mine built very similar output stage with double emitter follower output. He used smaller transistors in parallel (I think they were BD243), these were considerably less linear [Ic=f(Ib)] compared to 2SA1943/2SC5200. The listenings tests were made and CFP version was preferred (the circuit shown on my www). CFP version also did measure better than 2SE.
Thanks for your interests in RF intermodulation influence (residual spikes of D/A conversion), this really gives enormous sonic improvement to filter the spikes properly and effectively.
Pavel |
|
|
| dimitri |
| Alternatively, complementary differential has several limitations. As was pointed by NP the quiescent current in the VAS after complementary differential input stage has poor stability. For the reason of stability, differential pairs load are resistors, thus the VAS input nodes have low impedance. The ordinary Miller compensation will be ineffective, as the dominant pole is formed by the VAS output network. Without Miller compensation the VAS output impedance would be high so the nonlinear input current of the output stage might contribute to overall distortions. Extra precautions should be taken e.g. extra emitter follower after VAS. |
|
|
| john curl |
| I don't know what you are talking about, Dimitri. |
|
|
| mlloyd1 |
Neither do I :confused:
Are you referring to the "current mirrors as the input diff stage load" item that was talked about a short time ago in the discussion on Sloan's designs ?
mlloyd1
| quote: | Originally posted by john curl
I don't know what you are talking about, Dimitri. |
|
|
|
| AKSA |
Pavel,
Interesting; thanks for the response.
I know the linearity of the drivers and output devices is very important to sonics, as is Vbe and beta matching of like devices where output pairs are used.
I did the side by side comparison with 2SC5200/2SA1943 on both; one, a CFP (with 100R base emitter resistor on the outputs) and the other a DEF Type II with 150R/100nF between driver emitters.
There really was no comparison, with DEF being superior, particularly in engagement, whatever the hell that means.......
Your work on low pass input filters is confirmed by Jonathan, in fact. That's good enough for me. No doubt; unless the amp is intrinsically band limited elsewhere, and probably despite that, we should use them liberally!
Do you chime them in at say 3dB down at 50KHz?
However, polystyrene or mica - that is the question!
Cheers,
Hugh |
|
|
| fdegrove |
Hi,
| quote: | | No doubt; unless the amp is intrinsically band limited elsewhere, and probably despite that, we should use them liberally! |
Both approaches suggested LP filters in the GNFB loop as I understood it. But, wouldn't it be much easier on the amp to implement the filter right at the input?
| quote: | | Do you chime them in at say 3dB down at 50KHz? |
That would be a nice starting point. Personally I like it more higher and use the Miller capacitance effect to do the work for me...
Valves are easy for that, especially high µ triodes.
| quote: | | However, polystyrene or mica - that is the question! |
To my ears, silvermica caps. They're as close to the ideal cap one can get and sound great.
Cheers,;) |
|
|
| dimitri |
JC wrote: I don't know what you are talking about, Dimitri.
Self singled out several types of distortion:
Input stage nonlinear transfer characteristic
Input stage CMRR
Voltage Amplifier Stage nonlinear transfer characteristic
Non-linear VAS loading
Output stage large-signal nonlinearity
Output stage crossover
Output stage switch off
From the above, Complementary Differential has two times higher gain, it has even distortion compensation in VAS, probably the JFET input will have better CMRR.
I’m speaking about nonlinear VAS loading. In Single Differential design the almost universally adopted Miller compensation effectively lowers VAS output impedance. Thus DEF with that nonlinear input resistance works well.
In Complementary Differential VAS feeds from much lower collector resistor, and the loop gain inside Miller loop would be lower. IMXO Miller compensation isn’t effective, 1 just eats the bandwidth, 2 is the best. Also one can isolate output stage from VAS by TEF instead of DEF. |
|
|
| john curl |
| I will look into it, but I think that you are creating a special case. I cannot see this effect, if real, is very important. Was this something taken from another source? Doug Self for example? |
|
|
| jewilson |
The Miller effect is reduced by using a cascode pair or a current mirror on the diff amp. As for as CMRR goes if the hfe or Tranconductance is not matched in the diff pair the CMRR will get worse as the miss match increases.
