Post by: arconaut on November 21, 2006, 01:07:14 PM
In Eddie Ciletti's column in Mix this month, he makes the following statement:
"The 2520 is slow by modern standards - 2 volts per microsecond-but you'd never know it. The transformer's 1:3 windings ratio provides more than 8dB of headroom, and that increases the slew rate. You can apply this approach to any amplifier by taking advantage of 24 bit headroom, dropping the level 6 dB under the guise of maxmizing resolution."
I'm not sure I understand this - is he saying that because the transformer increases the output voltage by a ratio of 1:3, that a slew rate is essentially faster because at the transformer the signal rises to a higher voltage in the same amount of time that it takes for the signal to rise to a lesser voltage at the op amp?
And then in his last sentence there, he is presuming that you are sending your D/A's signal into an amp that has an op amp circuit with an output transformer. But is it true? Instead of "any amp", should he say "any amp with an output transformer" and perhaps also "any amp circuit with an op amp with a slow slew rate". What if your op amp has a really good slew rate? Then you wouldn't?
Noah
Post by: bruno putzeys on November 21, 2006, 02:12:22 PM
The second half of the quote a bit vague. As I interpret it, he then applies the same logic to any system where gain can be varied in several places, like an amplifier followed by an ADC. Dropping the gain before the converter and making it up in the digital domain indeed reduces slew rate (and other) distortion in the amplifier.
Actually, reducing the level in normal (class A) amplifier circuits always translates into reduced distortion, be it slew rate related or otherwise.
One rarely sees slew rate induced problems in modern op amp implementations (those aimed at audio at least). So, if slew rate is good, there's no need to be too concerned about it.
As a sideline I should note that slew rate does not relate well with distortion performance. Some op amps have fantastic slew rate specs, but slew rate induced distortion may become noticeable at much lower slew rates. Others may have a relatively slow slew rate, but don't distort much until you run into it.
Like one can have 24dBu headroom but THD starts to climb at +14 already, whereas another could only have an 18dBu headroom but still not distort discernably at +17.9.
An op amp with a degenerated bjt input pair will be much better behaved, slew-rate wise, than a jfet input op amp with the same transconductance (but no degeneration).
Post by: Larrchild on November 21, 2006, 03:06:42 PM
At 2v/us, T.I.M. would suck, otherwise.
But less feedback means more distortion rise earlier on.
Is that right, Bruno?
(And Welcome!)
Post by: bruno putzeys on November 21, 2006, 03:50:49 PM
An alternative interpretation of "less feedback" could be that the DC loop gain is intentionally limited to the value found at 20kHz. GBW remains the same, but loop gain at frequencies below 20kHz is lower. In circuits with non-negligible distortion, limiting the DC loop gain yields a sonic improvement (even though low-frequency distortion is clearly made worse).
Asymptotically (ie. neglecting any intentional or unintentional gain limitation at DC) the open loop gain of an op amp is determined by the internal integration capacitor (usually a feedback cap around a transistor in a miller arrangement) and the input stage's transconductance as:
A(f)=gm/(2*pi*j*f*C) (preferentially written as A(s)=gm/sC )
Solving for |A(f)|=1 yields
f=GBW=gm/(2*pi*C)
GBW is determined by gm of the input pair and the miller cap.
Slew rate limiting occurs when one of the input transistor cuts off i.e. when all of the emitter current comes out the other transistor. Slew rate works out as
SR=Ib/C
Again C is in the equation, but gm is not. Put simply, slew rate and GBW (and hence "the amount of feedback") are not directly related. Although gm of a transistor is directly proportional to emitter current, the use of degeneration (resistors in series with the emitters) allows SR and GBW to be set independently of eachother.
What makes slew rate distortion so annoying is that it occurs outside the feedback loop. When it occurs, no amount of feedback can correct it. Slew rate problems (and associated performance metrics like TIM) are not helped by the presence of more feedback. They are also not made worse.
Post by: Larrchild on November 21, 2006, 04:05:09 PM
Thanks!
Post by: dcollins on November 22, 2006, 01:31:44 AM
| Larrchild wrote on Tue, 21 November 2006 13:05 |
I gotcha. So poor slew performance is as ineffective at handling the feedback signal as it is, the primary signal. Therefore, more feedback will not create more (or less) TIM in a slower opamp. Outside the loop. Thanks! |
Isn't the solution to slew-rate limiting just to LP filter the input so that no possible rate-of-change will cause the NFB to arrive "late?"
A highly over-rated parameter imho.
DC
Post by: bruno putzeys on November 22, 2006, 03:24:48 AM
Also the amount of bandwidth limitation one can afford depends on how much accumulated phase shift one can tolerate.
Slew Rate as a figure of merit is indeed highly overrated. What we're interested in is what happens to the audio. An 18.5kHz+19.5kHz IMD test will tell us whether or not we've got a slew rate problem on our hands. If an amplifier performs well in such a test, we shouldn't care if slew rate is only just enough to reproduce this signal or if it's a thousand times higher. What matters is how linear the thing is when reproducing the fastest slewing audio frequency signal.
Post by: maxdimario on November 26, 2006, 03:20:06 PM
Post by: bruno putzeys on November 26, 2006, 04:06:04 PM
Post by: Ruairi O'Flaherty on November 26, 2006, 08:32:06 PM
| Bruno Putzeys wrote on Wed, 22 November 2006 08:24 |
An 18.5kHz+19.5kHz IMD test will tell us whether or not we've got a slew rate problem on our hands. If an amplifier performs well in such a test, we shouldn't care if slew rate is only just enough to reproduce this signal or if it's a thousand times higher. What matters is how linear the thing is when reproducing the fastest slewing audio frequency signal. |
Bruno,
could you elaborate on that test for an electronics beginner. We are looking at how the amplifier deals with the harmonic content created by the intermodulation distortion?
thank you,
Ruairi
Post by: bruno putzeys on November 27, 2006, 03:31:59 AM
If we wanted to measure distortion on a 20kHz signal, the second harmonic would be at 40kHz, the third at 60kHz and so on. This makes distortion measurements at 20kHz quite a pain.
The IMD test signal mentioned is the sum of a 19.5kHz sine wave and an 18.5kHz sine wave of equal amplitudes. If the amplitude of each is 1Vrms, the peak amplitude of the sum is 2.8V. The maximum slew rate will be that of a 19kHz signal of 2Vrms.
The second order product will appear at 1kHz. The 3rd order product appears at 17.5kHz and 20.5kHz. The 4th order product at 2kHz, the 5th order at 16.5kHz and 21.5kHz.
You'll see two bunches, one around the test tone (odd products) and one near DC (even products). If the distortion is bad (clipping or hard slew rate limiting) the two regions will overlap but you can still tell them apart because the even products are at integer multiples of 1kHz and the odd products at odd multiples of 0.5kHz.
Now nearly 40 distortion products will fit inside the audio band. If these distortion products are all very low, you can be sure the amplifier has no problems, including slew-rate wise.
This two-tone test has nothing "magical" that makes it more suitable for slew rate related problems, but it stresses the amplifier maximally and produces a lot of information. A 20kHz signal also stresses the amplifier maximally, but in a bandwidth limited system you don't get to see the result.
