Try the series resistor, if it gives any hints for improved performance and if its something worth to pursue (edited it to the previous post).
Joes current amplifier, or project from I linked on previous post, some others available that are relatively easy to build and try out. Shouldn't be too much hurdle to kind of get feel whats there audio quality wise.
Joes current amplifier, or project from I linked on previous post, some others available that are relatively easy to build and try out. Shouldn't be too much hurdle to kind of get feel whats there audio quality wise.
Last edited:
Hi Lars
[Why did I have a feeling you might show up?]
I realise of course that this was about mass limiting acceleration and bandwidth and that I was quoting a throw-away-line that was added about damping factor being a myth. I stopped using the myth word because strong language like that puts some people off; but still thought it was a myth. I have now gradually gone back to using it again and seeing it on the blog was a good thing.
But this raises something else that I am curious about. Yes we know that mass, acceleration and bandwidth are related, just as the blog describes, plus the myths about lighter cones being faster etc. The Purifi cone is not light mass and as long as sensitivity of the '6.5W08' is up around 88dB (and I have measured it and it is), then I have no issue.
But I note that ring of mass behind around the cone periphery. It is of course hidden from view in use. Whether on purpose or not, that added mass and the reduction in acceleration at the edge of the cone (we know flat response requires flat acceleration), that it has this benefit of reducing the radiating surface area starting near 1KHz and up. You can see that in the frequency response at off axis versus on axis (see below), and at 3KHz how little the response changes over a 60° arc. And they keep staying close together over 30° arc, even as the response falls around 4KHz (also helped by inductance I would imagine) and up. It is something I have not seen in any driver that I have measured, or other drivers for that matter. I can't help thinking that this was deliberate design feature, so I am curious whether it just worked out this way, or was there more to it?
I might add that I use the tweeter to fill in off axis limitations of the midrange driver. So it is something I am very conscious about. Hence I look very careful at the beaming in the upper midrange. Hence also liking waveguide tweeters.
Cheers, Joe
PS: Can we have that blog on damping factor myth coming soon? I suspect it will be Bruno writing it? It should be fascinating.
Hi Teemu
Yeah, absolutely! Do try adding a series resistor. I suggest an 18 Ohm value. I have a reason for mentioning that value and tomorrow will explain why.
But I have another suggestion, if the speakers are 8 Ohm, then put a parallel 8R resistor (it will get hot) and also try that. It will become a 4 Ohm load. The fact that this is a voltage source does not mean we are diluting any current-drive supposed benefits. It's not a current source, so it is something else.
Now it seems that Esa has seen some kind of light. In his book he pooh-poohed tube amplifiers, but seems to have softened his view, that even a moderate amount series impedance can be heard.
Tomorrow I will post something that Menno Vanderveen did at ETF17 held in Denmark and about 30min drive from where my sister lives, so she dropped me off.
Menno did this very curious demonstration. I will tell you about it tomorrow.
Cheers, Joe
PS: Just for the sake of some entertainment, I scanned the back of my well-worn copy of Esa's book (I was among the first to buy it and I still consider it an important book), the language should make you smile:
If there are any issues about posting this, let me know. I hope that it is allowed by the now widely recognised fair use doctrine.
I also definitely recommend buying the book!
I'll stop you there with an example, say you drive a speaker each from a Voltage and a current amp. You use EQ before one amp to make the response damping the same. What makes this work?
Like all concepts, negative drive impedance is a concept with many restraints and trade-offs involved.
The main point is, the static VC impedance is not simply resistive and stable, the notion of Re is a gross simplification. Actually, we have a complex impedance Ze as there is a semi-inductive component usually named Le and neither the Re nor the "inductive" component is stable. Re varies with temperature, and the non-resistive part varies with voice coil position, Le(x), and voice coil current, Le(i). So we have a very nonlinear and non-stable voltage-to-current transfer impedance and when we subtract a constant resistance from it to make it more effective, the overall nonlinearity increases, in relative terms.
On top of that, the VC, both as actuator and as sensor is not linear as well, as both depend on the BL product (the force factor) and BL varies again with position, BL(x) and current, BL(i). That is even true when a second coil is used for the sensor like with a dual voice coil woofer.
The varying BL enters the equation twice, whereas with current drive it enters the scenario only once, via F = BL * i. But there is an interesting compensation that can sometimes be had around for restricted frequency range. Assume BL has dropped because of BL(x), leading to less than required velocity. This also leads to less microphonic voltage generated which in turn leads to a higher compensation current than when the sensor were linear. When all things fall in place this can result in lower distortion over a larger excursion range, notably compared to pure current drive.
There are other schemes that try to extract the cone velocity with bridge circuits and those can even be made to track and compensate varying VC resistance but still suffer from the dynamically varying impedance. The only way for solid motional feedback is to use a completely independent sensor, not affected by anything happening in the main motor. In some special cases using the VC as sensor does work well, one example are high power large compression drivers used in a HiFi context where negative drive impedance at and below resonance reduces distortion significantly (below resonance a driver's motion is spring-dominated and when there is enough feedback even below resonance then a lot of distortion from the progressive spring stiffness can be reduced).