In addition, I do not see the point in trying to compensate for the effect at the second stage. I have never seen it work, however you can build a filter there.
Cool, Curl get runner up to new product of the year in StereoPhile.
Most be a Dam fine sounding amp.:cool::) |
|
|
| PMA |
| quote: | Originally posted by AKSA
Pavel,
I know the linearity of the drivers and output devices is very important to sonics, as is Vbe and beta matching of like devices where output pairs are used.
I did the side by side comparison with 2SC5200/2SA1943 on both; one, a CFP (with 100R base emitter resistor on the outputs) and the other a DEF Type II with 150R/100nF between driver emitters.
There really was no comparison, with DEF being superior, particularly in engagement, whatever the hell that means.......
Your work on low pass input filters is confirmed by Jonathan, in fact. That's good enough for me. No doubt; unless the amp is intrinsically band limited elsewhere, and probably despite that, we should use them liberally!
Do you chime them in at say 3dB down at 50KHz?
However, polystyrene or mica - that is the question!
Cheers,
Hugh |
Hugh,
I was thinking about our different experience with CFP x DEF. There might be one difference - my circuit is a true class A for 8 Ohms (Iq even higher than shown) and was compared to class A DEF. How about you, did you test class AB or class A output stages? In case of class AB output stage I believe that DEF would give better sonic results (remember the thread about CFPs about a year ago, do not want to go into details now).
Regarding filtration - it is about 90 kHz/-3db as a result. But not only input RF filter, but every stage is in fact a LPF (with RC defined pole). I do not use global FB, but the amp is divided into say 3 stages with local FBs [this is done for preamp and power amp as well], every stage is a LPF. This brought results, and you can hear it comparing versions with and without LPFs in individual stages. I should mention that the amp's construction must be done the way that does not allow the RF signal to bypass filtration.
Cheers, Pavel |
|
|
| AKSA |
Pavel,
I've just read through John Curl's Parasound JC-1 amplifier. WOW!! (this reads 'walk on water'!). Very, very impressive. Such a luxury to have three men behind every product; design, layout and component choice. Should work well!
| quote: | | I was thinking about our different experience with CFP x DEF. There might be one difference - my circuit is a true class A for 8 Ohms (Iq even higher than shown) and was compared to class A DEF. How about you, did you test class AB or class A output stages? In case of class AB output stage I believe that DEF would give better sonic results (remember the thread about CFPs about a year ago, do not want to go into details now). |
I used a Class AB bias to just less than 60mA.
I also agree that in Class AB, the DEF will perform better as short term oscillations would appear to be easier to control with base stoppers. In Class A, I would expect CFP to give better FR and damping factor, and possibly bias control too. It's a very attractive topology, plagued only by low current local instability. I first found this topology in an engineering text from the early sixties. The application note commented that Zout was around ten times less than a conventional emitter follower; I call it the monkey atop the elephant.
I should comment on multiple outputs of dubious linearity, like the BD243. Matching for Vbe and beta is important; it won't sound much good until the former is within 1mV and the latter within about 2%. This alone could explain a few of the observed differences.
I admit to looking carefully at the circuits on your site with admiration. Nicely done; some very deep thinking, and my thanks for this work. And yes, layout is crucial, particularly short feedback paths and shielding from RF input. I find your comments about RF filtration fascinating; it would seem best to place it outside the feedback loop at the input; this is an area I need to research more.
On Type 1 v. Type 3: To absorb tolerances and keep bias stable, it is indeed necessary to use considerable degeneration on the emitters of the VASs with Type 3 (fully complementary diff input stage). This pulls back open loop gain very considerably, where such artifice is not necessary with Type 1. In most cases, and even with resistor loading at the collector of the diff pair, the OLG of the Type 1 will be higher, leading to higher feedback factors. Whether this is an advantage is moot, of course. A factor in favour of Type 3 is the fact that two VAS devices are now driving the output stage, which improves drive power and better resists reactive reflections from the speaker back through the output stage.
