1M-F5 - A 1 MHz F5 project

[triode_al]

And that > 1 MHz claim . . . (yes, an F5 with o n e thousand kHz) bandwidth?
Easy at 8 ohm; with 4 ohm it drops to 910 kHz. And DC is also possible. My generator goes to 0,5 Hz - this is still not -3 dB.

audio -3dB no load 2,1 MHz.jpg
Surprise: A channel unloaded has a bandwidth of . . . an astonishing 2,1 MHz.
With 8 ohm it is 1,2 MHz, and with 4 ohm it is 910 kHz.
I'll gladly go for a lower bandwidth. This high bandwidth is a bit scary.

I do not have high frequency probes; I load the output with a 1,5 meter RG59 cable (300 pF) which might influence the result I see. The stray capacitance can be the reason the the peak on the screen; when going through the frequencies one does not see a high Q (maybe above 6 MHz?).

At these frequencies you will see slew rate limiting. And indeed if I run a square wave through, I will see limiting starting on the rising flank of the wave above 6/700 kHz, and soon it is a triangle.

I see a very small oscillation at 2,6 Mhz and a bigger one above 5,6 MHz (the highest my signal generator can go). This surely is due to capacitative coupling for instance of the lead wires and power supply interaction.
I connected a 2,7 nF silver mica across the 50 ohms feedback resistors: the bandwidth dropped to 650 kHz. Aklso, the unloaded square is now tightly controlled.

There is slightest ringing on the flanks (e.g. 300 kHz) more in one channel than in the other, probably dissymmetry in high frequencies.

audio 520 kHz last square 1.jpg audio square rounded by capacity.jpg

What I did was load a 470 capacitor across the output in (2). It shows the little series of straight lines, like a little slew rate (but it can't be slew rate as I see it).

Cures:
- I added a selected pair of 2,7 nF capacitors across the 50 ohm feedback resistors, selected pairs, tight matching is needed. Doesn't solve it.
- Maybe better : use input filter, 1k2 input loaded by 400 pF (low pass about 350 kHz). But this does not solve that smallest amount of ringing.
But neither helped.
- It could be balanced by adding a separate 300 pF gate/source capacitor on the 2SJ55 (because of its lower Ciss. Would this help?
- Or by adding slight different feedback caps.
Or implementing the trick in the version 2 F5 with a feedback fine-tuning pot.

Has anybody else bothered about this?
I decided to neglect this whole issue for the moment and . . listen.

[triode_al]

Because I have a electrostatic LS with a complex load the next test is very important: will the amplifier oscillate with a capacitative load? So I connected a capacitor across the output, and later used it with a damping resistor
Just a resistor (2n7 feedback cap):
audio 4 ohm, 2n7 fb.jpg

Overshoot with capacitative load, 8 ohms with 0,47 muF
audio 4 ohm + 470 nF.jpg
This is the ringing we saw earlier too.

foto overshoot 4 ohm parallel with cap 470 nF with a series 1 ohm resistor
audio 4 ohm parallel to 470 nF+ 1 ohm series.jpg
This is well damped.
This load can be made more complex with an inductor (i would have to look for the equivalent circuit of the Quad ESL57)

The little feedback capacitor of 2,7 nF stops the overshoot unloaded and reduces the ringing only a little bit. I would need 15 nF to really stop the ringing (I tried 21 nF but this was just a bit too much damping).


Distortion picture at 2,47 kHz
audio Distortion 2,9 kHz, 0,1%.jpg

Predominant second harmonic and down near the noise floor.
audio meter reading.jpg


Compare with Le Monstre distortion
The similar topology of Le Monstre with a cascode gave me 0,6% second harmonics. See my avatar.

[triode_al]

Because I have a electrostatic LS with a complex load the next test is very important: will the amplifier oscillate with a capacitative load? So I connected a capacitor across the output, and later used it with a damping resistor
Just a resistor (2n7 feedback cap):
Attachment 242038

Overshoot with capacitative load, 8 ohms with 0,47 muF
Attachment 242039
This is the ringing we saw earlier too.

foto overshoot 4 ohm parallel with cap 470 nF with a series 1 ohm resistor
Attachment 242040
This is well damped.
This load can be made more complex with an inductor (i would have to look for the equivalent circuit of the Quad ESL56)

The little feedback capacitor of 2,7 nF stops the overshoot unloaded and reduces the ringing only a little bit. I would need 15 nF to really stop the ringing (I tried 21 nF but this was just a bit too much damping).


Distortion picture at 2,47 kHz
Attachment 242041

Predominant second harmonic and down near the noise floor.
Attachment 242042


Compare with Le Monstre distortion
The similar topology of Le Monstre with a cascode gave me 0,6% second harmonics. See my avatar.

[h_a]

Really nice work!

Thanks for posting!

Hannes

[ichiban]

Originally Posted by triode_al

I decided to neglect this whole issue for the moment and . . listen.

And...? How does it sound.

[triode_al]

yesterday I moved it to the living room.

First impressions:

Pleasing.
Excellent. Much detail is transmitted, in depth.
Syllables and voices are having their own character, two singers in guet are clearly distinguished. No specific sound is projected (syllables).

What also surprises me is that the character remains stable when changing volume. [I first heard that on Hypex modules].
Low register is clear, the electrostats have the tendency to sound nasal; this is not [yet] heard either.

I cannot call the sound 'laid-back'.

[ichiban]

Originally Posted by triode_al

First impressions: Excellent.

[triode_al]

Second impression:
I was listening to Alfred Deller, accompanied by a pianoforte.
The details of a highly oversampling are now transmitted. Previously, the highs (in classical music) had the tendency to have a similar sound, now I can distinguish many high instruments and the complex notes of the instrument, also behind the players.
Differences between tracks are heard.
(Carla Bruni - on m'a dit) Small artifacts that the producer mixes in are heard, sound such as from the throat (coarse) can be heard.

Very very detailed, all due to the capacity to really dump a lot of current very fast in the electrostats.

My Meridian comes to life again!!!

[a.wayne]

9 watts is a lot of current ....

Fantastic write up Triode_al , very good details and information, much appreciated, Anyway of posting squarewave performance in the lower bandwidth, 20hz, 1 k, 20k, 40k, 100 K for eg...


regards,

[triode_al]

The amp does sound as if it is very powerful.
The analogy in impression with the 8W Le Monstre is clear now for me. It is a Monster too: 15 W.
I can calculate the peak current (the formula is Vrms/0,7=Vp); what I did do is check what the voltage was (via scope, DC input) and that gave me around 15 V on 20 V power lines when clipping starts in, if i remember correctly, with 8 ohm. I did't check the others, but surely they are high too,
2 ohms, 7 V RMS, 25 watt (10Vp=5A peak)
4 ohms, 9 volt RMS 20 watt (13Vp, 3,2 Apeak)

- I could do is take out the current limiter 'output' and connect that to a LED via a 2-3 K resistor: if it blinks I hit my max. Great test for a lab setting.

I'll make square waves from low frequencies too but I'm afraid this will have to wait three weeks (travel ..)
albert