Post by: maxdimario on November 27, 2006, 06:29:17 AM
what kind of circuits exhibit the least asymmetric slew rate?
Post by: bruno putzeys on November 27, 2006, 07:12:44 AM
It doesn't matter btw. What matters is how an op amp behaves when confronted with a worst case audio signal. Slew distortion sets on well before the op amp actually hits the limit. I'd rather not use an op amp with "super slew rates" if it already starts distorting at 20kHz.
In other words, don't head for the slew rate spec if you want to find out if an op amp is going to measure well or sound good. Actual THD at 20kHz says much more.
Post by: maxdimario on November 27, 2006, 12:55:19 PM
Post by: bruno putzeys on November 27, 2006, 01:18:49 PM
An op amp that produces slew rate related distortion tells you it's been incorrectly chosen for the application. Conversely, no normal audio product will have such problems, because the designer will have taken care not to select an op amp that runs out of steam below 20kHz.
Post by: maxdimario on November 27, 2006, 07:50:06 PM
Would the slew rate influence distortion in opamps considering their complexity and corrective circuitry?
I say this because corrective circuitry needs to be fast and stable in order to work ideally..generally speaking.
I see you talk mostly using OpAmps as reference.
Post by: dcollins on November 28, 2006, 01:29:54 AM
| maxdimario wrote on Mon, 27 November 2006 16:50 |
How about a discrete transistor circuit with relatively low slew rate but low(er) global feedback? Would the slew rate influence distortion in opamps considering their complexity and corrective circuitry? I say this because corrective circuitry needs to be fast and stable in order to work ideally..generally speaking. I see you talk mostly using OpAmps as reference. |
Max, do you like the sound of IC opamps?
DC
Post by: bruno putzeys on November 28, 2006, 06:47:05 AM
When I speak of op amps I mean what's inside them. Whether this circuit is executed using discretes or on a chip has no import on how the circuit is analysed. Most discrete amplifier circuits are op amps too.
Depending on the amount of abstraction, an op amp has several possible definitions.
1) (schoolbook) a circuit with two inputs and one output. The output voltage is plus infinity times the voltage difference between the inputs. Actual output in a given circuit is arrived at through limit calculus.
2) (practice) a circuit with two inputs, one output and a output reference (usually one of the supply rails, sometimes ground). The output voltage is the time integral of the voltage difference across the input times the gain-bandwidth (expressed in radians/sec), referenced to the output reference.
According to the first definition, an op amp is a P controller with P tending towards infinity. According to the second definition, an op amp is an I controller with I equal to the gain bandwidth product (in rad/sec).
There is also a third "definition" which is more like a recipe:
3) An op amp is a transconductance amplifier followed by a transimpedance amplifier, where the transimpedance of the latter is substantially capacitive.
In most op amps (be they integrated circuits or discrete circuits) there is no "corrective circuitry" other than the gain of the amplifier itself. In circuits employing local feedback loops, these loops are faster than the main loop as a matter of course, or the whole shebang simply wouldn't work at all. No op amp using "corrective circuitry" that is *not* stable would even make it off the lab table.
If the need arises (and when I have more time) I'll put up some op amp theory here.
All too often, I read blurbs (sales literature) suggesting that "stability" and "correction" and what have you are "difficult" and so on. I should warn against the hijacking of technical terms to be used by sales people who charge them with emotions.
It seems that what "the terrorist" is to politics, negative feedback is to audio. Something to be vilified, stereotyped and used to make anything acceptable by.
Post by: zmix on November 28, 2006, 01:27:53 PM
| Bruno Putzeys wrote on Tue, 28 November 2006 06:47 |
I should warn against the hijacking of technical terms used by sales people who charge them with emotions. It seems that what "the terrorist" is to politics, negative feedback is to audio. Something to be vilified, stereotyped and used to make anything acceptable by. |
Bruno,
I do hope that this forum can maintain a utility beyond invoking 'gear hypochondria' in the infirm, but as they say "a little knowledge is a dangerous thing".
Be prepared...
-CZ
Post by: bruno putzeys on November 28, 2006, 02:02:53 PM
| zmix wrote on Tue, 28 November 2006 19:27 |
I do hope that this forum can maintain a utility beyond invoking 'gear hypochondria' in the infirm |
An occupational hazard for anyone trying to make anything clear, I fear.
| zmix wrote on Tue, 28 November 2006 19:27 |
Be prepared... |
Uhh don't frighten me more than I already am
Post by: dcollins on November 29, 2006, 01:08:24 AM
| Bruno Putzeys wrote on Tue, 28 November 2006 03:47 |
It seems that what "the terrorist" is to politics, negative feedback is to audio. Something to be vilified, stereotyped and used to make anything acceptable by. |
In Audiophile land, I think this is largely true. No one advertises that they use high levels of NFB!
At the Hi-Fi show it's always fun to ask the designer that shuns any type of feedback if he uses emitter resistors....
If they only knew that NFB is not magic, but it's close!
DC
Post by: Andy Peters on November 29, 2006, 01:35:53 AM
| dcollins wrote on Tue, 28 November 2006 23:08 |
At the Hi-Fi show it's always fun to ask the designer that shuns any type of feedback if he uses emitter resistors.... |
Bingo!
Oh, those degenerates ...
-a
Post by: arconaut on November 29, 2006, 08:56:44 AM
By the way, the Ciletti quote is out of context. His article was about modifying gear, and that sometimes in trying to make something "better" you take away the character that made the gear cool in the first place.
In swapping op amps, one might say "this slew rate is terrible" and try to find a better amp. That's when he made that statement about slew rate and the output transformer, his point being that the slew rate didn't really matter that much in the grand scheme of things. I just wanted to understand that little detail - but I got so much more.
Thanks all.
Noah
Post by: dcollins on November 30, 2006, 02:10:46 AM
| arconaut wrote on Wed, 29 November 2006 05:56 |
In swapping op amps, one might say "this slew rate is terrible" and try to find a better amp. |
3000V/us has got to sound better than the primitive gear that made all the records we love today............
DC
Post by: maxdimario on November 30, 2006, 08:47:32 AM
As DC has mentioned I have never found an op-amp which could perform as well as a discrete circuit for critical amplification applications such as mic preamps.
When I have been forced to use IC's, I have noticed that on GOOD (expensive) high-slew rate opamps the feedback-induced artifacts which I hear quite clearly, seem to diminish.
of course 'fast' circuits are prone to oscillation and ringing and other HF anomalies, and there is more to an IC than specs..
wouldn't a faster overall slew rate improve phase-issues in the feedback path? even subtly?
on the other hand, you SHOULDN'T need super-fast slew to make an amp work well with audio... but I have found that in commercial IC op-amps the high slew rate tends to minimize typical opamp artifacts (ear-wise).
have you ever worked with discrete transistor designs ?(incl. discrete opamps).
By the way I believe that hi-fi salesmen are one of the lowest forms of life.
Audio electronics and hi-fi are two different things..
