BH delay schematics or let’s minimize relay contacts bouncing

For this article let's accept an axiom that warming up heaters well before raising B+ is a must for any good behaving thermionic valves based equipment.

The inspiration for this writing actually came from lots of lame designs seen on the Net. The last drop was a reasonably professionally written article describing a headphone amp using SRPP topology. Unfortunately there too in the final schematic the author proposed the following solution for delaying BH at the power-up (ugly smile added by me):

Bad example of a B+ delay implementation. C1=C2=2200uF; VT1 2N6038.

How does the hack shown above work? Not so good indeed.
Once the system powers up the voltage on C2 raises slowly. The voltage on the emitter of VT1 follows it being only some 2 Volts lower. At a certain moment in time it reaches the voltage at which K1 gets activated. So it does. Slowly pulling its contacts together. The contacts touch and bounce because there's just enough force to bring them close. In this circuit we obtain almost as much and as long bouncing as it was at all possible for that particular relay. This bouncing is very undesirable, especially in a high-voltage circuit. Unfortunately it's unavoidable, but we should strive to make it as little as possible - by raising the force that pulls contacts together to its max as fast as we can.
One more negative note about the schematic cited above. At the powering down most of the discharge current from C2 drains through the base of VT1. Probably the transistor will survive. But was it a clean solution? Not at all.

The next variant I picked from the TL431 datasheet by Texas Instruments:

Single-Supply Comparator With Temperature-Compensated Threshold

I had doubts right after giving it a first thought. There's no such thing as a usable comparator without positive feedback. It must either oscillate or produce analog output and not firm yes/no or 0/1 during the input transitions near its threshold.

Then I saw exactly this schematic deployed. It was found in an otherwise perfect collection of "preferred building blocks" for tube amps on a web site of one very respected author. There the author clearly realized the necessity of positive feedback but still saved on it in the final design:

Very doubtful example of a B+ delay implementation

After all I followed what the opinion leaders say and used a so-called "comparator" based on tl431 right in a nice-looking prototype without giving it a try on a breadboard. The failure of that "comparator" was easy to hear immediately: in that particular case the relay was connecting an audio signal, not power rails. Thus that bouncing was obvious right in the loudspeakers - a very uncomfortable thing I must admit.

Attention - heavy contacts bouncing!

It's worth mentioning that tl431 will not let the voltage on its REF pin go much higher than its threshold of 2.5V. Therefore in the schematic above one could use a capacitor rated as low as 4V safely.

After a bit of thinking (finally!) and quarter-an-hour bread-boarding I came up with the following solution. It utilizes complementary BJT's that form an analogue of a three-p-n-junctions structure also known as thyristor.

B+ Delay Schematic - works wonders! 🙂

The two 0.1uF capacitors were needed in order to prevent the schematic from latching in case the power supply rose V+ at very high speed: the "Miller" capacitance served to open transistors a bit and that was just enough to not have any delay. Even though it was happening only on the bench and never in the final design - I left the capacitors there for "just in case".

As drawn this delay schematic works nicely: the relay "clacks" firmly approximately one minute after it's powered up.

As usual there are some drawbacks exist in virtually any engineering solution:

• The proposed schematic is more sensitive to the temperature variations than the one based on tl431 was. But who cares if we still use an electrolytic capacitor for setting the delay?
• This nice power-up delay circuit will de-energize its relay only when the current through it drops to near zero. Again it's fine in the particular application where I used it.

Shall I ever need the two "flows" listed above to be fixed - I'll most probably return to a tl431-based circuit while adding some positive feed-back loop to it.