Head-Fi.org › Forums › Misc.-Category Forums › DIY (Do-It-Yourself) Discussions › The "ultimate" modification for the AKG K340?
New Posts  All Forums:Forum Nav:

The "ultimate" modification for the AKG K340?

post #1 of 10
Thread Starter 

Hello everyone,

 

I have this idea sitting in the back of my head for a long time, and didn't follow up mainly due to the lack of time as I'm preoccupied with other projects (some of which are related to stereo/headphones as well).

Eventually I've decided to share it with this forum and see what other member of the forum think about it, so that if there's some interest I will follow up on that - be it alone or with some other member of the forum.

 

The K340's are probably amongst the most heavily modded headphones ever. I've had lots of these, practically all of which I've modded with any possible mod, or at least so I thought :)

 

I was thinking, the AKG K340's have a small PCB with a cross-over which consists of a first order filter (resistor + a small transformer). Well, in speakers design there's a very large following of active cross-over, so why not do the same with the K340's?! :) Some would sat the inclusion of the transformer is just another reason to do this, since the transformer can increase distortion.

I was wondering if someone else did this already?

 

It shouldn't be hard to do, there's only the question of whether there's the will to do it. From all of the K340's I've had in the past/some I've modded for others, I have a few spare parts. So I can easily draw a schematic of the PCB version I have, as well as part value (including the voltage gain of the transformer). Obviously I will post anything I'll find out, but first I'd like to know what others think about it/if there's interest this.


Edited by KT88 - 9/16/13 at 1:29pm
post #2 of 10

If memory serves Aerius was talking about this at one time or another. 

 

The "crossover" is not exactly a crossover, as it does not have a roll off/roll in frequency. Its just a stepup transformer. 

 

At one time I had the thought to use a fairly conventional push-pull tube amp to drive the cans. Cap coupled off the plates to the electret driver, and transformer coupled to the dynamic driver. Still no real crossover, but it could be cool to do anyways. 

 

As for active crossovers, this could be really cool. Possibly expensive, but if you already had an electrostatic amp hanging around I see no reason not to try it. 


Edited by nikongod - 9/16/13 at 12:46pm
post #3 of 10
Thread Starter 

Well, such a small transformer will have a very high low-cut-off frequency, so it does act as a filter (and that resistor in series will make sure it acts as a filter even if the transformer isn't a limiting factor). I also remember discussions about the -6dB/octabe roll-off. There's also the question of do we really need a very high voltage amplifier? My initial thinking is that well, no. These drivers are not that large, and they only produce high-frequencies so they don't need to travel a large distance. This allows small dimensions and lower voltages. So now I've decided its so interesting, I'd check :)

 

I've connected the PCB to my signal gen, and the input + output to the scope (the electret wasn't connected for this simple test). Here are some observations:

10KHz sine wave input:


As you can see the gain is very low, just ~X5 (14dB). So we can probably drive it just fine with a high voltage "normal" amplifier.

 

As for the filtering, well, its has a very significant roll-off. I've measured the following:

frequency - Xgain

20KHz - X5.8

10KHz - X4.8

5KHz - X3.8

2KHz - X1.5

1KHz - X0.8

So it does act like a 1st order high-pass filter with the -3dB point somewhere around 5.5KHz (and I remember past discussion where 4KHz was mentioned by AKG in the brochure). Such a simple active-cross over can be implemented using an op-amp and some caps+resistors around it. It's just a 1st order filter. No need for many stages or notch filters or anything. An active cross-over can also be adjustable to allow different gain for each drivers, and fine adjustments of the cross-over frequency.

 

These measurements actually show it should be very easy to implement this. The great thing about it is that experiments can be done without damaging anything. Simply take the original PCB out and connect the ground of the electret directly to the ground of the dynamic driver (which is already done by the PCB), and add another wire from the electret to a second amplifier. As a first step this HPF can be even implemented in a passive network before this additional amplifier.

 

Edit: I've now measured the output voltage from my amplifier at above-normal listening level with the K340. Its just 0.7VRMS. This means we only need ~3.5VRMS to driver the electret - that's a swing of ~10Vpp. That's easily achievable with lots of headphone amplifiers out there. Heck, even my PimetaV1 can do that with the 24VDC power supply :)

 

BTW, I'd like to mention that I've measured the output of the PCB with a X10 probe to minimize loading. That's 10Mohm with a few pF in parallel (cant remember exactly what the capacitance is).


