0 Members and 1 Guest are viewing this topic. Read 2718 times.
Also, I would hardly choose a dht tube to produce the most "real life like sounds".CheersSteve
And I would hardly choose an idht beam/pentode tube for said purpose. Linearity is your friend... well a better one anyway, than gobs of feedback.
First, the intermodulation distortion (IMD) of any triode is approximately 2.8 times that of total harmonic distortion (THD). So at 5% THD, the IMD would be approximately 14.8% for each tube. Not surprising that SET amplifiers don't work well with many instruments/voices. At one watt output, there is virtually no difference between the tubes.
Secondly, we need to consider the entire amp design. Both tube types require a gain stage to boost the input signal voltage to drive the output tubes. Gain stage tubes as well as the output DHT and IDHT tubes produce HD as well as IMD.
Next, we need to investigate the amount of drive signal required for the 300b VS the 6550/KT88 etc. The 300b requires ~ twice as much signal drive voltage as the 6550. This means the gain stage of the 300b amp will produce much more THD and IMD than the gainstage of the 6550.
Next? When we have two stages of amplification, the gain stage and output tubes, the 2nd order HD combines with the 2nd order HD to produce 4th order HD, the 2nd and 3rd order HD also combine to produce 6th order, and the 3rd and 3rd order HD produce 9th order harmonics. Of course IMD is terrible because it mixes all the musical and non-musical signal harmonics together. IMD is not part of the musical presentation. So the 300b amplifier will produce at least 2nd order, 3rd order, 4th order, 6th order, and 9th orders of harmonic distortion as well as IMD. With 3 stages, there will be even higher orders.
However, the gain stage of the 6550 amp can be designed for virtually zero distortion. What does this mean? The result is that only the output tube is producing HD. So the 6550 amplifier will only have 2nd and 3rd orders of distortion, and much less IMD due to fewer harmonic orders present.
It is not surprising why SETs do not do well with multiple instruments, such as in a large band or orchestra.
Now that an audio myth has been busted, we can go back to the OP, that ICs do "sound" different.
With an analog interconnect system, it takes lots & lots of capacitance to roll off the high frequencies (that's mFd not pFd).The output stage will be unhappy with such a large capacitive load.Nope, 10 or 20 pFd doesn't change anything.
Don't you have any curiosity? When something sounds different that shouldn't, don't you wonder what went wrong?Nothing wrong with sounding like Hillary.
My comment was about the linearity of the amplifying device; I specifically did not mention SET amplifiers, but I’ll play along. SET’s, along with SEP’s are a subset of amplifiers with their own unique characteristics. Let’s stay on SE amps for a moment. This topology produces mostly 2nd order HD which in turn produces most of the IMD. It’s in the topology and not in the inherent nature of the DHT tube that gives rise to this characteristic. Pentodes in a SE config would give substantially more THD and IMD without the use of feedback. It seems a bit disingenuous to compare a DHT implemented as a SE amp to an IDHT implemented in a PP stage were the main culprit of IMD, namely the 2nd order HD component is canceled. Agreed
A typical drive signal for a 300B is around 50V RMS, while the 6550 working at a typical OP as found in the datasheets (either in UL or pent mode), requires an input of about 30V RMS. Low distortion drivers such as the 6SN7 and 6922 exhibit a modest rise in HD between these two values.
In addition, a good design will further mitigate this difference with the use of CCS’s or plate chokes as loads.
This flattens the load line, increases gain and reduces distortion quite substantially. Case in point, some time ago I sim’ed a 6922 using a CCS as a plate load and the rise in HD (which was mostly comprised of 2nd order harmonics) went from 0.6% to 0.8% between 30 and 50 Volts RMS. Hardly a big jump.
Wrong. In a two stage amp the stages are in anti-phase. This means that the 2nd order harmonic from the driver will be amplified, inverted and partially cancel the 2nd harmonic component of the final tube.
Competent SE designs use this ‘trick’ and try to choose operating points where at least the 2nd and 3rd HD components of the driver are proportional to those produced by the final tube. That’s where the term distortion cancelation in SE amps came from. A big advantage of said DHT’s is that they produce very little upper harmonic HD artifacts, unlike the IDHT’s. So again, the mechanism as you described here giving rise to IMD, is more prevalent in IDH tubes, where feedback saves the day.
As I showed above, the gain stage of a 6550, all things being equal, will at best be a marginal improvement over the one designed for a 300B.
