When your highly trained and heightened perception allows you to perceive the sonic difference between an amp biased using a vbe multiplier and a diode string, why not put it to other good uses like reading others thoughts, "seeing" through their words to pull out that deeper meaning.
Admit it Scott, you were sneering at the time.
Steve,
I`m sorry if you took my criticism in an unintended manner. Reducing switching and crossover distortion in class B is a tricky business. Numerous questions arise, for instance, I cannot see how it could be done without any load current control (proportionality between output current-input voltage). Until a more promising design is realized, I`d remove that ugly circuit (in sugar-coated wording) with collector-input and diodes and stick to pure diamond buffer.
I`m sorry if you took my criticism in an unintended manner. Reducing switching and crossover distortion in class B is a tricky business. Numerous questions arise, for instance, I cannot see how it could be done without any load current control (proportionality between output current-input voltage). Until a more promising design is realized, I`d remove that ugly circuit (in sugar-coated wording) with collector-input and diodes and stick to pure diamond buffer.
Well that sounds reasonable. And I will make the effort to keep this post completely “unambiguous” and “clear”.
As Andy pointed out above, renaming Rbias makes absolutely no difference – LTspice ain’t broke.
As for the behaviour of this circuit, I’m afraid it was adequately outlined way back in post 7 on the very first page of this needlessly protracted thread. Any “non switching” behaviour demonstrated under certain operating conditions is entirely due to the lack of a biasing resistor for the (consequently current starved) driver transistors and has absolutely nothing to do with the Krill biasing circuit.
It has been explained why the biasing circuit can’t work as claimed over and over and over and over again but ignored. I even made it quite clear in my (also ignored) post #79 how you can check for this fact for yourself. On this I do not particularly care to waste any more time.
As for the recent claims I have replicated your simulation results from posts 138 and 140 and attached the results below. I have not only replicated your “non-switching” results with the Krill biasing scheme, but also with the same output stage biased with an ideal voltage source instead - reiterating the point I made way back in post 8 in response to post 7.
The first picture is a screen shot of this simulation. The green trace shows the ballast resistor current for the complete Krill output stage, the red trace shows the ballast resistor current for the output stage with the ideal voltage source bias. Both stages have the same quiescent current.
As can be seen, both output stages behave virtually identically. However, current this low doesn’t matter much anyway – a uA or two is really nothing as far as a power output stage goes and this can be well and truly considered “off” – such a small current won’t do anything to mitigate any output stage non-linearity. The attached file entitled Krill_1pair.asc is the LTspice simulation file.
What you have done here is to exacerbate the cross-conduction issue by quadrupling (four pairs of output devices) the load capacitance presented to the current-starved driver stage. Since the stored charge is now four times as great the output devices in fact take so long to turn off that, even at 20Hz, the current never has a chance to drop below several mA.
This is nothing extraordinary and has absolutely nothing to do with the (claimed) operation of the Krill biasing circuit, but is entirely due to the lack of a biasing resistor for the driver transistors.
Want proof? The second picture is a screen shot of the simulation for the Krill output stage with four parallel transistor pairs, run in tandem with the same output stage biased with an ideal voltage source. As before, the quiescent current is identical in both cases. The green trace shows a ballast resistor current for the complete Krill circuit and the red trace shows a ballast resistor current for the output stage with ideal voltage source bias.
The attached file entitled Krill_4pair.asc is the LTspice simulation file.
As before, the behaviour in terms of minimum current is virtually identical and also as before, the ideally biased output stage actually performs slightly better in this regard – the current plateaus at a higher minimum for the non-Krill version.
In conclusion the “minimum current” behaviour shown in posts 138 and 140 is in fact just cross-conduction ultimately caused by a current starved driver stage.
The Krill biasing circuit does not operate as claimed or any differently from a standard biasing circuit as far as output device switching is concerned (for reasons already explained ad nauseam).
Factual statement: If you have any regard for technical proficiency you might want to reconsider the hosting of that write up for the Krill biasing circuit operation concocted by Hugh Dean – it is just plain wrong. Sorry, but that’s just the reality.
And anticipating the response that this post is likely to receive due to the dismal past history of the discussion, I’d like to point out semi-humorously in closing that feigned moral outrage, hand waiving, ridiculous parallels drawn with the persecution of Galileo or the inventiveness of Nicola Tesla, projection of ones personal technical illiteracy with insinuations that the Krill biasing circuit is inherently mysterious in operation or beyond human comprehension or can’t be properly simulated, or even an evasion in the form of an accusatory reply consisting of a dozen or so out-of-context quotes in random order, legalese fashion, won’t cut it as a legitimate technical response.
Attachments
I haven't claimed it is better, just different. I'm afraid I can't say much more than that without straying off into what it sounds like. That is subjective, and can not be proven (or disproved) over the internet.
No, I don't. Neither does Douglas Self, unless he has changed his mind in his latest book. May I ask why you think it is low? What would you consider a minimum value here?
For the rest of you, I did the simulation on the output stage only circuit that I posted earlier. I added a biasing resistor of 220 ohm to that schematic and ran the simulation again, at 20Hz. The minimum current flowing in any emitter resistor is essentially unchanged from the values without the resistor added. I hope none of you are heartbroken by this revelation.
If we don't feed it, maybe it will go away.
Avvv... Steve, you want me to start quoting the Krill thread
?0.1ohm requires tight matching and also requires a larger HS to avoid thermal issues. 0.15-0.22ohm is much easier to work with.