The point is that even seemingly small distortion like 1% (measured steady-state) can be already enough to introduce dynamic DC offset shift of the VC. When the stimulus ends (or changes), the dynamic DC offset changes and it does that move to the new offset position with the natural damping Q.
This can be measured actually, and set into comparison to schemes with higher damping (like standard voltage drive), the idea is to subtract out the linear part of the signal (which can be arbitrary, including music signal) and look at the residual. With pure current drive, you will see a lot of ringing at the resonance even when the overall residual may (or may not) be lower and cleaner. The same driver, with the same frequency response dialed in but under voltage drive has a different residual pattern with much less of the resonance frequency left. Overall, the results are often shocking for both cases, even at low signal levels. Speakers have ton's of distortions and some only show up with dynamic tests, like the nasty hysteresis distortion.
Exactly!
Current drive only addresses a few of many the error mechanisms in speakers (in particular, Le(x) and Le(i), and power compression) and only when those mechanism dominate then current drive will result in better performance, when restricted to the frequency range where it doesn't have negative side effects, that is.
Which boils down to something I wrote here, many years ago: For a specific driver in a specific acoustic situation and use case, you can experimentally(!) find a source impedance vs. frequency profile that gives the best performance compromise overall. If you are lucky, standard voltage drive is already that profile, or any other simple fixed output resistance (including infinity and negative values). More often than not, you will need some frequency dependence, typically sufficiently low impedance around resonance and as high an impedance you can manage to implement everywhere else.
Fun fact here: With AMT tweeters, current drive actually tends to prevent burn-out. When the pleatings are warping enough so that adjacent sections touch this partially shorts out the driver, reducing Re. With current drive the power delivered and to be dissipated is reduced (P = I^2 * Re) whereas with voltage drive power is increased (P = U^2 / Re).
Yes, and this is what I did in my main experiments when directly comparing current drive and voltage drive.
For a valid comparison, frequency response must be the same in both cases. Therefore I measured the impulse response of the driver's terminal current under standard voltage drive (at a reasonable operating point wrt level) and used that IR for EQing the current drive amp (with a DSP convolver) which established the exact same frequency response down to 0.1dB precision.
And, no surprise, to first and second order of precision both variants then sounded identical. The difference was in the "fine print" character and notably in the large signal behavior where things tend to diverge quickly.
I ask again - what are the benefits of using current drive on magnetostatic drivers? (An AMT is not a magnetostat)
I did a distortion test using an 18 ohm resistor in series with the bass panel of a Magnepan 3.6, which is 4 ohms resistive. When measured at the same SPL, there was absolutely no difference in either frequency response or distortion between voltage drive and 18 ohms series resistance.
If a current sense resistor is sensibly sized based on its temperature coefficient, it could be matched to the speaker to reduce or even eliminate the possibility of thermal runaway. So, if the speaker gets hot, so does the sense resistor and the power consumption doesn't spiral.
~~~
Unless maybe you accept that it's just a convention? Your model could just as easily be a current source with a parallel resistance.
~~~
Unless maybe you accept that it's just a convention? Your model could just as easily be a current source with a parallel resistance.
Which falls under what I wrote earlier: Experimentally find out what works and makes a difference and what does not.
I have no experience with magnetostats and have no idea whether or not they could improve with current drive as perhaps not even power compression is an issue here.
I'm coming back to this as this seems to be the root cause why Joe's miracle concept of "(adjustable) damping even with current drive" stirred up the pot so much.
I, as everyone else here it seems, thought he was referring to the driver being actually operated under current drive where there is no damping as we all agree. His initial secrecy about how he is realizing this, together with claims like "something completely new and yes undiscovered") only fueled things and admittedly triggered my "bull$hit detector", sorry @Joe Rasmussen about this.
But now we know he actually simply degenerated the current output amplifier with a paralleled impedance-shaping network. Seen as a black box from the driver's perspective (as that is what counts) this isn't current drive anymore, it's commonly referred to as mixed-impedance drive and could as well be established by shaping the feedback network of the amplifier circuit rather than using bulky large passives at the output -- both ways have been done for years if not decades as it is, in the end, pretty much trivial, no matter if we look at voltage-output or current-output "base amps" enriched with passives (in the XO or elsewhere) to shape impedance. For example, using a series inductor as the last element in front of a woofer or midrange or a series capacitor in front of a tweeter does the same thing -- shape impedance --, besides reducing level it also increases source impedance as seen by the driver, helping its distortion at the transition into the stop band. Purify's improved series notch to tame cone breakup rings the same bells, shape source impedance in a beneficial way.
A lot of buzz without need, generated just by some IMHO sloppy wording. Again, sorry for that from my side.
Last edited:
I think the core of it can be stated in just a few words.
The damping of a driver depends on the total source resistance its voice coil (as a generator) sees, and that includes the driver's own Re and any additional series resistance in XO coils and such. So we have always several Ohms here.
