It's very easy to set up a testing situation where the ears can be trusted - simply put on a recording which stresses the system in a key way - by, say, running it at high volumes, and for the mix to be very complex, with high levels of echo and ambience - where the distortion artifacts, the 'problems' are easy to hear - it's then trivial to pick the differences caused by various factors, the obvious distortion becomes your guide as to where "work needs to be done" ...
Last edited:
I don't undesrtand what the problem is?
On the other hand, I don't really know what "digital sound" is, to me something either sounds right to my ears, or it doesn't. The usual words used for that are harsh, spitting, etc, but I have heard the same problems with purely analog, so it's probably not that?
People can get carried away with running away from digital sound, you know. My own old Yamaha CDX 993 CD player is a good example. AFAIK, it was never sold in North America. Anyway, buillt like a tank, weighs in at around 22 lbs, two trafos, 4 6,800 uF caps for main filtering, the works. As is, out of the box, it sounds warm and lush, as few CD players I have encountered ever did.
Inside, the I/V stage, buffer stage and output amp are truly separate stages, each all on its own, and the buffer and output stages are done in discrete technology, no savings, no cutting of corners, somebody really wnet to town on them.
And all that is well and fine, until at some point I realized that all this warmth had a price tag attached to it - a certain loss of fine detail.
The point is, getting away from the usual somewhat sterile or clinical sound of many CD players can be overdone and lead you astray in the other direction.
And who's to decide? Your ears only, there is no other meritory judge. I think that's what happened to John - he heard something he liked, and note he did say he thought he was "usually" right.
So why do we now need to discuss his personal imressions?
Just for a change, let's not.
On the other hand, I don't really know what "digital sound" is, to me something either sounds right to my ears, or it doesn't. The usual words used for that are harsh, spitting, etc, but I have heard the same problems with purely analog, so it's probably not that?
People can get carried away with running away from digital sound, you know. My own old Yamaha CDX 993 CD player is a good example. AFAIK, it was never sold in North America. Anyway, buillt like a tank, weighs in at around 22 lbs, two trafos, 4 6,800 uF caps for main filtering, the works. As is, out of the box, it sounds warm and lush, as few CD players I have encountered ever did.
Inside, the I/V stage, buffer stage and output amp are truly separate stages, each all on its own, and the buffer and output stages are done in discrete technology, no savings, no cutting of corners, somebody really wnet to town on them.
And all that is well and fine, until at some point I realized that all this warmth had a price tag attached to it - a certain loss of fine detail.
The point is, getting away from the usual somewhat sterile or clinical sound of many CD players can be overdone and lead you astray in the other direction.
And who's to decide? Your ears only, there is no other meritory judge. I think that's what happened to John - he heard something he liked, and note he did say he thought he was "usually" right.
So why do we now need to discuss his personal imressions?
Just for a change, let's not.
That "loss of fine detail" can be recovered in those units. My CDX 1100 needed a lot of conditioning, over many days, to get there, but it was that CD player that first gave me convincing sound; it showed the way forward, and was my mainstay until recently. It was able to playback test tracks attentuated 60dB, not quite as clean of digital artifacts as recent units, but still retained its 'musicality' at these low levels.
It didn't fail ABX (or whatever is the equivalent in sensory research). Precisely because it was shown that sublimal messages DID have a clear and undoubtfull result, was it banned.
Although the advertising industry has found ways to influence you overtly and again there's no doubt it works.
jan
Fine Frank, but I really don't want to have rebuild each and every CD player I use. It was fun for a while, but that was like 20 years ago.
These days, I'm just a 61 year old fart who couldn't be bothered to do anything more than change an op amp or two.
It works only if you open the door, i.e. switch on your TV. These days, I don't have to do even that, my Sony Blue Ray player does that automatically when I press PLAY.
Regular TV I almost do not watch at all, only a good movie now and then. It's a nasty habit I managed to shake off many years ago. Besides, no TV program can come even close to you guys here.
What's f in the formula?
You seem to know the mathematics well, do we factor in settling time as well? What is an ideal settling time?
I've listened to op-amp's for years, playing around with them until the late hours of the night, which I really enjoy, I just wish I knew which exact parameters I'm actually listening to.
