A First Look at Current Source Amp Performance

In an ongoing effort to investigate, understand, and then produce a tutorial on current source amplifiers, I've done a first round of Bakoon HPA-21 measurements. One of the tricky things about investigating this type of amplifier is that there seem to be a variety of interpretations of this type of amplifier technology. Evidently, not all current source amps are equal.

For example, the Apogee Groove is claimed to have a circuit design that prevents changes to headphone frequency response with changes in its impedance vs. frequency response. I'll be doing some measurements of the Groove sometime soon.

Anyway, to establish a base line, and to make sure the model in my head had some grounding in reality, I felt the need to grab a little real data.

Single Tone Response
Headphone101_FirstCurrentSourceMeasurements_Tek_ToneResponseA pure current source amp (sometimes called a transconductance amplifier) takes a voltage signal in, and outputs a current signal.

Let's say we have a 1 Volt RMS 500Hz tone incoming voltage signal and set the volume control for 1 Amp RMS output. If we have a 1 Ohm load resistor, we will measure 1Vrms across the resistor. If we switch to a 2 Ohm resistor, the amp will output the same 1Arms current signal and we'll get 2Vrms across the resistor.

If an amp behaves this way, we can likely say that it's a pure current source amplifier. I hooked up the Bakoon so that I could place varying resistive loads on the amp to see if it exhibited this behavior.

In this case, I had a 500mVrms 500Hz tone going into the Bakoon, and set it up for a 1mArms output into a 150 Ohm load, which will give a 150mVrms across the load. I then put various other resistive loads to see if the output current signal remained constant.

You can see from the images at right that I got a fairly constant 1mVrms/1 Ohm response with all loads. It's off just a bit with the 600 Ohm and 16 Ohm loads. Probably a resistor tolerance issue with the 600 Ohm load, and cable resistance issues with the 16 Ohm load.

It is close enough though to say the Bakoon is acting as a pure current source amp.

Effects on Frequency Response with Varying Headphone Impedance
Some headphones have dramatically changing impedance with frequency. For example, the Sennheiser HD 800 has about 630 Ohms impedance at 100Hz, but has about 340 Ohms impedance at 2kHz.


The Sennheiser HD 800 has a wildly changing impedance response (purple plot) with frequency, varying from 320 Ohms to 630 Ohms.

In the plot above we can see the wildly swinging impedance of the HD 800. As we saw from the single tone experiment above, as we change the load impedance, we change the voltage across the load. This means that the HD 800 will get about twice the drive voltage at 100Hz than it gets at 2kHz. With a pure current source amplifier, the tonal balance of the headphones will be modified by the impedance curve of the headphones.


In the above plot, the blue traces are the HD 800 being driven by a traditional voltage output amplifier. The red traces are the HD 800 being driven by a current source amplifier. At 100Hz, the HD 800 is almost double the impedance relative to its impedance at 2kHz. It should therefor have about double the voltage drive from a current source amplifier. We can see in the above plot that the HD 800 is about 6dB (double in amplitude) at 100Hz relative to when being driven by a typical voltage amplifier. Science!!!

The HD 800 also has a rising impedance after about 7kHz. In the above plot we can also see, if we look closely, that the red plots start to increase above the blue plots above 7kHz.

The big impedance swings of the HD 800 is a bit of a rarity these days; most headphones are lower impedance and have less severe swings. Lets look at the V-Moda M-80 as a more typical headphone.


The M-80 hoovers around 33 Ohms impedance, and while it varies, it's no where near the HD 800. Let's take a look at the frequency response plots.


As you can see, there's a little difference in the bass due to the primary driver resonance hump centered at 40Hz, but not much difference otherwise.

Planar magnetic headphones have a virtually flat impedance response; I performed this test with the Mr. Speakers' Ether. You could just imagine a flat line, but why not show it.


This means that there should be very little difference between the Ether's frequency response with voltage and current output amplifiers. Here's the FR plot.


While the two response curves are nearly identical, they do seem to be somewhat different below 40Hz. I thought this might just be a difference between the output coupling of the two amps, but it doesn't appear in the other headphone plots. Not sure what's going on here. Feel free to brainstorm your ideas in the comments.

