I think it's great that you are trying to bring something different to market. With the 80s tube revival in the UK it took me while to realise the nearly everyone in UK was recycling the same circuit. Only Tim deParavcini was doing something different, and his EAR509 is still in production.
Jan, Don't have the paper, but you can download Rosenstark's book for free at his website: https://web.njit.edu/~rosensta/books/Feed_Back_Amps.pdf
Chapter 4 is on loop-gain measurement.
Starting on page 61 of the pdf, he summarizes his and Middlebrook's papers on closed loop methods.
Chapter 4 is on loop-gain measurement.
Starting on page 61 of the pdf, he summarizes his and Middlebrook's papers on closed loop methods.
Bruno does similar things with nested feedbacks in class D amps. But still, single ended tube amps have one unbeatable property: distortions always go down, and only down, when sounds decay. And it is what our perceptions expects from sounds of Mother Nature. That's why I went by this road...
The issue with SMPS is typically they are outsourced for cost, built too small with the heat generating switching devices mounted too closely to the input /output electrolytics that are cheap no name brands that dry up and bulge with the heat of the switching device and ripple current, and the glue used to prevent transport movement of the big caps combined with cheap board material means it's unserviceable and the device it's powering gets scrapped because of the penny pinching mentality in sourcing the power supply based on watts and cubic inches by accountants, rather than engineers. ( open a Roku table radio for a text book example of this).
Then there's the EMI filtering, all claiming to meet eN, FCC, VDE spec, that poor asssembly in cheap labour makerts leaves long untrimed leads radiating from the substituted lower cost or missing EMI components. Properly designed low noise, reliable switch mode powers supplies are possible, you have to be careful to examine a product dependent on them as much for the power supply design and execution as the audio circuit itself. Try asking for a screw driver to open your next considered purchase at Best Buy or local audio emporium.
Then there's the EMI filtering, all claiming to meet eN, FCC, VDE spec, that poor asssembly in cheap labour makerts leaves long untrimed leads radiating from the substituted lower cost or missing EMI components. Properly designed low noise, reliable switch mode powers supplies are possible, you have to be careful to examine a product dependent on them as much for the power supply design and execution as the audio circuit itself. Try asking for a screw driver to open your next considered purchase at Best Buy or local audio emporium.
Last edited:
Anatoliy, show me a set of trustworthy measurements of your SE tube amplifier. Sorry, an oscilloscope screenshot is not the one. After all those years of debates here with proponents defending their design approach mostly verbally, measurements are the only think that I count on in these virtual-world debates.
My truthworthy measurements, Pavel, are listening tests, and simple measurements of voltages, currents, and waveforms, that confirm that the things go exactly as I expected. I did plenty of measurements back in 1970'Th when designed guitar effect and musical synthesizers. I designed then also solid state power amplifiers, and feel by my gut what causes what kind of distortions, how they are reflected on waveforms, and how they sound. I can vouch for distortions that I am concerned about, but I do not play that THD games I played half a century ago. They are pointless.
This should not be a SS vs. valve argument. Nelson Pass has published plenty on SS amps with similar vanishing with output level distortion properties. Personally I think they should ban the typical THD + noise plots that hide all amplifier's low level distortion properties.
If your amp has a gain resistor to ground at invert input, you can bypass it with very large C value and the amp will go to full open loop gain. Use an excellent bipolar cap, perhaps. But a value so large as to have no affect on the OL phase response. DC offset/gain is unaffected.
THx-RNMarsh
Last edited:
I would not buy any amplifier or other electronic product without some basic test data on it..... then I would be interested in listening to it. The tests just verify it is a well designed product and has the adequate performance for my wants and needs.
A computer sound card or similar product can do most of the basic tests and more. Not expensive.
THx-RNMarsh
Last edited:
I agree. Otherwise, it looks like in a clinic, in order to get proper average temperature, some patients have to be killed, because others have some fever.
Mr. Wavebourn, to quote a classic DIYAudio monster ego, "leave your ego at the door" before further reading.