In my blind tests I can still hear IC's in the same way as I hear them sighted, so I've mostly set aside the illusion part but I have no idea what I'm hearing in the spec sheet, or if it's even in the spec sheet or in any dScopes at all, haha!
For example, the Cyril Bateman measurements of capacitors seem to be too err...... unique?... to exist in the standard functions of any dScope or in any spec sheet in the entire world.
That makes zero sense to me, but if assumed true it could mean there's a Cyril person for op-amp's as well? Just asking.
As far as I can see slew rates are anecdotal ( see video op amps later). The typical op amp style device which includes power amplifiers has two inputs . They are via differential long tail pair circuits ( LTP ) that resemble a see-saw balance with a fulcrum which might be said to be 0 volts signal input.
The inputs before dedication to task are identical in construction. My suggestion is that the input No 1 the signal is only mildly challenging. That even includes the less pure signal sources like Vinyl. Input No 2 receives the output signal which includes the transistors jumping and being asked not to hit either rail and any current limiter glitches. The waves into No 2 will be rich in RF harmonics so do demand the fast slew rates, to me the conjecture now looks real. Fourier series for even triangle waves will be rich in RF . My best guess is these glitches will be square cum triangle.That's my best explanation as to why we state something so often that never gets questioned. We did the sin of using our ears as evidence. That's the sin of all here until wanting to sell an idea.
If the feedback loop does let go of the output even briefly then stored energy in the speaker systems I suppose doesn't help things. I recently built the simplest possible speaker. I noticed it sounded great with all amplifiers , even the Audiolab that I can not stand . Alas not good enough to call it a good speaker to suit all tastes. The reactive elements in the speaker no matter how much we object are the sound we hear compounded by other pieces of wishful thinking of standard practice.
FET's often said to deal better with these problems at the input LTP. I would argue for different reasons at the output also which follows neatly into class D.
I spend my life looking at oscilloscopes . I recon if the Cdom ( dominant pole capacitor, see Douglas Self) is so hard to drive then to send 100 kHz in to produce full power should show something. Seldom much of note. Sometimes rounded off and asymmetrical. Rarely cataclysmic things. Try the same with a valve amp. The reason the valve doesn't mind is it has a set current and falls over if asking more. It is class A and often single ended. They sound fine because they are not switching or glitching. No junk feed into input No 2 if they have one. Interestingly if they do have feedback life starts to get much harder.
I built amps with single inputs instead of long tail pair ( LTP ). That is input No 1 base, input No 2 same transistor emitter. They are valued by some. To me they were not the answer. Double long tail pairs are a good answer as they simplify getting reasonable slew rates and symmetry. The only arrangement I have never tried in transistor is to sum the output with the input signal via input No1. It looks to be a better idea. I would say books that suggest one type of feedback better you should try them. I never got blown away by any of them . By that the measurements in the audio band never change very much and I did lots of listening. If you ground input No 1 of LTP and feed the signal in via the feedback capacitor if fitted you have the so called ideal situation . Inverting, summing and minimal phase shift. If you make that resistor 2 K most op amps will be happy to drive it and have a reserve of current. Up scale your feedback upper arm to suit, In my amps it is 1K 33 K , that can be 2 K 68 K. It should be the " bees knees ". I doubt it will be. Reason is the glitches are still there.
Why op amps should need high slew rates beats me. When driving headphones I suspect it will help. Op amps will drive 32 R loud enough. JFET input seems logical as they resist real RF better. Video op amps are often the fastest slewing devices and their slew rates seem just OK for the job they have to do ( interesting ). Some quote distortion and noise and hint at audio uses. Many look to be good MC pick up/ microphone amps as stage one using +/- 2.5 V rails. They often will drive 75 ohms. I recently made LM358 perform OK. I converted it's output to class A SE. I substituted it in my low distortion oscillator, it held up to 30 kHz. It put up a good fight and sounded OK into headphones. This is to op amps what a moped is to a Porsche. One thing I did notice about op amps is they have optimum gain . You will, have to find out for yourselves that one. Try 1 and 10 as starting points.
As said before the slew rate argument is the Phlogiston on electronics. I like everyone know it to be real and true. I just want to know it's real name?
