Current Control vs Voltage Control: A Large Scale Study

[RobL]

Current Drive

ESP Project Pages - High Output Impedance aka current drive. Does it make a difference?

sound-au.com

JSmith

Interesting article. Sounds like he’s not convinced current drive is actually advantageous or practical.

 

[TNT]

Yeah, but.... well.

//

 

[siarez]

Interesting article. Sounds like he’s not convinced current drive is actually advantageous or practical.

I read Rod Elliott’s article a few weeks ago. My takeaway is that it’s primarily about the practical challenges and tradeoffs of implementing high output impedance / mixed-mode feedback (stability, box alignment, damping, crossover, etc.). That’s useful, but none of those challenges are a show stopper. He also doesn’t meaningfully engage with the main claim current-drive advocates focus on: distortion reduction at the driver.

In my dataset which contains a variety of drivers, the distortion benefit is pretty consistent and, to me, compelling.
Of course there are limitations: more SPL points, and 2-tone/multitone tests would strengthen the case further. But even with those caveats, I think the results provide strong evidence that CD is very effective.

Also worth noting: some commercial designs (e.g., Kii) have used forms of current drive for years.

 

[charlielaub]

Current Drive

ESP Project Pages - High Output Impedance aka current drive. Does it make a difference?

sound-au.com

JSmith

Although Rod mentions distortion many times in that article, it seems to me that he is mostly focused on the change in the frequency response that occurs when you change from voltage drive to mixed mode to pure current drive. He never really talks about the harmonic distortion of the driver and probably does not have the equipment to measure that. He is more of an electronics guy IMHO. Seems to know that there are "some claims" out there about "improved sound" but the exact nature of that and how to measure it are maybe beyond his capabilities.

In contrast, the OP is primarily focused on HD and how the feedback based source impedance changes (reduces) the HD for many drivers. He mentioned that he uses FIR to completely flatten the frequency response, so that there is no level dependency between the drivers measured. That is a sohpisticated, loudspeaker-designer type approach to the topic.

 

[siarez]

I thought it would be interesting to see the effect of CD on a finished commercial speaker. So I got my hands on a pair of Monitor Audio Bronze 50 speakers.
I "converted" the speaker by bypassing the passive crossover on the woofer and connecting it directly to my amp in current driver mode.
I EQ'ed the woofer's response (with the on board DSP) to match the passive crossover as closely as possible.
I didn't not touch the tweeter circuit. It is powered by the other channel of my amp which I set to operate in VD.

I made measurements before (i.e. with the original passive xcrossover) and after conversion.
The measurements are taken from the full speaker from 12 inches away at 92dB inline with tweeter.
The measurement are done in a small room, so they not ideal, but we still see significant improvement from CD.

 

[charlielaub]

The result with the Bronze50 woofer shows the typical 10dB or so HD3 reduction under current drive in the midrange band. The HD2 reduction is likely IMO as a result of the feedback around the amp that you are using to generate the current drive.

 

[siarez]

The HD2 reduction is likely IMO as a result of the feedback around the amp that you are using to generate the current drive.

You can see the THD measurements of the amp I used for the measurements above in this post here.
The amp actually has slightly higher HD2 in CD (probably due to a slight imbalance in my feedback circuit). With that said, the amp has much lower HDs than the driver in both in VD and CD mode.

 

[ShadowFiend]

@siarez I wonder about the benefit of current drive on dome driver. Have you ever try current drive with either dome midrange or dome tweeter?

 

[popej]

You can see the THD measurements of the amp I used for the measurements above in this post here.

Results are very good, but I can't translate them to your actual use in presented measurements, where you applied like 20dB power less.

 

[siarez]

@siarez I wonder about the benefit of current drive on dome driver. Have you ever try current drive with either dome midrange or dome tweeter?

I have not tested any dome midrange drivers yet. But I have tested two tweeter, a Vifa/Tymphany XT25TG30 (a ring radiator) and a SB Acoustics SB26STCN (soft dome). Both showed the familiar improvements in harmonic distortion. I'm pretty sure a dome mid-range would also benefit the same way.

 

[MattHooper]

As long as I live, I will never truly understand electricity. All I know is that it lives behind the walls and God help us all if it gets out.

 

[KSTR]

As long as I live, I will never truly understand electricity. All I know is that it lives behind the walls and God help us all if it gets out.

 

[siarez]

Results are very good, but I can't translate them to your actual use in presented measurements, where you applied like 20dB power less.

I have THD+N vs. power results for the amp here that show 1W performance.

