DeathTrap 500™ - An Engineer's approach to the 300B SET

[tomchr]

Folks,

It's time to share... I've been working on this 300B design off and on for the past six months or so. Inspired by the TubelabSE and PowerDrive circuits I set out to design my own amplifier. I happened to meet a fellow tube enthusiast at work one day and he turned me on to the anode follower circuit, which I started to explore. I have tried many different things. High mu triodes, medium mu triodes followed by low-mu triodes, etc. In the end the anode follower based on a pair of 6J5 tubes won as it provides the needed gain and sounds the most neutral to me. Besides I have a good handful of the 6J5's. Might as well use them... A 6SN7 could be used as well.

I have arrived at something I really like. I wanted to call it the Engineer's Amplifier, but Pete Millett had already taken that name (bastard! ). As most of my prototype circuits are coupled together with clip leads and exposed wiring (NOT kid, pet friendly) I've started labeling my designs DeathTrap followed by the B+ voltage. In this case, the voltage on the reservoir cap is about 500 V. Hence, I bring you the DeathTrap 500™.

I do admit that the circuit is a tad over-engineered -- especially the power supply. But if I'm blowing $200 on a pair of tubes, I might as well go all in...

The bias supply is rectified with a pair of 1N4007 diodes. The main B+ supply is rectified by a Sovtek 5AR4. The long warm-up of the 5AR4 ensures that the bias supply has settled long before the B+ is applied. Both the B+ and bias supply are regulated (LM317/337-based). The B+ is set to 380 V; the bias supply to -180 V.

I wanted regulated DC supplies for the filaments to minimize hum and wasn't happy with the amount of power I was looking to burn in the linear filament regulators I had in mind. Hence, I chose to have the filament supplies regulated by a pair of switchmode converters. The switchmode regulators I use are 90~92 % efficient, hence, only waste about a watt...

The amp uses a modified Antek 2T350 power transformer and Edcor CXSE25-8-5k OPTs.

The only design choice remaining is whether to use a CCS on the source follower or not. To my ears, on my speakers, the CCS sounds slightly harsh but more revealing in the midrange. A resistor on the other hand makes the amp sound much more relaxed. There's still good detail in the mids and highs. As I can't make up my mind, I've left both options in the schematic. Once I get a PCB laid out, I intend to perform another listening test comparing the two options.

I really like the sound of this amp. I love the details on human voices and on metallic instruments. There's a lot of detail there. It's an amp that makes me want to listen to album after album.

I'm measuring about 1 % THD at 1 W into 8 ohm. The amp tops out at about 4.5 W (3 % THD).

Schematics and some pictures below. The supply PCB was done with toner transfer. I intend to make the amp board the same way.

Note that the voltage differentials in the power supply approach 1 kV. If you decide to build this circuit make sure you know what you are doing as this circuit could easily kill you. The reservoir caps store about 12 Joules...

~Tom

[tubelab.com]

The only design choice remaining is whether to use a CCS on the source follower or not. To my ears, on my speakers, the CCS sounds slightly harsh but more revealing in the midrange. A resistor on the other hand makes the amp sound much more relaxed.

There is no reason why you can't have both. Set the CCS current for a little less than you need, then add the resistor to make up the difference. I have done this for plate load resistors.

Some builders have gone to great lengths to design a nearly perfect (but complicated) CCS, while the other extreme still insists that a resistor or choke is the only thing to use. There is a lot of middle ground between the two extremes and sometimes it just sounds better. Some tubes just sound best with an imperfect CCS. Maybe your follower will too.

[ColinA]

First please pardon my lack of technical brains.
In an amp I am running I have 6J5 with 5v bias.
I am intregued with the LED bias.
On your schematic one tube has one LED & the other has two.
What bias does your set up give?

[tomchr]

Originally Posted by tubelab.com

There is no reason why you can't have both. Set the CCS current for a little less than you need, then add the resistor to make up the difference. I have done this for plate load resistors.

Reducing the output impedance of the CCS - making it imperfect. Yeah... That thought did cross my mind the other night. Thanks for reminding me.

I'm a little puzzled by the quest for the Ultimate Current Source as well. We all choose to geek out in different ways, I guess.

~Tom

[tomchr]

Originally Posted by ColinA

In an amp I am running I have 6J5 with 5v bias.
I am intregued with the LED bias.
On your schematic one tube has one LED & the other has two.
What bias does your set up give?

The bias points are noted on the schematic, though, I guess I should put the cathode voltages on there as well. If I recall correctly, I have about 1.8 V on the cathode of TU1 and 3.6 V on the cathode of TU2. The swing on the grid of TU2 is upward of 3 V peak, hence I need more than 3 V on the cathode for the tube to operate in the linear region. This led to the choice of two LEDs in series for TU2. Now, TU1 on the other hand only sees a 1~1.2 Vpeak swing, hence, only one LED is needed. The imbalance between TU1 and TU2 does set up a little trickle of current in R6, R7 which can be reduced by running TU1 with two LEDs in the cathode as well. But that led to higher distortion so I opted to go with the setup shown in the schematic.

The voltage you'll get across an LED will depend on the color (red, yellow tend to be around 2 V; green 2.4 V; blue 4-ish V; IR 4~5 V) and will vary slightly from manufacturer to manufacturer as well.

~Tom

[tomchr]

It turns out the signal source matters too (I knew that)... Rather than my $50 Phillips DVD player, I switched in my Arcam CD-17. Much, much better... Now the resistor wins hands down. The CCS simply sounds way too harsh.

~Tom

[Hearinspace]

If you are still into experimenting, Wavebourn's SVCS is well worth trying.

[tomchr]

In case anybody is interested in PCBs for this amplifier, I have a couple left that I'm selling at cost on my website (see my signature).