My other conjecture is class D works against much of class AB thinking. It works well enough and suggests it couldn't if the AB rules are right. In AB we might imaging things are bad, in class D they simply are bad. Send it up high and put a filter - 20 dB 400 kHz to solve the problems. I do wonder if we only need a tiny choke in our feedback loops as the cure to some AB ills? Air cored hand wound one. It is possible a wire wound resistor might work.
Hypex Electronics BV - UcD
Last edited:
In fact, I didn't do that - the raw quality was good enough to get me there, it was only a number of nominal things that I did internally; I found the key "ingredient" was the conditioning of the circuitry by leaving it running 24/7.
Yes, one slows down over the years, the onslaught of life takes its toll,
- but, my problem is that I'm a born troubleshooter, : if something is not quite right, and I know it can and should be better, then it really bugs me until I have at least managed a workaround ...
It works only if you open the door, i.e. switch on your TV. These days, I don't have to do even that, my Sony Blue Ray player does that automatically when I press PLAY.
Regular TV I almost do not watch at all, only a good movie now and then. It's a nasty habit I managed to shake off many years ago. Besides, no TV program can come even close to you guys here.
True. But if you think about it, there's hardly anything you can do, anyplace you can be, without being inundated with messages about what is good for you.
Go into town, look around you with this in mind.
Now, a lot of that you don't usually notice as you have pushed it down to your unconscious part.
Yet it still influences you.
Let's face it, the concept of free decision making in our so-called developed part of the world gets more and more imaginary!
Jan
It may have very well worked in reality, it may have been emotionally, physically, thoughtfully felt in different ways, however, they could not show if it worked or not.
"Subliminal advertising is banned in many countries, even though there is no evidence that it works. The term was coined in 1957 by market researcher James Vicary, who claimed that he had induced people to buy more Coke and popcorn by flashing brief messages during a film. The practice was banned in the UK, US and Australia the following year, even though other tests (undertaken in haste in order to complete them before the ban came into force) had failed to confirm the experiment, and in 1962 Vicary admitted that he had falsified the results"
QI : Quite Interesting
True, it's difficult to affix words to the way in which we experience audio and music. Let's set music aside, just audio is processed sonically, aesthetically, visually.
If it's subliminal at times, in some way, then it could be emotionally or physically processed as well, in theory.
There is some evidence learning towards that sound can, in some instances, be more efficiently processed in the visual cortex.
Neuroplasticity - Wikipedia, the free encyclopedia
The best example of subliminal is a keyword or words. You put in words that sound about right like " feedback loop security factor " 8 : 1 is 18 db and you have some evidence. It is suggested if another person you need to convince uses your words then you have won and should not use them yourself afterwards. These people often will be the promoters of your products. They are more often very methodical people who can never think fast enough to challenge a concept that is based on fact albeit loosely. The internet makes this even more powerful as bonding seems to be the key need of people on the net. Ironically more so when they don't meet. To be ultra successful, tell the truth.
Last edited:
^
Nigel Pearson, what did you mean about the really high Video op-amp slew rate that is "just ok"?
You said you've spent your life looking at oscilloscopes, what do you think about the Cyril Bateman papers on capacitors?
I noticed we are on page 1364 of this thread. What do you intelligent and clever people in this forum think about the LT1363 and LT1364 chips from Linear?
Nigel Pearson, what did you mean about the really high Video op-amp slew rate that is "just ok"?
You said you've spent your life looking at oscilloscopes, what do you think about the Cyril Bateman papers on capacitors?
I noticed we are on page 1364 of this thread. What do you intelligent and clever people in this forum think about the LT1363 and LT1364 chips from Linear?
Ahoy there, Nigel,
This is as far as I got so far. Attached is the latest version of the work in progress.
I cannot remember the changes from the last time, but there are quite a few. Some were made based on advice and guidance of Demian, most notably the input stage, which now uses standard FETs in a cascode configuration.
The VAS has evolved the most. As suggested by Samuel Groner, there are two semiconductors acting as a buffer stage. In case of Otala/Lohstroh amp, this was a classic differential amp, with its own CCS. In my case, these buffers are referenced to the ground rather than the negative power supply line; I tried bith versions and this one worked better for me.