This statement holds true for the gain stages viewed in isolation, because when looked at in the context of an amp, the 6550 has a big disadvantage. At typical operating points the 6550 has an in-circuit gain of around 11-12 while the 300B’s in-circuit gain is 3.5. That’s a big difference and it means that all the HD components from the driver will be amplified 12 times vs. 3.5 times for the DHT.
So as an example, even if we have a driver stage for the IDHT at half the distortion as compared to the DHT's driver, the final HD figure would be substantially to the disadvantage of the former. A 0.5% HD from the driver multiplied by 12 is 6% THD. A 1% HD from the driver of the 300B gives 3.5% THD. This fact alone tells us which tube places bigger constraints on the driver stage and it ain't the 300B!
Albeit a subjective opinion, there’s quite a large number of people that will disagree with you on this statement.
Of course these people will most likely be using very high efficiency speakers (over 100dB/W) where the SE amp is taxed very little and the distortion is kept low.
I’ll leave it to the readers regarding the ‘busting of myths’, but the fact remains that these 80+ year-old tubes are the most linear amplification devices in existence. As with anything, proper design and implementation falls on the competence of the designer.
What do you base your opinion on, please be specific? So what did go wrong in your opinion?????CheersSteve
from Linkwitz a long time ago:Let's put a little realism into high frequency loss due to long interconnects.I just measured 2 nF capacitance for a 5 meter RCA connector cable that I use with Pluto. That is 400pF/meter rather than 100pF/meter or less which is typical for interconnects.If the output impedance from the pre-amp is 1kohm (high) and the input impedance of the power amp is 10kohm, then the effective impedance is 10k//1k or 910ohm.With the 2nF cable this gives a lowpass 3dB cutoff frequency of Fc = 1/(2piRC) = 87 kHz, which is still far above the needed frequency range.
With a typical capacitance cable the cutoff would be at 350 kHz. If the passive pre-amp has 10kohm output resistance Fc would still be 70kHz.
But Linkwitz was working with 2000pF and your cables were under 40 pf that's more than a small difference.Lets not even get into that microsecond time delay thing. Those two superconductor guys got it all messed up.That may be the only mistake that J Neutron ever made.
Just to give a clearer idea of how formal science and the (incredibly rigorous) scientific process is conducted, I thought I would explain what went into publishing the two above mentioned papers that have apparently generated controversy among lay readers (interestingly there has been no controversy whatsoever in all the professional circles, which include audiologists, otolaryngologists, acousticians, engineers, and physicists ). An experiment has to be carefully thought out and then submitted as a proposal to an Institutional Review Board (IRB) and approved by them before it can even begin. Then optimum equipment, methods, and a multitude of cross checks must be developed (my papers give some details to help appreciate what went in). It takes about half a year to conduct each sequence of controlled blind tests. Consent forms (legally approved and certified by the IRB) must be signed. The results, analysis, and conclusions are then carefully considered and discussed with colleagues who are experts in their related inter-disciplinary fields; for this I went in person to various universities and research institutes and met with people in departments of physics, engineering, psychology, neuroscience, music, communications sciences, physiology, and materials science. After that the results and conclusions were presented at conferences of the Acoustical Society of America (ASA), Association of Research in Otolaryngology (ARO), and American Physical Society (APS). Seminars were also made at numerous universities and research/industrial institutions (please see the list on my web site). After each presentation, the audience is free to tear apart the conclusions and ask all possible questions. Eminent people such as presidents of the above mentioned societies and corporations were present at my presentations and engaged in the discussions. After passing through this grueling oral presentation process, written manuscripts were then submitted to journals. There, anonymous referees are free to attack the submission in any way they want. More than a dozen referees and editors have been involved in this journal refereeing process. Only after everyone is satisfied with the accuracy of the results and all statements made in the manuscript, are the papers published in the journals. The entire process took around 5 years from initial concept to refereed publications. (Note that an article in a conference proceeding does not go through the rigorous refereeing process of a formal journal. Essentially anything submitted there gets accepted for publication. Contents of books are also not rigorously refereed. When possible, reference should always be made to an original journal article.)I would like to add some other observations: (1) One should be wary of drawing conclusions based on an “intuitive feelings” or because something “makes sense”. This has its role in adding plausibility to the understanding but can sometimes be contrary to fact. Thus qualitative statements based on survival and evolution cannot lead to a quantitative estimate of temporal resolution. One has to gain a detailed understanding of the physiology of the ear followed by all the neural processing steps in the ascending pathways of the brain. This knowledge can take years to acquire. (I give some references below for further reading.) On the other hand something that cannot be