Edited by KT88 - 9/17/13 at 9:06am
post #4 of 10

Yepp, I remember the transformers being physically tiny, but I did not realize that they would roll off that quickly. 

 

Looks very doable, even with a simple cmoy-ish amp to drive the electret. 

Is the electret driven single ended or balanced? 

post #5 of 10
Thread Starter 

Yes, transformers are limiting. In that case it really is a first order filter as it has a resistor as well. Its an RL HPF :)

 

The electret is driven single ended. Its negative contact is shared with the dynamic driver.

I'd like to verify the voltage measurements, so I'll try and find the time tomorrow to hook up one of these electrets I have lying around directly to the signal-gen and feed in a 10KHz sine. I'll measure the amplitude required to drive it to a normal listening level and make sure its identical to what I got yesterday. It's just another verification step, I doubt we'll learn anything new from that.

 

And yes once more, that's exactly what I was thinking. Use a couple of 2-channel op-amps, one with a gain of 3 for the dynamic driver, and one with a gain of 15 and a LPF in-front (or in the feedback path, I'll see what I like better) for the electret.

 

Now there's also the question of how far we'd dare take it. Why not make it a second order cross-over and see what happen :veryevil: 


Edited by KT88 - 9/17/13 at 9:41am
post #6 of 10
Thread Starter 

Just did that measurement. 10KHz sine wave from signal gen directly into the electret, with the DMM in parallel measuring voltage and indeed it is audible from a few hundreds of mV's. 3.5VRMS as I calculated yesterday is much stronger - this is in open air a few inches from my ear so inside the headphones it'll sound even stronger.

post #7 of 10

Totally unsolicited, but how does this look:

 

(right click and open in a new tab/window)

 

 

I did something weird. Very weird... but I hope you like it.

 

So to walk through the circuit, signal enters from the left and goes through a standard non-inverting op amp. 

After this the signal splits and things get weird:

I generally like many headphones from high output impedance sources, so R4 is included to satisfy that preference AND isolate the op amp from the load. Since the signal for U2 is taken from this point a bit of isolation is beneficial. 

 

U2 is as an inverting op amp which is somewhat less conventional in headphone amps, but offers some cool points. C2 and R5 form a high-pass filter which could be used to set the roll off for the electret. Since U2 is driven by U1 small value resistors can be used around U2 with great success. Using 1Kohm for R5, and 3.3/4.7Kohm for R6 would not be out of order. R7 should be selected to balance the input offset currents if that applies to your op amp. R8 is kind of weird. The stock transformer no doubt has a highish output impedance - I have heard other headphones (the Taket piezo) that sounded absolutely dreadful when driven from a zero-ohm source, but totally delightful when driven from something with some source impedance and your chosen op amp may not be stable driving a purely capacitive load anyways. Maybe you just jumper this spot, maybe you put in a 5Kohm resistor... I would put in a dual 5Kohm linear pot and dial it in. 

 

Please note that the *signal* being fed to the dynamic and electret speakers are out of phase! The effort to swap the leads on the dynamic driver is nominal, and I think that doing this offers a benefit. Because the signals at the outputs of the op amps are out of phase they tend to cancel out the ripple fed back into the power supply - much like a balanced amp. It will not be perfect, and even a ****ty op amp has awesometoast PSRR but when I can find a possible improvement for free I tend to take it... I do not know exactly which way to wire the dynamic drivers for this! I would tune it by ear, or if possible with a speaker polarity checker tool commonly used for testing speaker crossovers. 


Edited by nikongod - 9/17/13 at 8:18pm
post #8 of 10
Thread Starter 

That's actually somewhat similar to what I was thinking as a first test option, not including R4, and with a lower value on R8. I know some people are fans of this trick, but so far I like my K340's with low impedance output stages much better and that would have been my first option. Obviously this can be easily added if someone likes it better.