As suggested by Demian, the input FETs are biased at just under 5 mA each; at first, I was falbbergasted and missed a heart attach by a narrow margin, this was unthinkable for me before that time, but I gulped and tried it, and man oh man, does it work, or what! Strangely enough, I found that the buffer FETs are better off at around 3.5 mA each. Since the voltage drop across them is just 3.2V, I can get away using standard 2SK170 FETs, as their 40V rating is meaningless in this spot, while the input stage FETs work at 10V, so their relatively high bias current is not a problem dissipation wise.
The last stage of the VAS works into a current mirror, as it did before, Not much change here, about the only one worth mentioning is the addition of a small value resistor to the base of Q21. Thorsten pointed out that using a transistor as I did was adding a leakage to an otherwise perfect current pump. As ever, I take hi comments very seriously, and tried with the said resistor added, and I did get somewhat better performance figures, a nick, but better.
The rest you know, my classic Locanthi three piece output stage. On your suggestion, I did add a small value resistor and a relatively large (2,200 uF) capacitor to the ground, to stabilize the driver and isolate it from the power devices.
The small value caps (10uF, 100 nF) you see sprinkled around is my own quirk. Experience has taught me that more often than not if the amp goes unstable at some point, it’s very likely due to inadequate bypassing, so I never save money on that, in fact, I’m rather generous in that respect.
The protection circuits for overvoltage/overcurrent have been pushed to the limit and have been worked out according to the manufacturer’s 50 mS performance data. After all, that is the the time delay of the circuit anyway. This allow each output device to pump out about 6.7 Amps in 50 mS peaks, or around 27 Amps for the whole kaboodle. The only person I know who will not be happy with that is Wayne, the rest of us should be just fine.
Speaking of supplies, they were, voltage wise, the ONLY constraint. I decided not to go over the magic 63V capacitor borderline, because that’s when prices start to soar. For the current amp, that is, the input stage and VAS are fed by stabilized +/- 61.5V lines, the current stage by +/- 56.5V (which, according to the manufacturer’s data, rises to just over 58V under the no load conditions – trafo regulation is given as 3% or less).
Which means that under dynamic conditions, the amp should be able to pump out peak power levels of 180/350W into 8/4 Ohms. A rather nice and healthy dynamic margin, I’d say, that’s about 2.2 dB over the nominal.
But remember, this is still work in progress, so for example compensation items (R15C16, C7 especially are as yet unknown). C13 and C14 were initially set at 10 pF, and C17 at 5.1 pF, and no instability was found so far. Thus, it is possible some of the compensation components may not even be needed, but we’ll see, to early to tell yet.
The values measured show promise, and then some. At full nominal blast (28.3V or 100/200W into 8/4 Ohms), under open loop conditions, 20-20,000 Hz THD is less than 0.3% into 4 Ohms and 0.15% into 8 Ohms. Under closed loop conditions, with 20 dB of global NFB, this is 0.03/0.015% into 4/8 Ohms. Open loop, full power bandwidth into 4 Ohms is 95 kHz, with the input limiting filter (corner frequency a nick above 200 kHz).
Closed loop shows THD of 0.015/0.03%, 20-20,000 Hz, 8/4 Ohms. Control measurements of THD at 100 kHz show 0.6/0.03% into 4/8 Ohms.
Now, for the „Nige Special“. At 1/2W (2.83Vrms) into 8/4 Ohms, 20-20,000 Hz THD is 0.005% and 0.008%, rather expected as the amp is in pure class A mode. No come-back distortion, as in poorly biased amps.
That’s as far as I got so far. Work continues.
This is as far as I got so far. Attached is the latest version of the work in progress.
I cannot remember the changes from the last time, but there are quite a few. Some were made based on advice and guidance of Demian, most notably the input stage, which now uses standard FETs in a cascode configuration.
The VAS has evolved the most. As suggested by Samuel Groner, there are two semiconductors acting as a buffer stage. In case of Otala/Lohstroh amp, this was a classic differential amp, with its own CCS. In my case, these buffers are referenced to the ground rather than the negative power supply line; I tried bith versions and this one worked better for me.
As suggested by Demian, the input FETs are biased at just under 5 mA each; at first, I was falbbergasted and missed a heart attach by a narrow margin, this was unthinkable for me before that time, but I gulped and tried it, and man oh man, does it work, or what! Strangely enough, I found that the buffer FETs are better off at around 3.5 mA each. Since the voltage drop across them is just 3.2V, I can get away using standard 2SK170 FETs, as their 40V rating is meaningless in this spot, while the input stage FETs work at 10V, so their relatively high bias current is not a problem dissipation wise.