I don't think it's the same thing. Mr. Putzey (sp?) uses high order compensation (N poles, N-1 zeroes, I think N=5) which indeed can greatly increase the amount of loop gain, resulting in lower distortions. The magic for such a high order design is not really the compensation network (which can be rather easily analyzed even without simulators, see the classic example of Cherry's NDFL for N=3) but the extra circuitry required to tame the resulting clipping behavior. When clipping occurs, the feedback loop(s) are effectively broken and hence the amplifier can severely misbehave, from ugly recovery to breaking into oscillations. The only systematic solution to this difficult problem, that I am aware of, is to detect clipping and add circuitry that kills the open loop gain in a controlled matter. Such circuity is, by it's very nature, yet another feedback loop that needs compensation, etc... The result of such an approach is usually a huge complexity, and if Mr. Putzey somehow made it to keep it within manageable limits (I haven't seen a schematic) that's a huge achievement, much more than using N=5.
Regarding tubes, I am ready to admit that tubes have nice clipping behavior and distortion profiles, both which are essentially due to their rather low transconductance (resulting in overall low open loop gains) and parabolic characteristics (rather than exponential as in bipolar devices) so the above facts still apply, solid state or hollow state. However, there's nothing extra that tubes can do compared to FETs, except contributing to the global warming.
But be it tubes or SS, thinking that splitting whatever amount of available loop gain over multiple feedback loops brings any major advantages is simply an illusion. It was long time ago mathematically proved that the end result of multiple feedback loops cannot be better than that of single global feedback loop, in fact it is, in 99% cases, worse (in terms of stability margins and linearizing effect). That's a fact, and anybody debating this is only further populating the already overcrowded audio mythology field.
The problem with this is that actual OLG can change a lot when you break the gain relationship between input and output signal voiltages. It also doesn't account for the effect of the feedback resistor shunt capacitor.
I am aware of such opinion. It is based on an analysis that does not include transfer curves of stages being linearized. Rough example I can give you right away, linearising crossover distortions versus saturation distortions.
I do not want to repeat Bruno who you probably misunderstood, we recently discussed his article on feedbacks. You can find it here. Yes, problems of stability have to be dealt with, but the point is not in instability that is just a by-product of applied feedback around stages with phase shifts. The point is, to make the resulting amplifier driving speakers to generate less audible distortions when it mis-behaves. And certain topologies, with certain nested loops, can be optimized for this major criterion.
Sorry, that's not an "opinion" it's a well established fact. For the rest, I would say the first paragraph is techno-babble, but it would be against the forum rules
.The second paragraph: "to generate less audible distortions when it mis-behaves. And certain topologies, with certain nested loops, can be optimized for this major criterion". Without a clear reasoning and a metric for this statement, it's as blank as any other unsubstantiated argument. And no, ears don't count if you want to have any technical consistent discussion. If you want to keep it at the "sounds good" level, then who am I to discuss your taste.
Admittedly, nothing drives me nuts more than otherwise qualified engineers, trying to persuade others about a "fresh" way to achieve audio Nirvana, that collides either with the first principles or with the last 100 years body of knowledge in EE.
Where is Mr. Putzey feedback discussion, regarding his feedback approach in D class?
Last edited:
You need to read Middlebrook. At frequencies where oscillations are a problem the input loads the output and visa-versa. Shorting the feedback guarantees you get the wrong answer.
I understand the thinking of nested loops and control circuits and such but from what I know you still need to close the loop at some point for the total loop controls to work. Now in electronic circuits I understand the basic concept and it seems to work the same way. Once you close the loop how is there a real difference between a simple NFB and one divided into subsections, can you truly get better linearization with multiple small loops or is the end result the same?
That's the whole point, there isn't. And BTW, control/feedback theory doesn't make any assumptions about the type and size of nonlinearities in the plant, so talking "analysis that does not include transfer curves of stages being linearized" is nothing but rubbish.
- Status
- Not open for further replies.