Therefore it doesn't matter at all whether the amp -- plus wiring -- has an output resistance of, say, 0.1 Ohms (a 4Ohms-DF of 40) or 0.0004 Ohms (DF of 10000), it is always a completely negligible part of the large total resistance available for cone motion damping. Very high output impedance does change speaker frequency response according to its impedance curve, of course, and if anything, that change is what is actually perceived rather than any alleged change in actual damping.
Guys let me ask you some questions. They seem to be related to the topic (which I have been reading very diligently for some time, although it is difficult for me to understand what you write, but it is a great reading!))
1) I found that my compression driver (RCF 940, 650-18k) sounds (much) better in the following sequence:
bigger cap (30uf) + parallel resistor 8ohm + EQ (best) > smaller cap (10uf) + parallel resistor 8ohm + EQ (medium) > L-pads, no differencе if [8-4 ohm + 5uf + EQ] or [4-8 ohm + 5uf + EQ] (worst).
It turns out that the closer to the direct control of the voltage amplifier driver (Lyngdorf, d-class), the better it sounds.
What I mean by better is a clearer, more open, dynamic sound, clearer highs (3-4k - 8-10k). I think that I can distinguish this in a blind test (it is unrealistic to conduct it, too many switches are required).
All responses were equalized to a flat frequency response. I didn't find a significant difference in distortion measurements (but my equipment isn't good for distortion measurements).
After I discovered this, I read a lot of comments that compression drivers with l-pad sounds worse than with autoformer or direct amp connections.
But shouldn't the series resistor in this situation (or the capacitor's resistance, if it's running high enough) be closer to controlling the current and give less distortion?
Is this something typical for compression drivers or will each driver/amplifier combo have different results?
2) If we talk about using an autoformer instead of an l-pad. How does it act with the driver?
It converts part of the voltage into current, as far as I understand, what does this change, is it a current or voltage driver?
What to expect, it can become even better with autoformers, or vice versa worse?
Sorry if the questions are not directly related to the topic.
1) I found that my compression driver (RCF 940, 650-18k) sounds (much) better in the following sequence:
bigger cap (30uf) + parallel resistor 8ohm + EQ (best) > smaller cap (10uf) + parallel resistor 8ohm + EQ (medium) > L-pads, no differencе if [8-4 ohm + 5uf + EQ] or [4-8 ohm + 5uf + EQ] (worst).
It turns out that the closer to the direct control of the voltage amplifier driver (Lyngdorf, d-class), the better it sounds.
What I mean by better is a clearer, more open, dynamic sound, clearer highs (3-4k - 8-10k). I think that I can distinguish this in a blind test (it is unrealistic to conduct it, too many switches are required).
All responses were equalized to a flat frequency response. I didn't find a significant difference in distortion measurements (but my equipment isn't good for distortion measurements).
After I discovered this, I read a lot of comments that compression drivers with l-pad sounds worse than with autoformer or direct amp connections.
But shouldn't the series resistor in this situation (or the capacitor's resistance, if it's running high enough) be closer to controlling the current and give less distortion?
Is this something typical for compression drivers or will each driver/amplifier combo have different results?
2) If we talk about using an autoformer instead of an l-pad. How does it act with the driver?
It converts part of the voltage into current, as far as I understand, what does this change, is it a current or voltage driver?
What to expect, it can become even better with autoformers, or vice versa worse?
Sorry if the questions are not directly related to the topic.
Horns seem very susceptible to strong diffraction effects, which can give tall narrow ripples in the frequency response, depending on how they're designed. It's not strictly non-linear distortion, but it can make for a very unpleasant ear-bleeding sound for some people.
One idea I came across was to vary the output resistance, so that 2 competing sets of ripples cancel out. In one set, the cone forms a 'node' that strongly reflects energy. In the 2nd set, the 'node' is moved behind the cone, so the resonances occur at a different set of frequencies. I'm not sure if this can be explored directly in Hornresp, but the driver T/S parameters can be altered.
One idea I came across was to vary the output resistance, so that 2 competing sets of ripples cancel out. In one set, the cone forms a 'node' that strongly reflects energy. In the 2nd set, the 'node' is moved behind the cone, so the resonances occur at a different set of frequencies. I'm not sure if this can be explored directly in Hornresp, but the driver T/S parameters can be altered.
In my case, I don't hear any unpleasant sound from RCF CD/horn combo (listeners also did not note similar effects). The horn certainly generates some diffractive and reflected sounds, but at least I can't complain about their audibility. To me, this combo sounds great (and at least not worse than popular similar waveguides like xt/xr1464)
The behavior at impedance resonances does change when a resistor or capacitor is used, characteristic peaks appear. After equalization, they are also inaudible. With a parallel resistor, their manifestation is almost not noticeable.
Therefore, I cannot explain to myself the audible difference with and without an l-pad, I do not observe any obvious signs in the frequency range or in distortion (perhaps I cannot measure distortion well). L-pad takes away some percent of pleasure