The last stage of the VAS works into a current mirror, as it did before, Not much change here, about the only one worth mentioning is the addition of a small value resistor to the base of Q21. Thorsten pointed out that using a transistor as I did was adding a leakage to an otherwise perfect current pump. As ever, I take hi comments very seriously, and tried with the said resistor added, and I did get somewhat better performance figures, a nick, but better.
The rest you know, my classic Locanthi three piece output stage. On your suggestion, I did add a small value resistor and a relatively large (2,200 uF) capacitor to the ground, to stabilize the driver and isolate it from the power devices.
The small value caps (10uF, 100 nF) you see sprinkled around is my own quirk. Experience has taught me that more often than not if the amp goes unstable at some point, it’s very likely due to inadequate bypassing, so I never save money on that, in fact, I’m rather generous in that respect.
The protection circuits for overvoltage/overcurrent have been pushed to the limit and have been worked out according to the manufacturer’s 50 mS performance data. After all, that is the the time delay of the circuit anyway. This allow each output device to pump out about 6.7 Amps in 50 mS peaks, or around 27 Amps for the whole kaboodle. The only person I know who will not be happy with that is Wayne, the rest of us should be just fine.
Speaking of supplies, they were, voltage wise, the ONLY constraint. I decided not to go over the magic 63V capacitor borderline, because that’s when prices start to soar. For the current amp, that is, the input stage and VAS are fed by stabilized +/- 61.5V lines, the current stage by +/- 56.5V (which, according to the manufacturer’s data, rises to just over 58V under the no load conditions – trafo regulation is given as 3% or less).
Which means that under dynamic conditions, the amp should be able to pump out peak power levels of 180/350W into 8/4 Ohms. A rather nice and healthy dynamic margin, I’d say, that’s about 2.2 dB over the nominal.
But remember, this is still work in progress, so for example compensation items (R15C16, C7 especially are as yet unknown). C13 and C14 were initially set at 10 pF, and C17 at 5.1 pF, and no instability was found so far. Thus, it is possible some of the compensation components may not even be needed, but we’ll see, to early to tell yet.
The values measured show promise, and then some. At full nominal blast (28.3V or 100/200W into 8/4 Ohms), under open loop conditions, 20-20,000 Hz THD is less than 0.3% into 4 Ohms and 0.15% into 8 Ohms. Under closed loop conditions, with 20 dB of global NFB, this is 0.03/0.015% into 4/8 Ohms. Open loop, full power bandwidth into 4 Ohms is 95 kHz, with the input limiting filter (corner frequency a nick above 200 kHz).
Closed loop shows THD of 0.015/0.03%, 20-20,000 Hz, 8/4 Ohms. Control measurements of THD at 100 kHz show 0.6/0.03% into 4/8 Ohms.
Now, for the „Nige Special“. At 1/2W (2.83Vrms) into 8/4 Ohms, 20-20,000 Hz THD is 0.005% and 0.008%, rather expected as the amp is in pure class A mode. No come-back distortion, as in poorly biased amps.
That’s as far as I got so far. Work continues.
Last edited:
With video op amps the speed seems to relate to the job they do. No safety factors. My feeling is that's the truth. There may be factors of design that say otherwise. Here is one selected at random.
A long time ago I presented the Cyril Bateman papers. It might have been on this thread. He taught me howto build an oscillator that to this day I never built ( Rosen style ).
The heading of this thread is Sound quality verse measurements. I have always been someone to build a machine that seems to me to have least number of pieces whilst having high performance. If it sounds better or worse than hoped the quest is to know why.
LMH6714MA/NOPB - TEXAS INSTRUMENTS - IC, VIDEO OP-AMP, SINGLE, 400MHZ | Farnell UK
A long time ago I presented the Cyril Bateman papers. It might have been on this thread. He taught me howto build an oscillator that to this day I never built ( Rosen style ).
The heading of this thread is Sound quality verse measurements. I have always been someone to build a machine that seems to me to have least number of pieces whilst having high performance. If it sounds better or worse than hoped the quest is to know why.