Jewilson: I have found constant current sources on the diff pair a mixed blessing. Better CMRR, yes, but much thinner sound. There is clearly a sonically pleasing quality in resistor drive to this stage, but PSRR suffers......
On Miller compensation: I have found it should be used sparingly; back off until unstable, then add about 25% to confer unconditional stability. Even so most SS amps with GNFB will kick up with highly capacitive loads like ES speakers. But it is palpable that reduction of Cdom always seems to improve sonics
Cheers,
Hugh |
|
|
| Fred Dieckmann |
"I have found constant current sources on the diff paira mixed blessing. Better CMRR, yes, but much thinner sound. There isclearly a sonically pleasing quality in resistor drive to this stage,but PSRR suffers"
:scratch2: Well...... that's fairly easy. The knob on the front of the amp can be called the "HD" dial or maybe even the "Dean Dial" :clock: :nod: |
|
|
| AKSA |
Brilliant Fred,
You're a bloody genius.......
Thank you!
Hugh |
|
|
| lumanauw |
What is TEF, DEF, CFP? Please insert drawing.
EF=Emitor Follower, CFP= Collector output? TEF=? DEF=? |
|
|
| bocka |
EF = emitter follower
CFP = complementary feedback pair
DEF = two stage (or dual) emitter follower
TEF = three stage (or triple) emitter follower |
|
|
| Bricolo |
triple emitter follower, double emitter follower?
So, just paralleled output devices? |
|
|
| bocka |
| No, it has nothing do do with paralleling devices. It's the number of darlington stages |
|
|
| hitsware |
Thanks !
WAY to complex for my tastes .....
I'm sure it's good BUT ....
No accounting for taste ....
I like 3 xsistors for the whole amp ...
But in the case of a dif amp front end
will go for 6 (total (whole amp) & max)
........Mr. Minimalist |
|
|
| Upupa Epops |
| of transistors for fully symetrical discrete opamp is eight - six is not enought :whazzat: , how you can see for example in every Bryston amps. |
|
|
| hitsware |
| quote: | Originally posted by Upupa Epops
of transistors for fully symetrical discrete opamp is eight - six is not enought :whazzat: , how you can see for example in every Bryston amps. |
Agreed, but I mean an opamp (not necessissaily fully semetrical)
.......mike |
|
|
| thanh |
| New idea: dual differential +complementation=dual complemetary differential . |
|
|
| slowhands |
| quote: | Originally posted by lumanauw
The first schematic (Kaneda) is single differential (TR1,2), with upper cascode (TR3,4). The second schematic (schdetail.gif) is dual differential (maybe I should say complementary differential) the upper is NPN (TR1,2), and lower is PNP(TR3,4).
Thanks for the example. This is exactly what my question is all about. Single differential (like kaneda) VS complementary differential (like schdetail.gif). |
This thread has wandered from the original topic, which is always fun, but the original topic is still worth some discussion.
In the Kaneta design, there is a second differential stage cascaded after the first, not just a simple VAS with current source. This second differential stage is cascoded and appears to drive a dual current source which may have current mirror action too due to the common current through the thermistor (not sure of value, which will determine the degree of mirroring).
It seems to me this second differential stage should provide open loop gain similar to the "complemetary differential" topology (6 db more than a simple VAS with CS), and perform about as well. However, that is not the prevailing opinion here. What am I missing? |
|
|
| lumanauw |
OK, what about this ?
There are 2 differential, T1 and T2 is differential, again differentiated by T4 and T5. T7 and T8 are VAS, T9 and T10 are current mirror.
This is dual differential, right?