Post by: bruno putzeys on November 30, 2006, 09:34:47 AM
The circuit I've settled on for the last 5 years was clearly worth the trouble. At around 20 transistors, it is neither simple nor minimalist. The reason why I designed this circuit was to get rid of the implicitly supply-rail-referred compensation in miller type circuits that are universal in IC op amps and almost so in other discrete op amps. PSRR of the standard miller circuit is zero if not for loop gain. I would go as far as to say that it's not worthwhile to build a discrete op amp if not to improve PSRR by moving the compensation reference to GND or to have more drive capability. All other performance specs on good IC op amps are already good.
Another way of improving PSRR on a miller type op amp is of course to increase GBW (PSRR of the "bad" supply rail equals loop gain). Usually though, fast op amps have other flaws. A fast op amp that is good on all counts is the new LM4562 . GBW is over 50MHz so PSRR (at unity gain) on the "bad" rail is at least 68dB at 20kHz (the +psrr plot at 15V appears to be in error).
Could you elaborate on how you deduce that certain sonic artefacts are feedback-induced and not caused by any other aspect of the operation of the circuit?
Post by: maxdimario on November 30, 2006, 11:00:06 AM
i'd love to but it's impossible to PROVE via internet.
I just basically spent 20 years listening.
a quick way of describing it is that regardless of tonal balance, freq response, percieved detail etc. there is a 'flattening' of the high-freq dynamics and percieved depth.
by percieved depth I mean the instruments which are up close are heard as being foreward or close and distant sounds and room reflections are heard as being distant.
sounds with a strong dry and clean percussive attack stick out a little more.
listening on speakers the effect is that the music is shifted foreward and 'in front' of the speakers.. into the room.
The non-tecnical response to low feedback amplifiers by people who have heard the results is that they can 'see' the musicians and instruments. This OBVIOUSLY depends on the quality of the source material and the recording techniques used.
so it has something to do with localization and percieved space.
when I think of localization etc.. I believe that the ears get their clues from something which is not tonally related but time-related..
in other words the correct reproduction of a transient attack and 'sharp' wave crests.
I've had arguments about this before...
Post by: maxdimario on November 30, 2006, 11:05:04 AM
in fact I believe that one of the main design or PRODUCTION advantages of using IC's is that they are built to be operated with cheap power supplies, but the power rail correction circutry creates problems with instability etc.
I can see why they would invent such devices..and why I would never want to use them personally..
Post by: Jim Williams on November 30, 2006, 11:36:40 AM
| dcollins wrote on Wed, 29 November 2006 23:10 | ||
3000V/us has got to sound better than the primitive gear that made all the records we love today............ DC |
Just might be. Pop in a LM6172 into a circuit using the tired 1975 vintage NE5532, albit with compensation and psu treatment and you will find that 3000 v/us slew rate part really does make a difference. It definately sounds much more open than either the LM4562 or the AD8599, although the noise is higher.
Post by: bruno putzeys on November 30, 2006, 11:49:06 AM
| maxdimario wrote on Thu, 30 November 2006 17:00 |
a quick way of describing it is that regardless of tonal balance, freq response, percieved detail etc. there is a 'flattening' of the high-freq dynamics and percieved depth. |
I am not asking what you like or dislike about the sound of certain circuits. I am asking why you believe this to be related to feedback.
| maxdimario wrote on Thu, 30 November 2006 17:05 |
in fact I believe that one of the main design or PRODUCTION advantages of using IC's is that they are built to be operated with cheap power supplies, but the power rail correction circutry creates problems with instability etc. |
What power rail correction circuitry are you referring to?
Post by: bruno putzeys on November 30, 2006, 11:56:44 AM
| Jim Williams wrote on Thu, 30 November 2006 17:36 |
Just might be. Pop in a LM6172 into a circuit using the tired 1975 vintage NE5532, albit with compensation and psu treatment and you will find that 3000 v/us slew rate part really does make a difference. It definately sounds much more open than either the LM4562 or the AD8599, although the noise is higher. |
Similar question: given that slew rate is not the only performance spec in which the 6172 is better than the 4562, why attribute the sonic advantage to slew rate?
Higher slew rates are found in op amps with greater GBW. As an example, I'll refer to PSRR once again. PSRR in a miller opamp equals loop gain. At frequencies above the dominant (real) pole, loop gain scales linearly with GBW and so does PSRR.
You can (easily) convince me that op amps with higher GBW will sound better, but that alone is not enough to draw the conclusion that specifically slew rate is involved.
Post by: zmix on November 30, 2006, 11:59:13 AM
| maxdimario wrote on Thu, 30 November 2006 11:00 |
Yeah! i'd love to but it's impossible to PROVE via internet. |
Max,
If you are going to continue to make the sort of bold assertions that you litter your posts with, you will need to allow your methods to be examined. As a bystander, it appears that you have no method at all, just a 'gut feeling' that what you want to believe is actually true.
Bruno has asked you to elaborate why you think these artifacts are feedback related, and not due to some other aspect of the design.
In other forums, when you have been asked direct questions you have abandoned the topic altogether.
Bruno is new here, he is a brilliant scientist and for the moment is accommodating all comers. Please don't waste his time.
Post by: Sahib on November 30, 2006, 02:15:21 PM
I apologise in advance for being out of topic.
| maxdimario wrote on Thu, 30 November 2006 13:47 |
By the way I believe that hi-fi salesmen are one of the lowest forms of life. Audio electronics and hi-fi are two different things.. |
Max,
I would take a great offence from this. I have a very good friend of mine whom is a hi-fi sales man. I don't think you realise that you are making the same subjective comments for which you look down on hi-fi salesmen. If I am remembering wrong then I apologise but you are the guy who said that you could anticipate the sound that a design would produce by simply looking at the schematics. Pleeeassseeee...
Finally, hi-fi is also a part of audio electronics.
Cemal
Post by: maxdimario on November 30, 2006, 02:35:24 PM
I take a circuit built with a variable feedback control..
I have built my own from scratch using both tubes and transistors..
...granted it is almost impossible to make a tolerable amp with zero feedback on say a FET amp, or a BJtransistor. usually tubes tend to do this better.
unvariably I hear the same thing in my circuits or different preamp models with different amounts of global feedback engineered in etc.
sometimes raising the gain on a preamp and applying a pad afterwards or before so that it is not working with high feedback will give better sound in this respect, although specs become worse for noise, thd and freq. response.
taking a sound which has the detail sucked out by passing it through a low fdbk (high thd) circuit will not 'revive' itself through a low feedback amp so this is proof that what I hear is not euphonic distortion. If this were true passing the sound through multiple stages of said amps would create an improvement instead of degredation.
obviously without feedback you are prone to problems with load mismatching, distortion artifacts etc. which are corrected by the feedback.
If the NFB is 'just enough' to get the amp to work properly and if the active components are good to begin with so that very little feedback is needed I find that the amp does not lose those elements of the sound which create the 'real space' illusion.
as far as the term 'corrective circuitry' I use it loosely perhaps, but what I mean is circuitry which is engineered in to improve a system which is not ideal.
I don't know if I am explaining myself in a complex enough manner.
basically I have just heard a lot of circuits: similar, different, opamp, discrete, transformer coupled, dc coupled, etc. etc.
hence my observations..
BTW some opamps sound better with less feedback but most need quite a bit.
