I really wondered how FM can achieved +/-2deg deviation from linear phase, and at the same time 36dB/octave cutoff slope. It mentioned Gaussian slope, does anyone have more information about it?phase_accurate said:
Maybe I mis-interpreted the meaning of "+/-2deg phase accuracy" in the FM newletter
It probably helps for the kind of crossovers I favor, the slopes are 2nd-order and the Q is 0.5 or less, thus synchronously-tuned 1st-orders in cascade would accomplish the goal. On the other hand, for the bass module, parametric EQ for room correction would be very desirable, and now we're talking about discrete or integrated-circuit op-amps. Then again, in the 50 ~ 500 Hz frequency range, the demands on the opamp are pretty modest, and they all have a lot of gain down there.
Hmm, in practical terms a professional parametric EQ + crossover would be fine for the 50 ~ 500 Hz frequency range, and a custom-designed buffer with synchronously-tuned HP filters would be appropriate for the ultra-high-quality HP path. This could be implemented with discrete transistors or a buffer-only variant of the Aikido. Maybe it's not as complicated as I've made it out to be - at least for my speaker. (Note that I'll be using passive crossovers for the mid/high crossover, partly because the power amplification levels are lower, and partly because the crossover parts aren't that large.)
Now if the mids or highs are horns with multichannel amplification, now we need more complex equalization, and the JE-990 or an Aikido-with-feedback might be necessary. Notch filters in particular are going to require feedback, unless you want to go nuts with custom-wound LC filters (which is pretty crazy territory, considering the inductances required).
Hmm, in practical terms a professional parametric EQ + crossover would be fine for the 50 ~ 500 Hz frequency range, and a custom-designed buffer with synchronously-tuned HP filters would be appropriate for the ultra-high-quality HP path. This could be implemented with discrete transistors or a buffer-only variant of the Aikido. Maybe it's not as complicated as I've made it out to be - at least for my speaker. (Note that I'll be using passive crossovers for the mid/high crossover, partly because the power amplification levels are lower, and partly because the crossover parts aren't that large.)
Now if the mids or highs are horns with multichannel amplification, now we need more complex equalization, and the JE-990 or an Aikido-with-feedback might be necessary. Notch filters in particular are going to require feedback, unless you want to go nuts with custom-wound LC filters (which is pretty crazy territory, considering the inductances required).
panomaniac said:
Huh. The Kaneda looks like a collection of unity-gain buffers, simple bipolar emitter-followers with FET current sources. Just cascoding the emitter-followers with another transistor on top of them would drop the distortion another ten times (I've measured this, the distortion reduction is real). Another more subtle benefit of a cascoded emitter-follower is the base capacitance drops by the beta of the top transistor, typically a 100x reduction. This substantially drops HF distortion, and results in a video-speed buffer.
The same cascoding method can be applied to the cathode-follower circuit of the Aikido, and I expect would give the same ratio of distortion reduction and HF extension. The biggest hassle is the requirement for multiple elevated heater supplies so the heater/cathode voltage rating is not exceeded. (This is typically +/- 100V.)
I dunno, this is where I have to gently disagree with Mr. Broskie. I see two identical tubes stacked on top of each other, two identical cathode resistors, no cathode-bypass caps, and the grid of the top tube is connected directly to the plate of the lower tube. That looks like a classical SRPP to me. A series-connected balanced circuit, though, is not really the same as true (parallel) balanced circuit, particularly when the load changes.
The gotcha with all SRPP's is the extreme sensitivity to loads, which dramatically changes the harmonic distortion spectra. John Broskie solves this by buffering the SRPP load, and at the same time injects a precisely-controlled portion of B+ noise into the grids at just the right locations, thus nulling out almost all PS noise. If the input and output tubes are the same, audio-frequency current flows through the entire circuit should be fully complementary, further reducing distortion.
The gotcha with all SRPP's is the extreme sensitivity to loads, which dramatically changes the harmonic distortion spectra. John Broskie solves this by buffering the SRPP load, and at the same time injects a precisely-controlled portion of B+ noise into the grids at just the right locations, thus nulling out almost all PS noise. If the input and output tubes are the same, audio-frequency current flows through the entire circuit should be fully complementary, further reducing distortion.