There is also more complex design, Complementari differential + dual differential, where T1+T2 has its PNP friends, T3+T4 and T7+T8 also have their PNP friends. This design has 4 differential and 2 VAS pairs, like this attachment mirrored vertically.
AKSA, could you tell a little, why you never find it right with complementary differential? J Curl insisted on complementary differential, but he always uses Jfets+stacked self-biasing. Never tell about bipolar complementary differential+ordinary ccs. |
|
|
| thanh |
| quote: | Post #257
New idea: dual differential +complementation=dual complemetary differential .
|
I don't know what D1,D2,D3 is.These aren't normal diode.I think I can never buy these.The image is too big.Can you convert it to jpg? |
|
|
| SkyChu |
| D1, D2 & D3 are CRD (current regulative diode). You can use small JFET to replace them ( you need match with Idss). |
|
|
| AKSA |
Lumanauw,
You ask me explain my religion!! Tricky, better put on your flame suit........:D
Fully complementary circuitry looks great on paper, I like it a lot.
However, any distortion it introduces is symmetrical, in a sense, 'balanced'. To many, this appeals. But symmetrical distortion on an audio waveform is the stamp of odd order harmonics; even order distortion is asymmetrical.
Thus, I like a single diff pair, as the voltage amplifier is driven in single end. This introduces even order distortion which is very well tolerated by humans, and while there is odd order certainly present, it is partially masked by the even order.
While the single diff pair doesn't look as nice as a fully complementary diff pair structure, it sounds better, to my ears.
I must stress this is my belief, and YMMV, and probably does. Everyone is free to pursue their own religion with passion and commitment; I'm pretty cool about this! ;)
Sampai pertanyaan saudara!
Cheers,
Hugh |
|
|
| mandat |
>lumanauw
I want to inform you that this amplifier was published by Matti Otala in 1973.
He has introduced new term of TIM distortion and theory which explained the reason of "transistor sound" of semiconductor amplifiers - SR parameter and depth of feedback loop.
A commercial version of this amp was manufactured by Electrocompanient. Otala co-operated with Harman/Kardon in projecting Citation power amps, too. |
|
|
| lumanauw |
AKSA, apa kabar istrinya, baik-baik?
My personal preference is also single differential. The sound it produces is just right, compared to complementary differential.
But for the VAS, I liked it push-pull VAS. (in this, I'm quite different with you).
Dual differential is another story. Like the Otala design, it gives more character to the sound it produces. Maybe it is because dual differential makes large open loop? Is that big feedback is not so bad?
Look at post #261. Q1 is giving distortion cancelation from left half to right half transistor. Q1 is also giving distortion cancelation with Q2. So there is "Double cancelation" works on full complementary differential.
Maybe what J Curl stressed is in Jfet differentials, because my experience is always with bipolar differential. For bipolar differential, I certainly likes single diff.
(too bad I cannot buy K389-J109 pairs here, other wise I would have experimented with full Jfet complementary differential, like J Curl uses)
This design has single differential, and pushpull VAS. The levelshifter is done by folded cascode. (leading to low openloop gain) The output of right differential is converted to drive lower VAS transistor with current mirrors. This is from AD817 opamp.
Can anyone help, if I want to make this design an audio power amp, what is the bias current for Differential, Folded cascode, and final transistor? The supply is about +/-50V. |
|
|
| mandat |
| I try to take part in this thread but I do not recognise what a kind of DIY society discuss here. I remember that pupil should keep his mouth shut when teacher is speaking. ;-) |
|
|
| AKSA |
Lumanauw,
Istri baik-baik, terima kasih, selalu tersenyum!
You asked me my thoughts on Sky-Chu's circuit; here goes!
I do not like the complexity, and particularly the large numbers of cascodes. I accept that with Jfets, cascodes are necessary because of the limited Vce ratings, but I have not found they add anything to the sonics, although they look great....... .
Current sources are difficult mistresses. Once again, they look great, but unless they are implemented very carefully, they can turn on you and destroy the sonics.
