Project Babybottle: The AC powered Quickie based headphone amp.

[mcandmar]

Choke load vs Constant Current Source:

Somebody raised a question recently on the sonic differences between the Hammond 156C choke load and the PJCCS plate load in a stock Quickie.  It sparked my curiosity so i took it a step further and implemented a Cascode Current Source as used in a number of Bottlehead amps, but configured for 4.25ma.

The first obstacle was the need of a higher B+ voltage for the CCS, using a bench supply i found 85v was the absolute minimum required, and moving up to 90v, 95v, 100v, 105v showed little to no advantage.  The reasoning behind trying these voltages lies with the available 7pin gas voltage regulators.  The basic options were,

75C1/0A3 75v 5-40ma
SG16P     80v 5-30ma
85A2       85v 1-10ma
90C1    90v 1-40ma
SG15P-2   102v 5-30ma
SG2P     104v 5-40ma
108C1/OB2 108v 5-30ma

The 85A2 is limited to 10ma which ruled it out as the total load was just over its limit. All the regulators under 100v glow a bright warm orange colour, and all the regulators over 100v glow with a purple/violet hue. They also have a slightly larger bottle, so purely for cosmetic reasons i opted for the 90C1 8)

The second obstacle was the raw rectified voltage from the mains transformer was below the ignition voltage of these regulator tubes so i switched the mains transformer from the dual 18v secondary to the next model up in the Talema range (RS0015P1-2-025K) which is identical in physical size, but with two 25v secondary’s. This now allows me to run everything up to the 108C1.

Sonic differences. Frustratingly i spent a day trying to get RMAA to work properly after upgrading my computer to Windows 10 and lost the battle.  Therefore any comparisons will have to be purely subjective.  The overall difference is a lot smaller than i expected.  What i noticed initially was a leaner sounding low end, but with time i have come to the conclusion it is just more controlled and less bloated, which may or may not be a good thing depending on your setup.  Mids and highs are so similar it is really splitting hairs trying to differentiate them. It is fair to say the overall presentation just seems a tad cleaner and tighter with the CCS loads, which is probably to be expected. As to which i prefer, i don’t really have a preference as of yet.  There seems to be pros and cons for both solutions, technically the CCS will perform better, but subjectively some people may prefer the sound of the choke loads.  How is that for an utterly useless conclusion?

 

[Paul Birkeland]

I went with a single 0A3 on the AC powered Quickie that I built, as well as using the C4S instead of the PJCCS.  Feeding the 0A3 with a C4S also has its advantages.

-PB

 

[mcandmar]

I went with a single 0A3 on the AC powered Quickie that I built, as well as using the C4S instead of the PJCCS.  Feeding the 0A3 with a C4S also has its advantages.

-PB

Interesting, what operating point were you running?  With the current config, 4.25ma and a 1.4k cathode resistor i have about 52v on the plate and 6.3v bias.  I used an old laptop with RMAA to get some idea what was going on and as you can see it was not happy with a 75v supply, but quickly starting to behave with another 10-15volt overhead.

Feeding the regulator with a CCS has me intrigued, never seen that done.  How much current do you pass into it?  My load is a little over 11ma, 4.25 for the plate, and 1.33ma for the LED string on each side.  Do you aim for half the regulators nominal current?  i.e. 20ma for an OA3.

 

[Paul Birkeland]

The first thing I did when I built mine was to heat the 3S4 with 3V, since I wasn't using D-cells and available voltage wasn't a problem.

It has been a while since I've even had that Quickie in my house, but I remember raising the plate voltage and current slightly.

I believe I hit the 5V bias/4mA/45V operating point.  That gives the possibility of allowing more signal to come in, and you're still going to run out of room on the 0V grid bias side of the operating point before you hit max voltage from the 0A3.  Do note that if you tried to do all this to a battery powered Quickie, you'd run into a gazillion problems.

For the CCS, you can use a single 0A3 with a cap bypass for both 3S4's.  If each 3S4 needs 4mA from a C4S, and you bias the C4S with 1mA for the LED's, then that's 10mA for both tubes, so you'd want to pass around 15-20mA of current through the CCS feeding the 0A3. 

I looked back a little bit, and it looks like I did indeed use the 105V regulator AC Quickie, which is likely why I didn't run into the issues you're finding. 

-PB

 

[mcandmar]

Interesting, thanks Paul.  Going to have a think about that CCS supply for the regulator.  I find it amusing that you have done all this already, i suppose you knocked it up in an afternoon too?

The next experiment i want to try is replacing the cathode resistor and bypass with an LED string. Four HLMP's should give me 6.3v which would be ideal.  Or is there some gotcha doing that with an output transformer?  Currently the primary is terminated before the cathode resistor to keep the capacitor out of the loop.

Cheers,

Mark

 

[Paul Birkeland]

It took a bit of time, mostly because of all the individual power supplies that had to be made.

With the LED biasing, I don't believe it matters where you terminate the primary since the LED's have little dynamic impedance.

 

[Paul Joppa]

The reason for using a cathode resistor is to compensate as the battery voltage drops. If you are using a line-voltage power supply, or even better a regulated high voltage, then LEDs are practical; in fact you can also parallel the two tube's heaters, requiring only one low-voltage supply.

I favor the 1.5v filament power, as it makes the tube more linear.

If you do make independent supplies for each tube's filaments, you can also explore filament bias (as used in the BeePre). It might be possible to share one supply and one bias resistor - there will be a bit of crosstalk between channels, but probably to little to hear.

 

[mcandmar]

Filament bias is an interesting idea i hadn't considered.  I actually run filament bias on my 4P1L preamp using the same Rod Coleman regulator boards, which performs wonderfully.  I will have to ponder about the pros and cons of Filament bias vs LED string bias.  ...hmm

I did opt to stay with 1.4v on the supplies so i could switch in 1S4 tubes.  It also never seemed right to me share filament supplies between DHT tubes, i much prefer the monoblock approach and keep them completely separate.  I know all this is probably a little overkill for a Quickie headphone amp

Cheers!