Post by: maxdimario on November 30, 2006, 02:42:28 PM
| Sahib wrote on Thu, 30 November 2006 20:15 | ||
I apologise in advance for being out of topic.
Max, I would take a great offence from this. I have a very good friend of mine whom is a hi-fi sales man. I don't think you realise that you are making the same subjective comments for which you look down on hi-fi salesmen. If I am remembering wrong then I apologise but you are the guy who said that you could anticipate the sound that a design would produce by simply looking at the schematics. Pleeeassseeee... Finally, hi-fi is also a part of audio electronics. Cemal |
right, well it WAS a bit toungue in cheek but I did spend a lot of time in my teens in hi-fi shops and I grew to hate that 'we know better' pushy attitude that MOST hi-fi salesmen working (with commission) would give me..
add to that the 'miracle knob' syndrome and... that's what I meant.
as far as HI-fi being anything other than consumer sales nowadays, I don't really see that happening.
there are nice passionate people out there I'm sure and I hope your friend is one of them.
Post by: maxdimario on November 30, 2006, 02:45:21 PM
| Jim Williams wrote on Thu, 30 November 2006 17:36 | ||||
Just might be. Pop in a LM6172 into a circuit using the tired 1975 vintage NE5532, albit with compensation and psu treatment and you will find that 3000 v/us slew rate part really does make a difference. It definately sounds much more open than either the LM4562 or the AD8599, although the noise is higher. |
That IS one of the better opamps.
instant gratification..
that was my 'secret mod' to shock my friends who had bought expensive CD players.. made a cheap CD player sound better than a high-end CD!
Post by: maxdimario on November 30, 2006, 04:29:08 PM
| Quote: |
you are the guy who said that you could anticipate the sound that a design would produce by simply looking at the schematics. Pleeeassseeee... |
can you tell what a harley will sound like by looking at it's engine?
can you tell what an ac30 will sound like by looking at it's schematics?
can you tell what a jcm 800 will sound like..
how about a neve,
an api?
a 50's tube preamp? a V76 a v72 a v77..
an st-70, a macintosh tube amp, an SET
when you've worked with many different circuits and tried mods on many different circuits you can tell what a circuit will sound like..
I can recognize the difference and I know that one type of circuit will be good for one thing, and another for something else.
it's NOT such a feat.
Post by: dcollins on December 01, 2006, 01:52:44 AM
| Jim Williams wrote on Thu, 30 November 2006 08:36 |
Just might be. Pop in a LM6172 into a circuit using the tired 1975 vintage NE5532, albit with compensation and psu treatment and you will find that 3000 v/us slew rate part really does make a difference. It definately sounds much more open than either the LM4562 or the AD8599, although the noise is higher. |
I dunno. We'll have to agree to disagree here, as my experience with super-slewing parts has not been good. One mans open is another mans irritating, I guess. And it wasn't oscillating.
What was Buddy Holly's slew-rate? That sounded pretty good!
I still use discrete opamps in some places, but unlike Max, I don't think the emotional content of music is damaged by an IC stage....
DC
Post by: bruno putzeys on December 01, 2006, 04:28:50 AM
| maxdimario wrote on Thu, 30 November 2006 20:35 |
Feedback related. I take a circuit built with a variable feedback control.. I have built my own from scratch using both tubes and transistors.. (...etc) |
I recognise this story, I've done similar batteries of tests. The result of this was that I've been "against feedback" for quite some time too. In fact, I've long preached precisely the same gospel as you do (uhh including that I could guess the sound quality by looking at a circuit).
What I've learned since is that "a little feedback is a dangerous thing".
Part 1: Free Harmonics
Imagine you've got a circuit with only a 2nd order non-linearity i.e. a second order transfer like f(x)=x+0.1*x^2. Next, you put an ideal 20dB gain stage before it and close the loop (P control only for the time being). The result will be a transfer which is neither linear nor quadratic:
Before feedback: y=f(x)=x+0.1*x^2
After feedback: y=f(10*(x-y))
(write out f()) y=10*(x-y)+0.1*(10*(x-y))^2
(expand) y=10*x-10*y+100*x^2-100*x*y+100*y^2
(collect for y) 0=100*y^2-(11-100*x)*y+(10*x+100*x^2)
Right. That's a quadratic equation. I don't need to solve it for you to see that there are going to be square roots in this thing. The series expansion of a square root contains an infinite number of second order terms. You start with a circuit with only a second harmonic and you end up with a full set of harmonics after applying feedback. What this means is that a moderate amount of feedback will have created higher harmonics which are more audible than the second you started with. When we apply more and more feedback, the new harmonics will eventually come down again as well. In the end we'll need to apply quite a bit more feedback before the net improvement in second harmonic outweighs the presence of a (now small quantity) of higher harmonics.
Part 2: Making matters worse
Clean P controllers don't exist. Neither the controller, nor the system being controlled has infinite bandwidth. A normal control loop is asymptotically integrating. A circuit that has 60dB of loop gain at 1kHz will have 50dB at 3kHz, 40dB at 10kHz and 34dB at 20kHz. The 3rd harmonic of a 1kHz test tone will be reduced by 50dB. The 10th by 40dB. As a result, integrating loop control will skew the "harmonic balance" towards higher harmonics. This in addition to the fact that the problem outlined previously will also become greater as loop gain declines.
The curious result of this is that an amplifier with moderate loop gain will sound more "natural" if the loop gain is knocked flat below 20kHz (ie. if the dominant pole is real and at 20kHz).
Conclusion: Don't Be A Whimp
We now have two good reasons why low amounts of feedback may not sound all that good. When using feedback, use tons of it. In op amp terms that means: use wideband op amps. Make higher-order loops if your math skills allow you to (ever seen an inductor in a degenerated input pair). If you can't get really high loop gains at 20kHz, set the dominant pole at 20kHz. In this case, however, do not expect total freedom from colouration. The best you can get is a nonintrusive sound, but which at least leaves the music intact (and which is sometimes euphonic).
Post by: Larrchild on December 01, 2006, 10:49:53 AM
Norman Crowhurst touched on this in the 50's and basically this takes examples of BJT, FET and Tube, and models them with no feedback and small amounts of NFB. Quite an eyeopener.
http://www.its.caltech.edu/~musiclab/feedback-paper-acrobat. pdf
Indeed, upon building the test circuits mentioned in the paper, I saw the "upper harmonic skew" Bruno just mentioned:
Note: NFB = "no feedback" (sorry) FB= degenerative feedback from 1k emitter resistor.
I think the moral is..that you are supposed to use large amounts of global feedback to drive that stuff into the noise floor or lower.
Post by: Jim Williams on December 01, 2006, 11:13:38 AM
| dcollins wrote on Thu, 30 November 2006 22:52 | ||
I dunno. We'll have to agree to disagree here, as my experience with super-slewing parts has not been good. One mans open is another mans irritating, I guess. And it wasn't oscillating. What was Buddy Holly's slew-rate? That sounded pretty good! I still use discrete opamps in some places, but unlike Max, I don't think the emotional content of music is damaged by an IC stage.... DC |
Many earlier high slew opamps have high THD as well. Many conclude that since "this" fast opamp doesn't sound so great therefore all of them must be the same. The later generations of fast wideband opamps have higher gain-bandwidth (some approaching 90 db in the audio band) which pushes down THD well below common audio grade opamps with their 60 db open loop gain at 10k hz. Some have mondo current output stages like the THS4051/4011 that are heavy into class A. Even the current miserly LM6172 (2.3 ma supply current) will output over 100 ma.
