µ-follower hybrid headphones amplifier
Nov 17, 2013 at 12:37 PM Thread Starter Post #1 of 4

00940

Headphoneus Supremus
Joined
Nov 6, 2002
Posts
4,493
Likes
47
Since I've been testing/listening to the amp since yesterday and everything seems fine, I thought I would document the build somewhere. It might interest someone else, who knows ?
evil_smiley.gif

 
The objectives of the amp:
 
-1- sufficient voltage and current for hd650 (aiming for 4Vrms as an overkill);
-2- low output impedance (under 10r);
-3- lowish gain;
-4- no global negative feedback (gnfb);
-5- low distortion with a decaying spectrum;
-6- no rare tubes or parts;
-7- OTL (see 6, good transformers aren't that common/cheap in Europe)
 
Design (or just one way to skin a cat out of many):
 
Gain problem: The main problem was to put together 3-4-5-7. Fact is, most tubes have an amplification factor (µ) which is higher than we would want. One can reduce that gain by loading the tube but that increases distortion. One could also use a transformer output to make use of that extra gain. But good transformers for high impedance headphones aren't plentiful. One could also reduce gain by using gnfb. But that was ruled out for the fun of the exercise.
wink.gif
More seriously, gnfb is nice if you have a lot of open-loop gain to be thrown away but then you get into stability troubles and the like. A last option was to use a low-µ tube to start with. Pete Millet went that way with his low-µ preamp. However, he used power tubes which are known to be kinda noisy (there are some complaints about that particular project around). So I picked an in-between solution and went for a triode trapped ecl85 (or compatile russian 6f5p), with a µ of about 8.
 
Topology: So I've got that beefy little tube which has to serve as voltage gain stage. To reduce distortion as a minimum, I want to load it as little as I can (which also means achieving a gain of almost µ). That means either a CCS on top or µ-follower. CCS on top means an extra stage for output (cathode follower to achieve -5-). µ-follower means I can fusion the output stage with the CCS. Since the ecl85 can endure 7W or so, I can even run all the output standing current through it. Makes thing easy, doesn't it ?
 
Output impedance: Not quite there yet. There's the output impedance problem laying around. Let's be realistic, a tube won't get us to 10r. A mosfet will though. Alan Kimmel goes at length in his classical paper about how nice mosfet are for that function, so let's follow his advice. The irf710 has a nicely low input capacitance, it's a classical choice. Sims give an output impedance of 5R
biggrin.gif

 
Power supply: I could have recycled the Tubecad reg I earlier used but Elvee just published a slightly better one, so I went for it. An irfbc40 serves as pass element. Btw, it's a beefy mosfet which, combined with the (very) high dcr of the r0080-036 transfo I'm using, allows it to survive the inrush current of the 220uf cap at its output. The cap would have to be reduced if another reg is used. For heaters, a simple zener reg is used. It's beefy  because I wanted to use 6f5p (925ma@6.3v each).
 
Output caps: I know this will be asked... yes, 40uf is a tad low. But honestly, comparing this amp to a dynalo (fully dc-coupled, from a fully dc-coupled dac), bass doesn't seem wrong nor missing. Headphones aren't much good under 30hz anyway. By the way, all the 20uF caps are big motor run mkp. They sound as good as any audiophile ones to my ears.
 
Schematic:
 

 
Building notes:
 
Nothing really unusual going on. Wire carrying HV should be 600V rated. Input cable should be twisted or shielded. Chassis must be earthed solidly and gnd tied to it only at one point. The schematic illustrates a star grounding topology, with the star ground at the gnd end of a cap. At the rca inputs, a small cap is soldered from gnd to chassis for hf protection. Fuse the primary of course.
 
Future developments:
 
I might add a pair of transformer at the output for low impedance cans. With only 25ma of standing current, there's nothing good for grado here. But I don't have low impedance cans for now, so I'll see later. A 4:1 or 5:1 would be good. The usual problem is to find such a transformer with low DCR at the secondary.
 
There's also room aplenty in the case for a volume control. Not really needed, I've one in each of my sources but I'll go for it later.
 
As usual with many OTL big thump at turn on. But the amp takes a while to warm up anyway (the 6f5p are awfully microphonic when cold). I'll add a delay when I get bored of disconnecting the headphones.
 
There's an easy way to tailor the sound of the amp, playing with the cathode bypass cap. Without it, THD increases significantly (yes, it's not like with a classic gain stage but the opposite) with a 300r load and so does output impedance (to 15r). I don't have the means to measure it precisely (no soundcard with good enough adc here
redface.gif
) but Ltspice gives an increase from about 0.015 to 0.15% thd at normal output levels. Spectrum isn't too affected though so 2nd harmonics will dominate. Currently, I'm listening with bypass caps (they're still in the mail) and I don't really feel the urgency to install them.
 
Yeah.. and of course there's the frontplate that needs a refresh. Case is an old Scott tuner.
 
Pics:
 
Some ugly pictures taken with a phone. I'll post decent ones once the casing really finished.
 


 
Dec 5, 2013 at 9:37 AM Post #2 of 4
A pair of changes:
 
- in the heaters supply, the two 2R2/5W resistors are removed (jumpered) and replaced by a 1R5/15W resistor before the diode bridge. It significantly reduces the current spikes in the transformer and in the first cap after the diodes (a DC supply for heaters is much harder on the xformer than AC heating). It dissipates less heat doing so too.
- I've added an external box with 100V line transformers for low impedance headphones, with 5:1 and 10:1 taps. A 32r headphones will reflect as a 3200 load, reducing gain to about 0.8.
 
Dec 5, 2013 at 3:01 PM Post #3 of 4
After messing around with rmaa drivers issue for a while, I checked the frequency response of the xformers in the good old way, with a signal generator and an oscilloscope. 
 
Testing the 5:1 tap with a 330r load, the FR is pretty much flat from 50hz to 20Khz. Under 50hz, there's a slight boost, with a maximum of 2.9db around 25hz. Nothing worrying.
 
Dec 13, 2013 at 9:59 AM Post #4 of 4
I'm pretty sure it's not of utmost interest for many but still... I got an old desktop to run rmaa with a better soundcard and got some indicative results.
 
THD @1K, about 1v peak into 300r gave me a thd figure of about 0.028%. Interestingly, the xformers aren't that bad thd-wise. THD figure with the xformers (same output level) is a bit less at 0.026% (thanks to reduce loading). Most of this is 2nd and 3rd harmonics,
 
As a baseline, the soundcard in loop is 0.0047%.
 

Users who are viewing this thread

Back
Top