Bottom line: these plots show clearly that current source amplifiers will effect the frequency response of headphones that have significant impedance swings with frequency.

Transient Response
As you can see, current source amplifiers have the potential of not working well for headphones with varying impedance response. One area of performance where they may excel is transient speed. We're going to look at response when driving the leading edge of a square wave. The thing to get a grip on to understand what's going on is the relationship between voltage and current at the output when driving an inductive load.

When you drive current through a wire, a magnetic field forms around the wire. An inductor is a coil of wire—like the voice coil of a headphone. When current flows through and inductor, the magnetic field around each winding adds to the whole forming a larger magnet.


Here's the important thing to know: It's the current flowing through the coil that makes the magnetic field, not the voltage across the inductor. When, at the leading edge of a square wave, the amplifier output rises in voltage almost instantaneously, but the current going through the coil does not rise immediately. This is because it take some time for the current to build up the magnetic field around the coil. So, the current lags behind the voltage when driving an inductor. While the voltage output from the amp may be very fast and responsive, the speed of the current output will be dependent on the reactive impedance of the coil and the output impedance of the amplifier.

In a current source amplifier, the output current will follow the incoming voltage transient exactly and will force the current through the coil immediately. But in so doing, the output of a current amplifier will have a very high voltage spike as it deals with the initially very high reactive impedance of the voice coil.


In the diagram above, you can see that in a voltage drive amp the current (green) lags behind the desired voltage waveform (red). With a current drive amp, the output current (and therefor motive force of the voice coil) will achieve it's desired level immediately—but the voltage will have a significant spike on it.

In an effort to verify this characteristic, I set up a current and voltage amp (Bakoon HPA-21 and Simaudio Moon NEO 230HAD) each driving one channel of a pair of headphones. I then set up an oscilloscope to look at the voltage of the two types of outputs simultaneously to see if it looked like the red voltage traces in the above illustration.


The o'scope display above shows the voltage signal from the two amps driving the Mr. Speakers Ether. The top trace is the current source amp, the bottom is the voltage amp. Please note the timescale of 1uSec/division—while the voltage rise time might look slow, it's actually quite fast.

You can see that the voltage step with the Moon is a simple voltage rise. But the step from the current amplifier (top) wiggles around quite a bit. I can't measure the current waveform, but it probably looks quite like the voltage waveform on the bottom trace. The reason the transient shape of the voltage on the Bakoon wiggles so much is because the voice coil, and its movement through the permanent magnet of the driver, is reactive and is "fighting" against the current being pushed through it. I can assure you that with a voltage amplifier, if you were to look at the current output from the device you would see similar wiggling about.

When driving a reactive load, you can control the voltage, or you can control the current, but in either case the uncontrolled parameter will do some wiggling about.

Here's the V-Moda M80.


You can see in this case that the V-Moda voltage waveform from the Bakoon has larger swings than the Ether. This is probably because the long zig-zag conductor traces on the diaphragm on the Ether has less inductive reactance than the tightly wound voice coil of the M-80.


And here's the HD 800...I'm not exactly sure what's going on here except to say it might be much more inductively reactive than even the V-Moda, but I'm not sure.

Regardless, the big question is whether or not the current source amp is able to product faster transient response at the ear. So I set up a similar test, but this time looking at the acoustic waveform from the headphones. Top traces are the Bakoon, bottom are the Moon.


Here's the Ether. Remember this is the left and right channels being driven simultaneously, so there are some slight differences due to being mounted on separate ears. You can see that while there are some small differences between channels, there is no obvious difference in the speed of the leading edge, or the amount of control of the signal. Both traces are fairly similar.


The V-Moda delivers some difference in response. The changing angle of the trailing edge is more about frequency response than speed. The rise of the leading edge is about the same. The Bakoon has markedly larger swings after the initial spike than the Moon. My guess is this might be related to the better damping factor of the low output impedance Moon compared with the very high output impedance of the Bakoon.


The HD 800 also shows no remarkable difference in the transient speed. While the shape of the wave form is slightly different due to the frequency response differences imposed on the HD 800 by the current source amp, the magnitude and shape of the wiggles is roughly the same.

