Push-Pull Class A Amp using type 5687 dual triodes!!!

Last Update - 3/11/96

Here is the schematic for one channel of my amp:

Introduction

This stero amp will give you 4-8W/ch of zero feedback triode power for less than a pair of Chinese 300B's ...if you know where to shop : )

For those of you who are unfamiliar with this lil' tube here's the poop:
The 5687 is a 9-pin miniature twin triode (a la 12ax7, 12au7, etc - but with different pin-out). Its specs are quite interesting:

mu = 16
gm = 5400uS
rp = 3k
plate dissipation: 4.2W each unit or 7.5W total (3.75W x 2)

After playing with the curves (so kindly reprinted in The Parts Connection Catalog) I realized that you could get some juice out of them (these tubes were apparently designed for high current pulse application - hence the high perveance).
Parallel the two sections and you start to approach the rp of a 45. The higher mu and gm mean that its easier to drive.
No way it could sound as good, tho'!?
Without getting my hopes up - I decided to give it a go!

Design Development

At first I tried to squeeze as much power as possible out of the 5687's. I ran the a breadboard circuit from a 350V external ps at a standing dissipation of approx. 3.5W per triode section (315V @ 11mA). I used cathode bias for stability (and so i could use a larger grid resistance). At first I used 220k grid resistors but soon had to lower them to < 100k to avoid thermal runaway. This worked pretty well, even with 12ax7's driving this grid load.
Just to test the theory, I decided to use one half of each 5687 as a cathode follower (direct coupled to output) to drive the other half (thus buffering the 12ax7 input stage). Measurements suggested no real improvement in drive signal linearity. Only noted the requisite decrease in output going from parallel output triodes to singles (not so significant given the output high load resistance (12k p-p).

---

Design Details

Input/Driver stage

The input stage is comprised of a both halves of a 12AX7 dual hi-mu triode in a differential amp configuration with a 1ma constant current cathode sink. Once again, I'm using field-effect (constant-current) diodes for simplicity. The current source/diff amp approach was chosen for good power supply rejection. In this way I could simplify the power supply design and filtering requirements. The constant current diode cathode is returned to a -18V supply (full-wave rectified a small 12V xfmr) instead of ground in order to avoid the non-linearities about the pinch-off region (Vpo ~= 1.5V).
The stage gain is about 20 times into a 100k following grid.

Output stage

Parallel Push-Pull 5687 sections using common cathode resistor bias.
This means one tube per side (push-pull). I saw this as a neat alternative to triode connected EL84 (same number of tubes). Of course, the 5687 draws 0.9A @ 6.3V filament current!
In the final version of the amp my B+ was only 265V. Cathode bias using a common 250 ohm 1W resistor yielded -12V bias, giving 12mA per triode section, and about 3W plate dissipation. This lower plate dissipation allowed me to increase the grid resistors to a more acceptable 100k.
Output Transformers are stolen from a Fisher console amp using ECL85's. Impedance ratio is 6K:4ohm, but I'm using 8ohm speaker for 12K plate to plate load. I calculated "optimum" load for paralleled 5687's to be about 3k plate-to-plate. Hence, power is down considerably from what is could be. For example, 8.5W with 3k load vs. 6.0W with 12k load. I'd like to try a 6k:8ohm OPT which would give 8.0 W or so - but alas I can only work with what I got :(

Power Supply (not shown)

Power supply is as simple (and cheap) as it gets. A 110:220VAC isolation xfmr - to a bridge rectifier (1N4007's) - to a 200u/330v photo-flasher - to a 75ohm WW Resistor - to another 200u/330v photo cap. That's it!
B+ is about 265VDC and only drops to 257VDC or so at max sine-driven output power.
Ripple at B+ is a little less than 200mVpp. This results in less than 10mVpp of output noise/hum thanks to push-pull output topology and differential driver.

Parts Selection

El Cheapo Grande!
I bought a mess of 5687's from Angela. The buck-a-piece blowout jobs were sold out so i *sprung* for the good 'uns. (Philips) @ $20 per 10 pcs. I picked up the isolation transformer for $3.50. Tube sockets are all the cheezy wafer type. Aside from the scrounged OPT's, the single most expensive part was the Radio Shack 12V-0-12V / 3A filament xfmr at 10 bucks.

Construction Details

I mounted all xfmrs in a line along the longer edge of an old 14 x 9 chassis bottom. I pop-riveted the tubes sockets to a 5.75 x 7 galvanized "TECO" plate ($1.99 at Home Depot lumber section). The sockets' mounting holes lined up perfectly with the TECO plate perforations. I mounted all components between the socket terminals and strategically placed terminal strips (from Radio-Shack). This Assembly is attached to the lower chassis via standoffs. A wooden cover (with the appropriate cut-out for tube protrusion) is on the drawing board.

Performance

Calculated power output should be about 5.2W for 12k load (8ohm spkr). Measured output is 5W at clipping. Noticeable distortion appears at around 15Vpp (3.5W) or so.
Overall gain is about 18dB.
Input sensitivty is about 800mVrms for full power.

---

Tweaks

I tried various input tubes: 12ax7, 5670 (cute!), 12au7, and 12av7. only the 12ax7 really have enough gain to drive into hard clipping from a CD output. The 12av7, however, did a good job of driving the amp as loud as I'd ever need (thru my 90dB speakers). It had the added attraction that it fit in with the Econo-Mojo motif. Angela was selling them for less than a buck-a-piece :) . Except for the 5670, all were interchangeable. So I'm sure some 'sperimenting is in the cards.

Listening Impressions

So far this amp sounds very good. I can't stop saying "No Way!". Tighter-than-you'd-think bass. Too good to be true! Somebody wake me up, I must be dreaming! Or maybe this lil' amp will wake someone else from my Nightmare - spending $1000's on an amp.

Email your comments/suggestions/critiques/questions to Bob. D. at robert.m.danielak@lmco.com

This page, and all contents, are Copyright (C) 1996 by Bob. D.