One problem with these parts is they reveal. A harsh top end is not a result of these parts, it's from downstream and now you are hearing it without filtration effects. Fix the source of the irratations and all is well. Just don't expect Pro Stools to do this, it won't.
To those who still dismiss this new technology, I say try them yourself. Report back your findings, just don't dismiss something you haven't heard or tried yet. Everything sounds great until you hear something better. It's our job to do that.
BTW, I believe Buddy Holly's slew rate was 12 v/ us. That's what a 12AX7A can do. Pretty good compared to your average 5532/5534 opamp at 6 v/us. I still believe he sounded better live. Great artists have a special something that no microphone can encode. This I've learned working for Stevie Wonder. If you all could hear him without a mic, it's amazing. All recording gear takes away something. I can only imagine the presence and clarity if Buddy were to record with some of this new technology and I don't mean Pro Stools.
Post by: Larrchild on December 01, 2006, 01:11:47 PM
| Quote: |
BTW, I believe Buddy Holly's slew rate was 12 v/ us. |
Not through the input iron and output iron and magnetic recorders of the day.
Hardly anyone liked that custom Cherokee console with video opamps in it in the 80's if I recall. Hmm.
Post by: Andy Peters on December 01, 2006, 06:54:10 PM
| Bruno Putzeys wrote on Fri, 01 December 2006 02:28 |
When using feedback, use tons of it. In op amp terms that means: use wideband op amps. Make higher-order loops if your math skills allow you to (ever seen an inductor in a degenerated input pair). If you can't get really high loop gains at 20kHz, set the dominant pole at 20kHz. In this case, however, do not expect total freedom from colouration. The best you can get is a nonintrusive sound, but which at least leaves the music intact (and which is sometimes euphonic). |
Just to clarify: "tons of feedback" could mean low closed-loop gain?
-a
Post by: maxdimario on December 01, 2006, 08:26:33 PM
| Quote: |
BTW, I believe Buddy Holly's slew rate was 12 v/ us. That's what a 12AX7A can do. Pretty good compared to your average 5532/5534 opamp at 6 v/us. I still believe he sounded better live. Great artists have a special something that no microphone can encode. This I've learned working for Stevie Wonder. If you all could hear him without a mic, it's amazing. All recording gear takes away something. I can only imagine the presence and clarity if Buddy were to record with some of this new technology and I don't mean Pro Stools. |
excellent points.. this is exactly what I've been trying to put across.. especially preserving the live quality of a great artist.
stevie wonder will sound best without a mic, and therefore the presence needs to be preserved as much as possible..
there is a very 'fine' and etherial presence which can be felt being in front of an artist which some circuits capture better than others.
with a non-artist.. you can camouflage with 'euphonic distortion' and perhaps improve the result, but with a true artist you must try and get the essence of the performance..
it only takes a bit of smearing on the highs or phase distortion and you lose the ....feel...
good sound reproduction with a clear and open high end and midrange makes you emotionally alert to the sound therefore you become more RECEPTIVE to the emotion of the performances.
dull splatty highs make the music sound unreal, therefore not worth taking seriously from an emotional standpoint.
Post by: maxdimario on December 01, 2006, 08:44:09 PM
| Quote: |
When using feedback, use tons of it |
I am not a mathematician, so I cannot dechipher what you stated, although I do not doubt it's validity.
From a philosophical point of view it is better to use a system which has either no feedback and therefore exhibits all the strong, good characteristics of a no-feedback design, or if you're going to use feedback, you might as well go all the way and emphasise the good positive aspects of that high-fdbk design.
low feedback amps have typical distortion artifacts and positive qualities...and are cost-inefficient.
high feedback amps have their own subtle distortions.... even though the feedback is a corrective device to effectively cancel distortion, and the distortion can be made very low because of this correction. they are also very cost-effective
obviously you would aim to exploit the strong points of either type: high or low fdbk.
No-feedback designs are almost impossible to realize 100% because active components distort naturally..
the balancing act in getting the least amount of feedback includes the effort of making feedback as 'local' as possible and avoiding network feedback altogether if it can be done..
for this reason an amplifier with 2 active stages which uses low feedback will represent a more ideal 'low feedback' model than a low-feedback amp with 6 or 12 stages and global (network) feedback..
in opamps, which usually have many active components low feedback therefore is unapplicable.
Post by: Jim Williams on December 02, 2006, 11:14:10 AM
| Bruno Putzeys wrote on Thu, 30 November 2006 08:56 | ||
Similar question: given that slew rate is not the only performance spec in which the 6172 is better than the 4562, why attribute the sonic advantage to slew rate? Higher slew rates are found in op amps with greater GBW. As an example, I'll refer to PSRR once again. PSRR in a miller opamp equals loop gain. At frequencies above the dominant (real) pole, loop gain scales linearly with GBW and so does PSRR. You can (easily) convince me that op amps with higher GBW will sound better, but that alone is not enough to draw the conclusion that specifically slew rate is involved. |
Slew rate is one factor. It's also a reflection of the speed and bandwidth of the process transistors. I had an interesting discussion with an engineer at National Semi about the 4562 and other fast opamps from them like the 6172, 6171, 7171, etc. They claim crossover distortion is reduced in high speed opamps because the transistor's switching frequncy is so fast the zero cross point is of a shorter duration.
I've found that fast opamps with decent slew rates which have very high open loop gain at 10k hz, around 85~90 db, sound better than opamps with a 3k+ slew rate and 60~70 db open loop gain at 10k hz. Sonically, one will hear low level transients and details missing from lesser opamps without the huge open loop gain in the audio band.
Post by: Jim Williams on December 02, 2006, 11:20:38 AM
| Larrchild wrote on Fri, 01 December 2006 10:11 | ||
Jim Writes:
Not through the input iron and output iron and magnetic recorders of the day. Hardly anyone liked that custom Cherokee console with video opamps in it in the 80's if I recall. Hmm. |
Yes there was transformer slowing, it was an example. The slew rate of analog tape at 15 IPS is way worse. As I mentioned before, early high speed opamps had lot's of distortion. Still, some revere them. Consider the worshiped Ampex ATR's with their 1971 vintage LM318 70 v/ us slew rate opamps, folks love these things. I would believe that Cherokee console would sound lovely today with some of these improved opamps refitted in it.
Post by: maxdimario on December 02, 2006, 11:37:53 AM
Post by: Larrchild on December 02, 2006, 02:06:19 PM
| Quote: |
Consider the worshiped Ampex ATR's with their 1971 vintage LM318 70 v/ us slew rate opamps, folks love these things. I would believe that Cherokee console would sound lovely today with some of these improved opamps refitted in it. |
Interesting factoid regarding the ATR fast-slew opamp circuit people do love. The Cherokee example was used because I remember it as being the first time someone went "Let's build a desk with fantastic bandwidth and slew-rate" and after the fanfare, most people promptly went back to the A Range Trident. Perhaps the idea was not mature yet, and what you say is: "Give fast opamps a listen these days".