First, it does seem that the Bakoon is a pure current source amplifier. It's going to be interesting to compare this traditional transconductance amp with the Apogee Groove, which includes circuitry intended to rid it of frequency response changes due to varying headphone impedance curves.

Second, boy I wish there was some hard evidence the Bakoon delivered faster transient response...but there doesn't seem to be so far. I'll keep poking around more though, you never know when something will turn up.

thune's picture

Current source (very high source impedance) drive of dynamic drivers can reduce some harmonic distortion and eliminate thermal compression by bypassing the mechanisms that generate them.
Here is the Klippel large-signal model for a dynamic driver.

The "voltage" (v) that drives the mechanical-system model on the right, is the force-factor [Bl(x)] times the current in the electrical-system loop on the left (i). If you force the current on the left with a current source amp, then the voltages produced across the elements on the left (and their non-linearities) become irrelevant, the current in that loop is solely determined by the amp.

Supposedly the benefit of current drive is reduced mid-band distortion, especially for simple dynamic drivers that do not have copper or aluminium caps/sleeves/rings [ which is most or all of dynamic headphone drivers.]

Tyll Hertsens's picture
Hadn't thought about it, but your comment makes me think I should run distortion plots as well.
thune's picture

--The "voltage" (v) that drives the mechanical-system...
should read
--The force/"voltage" that drives the mechanical-system...

johnjen's picture

This is thought provoking in multiple ways.
These sorts of investigations always help to shine more light on the relationships between the related gear and its performance.

But it has always been my understanding that an amp should NEVER dynamically change the FR of the signal that is presented to the drivers, regardless of the characteristics of the load itself.

And most dynamic drivers all have variable load impedance vs. frequency. And if a crossover circuit is introduced into the circuit this usually introduces additional impedance fluctuations.
And granted headphones rarely have crossovers, but in 'extreme' cases (HD800's, HD650's etc.) where there is a large impedance shift these amps would seem to add 'color', by definition.

And if this 'color' were considered to be 'desirable' would this be along the lines of tube 'colorations'?

It's also interesting, at least to me, that voltage amps which have a very low output impedance (well under 1Ω) is their way of 'providing' current dumping capabilities.

This also brings up the idea of a 'transmission line' (the wires) which feeds the signal from the amp to the drivers, as being more of an issue for voltage amps than current amps.


More food for thought indeed!

Like I like.


xnor's picture

You can simply add a resistor in series with the load to achieve a similar effect with any "voltage source" amp that has high enough gain.

This "coloration" is basically load-dependent equalization.
If we just look at that, then it doesn't matter if there are tubes in the amp or not. It's the output impedance that causes this interaction with the load.

Phoniac's picture

of their marketing idiots, throwing around intentionally misleading phrases more and more. What made you think that the Groove is a constant current amp? Ah, I see:
'Constant Current Drive™ provides smooth frequency response with any headphones'

> For example, the Apogee Groove is claimed to have a circuit design that prevents changes to headphone frequency response with changes in its impedance vs. frequency response.

The reality will be that this amp simply uses feedback to reach a high damping factor - as so many others do as well.

While I don't own one the specs from the Apogee website speak numbers. The Groove wil provide 86 mA at 30 Ohms, and 8.1 mA at 600 Ohms. There goes your 'Constant Current Drive'...

thune's picture

I'm dubious about the supposed "impedance sensing" part only. Measurements will cut though all the vague speak and definitely give some answers. I'm looking forward to them.

However, the numbers you provide above are derived from the max power numbers, and it is certainly reasonable that the device is power/current or distortion limited into 30ohms while at 600ohms the device is simply running into the maximum voltage output of the device ~4.9volts_rms. I wouldn't use these numbers to support any claims about the mode of the output. [BTW, it would take over 50volts_rms and 4 watts to put 86ma into 600 ohms (clearly unreasonable).]

wink's picture

Clearly not.

The paradigm should always be:-

"Go big, or go home.....!"

Bill Brown's picture

I think that 1) the Groove approximates a current source and 2) does not have any circuit correcting the frequency response (I think this is a misunderstanding of marketing terms)

1) the site lists 225mW into 30 ohms and 40mw into 600 ohms, my calculations of the currents are 87mA and 82mA, respectively (this being almost 5V into 600 ohms!)

