White Cottage

Having finally assembled the board for the 2nd generation Valve Analyser (based on a single Arduino Mega) I started having a play with it. My initial attempts to coax it into life were somewhat hampered by the fact my USB cables were duds – I tried two of them so concluded at first that the problem was with the board! However, equipped with a working cable I was soon up and running and the Mega is certainly a very capable board. Moving to the Mega offers more flexibility than the Nanos because it has more PWM outputs and also more ADCs – this enables:

• The use of PWM to manage the current through the charge and discharge resistors
• Additional anode/screen current sensing ranges to improve measurement accuracy for signal (rather than power) devices

For the latter improvement, I needed more rail-to-rail op amps for the current sensing circuitry so I opted for the LM6144 (there aren’t many options in DIP packages) – I sourced these from China but was hopeful that they would be genuine as they had different batch numbers. However, the resultant measurement performance was awful* so I strongly suspect that they are just some very cheap quad op amp. On reflection, it occurred to me that, rather than using a hard to find rail-to-rail device with a single ended supply, it would be better to simply create a negative supply such that the op amp would have no difficulty getting its output all the way down to 0V. The new board has a buck converter (adjusted to give a supply voltage of around 7.5V) on it to power the Arduino from the 24V heater supply, and it is a simple matter to use a MAX1044 or ICL7660 to invert this. I had a few other minor errors to correct so I have now created a second iteration of this board and am awaiting production. With the split supply for the sensing amps I can now use TL084 devices which are cheap, easy to find and have a very low offset voltage, thus making them ideal in place of the troublesome LM6144s.

* With a bit of subsequent testing, I got more sensible performance by changing the configuration of the amplifiers. Initially, I had simply used the current sense voltage to drive to inverting amplifiers (upping the input resistor on both to minimise the load). I then tried making the second amplifier non-inverting and putting it after the first amplifier. With gains of 2 and 3 for these stages, the sensitivity becomes x2 and x6 – which corresponds to a full scale of around 60mA through the sense resistors for the original configuration and 20mA for the new more sensitive range. While I remain unconvinced about the amplifiers, this approach certainly works in terms of increasing the measurement resolution and the resultant measured curves for a 6N2P were much smoother than with the Nano based board.

Additional improvements from the first version include:

• 1K protection resistors on the ADC inputs (the ATmega2560 has diode clamps onboard but the resistor then provides ample current limiting without compromising the performance)
• PSU decoupling on all the amplifiers
• A low-voltage sense network run into an ADC (so that the Arduino can confirm that the heater circuit is adequately powered)
• An optional bleed resistor (>470k) across the HT to allow the HT filter capacitors to discharge when not in use

Fingers crossed, after Christmas I’ll be able to try the new board…