There are large numbers of devices in this circuit, some of them quite costly, and this goes against my religion. I have built such amplifiers, and generally found that complexity does not improve the sonics. The solution to good sonics is a simple circuit, very carefully chosen operating points, and great care with component selection. Anything else is all too often academic self-indulgence.
Further, this is a voltage amplifier only, and if we are talking of power amplifiers, which usually include a nfb loop, this schematic is only half the picture. Nfb introduces all sorts of issues which must be considered in totality. And all too often nfb is necessary!
Unfortunately I haven't the time right now to explain precisely why largely because I'm not actually too sure why these things are problematic. But I am convinced that simplicity is the key.
Try an AR Bailey amplifier from the sixties, using modern semiconductors. You will be surprised! Or perhaps Le Monstre, from Jean Hiraga.
Cheers,
Hugh |
|
|
| lumanauw |
Here's lesson I capture from you| quote: | | Current sources are difficult mistresses. Once again, they look great, but unless they are implemented very carefully, they can turn on you and destroy the sonics. |
CCS can be replaced by R+bootstrap. What is the meaning of Bootstrap? Putting C between 2 R's is bootstrap?| quote: | | and generally found that complexity does not improve the sonics |
| quote: | | The solution to good sonics is a simple circuit, very carefully chosen operating points, and great care with component selection. Anything else is all too often academic self-indulgence. |
Operating points, that is very important. I've heard difference from only 1mA difference in bias.
Is there any guideline for component selection? Is this more about the brand? Or it is completely trial and error?| quote: | | But I am convinced that simplicity is the key. |
Passlab designs are simple, but they do perform better than more complex designs.
I've got 1 question about differential. Why is that in non-feedback amp design, the differential is almost always using full complementary differential? Is it imposible to built non-feedback audio power amp using single differential? |
|
|
| lumanauw |
This is silly question, but I want to know why. Is there any opamp or chipamp designer here?
Why is that every topology inside opamp and chipamps always using single diffential? |
|
|
| jcx |
first, i'm assuming you really mean parallel complementary differential input stages - there are in fact many series dual differential stage op amps, particularly high precision types with high open loop gain
maybe because their is no objective, salable datasheet spec that is improved by complementary differential inputs?
op amp designers are driven by cost/yield/process complexity vs measurable performance
early op amp designers didn't have reasonably complementary devices, only since the 90's have good complementary processes been broadly available
current feedback op amps usually have symmetric input stages and some modern low voltage rail-to-rail input op amps do use dual complementary input diff pairs
Walt Jung's "Op Amp Applications" book is a valuable reference work of op amp design history and shows examples of many of these devices |
|
|
| mlloyd1 |
I'll take a stab:
hmmm ... let's see ... it's cheaper :-)
mlloyd1
by the way, there were some harris high speed op amps some years ago that used complementary differential input stages. Wish i could remember some part nubers, but for some reason, I can't at the moment.
| quote: | Originally posted by lumanauw
This is silly question, but I want to know why. Is there any opamp or chipamp designer here?
Why is that every topology inside opamp and chipamps always using single diffential? |
|
|
|
| thanh |
Lumanuaw!The distortion isn't cancelled by Q1 and Q2.
I simulated Slone'topo .The sim say that the distortion still alive.
I built a Slone amp which use collector resistors instead of current mirror.
I didn't succeed in current mirror.I don't still know what reason is.
Advantage of Slone's topo is the cacellation of voltage offset.It can't cancel distortion. |
|
|
| fab |
| quote: | Originally posted by thanh
...I built a Slone amp which use collector resistors instead of current mirror.
I didn't succeed in current mirror.I don't still know what reason is. |
If you are refering to Figure 7.6a of Slone book, you are not the only one with that problem. I have not built it but I have discussed that with mcp and because of use of current mirrors for loads in first stage there is an imbalance for the dc operating points (VAS: Q11 and Q12) of the amp resulting sometimes in the output swinging to either positive or negative rails. Q14 and Q17 impose a maximum collector current for the VAS stage but it is only there for protection not normal conditions. |
|
|
| AKSA |
Conventional current mirrors take no account of the bias current of the VAS. Consequently, they balance only the currents into the current mirrors; not the currents coming out of the diff pair.
