Not a surprise

What I suspected it was going to happen, it did in the end. Although a bit premature and in a bad time. I’m expecting today a friend to come around for a listening session and having no amp wasn’t an option.

To cut a long story short, the Salas SSHV2 shunt regulator has been playing silly buggers for a long time. Since I upgraded the output transformers and readjusted the bias, it looks like I was operating it at the verge of its abilities. The CCS was running at 80-90mA and somehow the stability of the shunt regulator was compromised. Initially was a periodic lost of regulation during warm up, this created an annoying “pop” now and then,  later I decided to replace it with a new SSHV2 and blew a pair of DN2540 after the regulator failed to set the output voltage randomly. It worked fine on the test bench, however there is something on my system which is disturbing / interfering with the regulator or the regulator isn’t stable enough at the hot operating conditions I was submitting it to.  I have nothing against the SSHV2, in fact, I use it extensively in my preamps. However, I think I’ve found the limit at which it can safely operate. The additional drawback of the SSHV2 is its temperature stability. It’s not great as it drifts when temperature rises.

So the regulator went busted on Thursday evening and I was running out of time. Only Friday was available to fix the amp. Luckily, I was on holidays this week and had the time to fix this, but unfortunately this diverted my energies and time from the 300B amp 🙁

The obvious choice would be a series-feedback regulator like the one I used many years ago for a 45 Amp. It does work but you need to watch out for frequency stability (HF) and also although its PSRR is great, it sounds awful in SE:

The working principle of the above is quite simple and it has been covered extensively by many out there. The pass FET (M1) voltage is set by the error amp formed by Q1. The reference voltage is provided by the zener string D2-D5 which its noise is tamed by the filtering cap C1. Other methods to avoid the zener string are possible a better error amp like an op amp but increases complexity. I have done a very succesful implementation for this in this 600V supply. The output voltage is sensed from the resistor divider R3 and R4 and fed back to the error amp. The Q1 will do the work to stabilise the output voltage at all times. C4 is used to improve the frequency response of the error amp.

In this case, I wanted to stay away from a feedback regulator. I decided to go for a passive regulator / voltage stabiliser based on a cap multiplier circuit and a pass element. The final circuit designed in an hour and built and tested in 4 hours is shown below:

This is not a perfect circuit and is not a fantastic regulator, but it does the job quite well and is robust. The Achilles’ heel is the temperature variation. Let me go through the circuit design first. The voltage reference is defined by the cascoded depletion FET pair (M1 and M2) which sets a precise current through R2. This produces the reference voltage. C1 provides the filtering of this supply. Its location is key at this point as it effect will be multiplied by Q4. Since M1 and M2 forms an CCS, the voltage will rise slowly and linearly at start time whilst C1 is charged up. You may need to play with R6 and the trimpot “Rset” to achieve the right voltage setting range depending your needs.

Q4 is a simple emitter follower with a tail CCS formed by Q3, R4, D1, D2 and R8. This tail CCS has a current set by R4. Just a few milliamps that will drive the gate of the pass element M3. R5 is a simple current limiter which protects M3. Is set to about 150mA. When start to operate D3 will conduct and sink the CCS current out of the M3 gate therefore regulating the output current. The price paid here is that R5 contributes massively to the output resistance of this regulator.

The good thing is that the response of the regulator to input voltage variance is very good (43mV/V) for a HV supply. As I said earlier, the weakness of this circuit is the temperature drift of the depletion FETs.

Despite setting the CCS to operate somewhere around the most stable part of the curve shown below from a temperature perspective (i.e. 500-600μA) the variance is noticeable.

 

At least the simulation shows 100mV/°C drift which results in a 1V or 2V variance at warm up. This may not be a problem in most of the cases when 0.3-0.5% variance is insignificant for the valve amplifier. In this case, this will work in a DC-coupled amplifier and voltage stability is key. A workaround suggested by Rod Coleman is to use a bottle cap filled with epoxy on top of the LND150s. This will add thermal mass to them and minimise the drift. I haven’t tried this yet, but certainly will.

The good outcome this time was that I was able to fix the amp in less than a day. After a 5-10 min warm up time, the regulator achieves 257V reference to feed the 46 driver stage. It provides about 60-70mA plus 7mA consumed by the regulator. After a couple of hours of playtime I can notice a 1V drift which doesn’t impact significantly the bias point of the output stage. It does the job and very well! Sound-wise, given is a passive regulator, I can’t notice any impact at all. It sounds brilliant as before.

Hope this experience is helpful to some of you out there!

Ale