Head-Fi.org › Forums › Misc.-Category Forums › DIY (Do-It-Yourself) Discussions › Twilight Tube-MOSFET mu-stage hybrid headphone amplifier
New Posts  All Forums:Forum Nav:

Twilight Tube-MOSFET mu-stage hybrid headphone amplifier - Page 2

post #16 of 35
4/3/08 at 5:18am
Thread Starter 
Holco.

Ok, an update:

Gradually feeling the amp out, I get the sense that the treble is slightly attenuated. If I had to guess, 1-2 dB at 20 kHz.

Thing is, according to my calculation this really shouldn't be the case.

The triode has a Miller capacitance of about 80 pF. (4pF x 20) The input capacitance on the MOSFET is about 90 pF, meanwhile the output impedance of the triode is about 2 kohms. By my estimate, the bandwidth should extend to a 3dB point of 200 kHz, even with the grid and gate resistors at 3.3k.

Would anyone care to make a better stab at calculating the frequency response?
Reply
post #17 of 35
4/6/08 at 10:56am
Thread Starter 
I tested the bandwidth Twilight amp. The -3dB point is 320 kHz. The -1 dB point is 160 kHz. There is no rolloff at audible frequencies. Nonissue.
Reply
post #18 of 35
4/6/08 at 11:08am
What a nice unit, wish it was commercially available. Or at least as a kit.
Reply
post #19 of 35
4/6/08 at 1:38pm
nice design! gotta put this on the long list of designs to build and try! great work man, very very good job, congrats!
Reply
post #20 of 35
1/30/09 at 9:25pm
Thread Starter 
Rightmark Data
Taken with Onkyo SE-200PCI soundcard. The noise baseline at about -125dB and the channel imbalance of 0.25 dB in the attached plots originate from the input stage of the soundcard, not the Twilight amplifier.

The soundcard output volume was 100, +10 dB
The Twilight volume was set to -10 dB overall amplifier gain (normal listening position)
The soundcard line in volume was set to 44, +0 dB.

This means the output signal from the amplifier is 1:1 of what Rightmark thinks it is and the dB scale should be accurate.

The Twilight amplifier was output loaded with 330 ohm resistors as a substitute for the headphone load.

Things to note:

Frequency response (-1dB) is 30 Hz to 40khz. The bass response is worse than I expected, and is not dependent on the output loading, so the size of the output coupling caps is not the cause.

The noise is at least -80 dB, and that's mostly 60 Hz filament hum.

As best as I can work out, from other test results not shown - and I admit there's a bit of guesswork involved here - the THD at 5 mW output power is 0.1%, while at 0.5 mW output power (typical music max) it's about 0.01%. I had designed for 1% THD+N at 45 mW, so these numbers seem in the right ballpark.

The actual channel imbalance is 0.1dB.
LL
LL
LL
LL
Reply
post #21 of 35
1/31/09 at 12:29am
Quote:
Originally Posted by rjm003 View Post
Frequency response (-1dB) is 30 Hz to 40khz. The bass response is worse than I expected, and is not dependent on the output loading, so the size of the output coupling caps is not the cause.
Did you try bigger cathode caps? 22uf seems to be small.

I agree with the others in this thread: VERY nice work!
Reply
post #22 of 35
1/31/09 at 10:37am
Thread Starter 
There are three RC high pass filters in the circuit which limit the bass response.

The output coupling cap (100 uF) against the output load (300 Ohms)

The cathode bypass cap (22 uF) against the cathode resistor (2200 Ohms)

The capacitor linking the triode plate to mosfet gate (0.1 uF) against the parallel load of the bias resistor 470 kohms and the mosfet gate (assumed to be very high at low frequencies)

The 3dB point of these pairs (1/2piRC) is about 5 Hz for then first set and 3 Hz for the last two. i.e the headphone load should be the dominant factor in determining the bass response. Apparently it's not.

It seems like I'm missing something in that the load at least one of the capacitors sees is far lower than my estimate. For example, if the MOSFET gate was only 100 kohms, that would explain it...
Reply
post #23 of 35
1/31/09 at 6:44pm
I think you may have oversimplified the calculations for sizing the cathode bypass cap.

You need to calculate based on paralleling cathode resistance (Rk) of the tube with the cathode resistor.