The Mic issues and HD2
I did not go into details of this in my first post, but it could explain why some people do not observe HD2 improvements with current drive.
Initially I measured all the drivers with an Earthworks M23. When going through the measurements I realized HD2 hovers around -55dBr for most of them.
This tipped me to doubt the microphone, further tests quickly showed the mic was at fault, and that M23's HD2 was dominant even at sub 90dB levels.
What was confusing was that HD3 and HD5 were captured pretty accurately even at 105dB SPL, but HD2 was extremely level dependent.

So I started searching for other mics. All the small affordable measurement mics that I tested had this flaw, and were unfit for HD2 measurement.
Two small diaphragm mics that performed well were Neumann KM 183 and NTI M2010. Both KM183 and M2010 are in a completely different league compared to the M23 - in this regard. I could not afford the KM183, but I found a used NTI at a good price.

Below you see the M23 and M2010 measurements of the same driver. The sweeps were at 105dB SPL.
We see that HD3/5 both match pretty well between the mics, but HD2 is vastly different.
I have a suspicion that even M2010 may be exaggerating the Purifi's HD2 here.

The only reason I noticed this issue was because I tested a large number of drivers including some low distortion ones like the Purifi in a very controlled setting.
I suspect most people who dabble in current drive don't notice that their microphone could be a problem for HD2 measurements.

 

[KSTR]

It is a sad fact that "measurement" microphones seldom have any meaningful distortion specs and we have to find this out on our own. The best test is to use two drivers each playing a sine at a different frequency and look for the intermodulation products which can only come from the mic. Often, with phantom-powered mics, notably ones with back-electret capsules with integrated FET buffer, the buffer and level shifter circuit is the culprit, not the mic capsule proper.

 

[AnalogSteph]

Often, with phantom-powered mics, notably ones with back-electret capsules with integrated FET buffer, the buffer and level shifter circuit is the culprit, not the mic capsule proper.

I would even be willing to go out on a limb and say "always". Condenser capsules tend to be quite decidedly dominant H3, while an ideal JFET common source amplifier will generate pure H2. It takes considerable degeneration (see Linkwitz mod which converts common source to source follower) or other trickery (see John Conover's technique) to tame the H2 in common electret capsules.

Mics with little space for electronics may also err on the side of less circuit complexity and stick with something single-ended, while this is much less of a concern for classic side address LDCs which regularly ship with "Schoeps-style" circuitry. I would generally expect better linearity from a cheap LDC than a cheap pencil condenser.

Side note, if the M23 hits -55 dB H2 at 105 dB SPL, one would expect 0.5% at 113 dB and 1% at 119 dB. Not terrible, not great. Either the spec of 140 dB "Peak Acoustic Input" means "never subject it to more than that" or "highest level with some semblance of linearity" (neither of which would be considered standard - you generally see a level spec for either 1% or 0.5% THD), or the mic under test may not be meeting spec for whatever reason. (Bad solder joints, tarnished contacts, ...)

PS - The M2010 is specified for 145 dB @ 3% THD. Conservatively assuming that this is still dominant H2 (which it very well may not be), this extrapolates to 0.03% (-70 dBr) at 105 dB. Which in turn would be quite in line with measured results.

 

[bmc0]

Condenser capsules tend to be quite decidedly dominant H3

Which microphones are you referring to? Most measurement/laboratory mics have a single diaphragm and single backplate, so H2 should dominate:

(From the B&K Microphone Handbook, Vol. 1 [BE 1447–12], chapter 2, page 42)

 

[BerndH]

If you are asking about the no-EQ harmonic distortion, below is the HD from an uncompensated sweep. If you compare it with the one from my first post, you can see CD performs worse at low frequencies here and worse than VD, largely because the driver has high impedance near resonance so it is being push much harder there.

Big respect for speaking out!

Best regards
Bernd

 

[siarez]

I got around to do multi-tone testing of the Monitor Audio Bronze 50. (My measurement is similar to Klippel's here.)

I converted one of the speakers using my amp to be fully active. i.e. one channel for the woofer and one channel for the tweeter with DSP crossover/EQ to match the passive crossover.
The other speaker is unmodified.
I developed a Python script for QA403 to make sure the tone amplitudes match between measurements.
The plot below shows the spectrum of the recorded signals from the mic for both speakers.
Blue is the unmodified passive speakers and orange is the bi-amped current driven speaker.
If you look closely you see that the amplitude of the stimulus tones matches perfectly (except for the few tones above 15kHz)

To make the differences easier to see, I wrote some code to remove the stimulus tones (leaving only noise and distortion), and then apply smoothing.
That is the plot below:

We see significant improvement in the midrange where our ears are the most sensitive.