I am skeptical that any active solution would be as transparent as a passive line-level XO for the high pass. A passive second-order is pretty easy to do. I'm using the Marchand XM46SB passive XO (2nd-order 80 Hz HP), with a Marchand XM46 active (4th-order 80 Hz LP). I replaced the paralleled caps in the XM46SB with a single Auricap.Lynn Olson said:
http://www.marchandelec.com/xm46.html (bottom of page).
If you want cascaded 1st-orders, there's this:
http://www.t-linespeakers.org/tech/filters/passiveHLxo.html
Though I respect FM for their serious engineering and build quality I assume that they exaggerate a bit.
I don't think that it is a true phase-linear crossover because these can't be done with just a pair of steep high- and lowpass filters in the analog domain.
Lynn:
Since you prefer 2nd order filters of quite low Q I dare to ask why you don't simply design your amp with this response ? Or am I missing something ?
Regards
Charles
Since my amps are all-transformer-coupled this is actually quite awkward. Transformers require a reasonably low source impedance, and behave strangely with a parafeed (capacitive) source impedance. If my amps were conventional RC-coupled vacuum-tube or DC-coupled transistor amps it would be trivial to use a small coupling cap and matching R and that would be all there is to it.
As it is, though, with the Amity and Karna having input transformers, unless I want to completely re-design the input section (perhaps a Mullard circuit with a transformer output section), the simplest route is a separate linestage with the highpass function built-in. An Aikido, for example, with an RC in between the input and output section, which would then set the highpass function.
Hmm - even that has issues, since the output cap has to be quite large in order to avoid subsonic peaking with the input transformerof of the amplifier. Replacing the output cap with a transformer on an Aikido circuit is not a trivial project, since the cathode of the follower is sitting at 150V, and avoiding DC going through the transformer requires the primary also be at 150V and the other side of the primary must then voltage-track the 150V, as well as exhibiting a low impedance at audio frequencies.
As it is, though, with the Amity and Karna having input transformers, unless I want to completely re-design the input section (perhaps a Mullard circuit with a transformer output section), the simplest route is a separate linestage with the highpass function built-in. An Aikido, for example, with an RC in between the input and output section, which would then set the highpass function.
Hmm - even that has issues, since the output cap has to be quite large in order to avoid subsonic peaking with the input transformerof of the amplifier. Replacing the output cap with a transformer on an Aikido circuit is not a trivial project, since the cathode of the follower is sitting at 150V, and avoiding DC going through the transformer requires the primary also be at 150V and the other side of the primary must then voltage-track the 150V, as well as exhibiting a low impedance at audio frequencies.
Lynn Olson said:
Hi Lynn
Something has been bugging me for years re SRPP, especially as you describe here. I do see that the PP part is push-pull AC wise as the cathode resistor of the top element creates a similar AC signal that is fed to the grid of same. Effectively two stages, not one, again, AC wise. Broskie, from memory, has covered that.
But I see something else. Assuming zero grid current, then the top half is biased as a current source and so it the bottom. Am I the only one seeing that? Now, DC wise, what is the result? Two current sources in series? Being fed from the same B+ I can see it hunting for some kind of equilibrium and if you cannot get DC stability, then how can you get AC stability? What's more, to me it sounds that way. SRPP are vague, maybe tubelike (not in the best sense of the word) but not tight. Taking the AC from the top cathode doesn't seem to change that much, even if supposed to lower Z.
A certain mutual acquaintance suggested to me a differential SRPP (current sourced, long pair). I seem to recall measuring output Z well over 1K. I didn't like it all that much, rebuilt it as a simple triode gain stage, then used the second available triode as a cathode follower (triode gain stage now seeing very high Z) , added a SS (fet) current source, bootstrapped another fet follower on top (you know the idea I'm sure), the output Z went down to 100 Ohm.
Sound? One sounded lose and the other tight and dynamic. Made my report and hence since Allen refers to SRPP as Miss Piggy. Not sure what that means? All bluster and no substance? Or love affair gone awry? Premature... ?
Joe R.
After reading about Rod Elliot's experiment with THS6012 (http://sound.westhost.com/highspeed.htm) I can't help thinking that an active xo using a that (or a simliar line driver) would be an interresting exercise.
ion said:
I am hoping that this thread may lead to a non-opamp solution, and I include discrete (non-IC) opamps as well. Read: no negative f/b. No disrespect to Rod - we occupy the same city - he is a conservative in audiophile matters. His website is not without value and his efforts are to be applauded.
Joe R.