Rk=(Rl+Ra)/(MU+1)

Rl=plate load.
Ra=plate impedance

Now that thats said, Im not 100% sure what the plate load actually is The mosfet's source looks like an infinite load, but I think that the tube also sees the path through the plate resistor and the 1.5Kohm resistor as a load in this case.

What is the voltage gain of the amp ignoring the volume control? I wasn't sure from the RMAA post. Anyways, if the gain is 20 or so, Im probably wrong about the cathode cap. If the gain is about 5 or 7, Im probably right.
Reply
post #24 of 35
1/31/09 at 9:30pm
Thread Starter 
I forgot to list the amplifier gain above. It's 23 dB, or 14. The mu of the 6CG7 is given as 20, so there is a bit of loading going on, definitely.

It's been awhile, but iirc the plate load is the small plate resistor 3k3 multiplied by about 20 owing to the mosfet transconductance of 0.05 S, about 67k I estimated. The plate impedance - again it's been awhile - is around 5k I think.

It rather looks like the plate load is a fair bit lower than I though it is, presumably because I messed up my calculations somewhere.
Reply
post #25 of 35
2/1/09 at 12:48am
Thread Starter 
Back with some datasheet numbers:

Forward admittance of 2SK213 mosfet: 0.1 S

For a 6CG7 at 1mA 50V

Transconductance gm: 0.001 S
Plate resistance, rp: 16 kohms
mu: 18.5 u

Datasheet typical operating conditions; 90 V power supply, 100 kohm plate load, 100 kohm load, 2.6 kohm cathode resistor, the gain is listed at 14.

Following nikongod's formula:

Rk=(Rl+Ra)/(MU+1) [Rk usually the physical cathode resistor, so lets call this quantity just R for now]

R = (3.3k*1/0.1 + 16k)/(21) ~= 50k/21 = 2.5k

R||Rk = 2.5k||2.2k ~= 1.2k ... about half the value of Rk alone. So I should increase the cathode bypass capacitance to at least 47 uF, probably 100 uF to be safe. The voltage rating can be very low since they never see more than 3 V, so the high capacitance value isn't a problem.
Reply
post #26 of 35
2/2/09 at 6:00am
Thread Starter 
"Following nikongod's formula:

Rk=(Rl+Ra)/(MU+1) [Rk usually the physical cathode resistor, so lets call this quantity just R for now]

R = (3.3k*1/0.1 + 16k)/(21) ~= 50k/21 = 2.5k

R||Rk = 2.5k||2.2k ~= 1.2k"

That's enough to move the 3dB point from 3-4 hz to 7-8 hz, but not enough to reproduce the 20 Hz f3dB measured.

The gain of 14 allows us to estimate the effective Rl value. It is 37 kohms.

The actual plate resistor though is only 3.3 kohms. If you use that value, rather than 37 kohms, in the formula above, it does reproduce the measured 20 Hz f3dB point as the effective cathode resistance is 920||2200 or only 560 ohms.

I'm not sure if that's right, but the amplifier behavior correlates with Rl << Ra. Either Rl is the headphone load of 300 ohms, or the plate load of 3.3 kohms, or some other small value, but it does not seem to be more than 3.3 kohms.
Reply
post #27 of 35
2/9/09 at 11:52am
Thread Starter 
Increased cathode bypass capacitors to 100 uF.

Frequency response is unchanged. -3dB at 20 Hz, loaded or unloaded. Double checked this by measuring the output with a voltmeter, just in case there was anything odd going on with the soundcard inputs.

Next thing to try is increasing the capacitance to the MOSFET gate I guess.

I must admit this whole thing has me really confused.
Reply
post #28 of 35
2/9/09 at 1:47pm
This is also a long-shot, and a weird guess:
can you test the soundcard into a 10K ohm load?

Its possible that the soundcard output is cutting out the low end I guess.
Reply
post #29 of 35
2/9/09 at 2:18pm
Where can I those enclosures? Damn that is nice.
Reply
post #30 of 35
2/9/09 at 2:39pm
Quote:
Originally Posted by olblueyez View Post
Where can I those enclosures? Damn that is nice.
really uncool.
Reply
New Posts  All Forums:Forum Nav:
  Return Home
Head-Fi.org › Forums › Misc.-Category Forums › DIY (Do-It-Yourself) Discussions › Twilight Tube-MOSFET mu-stage hybrid headphone amplifier