Consequently, the diff pair is not in balance and residual distortion and overload margin are compromised. Listening tests tell me this is very significant sonically.
A resistor in the working arm of the diff pair, with the other arm taken to rail, is a better arrangement. But you must match the beta and the bias networks of the diff pair transistors (and carefully dimension the arm resistor) for this to work. I routinely get better than 2% balance with no special technique here, measured from the voltages across the (equal) bias resistors.
Cheers,
Hugh |
|
|
| fab |
AKSA,
Who are you responding to with your last comment? I believe that your comments are not related at all to my last post since there is no obvious relation. |
|
|
| AKSA |
Hi Fab,
Yes, you are right, I was responding to Thanh's comment here:
| quote: | | I didn't succeed in current mirror.I don't still know what reason is. |
Hugh |
|
|
| lumanauw |
Some of opamps have incredible specs. Very high slewrate, very high bandwith etc.
1. Does anyone has built audio poweramp following the topology of these "superspec" opamps?
2. These opamps, in their datasheet, are not supposed for audio. What will be the sound like, if we use it for audio, like making preamp with it? Is the topology for audio opamp has certain topology, that differs from these opamps? |
|
|
| jewilson |
Andy, your correct I still have some samples left from the 80's and never used them. One of them even in a 14 pin package. At the time I was into using PMI OP16 and 17's and OP37.:) |
|
|
| thanh |
| I want to build a JFET amp but i can't buy any small power JFET in my country.I'm using a BJT amp. |
|
|
| GRollins |
You're going to have trouble finding power JFETs anywhere in the world.
Power MOSFETs are what you need, and should not be too hard to find.
Grey |
|
|
| john curl |
| Many advanced audio designers use techniques similar to or better than the concepts described in IC designs in previous threads. Often, we invented them first, such as the complementary differential input stage. |
|
|
| Bricolo |
| quote: | Originally posted by john curl
Many advanced audio designers use techniques similar to or better than the concepts described in IC designs in previous threads. Often, we invented them first, such as the complementary differential input stage. |
Mr Curl,
Since we're talking about input stages here. I don't remember on which thread, you mentionned that paralleled jfets was a bad idea (something related to (non linear) input capacitance)
Can you tell me more about this? I re-read Borbely's 2 articles about jfets yesterday, and it seems that he doesn't agree with you on this point. |
|
|
| john curl |
| I doubt that Erno and I have too many differences of opinion. We have known each other since 1975. We have compared notes. |
|
|
| Bricolo |
| quote: | Originally posted by john curl
I doubt that Erno and I have too many differences of opinion. We have known each other since 1975. We have compared notes. |
Maybe I misunderstood what he wrote about that.
Here's an extract from the 1st article, on page 4 (right column)
http://www.borbelyaudio.com/ae599bor.pdf
"However, even without RS, the
noise of a single K170 is not low enough
for MC pickups. To achieve lower noise,
you can parallel several of these devices." |
|
|
| GRollins |
Aye, lad, but note that Borbely is talking noise, whereas John is referring to distortion. Two different things. Just one of those annoying tradeoffs that nature is fond of. Yes, THD and noise are often lumped together in measurements, but that's not to say that they are the same thing. In other words, both John and Erno are right...from different points of view.
Nature cannot be forced, but she can sometimes be seduced. You could cascode the paralleled JFETs to make the Gate capacitance less of a problem, but then you're getting into another topology.
Grey |
|
|
| fdegrove |
Hi,
| quote: | | You could cascode the paralleled JFETs to make the Gate capacitance less of a problem, but then you're getting into another topology. |
Yep, lower noise...far higher gain and a ton more distortion....