2) I had for a time a Groove and I heard and measured the expected rise in the bass based on the impedance variation as would be expected from a current source.

3) Because of the current it will have no problem getting very loud into high impedance headphones, and it sounds very, very good (I corrected the calculable frequency response deviation with a good quality EQ plug-in)

I have a growing fondness for current drive. For headphones with a reasonably flat impedance I would never look back (the Groove made original Sennheiser Momentums sound the best I have ever heard them, very nice).


Bill Brown's picture

I should have mentioned this was into my preferred HD-650s.

Bill Brown's picture

I thought that Phoniak had misplaced a decimal point but it was ME! It IS 8.1 mA into 600 ohms, not 81. My humble apologies!

Now I am puzzled, as I definitely measured more voltage into the impedance bump of the HD650s.

xnor's picture

So it's a normal amp with a nonzero output impedance. No big deal.

Phoniac's picture

Yes, marketing is horrible, it seems. But the device itself is phantastic. Price to performance ratio and usability are definitely unique. This is one of the best things Apogee has built in years (IMHO). So they don't need such shitty advertisement/marketing at all, just stay with the facts is more than enough!

Phoniac's picture

Bill, I have meanwhile found statements of Apogee personel in some forums, and it seems you are right with 1, the Groove is indeed a current source amp - somehow. This logically means it will not show a flat frequency response over varying impedance, as you have experienced.

If so then 2) must be true as well and marketing is still to blame.

Hopefully Tyl will measure this device soon to reveal the truth.

ashutoshp's picture

to me at least. Could be wrong but I see you didn't have a hearing test to go with your notion that current source amps like the Bakoon are supposed to be better at transient responses. Do listening test show this?
On a separate note, maybe its just how you are used to looking at plots but I am used to looking at far noisier plots and (sometimes) interpreting them but in your case, those transient responses look VERY different indeed. I think you were expecting a "massive" response and didn't see one but if you tried to make them "smoother" (i.e., more linear), to get at their slopes for instance, you might see something more clear popping up.
Next you could try to see if this goes with what you can "hear", if at all. Only then should you discard your hypothesis.
This takes me back To Mike Moffat's post on headfi recently and what the human ear perceives.
Good stuff. I look forward to more.

tunde's picture

Am just curious about what the best amp for a newbie who just wants decent sound?

kevin gilmore's picture

tyll, get one of these

DC to 50mhz.

Probe is fragile, so don't ever drop it.

Ducatista47's picture

Have you seen Nelson's paper? I own his two current source amps, F1J and F2J, and tend to believe his take on these matters. Speaker oriented article, but informative. And yes, the F1J is a pure current source amp.


elmura's picture

The higher response below 40Hz on the planars driven by the Current Source appears to be due to rising capacitance. If that's the case, then the current source benefits even planars.

Pion's picture

The start was promising. Then the rest of 95% you can scrap it completely.
Fact is, if you want to link acoustic response of a magnetodynamic device (headphone) to anything, something that you can measure, then better start measuring current through the cans. How to do that with a voltmeter or an oscilloscope is a known method: use a so called shunt resistor (google it and then read some real electronics). For a pure current our amplifier measurements it should not matter if you choose 1Ohm or 10Ohms. For a voltage-out amplifier and using HD800 the shunt resistor can also safely be either 1Ohm or 10.
Ok, now read those shunt voltages and tell us again what conclusions you have. Better off, measure also the acoustic responses of the cans in the two amplifier situations.
The cans can normally do 200mW. Choose 10mW for safety of your ears or your expensive cans. Consider the highest impedance possible of the cans, something like say 1Kohm. Then current should be set at 3mA. Then if you use R_shunt 1Ohm is better to read out the voltage after a low-noise preamplifier (A=300, BW>1MHz) and you will have an output voltage of ca. 1V. This should be large enough to drive your spectrum analyzer (max value).
I do not say this is the best setup, but is a far better idea (= scientifically correct setup). Then analyze (always plot only the shunt currents, for reference compare with voltage signals from reference microphone in acoustic measurements) and then draw again your conclusions. If you want to contest my opinion, welcome, here or you have my email. Thanks!