So I will!
Post by: maxdimario on December 02, 2006, 07:49:57 PM
fast opamps will not stand certain approximations in circuit-board and power supply designs..
Post by: Larrchild on December 02, 2006, 10:25:21 PM
Post by: dcollins on December 03, 2006, 12:24:11 AM
| Larrchild wrote on Sat, 02 December 2006 19:25 |
Yet, for 50+ years, video engineers have amplified and mixed 5 Mhz signals thru a gaggle of routing, and managed to keep em stable and phase-coherent, so people do it all the time. |
RF actually has to work. There are no Radiophiles, telling you the wire has some magic property previously unknown to Science.
Fwiw, when we did blind tests of IC opamps, this one seemed to win.
http://www.datasheets.org.uk/pdf/mc34081-datasheet/mc34081-d atasheet.html
Nothing special in the specs, slow and relatively noisy by modern standards. Can't drive 600R, etc.
DC
Post by: Bob Olhsson on December 03, 2006, 08:10:41 AM
| dcollins wrote on Sat, 02 December 2006 23:24 |
...RF actually has to work. There are no Radiophiles, telling you the wire has some magic property previously unknown to Science... |
I attended a SMPTE meeting in San Francisco about one of the first digital HDTV facility installations. Apparently some of the vendors hadn't felt they had enough digital expertise so they went out and hired some digital audio experts to help with their interface and wiring design. Soon after the facility went into operation, experienced producers began pointing at their monitors horrified wanting to know what's THAT!
It turned out to be jitter caused by techniques that are commonly used with digital audio. The "experts" had thrown something together without really understanding what they were doing and no longer had ABX tests to hide behind when somebody complained.
Post by: dcollins on December 03, 2006, 11:47:56 PM
| Bob Olhsson wrote on Sun, 03 December 2006 05:10 |
It turned out to be jitter caused by techniques that are commonly used with digital audio. The "experts" had thrown something together without really understanding what they were doing and no longer had ABX tests to hide behind when somebody complained. |
Ok, but I bet those "experts" solved the problem using engineering, not by painting the camera lens with a green felt pen to cure the j*tter!
DC
Post by: bruno putzeys on December 04, 2006, 03:45:22 AM
However, low closed loop gain does not necessarily imply high loop gain because you also need to know the open loop gain.
If you want to stand a reasonable chance at transparency (ie. no audible difference between input and output), shoot for a minimum of 60dB of loop gain at all audio frequencies. With normal circuit techniques open loop THD figures of 0.1% are realistic so the end result will be in the order of 0.0001%.
If the circuit you're designing has a lot of gain (mic pre), don't compensate your amps for unity gain. Consider distributing the gain among the stages and if possible run a higher order loop around the whole thing. This will make the compound gain of all stages available to fight the distortion of all stages at once. Works like a charm, and sonically it proves that minimalism is a trivial pursuit.
Post by: maxdimario on December 05, 2006, 09:08:06 AM
I could 'prove' the opposite with listening tests, as i have done before... to people who know nothing about pro audio as well as people who mix.
Post by: bruno putzeys on December 05, 2006, 12:27:15 PM
Post by: maxdimario on December 05, 2006, 01:49:10 PM
but that something it adds is negligeable when you weigh it against what it does NOT take away.
So there is no need to have such low distortion if it causes the music to lose that immediacy which you get by standing in front of an artist.
once you reduce distortion artifacts to -60 dB by using feedback it's obvious that the amp will sound 'neutral' in the sense that you will have very low thd and ruler flat response amongst other things.
unfortunately, there is a fine element which dissapears. Basically the 'liveness' of which I was writing about in the posts above.
Like I said.. it can't be proven over internet.. at least by me.
Post by: Andy Peters on December 05, 2006, 01:51:28 PM
| maxdimario wrote on Tue, 05 December 2006 07:08 |
That is a matter of opinion. I could 'prove' the opposite with listening tests, as i have done before... to people who know nothing about pro audio as well as people who mix. |
Hey, Max,
Can you PLEASE learn how to use the "QUOTE" button when replying to a message?
Every single post you've made in this thread is a reply to the first post, and since you never quote, I (and others, I'm sure), have NO IDEA which post you're actually replying to.
Thanks.
-a
Post by: maxdimario on December 05, 2006, 03:37:43 PM
Post by: bruno putzeys on December 05, 2006, 03:38:51 PM
| maxdimario wrote on Tue, 05 December 2006 19:49 |
once you reduce distortion artifacts to -60 dB by using feedback it's obvious that the amp will sound 'neutral' in the sense that you will have very low thd and ruler flat response amongst other things. unfortunately, there is a fine element which dissapears. Basically the 'liveness' of which I was writing about in the posts above. |
An amplifier that takes something away, anything, by any description, is not transparent either. If an amplifier is undetectable it neither adds nor takes away.
Post by: maxdimario on December 07, 2006, 12:13:48 AM
the source must be straight from a mic or close to it, and not from a cd, in order to tell..
my favorite test circuits have been mic pres and turntable pres with 50-60's records because of the simple recording path etc.
I never heard a circuit that was undetectable, can you quote me an example?
Post by: dcollins on December 07, 2006, 12:44:19 AM
| maxdimario wrote on Wed, 06 December 2006 21:13 |
define 'undetectable'. |
Undetectable by who? Or is that what you mean?
| Quote: |
I never heard a circuit that was undetectable, can you quote me an example? |
Max, what about the AD-797? Looks pretty good on paper.
http://www.analog.com/en/prod/0%2C2877%2CAD797%2C00.html
It's "Ultralow Distortion!"
DC
Post by: bruno putzeys on December 07, 2006, 02:49:39 AM
Post by: Jim Williams on December 07, 2006, 11:08:31 AM
| dcollins wrote on Wed, 06 December 2006 21:44 | ||||
Undetectable by who? Or is that what you mean?
Max, what about the AD-797? Looks pretty good on paper. http://www.analog.com/en/prod/0%2C2877%2CAD797%2C00.html It's "Ultralow Distortion!" DC |
That part has been out for years. It's pretty bland sounding. They have a new one, the AD8599 dual which has similar noise specs but greater gain bandwidth. The new BurrBrown OPA211 is also similar in noise, 1 nv/hz/sq and is easy to work with and it has a 55 mhz gain bandwidth, greater than the 797 without the stability issues. A dual version is also in the works.
So now all these manufacturers have a new arsenal of quality opamps to replace the tired old TL07X and 553X opamps. Don't expect to see them. With low bit audio and compromised sonics being the norm, giving up on a 15 cent opamp for a $2 opamp isn't going to happen. I expect these new devices to be used in medical and industrial applications, not audio. The LM4562 is going to be a dissapointment for National which made the extra effort to release the parts in a 8 pin dip and TO-99 can. The rest of these are surface mount only.
Post by: bruno putzeys on December 07, 2006, 11:30:44 AM
The description "bland" fits my experience of the 797 quite well.
Post by: zmix on December 07, 2006, 01:58:41 PM
Interesting that you have brought up both the MC340XX and the AD797 series. I've had extensive time listening to each. Recently I had one of my power amps break down, it's a boutique solid state amp, over designed, current sources on the current sources sort of thing. It turns out that the one opamp in the circuit latched up, sending 60 volts DC to my woofer.