MC carts aren't too sensitive about the input C so paralleling selected 2SK170s makes alot of sense.
Cheers,;) |
|
|
| lumanauw |
Mr. Curl
I really want to read your writings about power amp. Where can I read them? |
|
|
| GRollins |
Frank,
I approach cascodes the same way I approach current sources. First I ask myself if there's any other reasonable way to get the job done. If not, I'll toss in a cascode. Trouble crops up when people assume too quickly that the answer is that, yes, the cascode (current source/current mirror/etc.) is indeed the one and only way to get the job done. After a few times of "thinking it through" in this manner, the answer becomes "well, of course I'll use a cascode." It's the 'of course' that I'm always crusading against.
That said, why don't we cascode a couple of 2SK170s with a 6922 or a 6SN7?
Might be fun.
Grey |
|
|
| andy_c |
| quote: | Originally posted by jewilson
Andy, your correct I still have some samples left from the 80's and never used them. One of them even in a 14 pin package. At the time I was into using PMI OP16 and 17's and OP37.:) |
Yep, I used to use the HA-2539 in radar baseband circuits back in the Reagan era, around '83-'84 :) . |
|
|
| john curl |
| Any published philosophy of my amp design is on the Parasound.com website. |
|
|
| thanh |
| join curl!Your site hasn't got any schematic which i can see |
|
|
| fdegrove |
Hi,
Grey,
| quote: | | That said, why don't we cascode a couple of 2SK170s with a 6922 or a 6SN7? |
This has been and still is being done for phono stages; using a 2SK170 or similar as the input and half a 6922 on top of it.
You can do this in single ended or differential just as easily.
You can also come up with a number of variations on this theme depending on what it is you're looking for...
In tube parlance one could compare a FET to a triode whereas a cascode of triodes would look like a pentode minus the space charge noise.
I wasn't condemning cascodes in any way, just pointing to the fact that you're looking at a topology with totally different characteristics than a simple FET alone has.
Best, ;) |
|
|
| GRollins |
Yep, that's why I noted the difference in topology. Comparing apples and oranges makes things more difficult.
Grey |
|
|
| Bricolo |
| quote: | Originally posted by fdegrove
Hi,
Yep, lower noise...far higher gain and a ton more distortion....
MC carts aren't too sensitive about the input C so paralleling selected 2SK170s makes alot of sense.
Cheers,;) |
why would cascoding increase noise? |
|
|
| fdegrove |
Hi,
| quote: | | why would cascoding increase noise? |
Did I say it does?
Cheers,;) |
|
|
| Bricolo |
| quote: | Originally posted by fdegrove
Hi,
Yep, lower noise...far higher gain and a ton more distortion....
MC carts aren't too sensitive about the input C so paralleling selected 2SK170s makes alot of sense.
Cheers,;) |
maybe I didn't get it.
was the "more distortion" still due to the paralleled jfets, and not to cascoding? |
|
|
| fdegrove |
Hi,
| quote: | | was the "more distortion" still due to the paralleled jfets, and not to cascoding? |
First of all noise and distortion are two distinct items....
That said, putting two devices in // should noticeably lower the noise.
Putting those in cascode configuration will give you way more gain, hence much more distortion too.
Note also that putting none //ed devices in cascode will yield roughly the same gain but at the expense of a much higher noise penalty.
There's more to the picture than what I'm saying here but it should give you some idea.
As usual, no such thing as a free lunch.
So, to answer your question, the "more distortion" is coming from the cascode topology, not the //ing.
Cheers,;) |
|
|
| Tube_Dude |
| quote: | Originally posted by fdegrove
In tube parlance one could compare a FET to a triode whereas a cascode of triodes would look like a pentode minus the space charge noise. |
Hi
In fact a Fet look more like a pentode, because de drain circuit behave as a not perfect CCS ,the same as the pentode.
A triode with is internal feedback, has a much lower intrinsec anode output impedance .
Cheers |
|
|
|