 

The out of phase issue isn't a real problem. You simply connect the electret out of phase as well, and problem solved, so that's not a problem at all - no need for a speaker polarity tester for that. In that case two wrongs really do make it right :D

I think connecting the electret backward is the right way to keep both drivers in phase, and not the dynamic driver. Connecting everything in a reverse phase can sound strange.

 

As for the stability issue with driving a capacitive load, its a non issue in this case. The capacitance of this driver is very low, just 500pF (I've measured - lucky for me I have a few electrets sitting around so I can measure this without taking apart my headphones). This value of capacitance can be driven directly from the amplifier output with many opamps, or with a very low value resistors with others. BTW, the resistor value in the PCB is 680R, the DC resistance of the transformer is 260R (entire transformer, from the high voltage output to ground). So the resistance seen by the electret is ~900R, and that is the maximum value we should consider using IMO.

 

I guess next step is to built the amplifier, and to mod the headphones to start some listening tests.

Now I just need to find the time to do that. I can easily make a small PCB that will replace the original PCB and reverse polarity as well and then cut it on my CNC. But that means I have to sit down and put the new Y-axis motor in place, which I can't find the time to do. And then there's the soldering of this amplifiers, who has the time for that?! :(

It'll probably take me some time to do, a few weeks at least unfortunately - as I really am short on time in the near future with school and work stuff. Any chance anyone else (such as you nikongod :)), want to help with that? Any help would be great, be it doing the entire test mod or just building the amplifier and then send it to me/I'll send you the PCB. This will surely make things faster.


Edited by KT88 - 9/17/13 at 10:04pm
post #9 of 10
Quote:
Originally Posted by KT88 View Post
 

As for the stability issue with driving a capacitive load, its a non issue in this case. The capacitance of this driver is very low, just 500pF (I've measured - lucky for me I have a few electrets sitting around so I can measure this without taking apart my headphones). This value of capacitance can be driven directly from the amplifier output with many opamps, or with a very low value resistors with others. BTW, the resistor value in the PCB is 680R, the DC resistance of the transformer is 260R (entire transformer, from the high voltage output to ground). So the resistance seen by the electret is ~900R, and that is the maximum value we should consider using IMO.

 

I guess next step is to built the amplifier, and to mod the headphones to start some listening tests.

Now I just need to find the time to do that. I can easily make a small PCB that will replace the original PCB and reverse polarity as well and then cut it on my CNC. But that means I have to sit down and put the new Y-axis motor in place, which I can't find the time to do. And then there's the soldering of this amplifiers, who has the time for that?! :(

It'll probably take me some time to do, a few weeks at least unfortunately - as I really am short on time in the near future with school and work stuff. Any chance anyone else (such as you nikongod :)), want to help with that? Any help would be great, be it doing the entire test mod or just building the amplifier and then send it to me/I'll send you the PCB. This will surely make things faster.

 

The output impedance of a stepup transformer is the source impedance times the turns ratio squared. 

680ohm*(5^2)=17Kohm. 

Ideally you should include the DCR of the coils ( ((680+primary)*(5^2))+(secondary DCR)=Zo ), but this is close enough. 

This may all be covered by the high-pass filter before U2, but it is still good info to keep in mind. If you plan for possible problems early on they are very easy to deal with if they do pop up. A ~330ohm resistor wont interact with the electret at all, and will keep any op amp stable here. 

 

I have not actually built anything in way too long... hmmmm. I will see. 

post #10 of 10
Thread Starter 

Right, right, I completely forgot about the N^2 factor for the impedance transformation :D

All my measurements are DCR, and not impedance at high frequency.

 

330R sounds like a good compromise, its low enough to have no effect on sound, yet sufficient to keep it stable with no problems. Its possible to add some feedback to around this resistor and add a small cap back to the input from before it. This will keep the amplifier stable but make sure the feedback at the audio frequencies is taken directly from the electret. But this is too "complicated" for a first go.

 

I will update once I have time to do this, but it might take quite a bit of time. If you can find the time to build a basic circuit (no need for a PS, I can hook this up to one of my lab PS's which goes up to +-40V) let me know and we'll give this idea a real test run.

New Posts  All Forums:Forum Nav:
  Return Home
Head-Fi.org › Forums › Misc.-Category Forums › DIY (Do-It-Yourself) Discussions › The "ultimate" modification for the AKG K340?