Hi Lynn
I have had a chance to use / play with / listen extensively to several “speaker controllers” as a result of my work. More recently, I had been using a BSS366 omnidrive and Behringer at home but have a new favorite. I am using a Xilica XD4080 now, which is a very nice unit, much nicer and better built than the Behringer, cheaper, and better sounding than an omni drive AND it also has linear phase filters for those who want to utilize them.
Anyway, it is very nice and doesn’t seem to have an effect or sound of its own (unlike the 366 especially and the “B” to a lesser degree).
Don’t forget that bumping your signal up to pro levels is part of getting the most from all balanced pro gear.
Best,
Tom Danley
DanleySoundLabs.com
I have had a chance to use / play with / listen extensively to several “speaker controllers” as a result of my work. More recently, I had been using a BSS366 omnidrive and Behringer at home but have a new favorite. I am using a Xilica XD4080 now, which is a very nice unit, much nicer and better built than the Behringer, cheaper, and better sounding than an omni drive AND it also has linear phase filters for those who want to utilize them.
Anyway, it is very nice and doesn’t seem to have an effect or sound of its own (unlike the 366 especially and the “B” to a lesser degree).
Don’t forget that bumping your signal up to pro levels is part of getting the most from all balanced pro gear.
Best,
Tom Danley
DanleySoundLabs.com
Joe Rasmussen said:
What about the back link in a Sallen - Key circuit - if we make a unity gain buffer without global feedback (any buffer has, I think by definition, at least local feedback) would you then be happy?
If you entirely disallow multiple paths of the sort needed for Sallen - Key, you either need inductors (and crazy big ones) or restrict yourself to filters that are realisable as a product of first order sections - in other words all poles are strictly real. If you allow Sallen-Key type circuits, you can then get imaginary parts to the poles.
I have a buffer circuit to offer (never built it yet, but simulates very well), which has no global feedback, but is pretty linear, good drive and good PSRR.
Simulated a buffer that doesn't use NFB and which minimises DC-offset (it will still not be zero) by the use of two subsequent complementary emitter-followers (Q1 and Q6).
Both followers are loaded by current-sources and their Vce is kept constant (as commanded by Lynn).
Got k2 of -100 dB into 600 Ohms at 1 V peak signal level. Maybe someone else will get better resuslts with more modern transistors and better dimensioning.
Thought this could be of use to someone. The disadvantage is that it needs almost as many transistors as a discrete op-amp.
Regards
Charles
Edit: forgot to mention that the load is R5
Both followers are loaded by current-sources and their Vce is kept constant (as commanded by Lynn).
Got k2 of -100 dB into 600 Ohms at 1 V peak signal level. Maybe someone else will get better resuslts with more modern transistors and better dimensioning.
Thought this could be of use to someone. The disadvantage is that it needs almost as many transistors as a discrete op-amp.
Regards
Charles
Edit: forgot to mention that the load is R5
Attachments
Just checking I am reading this correctly - Q2-Q3 form a "ring of two" current source for the main load. Q4 and Q5 ditto for the cascode bias circuit.
It looks to me like Q1 is running at very low Vce - about 0.7V (essentially VBe of Q6). What about putting a resistor (or maybe a diode) in the emitter of Q6, so that Q1 has enough voltage on it to get properly turned on, and to bring Cob down a bit?
It looks to me like Q1 is running at very low Vce - about 0.7V (essentially VBe of Q6). What about putting a resistor (or maybe a diode) in the emitter of Q6, so that Q1 has enough voltage on it to get properly turned on, and to bring Cob down a bit?
A question for Lynn, or other people who use or plan to use tube amps in the treble.
I assume that removing the bass content makes life much easier for the transformers. Is there any mileage in re-optimising around lower inductances, and maybe smaller cores, so that the leakage inductances and stray capacitances are reduced, improving HF performance? In other words, is the optimal transformer design for say 20Hz - 20kHz different to the best design for say 600Hz and up?
Clearly, given existing amps, one won't make these changes gratuitously, but is it a potential place to get some extra performance?
I assume that removing the bass content makes life much easier for the transformers. Is there any mileage in re-optimising around lower inductances, and maybe smaller cores, so that the leakage inductances and stray capacitances are reduced, improving HF performance? In other words, is the optimal transformer design for say 20Hz - 20kHz different to the best design for say 600Hz and up?
Clearly, given existing amps, one won't make these changes gratuitously, but is it a potential place to get some extra performance?
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.