Go Back   Home > Forums > Blogs > ballpencil

Rate this Entry

Not Another Gosh-darn EL84 Amplifier!

Posted 17th September 2016 at 06:32 PM by ballpencil
Updated 26th March 2017 at 06:01 AM by ballpencil

.. or NAGA84 for short.

That familiar feeling.. All tube enthusiasts will have a hard time resisting the urge to build an EL84/6BQ5 push pull amplifier, at least to hear for themselves what makes this tube so well-known (and overpriced).

Lurkers of the tube section most probably have already seen the Baby Huey wiki and thread. With such unique local feedback scheme, it's unsurprising that it garners some avid follower. As amazing as the feedback scheme may seem, there are few things i prefer to do differently on that amp. The most prominent thing would be NOT to use a triode for the input/driver stage as the rp of a triode will change with the signal swing and cause inconsistent feedback ratio. That, and the fact that it doesn't measure really well under simulation, motivated me to do my own version of the EL84 push pull amp.

To remove the triode from the feedback equation, there are two options i can think of:
1. Use pentodes, obviously.
2. Cascode.
From the options above, i prefer the second one because we get to keep using the same input tube type (i.e. ECC803 which is the fancy version of the ECC83/12AX7). I understand 12AX7s don't make the best cascode with its petty gm but we can easily increase the anode load impedance to get the same gain we would if we use high gm triodes.

Unfortunately, cascoded 12AX7 also comes with its own drawbacks. First, it's not very linear as the lower 12AX7 is working with near vertical load line (it's driving a low impedance node: the upper device's cathode/emitter/source/whathaveyou). Second, if you insist on using tube-to-tube cascode, you will have to deal with raising the heater supply to keep Vhk within the limit and we might complicate the power supply requirement as it's likely that we need another power supply just for the input/driver cascode.

I'm not really concerned with the first drawback as i can iron it out using global NFB. The solution for the second drawback is obvious: use solid state for the upper device, which in my case is an NPN BJT. *gasp* Come on.. you see how many solid state devices on Baby Huey? A few more won't hurt, right?

So.. how does NAGA84 look? Here you go:
Click the image to open in full size.

Some notes:
1. Gain is about 19dB. It takes about 1.3Vrms to reach clipping at around 15W.
2. Pentode mode for the OPS is deliberately chosen as it provides some advantages. First, we no longer need UL tap on the OPT which is made possible by the additional loop gain from using cascodes. Second, with no Miller effect, we can increase the grid stopper resistor value as high as 47K. This will then minimize grid current and blocking distortions.
3. The IRF740 follower unburdens the cascode (and the local feedback ratio) from the OPS' low grid leak resistor value which is preferable for long tube life. Again, a few more SS won't hurt..
4. During long, sustained crescendos the output tube rectification effect will charge up the cathode bypass capacitor, throwing off the bias voltage. Worst case scenario, the EL84 will completely cutoff causing unpleasant crossover distortion. D1 and D2 zeners will prevent this. We generally choose the zener breakdown voltage to be 1-2 volts above the idle cathode voltage. We can also put an LED in series with the zeners for some funky warning that we are venturing too deep in class B.
5. We need only two power supplies: -12.6VDC for the tube heaters and input differential's CCS tail (12V/5A SMPS is cheap!) and +310VDC for the high voltage part. Simple, no? With such symmetrical topology, we can even get away with simple CRC filtering for the high voltage as the CMRR will keep any hum from escaping to the speakers. We can copy Baby Huey's CCS and bias block circuit. Just keep in mind that the input differential is drawing 1mA per side, instead of 0.5mA.
Attached Thumbnails
Click image for larger version

Name:	naga84.jpg
Views:	991
Size:	168.2 KB
ID:	2069  
Attached Files
File Type: asc EL84 NAGA84.asc (8.0 KB, 159 views)
Posted in Uncategorized
Views 632 Comments 0
Total Comments 0



New To Site? Need Help?
Copyright 1999-2017 diyAudio