understood or explained (at the moment) isn’t necessarily false. It can be dangerous to dismiss claims just because they don’t make sense. Science should deal with properly authenticated facts. (2) Listening tests can be notoriously unreliable unless properly designed. This is why the proposal and consent forms for tests on human subjects have to approved by the IRB, otherwise no journal will consider an article for publication. The tests have not only to be blind but also must be absent of extraneous cues (such as the switching transients discussed in my papers). I would therefore be wary of informal listening tests conducted at home – these can be useful in helping you decide which component works better in your system but not rigorous enough to establish a scientific fact. (3) There is an erroneous statement in one of the forum posts “Such temporal resolution depends on the "coincidence detector" circuitry of the medial superior olive … mostly effective below 3kHz.” Actually the bipolar cells in the MSO (medial superior olive) encode relative delays between right and left ears which are used in azimuthal localization (left-right location determination). This has nothing whatsoever to do with the monaural temporal resolution being discussed. Coincidences between different frequencies arriving at each ear are encoded by octopus cells (which act like synchronous AND gates with a huge number of inputs) located in the PVCN (postero-ventral cochlear nucleus). This slew-rate information from the octopus cells then feeds bushy cells in the VNLLv (ventral subdivision of the ventral nucleus of the lateral leminiscus) which contributes to elements of pattern recognition.I hope this clarifies the meaning of temporal resolution in the context of sound reproduction systems. For further insight into psychoacoustics and the neurophysiology of hearing, I can recommend the following books: (1) “The psychology of hearing” by Brian Moore(2) “Integrative functions in the mammalian auditory pathways” by Oertel, Fay, and Popper(3) “Neuroscience” by Purves et al. I have personally met with and discussed my results with the authors of the first two books. All of these books are used as texts at universities. The last one is used in introductory neuroscience courses and is relatively easy to read.Sincerely,Milind Kunchur*********************************************************************************Milind N. Kunchur, Ph.D.Professor of Physics
Secondly harmonic structure is mostly determined by the load on the tube, and the tube itself. Any triode can be manipulated to produce mostly 2nd harmonic distortion by how it is loaded.
Third, ALL tubes, including triodes, produce 3rd harmonic distortion and higher, check the graph below. True, with PP 2nd harmonics are cancelled, depending upon the balance etc. However, that does not eliminate the intermodulation problem caused by the gain stage and the 3rd and higher harmonics. Again see the graph below.
Big difference. P-P, 50vrms equates to 141 volts peak to peak, 70.5 volts peak. 30vrms equates to 84 volts peak to peak, or 42 volts peak. The distortion rises rapidly with increased signal amplitude. The JJ E88cc produces much less distortion than any 6SN7 type tube. In a preamplifier, it is ~1/9th, or ~-19db lower, in a typical common cathode circuit. Even a regular 6dj8 can be designed for less distortion. By the way, using a computer program is not going to get you where you want to go.
A plate choke is the last part one should use in a design. The impedance varies with frequency, the load line becomes elliptical, distortion increases. The gain is not flat. To keep phase shift between stages minimal, the frequency response (FR) needs to be at least 1/10th the lowest frequency reproduced, -3db. An inductor is not going to make it if one wishes an accurate/natural copy of the input signal.
Simulation is not a good method, I actually measure. See above for other problems with inductors and the load line. It certainly is not flat through out the frequency range needed.
Secondly, using a poor circuit will give a very poor distortion measurement, as your distortion figures corroborates what I stated in my previous post; two stages of distortion with the 300b. I use a gain stage that produces less than 0,05% HD at 50 volts peak, or 100 volts peak to peak. That is being conservative as I list my entire amp's total THD at .05% at 1 watt.
Who told you that? We have problems wsith that statement. The 1st problem being higher orders are produced. Secondly, each tube will have wildly varying harmonics, and their individual amplitudes vary, as shown, from Eimac. Besides that, the distortion of the 1st stage is usually quite low vs the output stage, so cancellation is minimal. If someone increases the distortion of the 1st stage in an attempt to cancel 2nd order distortion, then higher order harmonics with greater amplitude will also result. I think someone(s) has been feeding you a line. Now check out this graph from Eimac.See how the harmonic distortion amplitude constantly changes with signal level. Notice the graph is for 3rd, 5th, and 7th harmonics of a perfect tube. Most, including DIYers, have been duped. See above.
Those 300bs and 6SN7 types must be out of this world if they don't produce any distortion products higher than 2nd order.
As we have seen, that statement is completely false, except for those manufacturers who do not understand how to properly design an amplifier. That is completely, 100% false. See the above Eimac graph. The only thing one will obtain is sporadic, limited cancellation at best. Gain does not factor into the equation at all, none. The real problem are the IMD products created, because they are dissident, not part of the music. Rid the first stage of HD and the amp output will have much less IMD.