Where is the data? This fits your standard pattern whenever you make a claim. That pattern is one of the following three situations:
1) Provide a simulation file, but with no simulation results, such as a graph, corresponding to that specific simulation file.
2) Provide a graph of the simulation results, but without the corresponding simulation file.
3) Provide neither, just a claim.
So we have case 3 above here.
Anyway, I did a simulation of exactly the situation you just described, stepping the driver emitter resistor from 220 Ohms to 1e12 Ohms. That's done with the statement:
.step param Rdrv list 220 1e12
The graph is attached below, along with the exact simulation file used to produce it. Anyone interested can run the results and see for themselves. It clearly shows the absence of driver emitter resistor is what causes the soft switching.
Attachments
Avvv... Steve, you want me to start quoting the Krill thread
0.1ohm requires tight matching and also requires a larger HS to avoid thermal issues. 0.15-0.22ohm is much easier to work with.
You'r saying I did claim it is better? I don't remember doing so, other than how I felt it sounded. That is my opinion, and should be taken as only that. I admit it is very possible I said some things I don't remember, or remember incorrectly, from my early days here. Drugs will do that.
I agree with you on all counts concerning emitter resistors.
For some reason I can't open your file. Can you post the schematic you used?
Concerning your points 1, 2 and 3, are you suggesting I have been lying about my results? I stated quite clearly which schematic I was using. You provided the simulation file. I added the duplicate parts and changed the resistor values to match the schematic I was using in my sim program, and that I had posted here. I think I have been very open about what I was doing and how I was doing it.
What is the procedure for posting my simulation file so "anyone interested can run the results and see for themselves" from LTspice?
To me, "sounds better" is the Godwin Law for Audio. I can't debate past this type of arguments.
Syn08,
I can see your argument here, but the clear implication is that measurement of distortion is the ultimate assessment of audio quality. Sadly, it is not, because people's tastes do vary, like food. After all, tubes distort pretty badly, but many like them.
At some point, there needs to be agreement that both measurements and subjective listening are merely opposite points on the spectrum. Both must be addressed - somehow - but the population do have their preferences, however confused, and the market is thus highly varied. To me, and many I suggest, total reliance on distortion figures is a false dawn and eventually people will always say, 'how does it sound'.
It is exasperating, but the correlation between THD20 and Godwinesque 'good sound' is very tricky. Have you any suggestions about accommodating 'good sound'?
Andy, don't be absurd, no one suggests you are lying, but people engaged in strong debate often present the most salient facts - to them - without considering others which might also be salient but contradictory. Why are you so sensitive? And what is the point?
Hugh
I can see your argument here, but the clear implication is that measurement of distortion is the ultimate assessment of audio quality. Sadly, it is not, because people's tastes do vary, like food. After all, tubes distort pretty badly, but many like them.
At some point, there needs to be agreement that both measurements and subjective listening are merely opposite points on the spectrum. Both must be addressed - somehow - but the population do have their preferences, however confused, and the market is thus highly varied. To me, and many I suggest, total reliance on distortion figures is a false dawn and eventually people will always say, 'how does it sound'.
It is exasperating, but the correlation between THD20 and Godwinesque 'good sound' is very tricky. Have you any suggestions about accommodating 'good sound'?
Andy, don't be absurd, no one suggests you are lying, but people engaged in strong debate often present the most salient facts - to them - without considering others which might also be salient but contradictory. Why are you so sensitive? And what is the point?
Hugh
I just downloaded it from my post to my hard drive just fine. And it opened and simulated just fine. Not sure what's going on there. I'm not going to allow myself to get into a situation where there's a possibility that what you simulate and what I simulate are different, because you will play that for all it's worth. That's what you've been doing in this entire thread.
Yes.
Beg to differ.
Simple. Just do what everyone else who has been posting LTspice simulations has done for years. Post a graph in GIF or PNG format of the simulation results, along with a ZIP file of the LTspice simulation itself, preferably with the necessary model files included. I keep my LTspice installation unchanged from the standard download (with no models added to the standard installation), and post all the models that aren't in a standard installation in a separate file. Then, in the simulation file, I use a statement such as:
.include KSA_MJLmodels.txt
Don't put words in my mouth, Hugh. All I'm saying is there's no objective difference between the Krill output stage and a classic triple EF.I can't accept "sounds better" without a "because..." And the burden to fill in the blanks is on those making the claim.
About people preferences and biases, I can affort not to care about. Not so sure about those running an audio related business.
About people preferences and biases, I can affort not to care about. Not so sure about those running an audio related business.
That is precisely why I have avoided discussions of "how does it sound". I will say however, if i didn't think it sound good, I would have gone with something else. And yes I know I will be biased in favor of something I designed and built.
Steve, I'm the trader I believe....
Whoever would dare put words in your mouth, Syn08? It seems you have transcended reality, and are now into purist, hyper reality!
What about talking about the various harmonic artefacts in detail? These can be extracted from many simulators. Do you think a monotonic decreasing harmonic spectrum is appropriate?
Hugh
Whoever would dare put words in your mouth, Syn08? It seems you have transcended reality, and are now into purist, hyper reality!
What about talking about the various harmonic artefacts in detail? These can be extracted from many simulators. Do you think a monotonic decreasing harmonic spectrum is appropriate?
Hugh
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.