I peeled off the heatsink on the opamp to discover it was an AD797. I replaced the device in each channel with an op amp I prefer the 'sound' of and viola, every annoying harsh glaring quality of the poweramp was gone... not replaced by euphonic coloration, mind you, just lacking the harsh edgy quality I've endured all those years.
CZ
Post by: Terry Demol on December 07, 2006, 05:30:22 PM
| Bruno Putzeys wrote on Thu, 07 December 2006 16:30 |
Well I certainly hope they will keep making the LM4562 for some time. Even though I've not yet done an in/out comparison it's been sounding better here than anything it replaced. The description "bland" fits my experience of the 797 quite well. |
WRT 'bland' ditto here.
So Bruno, how can -120dB across the audio band sound 'bland'?
Or are we just looking for 'not bland'?
Another experience with the venerable 797 was with mic pre's.
Fully balanced, fixed gain of around 100, head amp for ribbon mic
used for stereo blumelin recording of small classical ensembles.
I didn't use compound stages as the 797 has good OLG
for closed loop gain of 100.
A balanced discrete zero feedback circuit was much better.
We were -not- looking for colour, and we had the live source
for reference. I was frankly shocked how transparent it was.
cheers
Terry
Post by: Dan Kennedy on December 07, 2006, 07:22:31 PM
Long time, no see.
Zero feedback, as in loop, or local (degenerative) or none of either?
Post by: bruno putzeys on December 08, 2006, 04:05:45 AM
| Terry Demol wrote on Thu, 07 December 2006 23:30 |
So Bruno, how can -120dB across the audio band sound 'bland'? Or are we just looking for 'not bland'? |
That's always the question. Like the most neutral speaker is bound to lose in a quick shoot-out, the most neutral amplifier is also generally found to be rater middle-of-the-road. Only when one listens longer does one hear that the "correct" amplifier does deliver all the goods, it only doesn't emphasise any of it. So to me, "bland" is not always an indication of a problem.
The 797 seems a different shade of bland though. I don't pretend to know how an amp with these figures could sound anything but "totally not there" - I hope to find out some day. I do have some indications though. In a noninverting circuit, several input stage error sources (small though they may be) are outside the feedback loop. One unbalanced line amp I've been toying with recently consisted of a pot followed by a discrete 10x gain stage. High frequency (6kHz) THD was markedly dependant of the wiper position, even if the input amplitude was adjusted to get the same output each time. As could be expected, THD was worst when the pot was half-way (we're talking about -106dB at 4Vrms). The impedance mismatch at the two amplifier inputs made the effect of the nonlinear input impedance of same stand out.
I then rewired the circuit to inverting, placing the pot in the feedback circuit and grounding the noninverting input. Even though this results in variable loop gain, the distortion became nearly constant (around -120dB) up to +20dB gain. Since both amplifier inputs are held at ground potential, the input capacitance is not modulated.
This line stage is now as undetectable as you can get in an unbalanced setting. In its previous state it was, well "bland", and there was a clear loss of detail when it was inserted.
I plan to do some further experiments with IC op amps to see if this finding can be extended to explaining some of the extreme colourations in IC op amps that nevertheless have good specs. In another context, I have been using AD797's and OPA627's as the first stage in an integrator chain (discrete 1-bit converter) and the difference was extremely small. Again, the inputs were used as a virtual short. Anyone who has toyed with these 2 op amps will attest that in normal gain stages there are no two more different sounding op amps around.
Whether that counts as an explanation is still an open question, but these two bits of anecdotal evidence suggest that inverting (virtual-short) circuits are less sensitive to these colourations, giving some indications as to where they come from.
Post by: Dan Kennedy on December 08, 2006, 08:19:11 AM
Post by: dcollins on December 09, 2006, 01:49:01 AM
| zmix wrote on Thu, 07 December 2006 10:58 |
I peeled off the heatsink on the opamp to discover it was an AD797. I replaced the device in each channel with an op amp I prefer the 'sound' of and viola, every annoying harsh glaring quality of the poweramp was gone... not replaced by euphonic coloration, mind you, just lacking the harsh edgy quality I've endured all those years. |
That's what I'm getting at.
Barring abject failures or dumb implementations, how can one opamp be both "bland" and "edgy?"
What did you replace the AD797 with?
An MC34081?
I swear there is some way to quantify these differences nowadays, but there is obviously nothing wrong with running the one that sounds better!
The 797 is also prone to "break into song" in a circuit that was working fine a minute ago....
But I think it can work.
Am slowly, slowly, turning to discrete versions....must resist....
DC
Post by: bruno putzeys on December 09, 2006, 03:26:06 AM
| dcollins wrote on Sat, 09 December 2006 07:49 |
Barring abject failures or dumb implementations, how can one opamp be both "bland" and "edgy?" |
I would indeed strongly suspect that something else (as in "look at the system level") is going on here.
Post by: maxdimario on December 09, 2006, 08:54:16 PM
makes sense to me. clean with no balls and a plastic-sounding top end.. plastic as in plexiglas.
that is the sound of a lot of clean opamps.
Post by: Terry Demol on December 10, 2006, 05:20:10 AM
| Dan Kennedy wrote on Fri, 08 December 2006 00:22 |
Hi Terry, Long time, no see. Zero feedback, as in loop, or local (degenerative) or none of either? |
Hi Dan,
Likewise.
ZFB as in no glob, no interstage, no local (as in cfp etc),
just degen.
You know I know that degen is, strictly speaking, not ZFB
Cheers,
Terry
Post by: bruno putzeys on December 11, 2006, 04:01:42 AM
The reason why many don't see the feedback circuit is:
*Because they don't get that the "input terminal" is the voltage difference between b and e and the "output terminal" is the current through c and e.
*Because feedback is evil whilst emitter followers are not so emitter followers are not feedback.
Post by: Terry Demol on December 11, 2006, 10:54:37 AM
| Bruno Putzeys wrote on Mon, 11 December 2006 09:01 |
I might spend a thread on this but in short the only way of properly analysing an emitter follower (and hence a degenerated cec) is as a feedback loop. Redraw the transistor as a 4-terminal block with voltage inputs and a current source output. This is a transconductance amp. One end of the current source is tied to a load and to the inverting input for feedback. Feedback is unity gain so loop gain = transconductance * load impedance. Tadaa... The reason why many don't see the feedback circuit is: *Because they don't get that the "input terminal" is the voltage difference between b and e and the "output terminal" is the current through c and e. *Because feedback is evil whilst emitter followers are not so emitter followers are not feedback. |
Yes, I completely understand this
I think most designers do and it has been discussed a lot over
the last few years. It's really a terminology thing to describe a
particular topology.
cheers
T
Post by: bruno putzeys on December 11, 2006, 04:08:05 PM
That's it, innit? Cathode followers aren't evil so they can't be feedback...
sheesh.