As mentioned earlier, the gain of each tube has no bearing whatsoever over the distortion figure; it is the actual distortion figures of each stage that count. If the gain of the 1at stage is .1%, and the 2nd stage gain is 10, that 0,1% distortion figure does not rise by 10 times to 1%. You see, the signal level also rose by 10 times, the distortion remains at 0,1%.
Output distortion of two stages.The harmonics of two stages multiple, 2nd harmonic of the 1st stage and 2nd harmonic of the 2nd stage will produce 4th harmonic. 2nd and 3rd will produce 6th order harmonic, 3rd and 3rd will produce 9th order harmonic.5th and 2nd will produce 10th order harmonic. 5th and 3rd will produce 15th order harmonic.By the way, I already use a gain stage that has at least 1/12 the THD (-21.5db) of your CCS gain stage example. Using the correct design is all important in reproducing accurate/natural music. I have tested my amps for sonic accuracy with actual sophisticated listening tests, so I know. Using a computer program will not help you much.
Just because a "number of people" do not understand principles of audio designing does not mean others do not understand.
Actual setup listening tests to identify how accurate an amp or preamplifier is does not lend itself to being subjective. If their opinions are subjective, so be it. If you wish to make this a subjective vs subjective, it would seem to me that your initial post has become naked.
So what spl can be obtained and still be clean, with 1 to 3 watts audio output? Need a subwoofer? How much compressed air distortion in the horn throat. Afterall, compressed air is not linear. Lot's of concerns when it comes to high efficiency speakers.
Again, my amp has less than .05% THD at 1 watt, not just the first stage.Sorry, but the JJs E88cc are some 19 db less THD, at least 1/9th the distortion, of any 6SN7 type etc. The output tubes I use also have very low distortion. I ought to know, I have been testing/electronics and tubes for some 55 years.
I know mine are near perfect, if not perfect, and I am using KT-88s.
I so much as said this, by my mention of using constant current sources to flatten the load line and reduce distortion to a minimum.
The sim was just that, an investigation in trying to gage ball-park performance of different drivers in different configs. Never built or measured said circuit. The 6922 is almost identical to the 6dj8 but there’s no accounting for the computer model’s accuracy. This I am fully aware of.
The plate choke was mentioned along with CCS loads as a means to a flat load line and its resultant benefits. I grant you that cascoded BJT/MOSFET devices and gyrators might give better results and they’re cheaper to implement too. As far as choke phase shift goes, it can be a bit of a problem, but mainly when global feedback is used; otherwise not much of an issue. You surely realize that coupling and cathode bypass caps of which there many in most amps, all introduce phase rotations and that their effects are additive. Right?
Not much of an argument here. As I answered this above, simulation is for preliminary investigation of a design, while measurements are essential in validating the design.
Again, it wasn’t a ‘design’, it was a single stage set-up (with a resistive load) and it stopped there. Kudos to you (and I mean this sincerely) for designing a driver stage with such a low HD figure. I hope though, that the irony is not lost on you, of the fact that such a good driver doesn’t really matter much in an amp that uses global feedback. Because the fact is that even a driver with 10 times higher distortion, will be swamped and have its non-linearities corrected by said feedback.
I was careful in saying that the second and maybe the third harmonics can be reduced with this cancelation method and I also never claimed total cancellation, just reduction – big difference. Higher order products will, as you say, be created - never said anything to the contrary here either. But I fail to see your point here, simply because I don’t buy into your argument of ‘substantially increased’ driver distortion in the case of the DHT. Based on what I’ve seen from tests of drivers at both 50V and 75V peak one way swings, the term ‘substantial increase’, does not hold true.
Both power tube types will produce these higher harmonics in combination with the driver as per the Eimac text you cited – this I never denied.
However, and I’ll keep repeating this, the DHT has the advantage of less higher order harmonics created within its own gain-stage.That’s actually true for all distortion components –from the 2nd to however high you want to go.
Having said all this, the main burden that the driver of a DHT faces (and which surprisingly you haven’t mentioned), is the driving of the triode’s shunt capacitance. The charging and discharging of the Miller capacitance is the cause behind most of the mediocre sounding DHT amps out there, rather than the exaggerated claims of ‘greatly’ increased distortion due to the driver. Overcoming C-Miller requires both a low impedance driver AND high currents. An e88cc running at 5-6mA will not cut it, even though its output impedance might be relatively low and its distortion even lower.