LMH6714MA/NOPB - TEXAS INSTRUMENTS - IC, VIDEO OP-AMP, SINGLE, 400MHZ | Farnell UK
I told you as much when I said that this is in many ways a reworked Otala/Lohstroh amp. Not a copy, some things have been done differently, and especially the CCS for the input stage. If there is a better CCS, I don't know of it, but would love to. I estimate its CMRR at about 92+ dB, and in my case, the base line has already been knocked downwards by the regulator filtering and the preceeding (not shown) power line filter, which kills a lot of junk before it even enters the amp.
I never stopped reading up on the subject ever since I got hold of the initial 1973 article. At the time, I didn't understand much of it, but it was clear to me that this was the way to go. I never looked back since.
And I did have help and support from friends, I am pleased to say. Thorsten Loesch never stopped for the last 11+ years, and Demain Martin (1Audio) was more than generous in offering a helping hand, he actually supplied 17 notes on it. Many were taken as is, a few I wonder how I missed seeing that myself when it's so obvious (but much is obvious in hindsight), and some I am still working on.
This forum can REALLY be helpful, there are quite a few very good people here, good as both engineers and people.
BTW, not "Degan", but "Dejan".
Or, if you prefer, my assumed British name of "Danny". Even if US servicemen's kids called me Danny years bfore UK, while in Ankara, Turkey.
Dejan I said I was dyslexic.
Now look at what my name means.
Nigel /ˈnaɪdʒəl/ is an English masculine given name. The name is derived from the Latin Nigellus. This Latin word would seem to derive from the Latin niger, meaning "black"; however this is thought to be an example of an incorrect etymology created by French speaking clerics, who knew Latin as well, to translate the Norman first name Neel in the Latin written documents. Indeed, the Latin word nigellus gave birth to Old French neel (modern nielle), meaning “black enamel” (same word as niello) and it explains the confusion, because the clerics believed it was the same etymology as the first name Neel, spelled the same way
Now look at what my name means.
Nigel /ˈnaɪdʒəl/ is an English masculine given name. The name is derived from the Latin Nigellus. This Latin word would seem to derive from the Latin niger, meaning "black"; however this is thought to be an example of an incorrect etymology created by French speaking clerics, who knew Latin as well, to translate the Norman first name Neel in the Latin written documents. Indeed, the Latin word nigellus gave birth to Old French neel (modern nielle), meaning “black enamel” (same word as niello) and it explains the confusion, because the clerics believed it was the same etymology as the first name Neel, spelled the same way
Dejan (Serbian Cyrillic: Дејан; [dɛjan]) is a South Slavic male name derived from the Old Slavic word dejati (дѣятъ[citation needed]>дејати), meaning "action; to act, to do".[citation needed] It may also be related to the Latin deus, meaning "god".[citation needed] It is very common in Serbia, and also common in the republics of Montenegro, Bosnia and Herzegovina, Macedonia, Croatia, Slovenia and Bulgaria.
The given name Dejan and the surname Dejanović is unusually widespread in the Serbo-Croatian linguistical region, and it is first mentioned in 1349, in Serbia.[1] Dejan (fl. 1346-1356) was a Serbian medieval magnate in the service of Emperor Dušan. The form Dejaniš is also found in the 14th century.[1] According to medieval texts, Dejan was widespread in Serbian lands,[2] and it is frequent in Serbian epic poetry.[3][4] The name is among the most common in Serbia: It was the fifth most given name in the period of 1961–1970 (at 5,61%[5]); first in 1971–1980; ninth in 1981–1990.[6]
The given name Dejan and the surname Dejanović is unusually widespread in the Serbo-Croatian linguistical region, and it is first mentioned in 1349, in Serbia.[1] Dejan (fl. 1346-1356) was a Serbian medieval magnate in the service of Emperor Dušan. The form Dejaniš is also found in the 14th century.[1] According to medieval texts, Dejan was widespread in Serbian lands,[2] and it is frequent in Serbian epic poetry.[3][4] The name is among the most common in Serbia: It was the fifth most given name in the period of 1961–1970 (at 5,61%[5]); first in 1971–1980; ninth in 1981–1990.[6]
- Status
- Not open for further replies.