Post by: maxdimario on December 11, 2006, 07:34:29 PM
Post by: dcollins on December 11, 2006, 08:49:58 PM
| Bruno Putzeys wrote on Mon, 11 December 2006 13:08 |
That's it, innit? Cathode followers aren't evil so they can't be feedback... |
Reminiscent of the quiz question "If my emitter-follower has no NFB and a gain of less that one, why is it oscillating!?"
DC
Post by: Andy Peters on December 12, 2006, 02:52:03 PM
| maxdimario wrote on Mon, 11 December 2006 17:34 |
i hate cathode followers. |
To what message are you replying?
-a
Post by: zmix on December 12, 2006, 03:14:41 PM
Post by: Larrchild on December 12, 2006, 04:41:00 PM
But it's there.
Post by: bruno putzeys on December 13, 2006, 06:36:23 AM
| Terry Demol wrote on Tue, 12 December 2006 00:19 |
You might find it interesting that the cathode follower has a reputation in the tube fraternity for sounding bad. Apparently loss of detail ( I think). But I'm sure you have heard plenty of tube stuff from Mr Tent and others. |
(Max gladly chimed in to confirm the CF's reputation)
The same tube community also has people who know how to build cathode followers that don't sound choked. Simply cascode another triode on top of the output tube to insure constant voltage operation, and replace the cathode resistor by a current source or use a negative supply to obtain constant current operation (apart from the output current).
An alternative to the cascode is to use a pentode with g2 boot-strapped to the cathode.
Post by: maxdimario on December 13, 2006, 09:11:49 PM
this will not yield as 'low' of a source impedance as in CF, but it does not have the load-dependent shortcomings of a CF.
In cathode followers the gain of the tube is being used 100% to compensate for voltage distortions due to the load. you also have asymmetrical current draw.
2-tube totem-pole-type circuits work better but are more complex and still rely on voltage feedback.
but better is to use a tapped inductor or... a transformer
this way you use the energy flowing through the tube in a more efficient and symmetrical manner, with the actual work of the tube being transferred to the load ideally.
Post by: eddieaudio on December 27, 2006, 03:59:05 AM
My apologies for taking so long to jump in on this topic, due to the holidaze... I thought I'd take a moment to explain myself...
When I've got a box on the bench, one of the first things I do is run a sine wave through and look at the overload characteristic - no load and 600-ohm load - and then apply a square wave under similar conditions. These two tests quickly tell me if there are any obvious problems (like a bad class AB output stage, bad caps, transformer ringing etc.). And, since I know what the rise time of a square wave is supposed to look like, I tend to investigate when things look a little "slow," if only to satisfy my own curiousity.
Take, for example, the 100k output pot on the 1176 (attached). I was scoping things out when I noticed a difference pre and post pot - keeping in mind the vulnerability to cable capacitance. UREI used a generic shielded wire from the pot wiper to the amp. For the hell of it, I tried a higer grade, low-capacitance belden cable and realized a slight improvement. I kept the original wire, just in case the customer objected, but he was happy and gave me other work.
In the studio, by contrast (and, you could even say, to contradict myself) I find that lots of things are too bright - or brittle - and while I wish I had an arsenal of ribbon mics to compensate, I make do with what's available.
One day I was recording sax and chose to use a modified Pultec MB-1. The circuit is very similar to the Altec 1566, but with negative feedback, so adding a feedback control allowed me to hear it both ways - and to demonstrate the effect feedback has on frequency response, the overload characteristic and square wave response to my studio maintenance class.
Now, understand I like the way 78-era recordings make the reeds sound and when the sax mic was opened up it was NOT what I was looking for so I tried the MB-1. Initially, it to had too much detail, so I turned the feedback OFF and attenuated the secondary of the input transformer to compensate for the additional gain. The "defocused" sound was an improvement to me and to those listening.
My point here is not to say whether feedback is good or bad - for opamps, it's obviously necessary. And, for opamps, I do prefer mine to be neutral. ALL of you have detailed how challenging that can be, both technically and psychologically.
In the case of the UREI LA-4, for example, I think we can all agree that the 4136 is neither transparent nor does it color the sound in a complimentary way AND that almost any newer opamp would be an improvement. Especially in this case, when a transformerless circuit degrades a squarewave like the stock LA-4 does, I tweaked it until it looked more like the input. And, I might add, not from an expensive piece of test gear but from a $60 battery powered oscillator.
In this digital era, we have come to rely upon / expect a wider range from our sonic palette and this is the reason there are so many options from so many manufacturers. On one hand, there is a definite need to pursue the more realistic path of high fidelity, of precise capture and reproduction. On the other, I think simple discrete circuits without feedback have an equal but opposite contribution to make.
Our ears are all different and so pleasing them at the same time is not an easy task. I rely on my ears, but I also take stock in what I see on the 'scope and the distortion analyzer - we've got to at least attempt to correlate these things - me in my crude way and others with their greater depth and arsenal of tools.
My hat is off to Bruno and anyone who can correlate the math with what we see via schematic and test equipment, as well as what our ears tell us. I also apologize for whatever shortcomings my columns may have. They are not easy to edit and I am often late in getting them in on time, making it harder to zoom out and get a better perspective (until they appear in print, when it's too late). I am thankful to all those who take the time to read and attempt to make sense of them. I also welcome your criticism and feedback, positive and negative.
Best wishes to all in the coming year.
sincerely,
eddie ciletti
Post by: bruno putzeys on December 27, 2006, 10:10:34 AM
| eddieaudio wrote on Wed, 27 December 2006 09:59 |
I also apologize for whatever shortcomings my columns may have (...) |
I've only recently started to appreciate how hard it is to strike the balance between technical exactness and readability. So far most times I tried writing something for a wide audience I got woefully lost somewhere between the facts and useful copy. So, there's no need to apologise!
Post by: maxdimario on December 28, 2006, 01:13:04 AM
| Quote: |
Now, understand I like the way 78-era recordings make the reeds sound and when the sax mic was opened up it was NOT what I was looking for so I tried the MB-1. Initially, it to had too much detail, so I turned the feedback OFF and attenuated the secondary of the input transformer to compensate for the additional gain. The "defocused" sound was an improvement to me and to those listening. |
This is what I used to think as well about 10 years ago.
with more experience in the matter, and after building some very clean sounding amps with superior tubes, I realized that detail is not that important compared to artifacts induced by NFB.
the fact that the sound goes out of focus (even though it may be a desireable 'effect') is not usually what is good about low fdbk amps..
you can build low feedback amps which are 'in focus' and only then do you realize that some of the information which the ear needs to hear to identify a sound as 'real' gets lost in higher feedback amps.
this is especially true in circuits with high 'global' or network feedback..
Post by: eddieaudio on December 28, 2006, 06:21:28 PM
Even after using the MB-1, I still pulled out alot of upper midrange when it was time to mix. Here's a link so you can hear what I was going for.
http://www.tangible-technology.com/ipr/AP-292/summer_06/day- 8/saveYourLove4me_proc.mp3
There are two saxes, one live and one overdubbed. The live sax is a combo of a Royer and a Sennheiser e-609. For the overdubbed vocal and Sax, a Neumann M-149 was used, amplified by the LA-3 and the Pultec, respectively. The rough mix was done on PTLE running on a sony VAIO lapdog. The verb is both real and soft - the overdubs were recorded in a reverberant space.
ec