Now, before you come back to me citing calculations for slewing, that indicate slewing to not be an issue going past 100 KHz with even a 12AX7 as the driver, I'll say yes, but the subjective results are really not in line with the theory here.
You are overstating the contribution of the higher order harmonics in the creation of IMD. IMD is mostly created by the low order products that are usually much larger in amplitude as compared to the higher order ones. In properly designed amps/stages, the higher components are much below the low order components (those HD products from say the 2nd to the 5th harmonic). These higher harmonics are in the order of -40 to -60 dB or even less, below the 2nd and 3rd components. Their contribution to IMD is simply too small. That is a fact. That’s not to say that they are unimportant; quite the contrary. Higher order HD products are highly dissonant to the human ear and their elimination should be a top priority. But I’ll repeat, not because of their capacity to create IMD in any meaningful way in a competently designed amp.
I’ll use your own words: that is completely, 100% false. You are conveniently ignoring the anti-phase term with grave consequences in your ‘reasoning’. Would inverse phase be a better term?
If the cancelation of two opposite in phase signals is bogus, then you’re basically negating the concept of inverse a.k.a neg feedback. You do realize that both Schade feedback and the anode follower use this exact mechanism, right? That’s local feedback from plate to grid using that very same inversion inherent to the operation of the common cathode stage. Are those schemes a figment of my imagination too? Don’t think so.
I will not disagree in that listening should be the final arbiter.
This has nothing to do with the understanding or non-understanding of any principles. The logical conclusion to your above statement, would then be that non-technical audiophiles cannot discern good sound from bad. Subjective opinion on sound is based on the sensory and mental perception of the individual and definitely not predicated upon the subject's understanding of any “design principles”. This should be self-evident.
You are walking a fine line here IMHO, with your emphasis on THD numbers and your attempt to link this to perceived sound quality. If so, what would this say about your amp when compared to a solid state design that has .0005% THD, a number that’s two orders of magnitude better than your amp? How would you answer someone that told you that XYZ SS amp sounded better than your amp because it has 100 times less distortion than your design? I would reconsider the above statement if I were you.
Don’t know where this last sentence came from. Seems kind of arbitrary, as I never so much as tried to make this into a subjective vs. objective sort of argument. In addition, if you think that actual tests take the subjectivity out of the equation, then with all due respect, this is a delusional stance. This is a whole other topic which I don’t care to discuss at this point, so we’ll just have to agree to disagree on this.
Again, speakers are another subject onto themselves and so I’ll just briefly touch on this topic. When we talk about high efficiency speakers, it’s mostly about horns (although there are notable exceptions) and with sensitivities above 100dB/W it’s almost exclusively the domain of horns. I’ll say this: no matter what the faults of horns may be, they have substantially lower IMD distortion than direct radiators.
Don’t have to spare my feelings as far as the 6SN7 is concerned. I never claimed the 6SN7 to be superior to the JJ E88CC although I’ve seen it being superior to other 6922/6DJ8 tubes. I only claimed that these two tube types were low in distortion. I can’t vouch for this since I never personally performed any meaningful tests on the 6SN7, but Lynn Olson who has, claims that this tube produces exceedingly low upper harmonic products; something to be mindful of as I’ve stated prior.
I have no problem with this statement – it’s your perception and I take it at face value. What I objected to originally was your statement implying that DHT’S are sub-optimal for a world class design, or something to that effect. Mind you, I am not wedded to any particular DHT - I currently don’t have an amp that uses one, but nevertheless, I don’t like subjective opinions being stated as facts.
While I think that it was good to split the posts about tube circuit design from an interconnect thread.All of my posts were about interconnects, I have no interest in tube circuit design.
This is an excellent discussion worthy of consideration and can be debated until the cows come home. However, in the final analysis, it really makes no difference to the end user of an amplifier.In my nearly 60 years of working with tube audio I have found that it is always the listener's ears which make the final determination of an amp design. If I had a nickel for every time a person actually preferred more than less distortion I could have retired 40 years ago. Since my work has always been custom (for local audio enthusiasts), and having access to their systems, I've had the luxury of involving them in the design process. Imagine going nuts trying to get the distortion figures of an amp to be exceedingly low only to be told that they preferred the sound when the distortion was higher! And, obviously, it's always about the synergy between the amp and a given pair of speakers. One only needs to look at the very popular Darling 1626 amp by Danielak as an example. In spite of its high distortion, it is one of the best loved SET designs out there. Regarding using fb with SEP designs, I have never liked or used global fb. I've found that much better results can be obtained by using some voltage fb at the output tube(s) only. Just my personal bias (pun intended) on this issue.
