Skip to content
May 5, 2013 / consort3

Darlington Pt 3 Serendipity

The picture at the end of Darlington Pt 2 showed the device heatsinks of the SAP16 version of the darlington amp. The SAP heatsinks are very impressive when you compare them with the smaller TO247 devices. They make up a wall of heatsinks leaving a problem of where to mount the thermistors. I thought of bonding them to the collector lead (especially as there is an inverted U channel in the moulding which could have been designed for it) So now a 7 lead device, I claim the record for the number of legs on a transistor! This makes the design independent of the thermal mounting arrangements. The previous mounting of the thermistor on the heatsink relied on low thermal resistances using mica washers and heatsink compound, which though very effective is messy.
I have tried 3 designs, lite, medium and heavy. The lite uses BDX33/34 output power transistors, the medium uses TIP142/147 and the heavy uses SAP16P/N. I used Veecad the stripboard layout tool. Using this and the paper templates cut down the errors. The main error is leaving a wisp of connective track when you cut the isolation holes in the strip. Best to use an eyeglass to check. Although a netlist import is available with Veecad, I did not bother. However it is easy once you have a layout to tweak it for different transistors.
When trying a board for the first time I use a bench supply set with a 100mA current limit and slowly increase the voltage monitoring the current. This catches any problems as typically any fault causes a massive increase in current!
I got around to measuring the distortion on the lite version and was pleasantly surprised that the high frequency distortion was below 0.1%. The spice simulation indicated that the high frequency distortion would be about 0.5%. The other pleasant surprise was that it was at a low quiescent current (or bias as it is known) of 15mA. Both figures are better than the TDA2030 so this would be a discrete component upgrade to that device. Profusion PLC and Tayda now sell that device for less than £0.50 so the design is not cheaper.
There is no DC connection between the op amp output and the actual output. The op amp DC output voltage is a measure of the balance between the bias current of the NPN and PNP output transistors. If the output was at zero volts this would imply balance. If the NPN has more bias than the PNP the output will be negative and vice versa.
Using this reasoning I found that the resistors R13 and R30 for the PNPs should be 130 or 140 ohms rather than the 120 ohms for the NPNs. The PNPs need slightly more voltage than the NPNs to turn them on, it seems.
The other idea is to bias the output devices to just below their threshold on the opamp quiescent current so that the control current through R10 and R28 is low. This has the desirable effect of making the overall bias less dependent on supply voltage. It is just about good enough to dispense with the series darlington regulator transistors. At the moment I run them, but without the zener defining the voltage. Also I found it more convenient to use 100k variable resistors for R10 and R28 rather than select on test (S.O.T) resistors.
With the high cost of toroidal transformers, I have tried using switch mode supplies and this helps since they produce a defined voltage and they are cheaper, also the series darlington regulators are not needed. I found that when the series pass regulator transistors actually regulate, the heat dissipation upset the delicate thermal sensing of the amplifier. Perhaps they need to be thermally isolated from the output darlingtons.
The design in Part2 seems atypical in that the ST TIP142 needed a 560 ohm base stopper. The other designs are using 12 ohms. The device may have a good frequency response because its SOA is nothing to write home about. The 2 being usually mutually exclusive, an exception being Sankens triple diffused devices.
The thermistors which temperature stabilise the design are bonded to the collector leads of the output devices using epoxy adhesive. I fix the thermistor position by temporarily soldering its leads to the transistor leads. I then turn the device upside down and using a cocktail stick run a bead of adhesive onto the collector lead. The adhesive runs down onto the thermistor. Hold the thermistor while bending its leads as the joint will not survive that mechanical stress.
With the output transistors on a heatsink , to check the thermal stability, I use a bench supply to monitor the current. Then apply full power into a 8 ohm load for a couple of minutes The current when under load should not increase much with time. Remove the load and the current should go back to the original or up to plus 20% straight away.
The lite version uses 4 off one ohm resistors in parallel to create a non-inductive 0.25 ohm emitter resistor.
Working on other circuits (the twin T oscillator) with the NE5532/4 I found that they will work at low voltages. This may explain why there is little or no switch on/off thump with this circuit as the working condition is swiftly established at switch on and then hangs on until the supplies are almost disappeared at switch off.

The above was going to be my last words on this theme but an Audiopro circuit was posted on Diyaudio which showed an op-amp based push-pull Vas.  I believe Garza in 1973 published a similar circuit although his has a quasi complementary output stage. The Audiopro circuit enables you to have a conventional Vbe multiplier for thermal compensation at the expense of 2 extra transistors compared to my offering.
Update:
I just realised that the Audiopro circuit had probably been developed from the Kuroda circuit with the V-fets output devices replaced by a Darlington type circuit. Notable additions are the 220pF caps on the push-pull drivers.
A development of this is to replace the drivers with current mirrors discussed in this thread:
http://www.diyaudio.com/forums/solid-state/255236-opamp-based-power-amp.html

The amplifier in the thread is based on Jan Diddens Pax amplifier featured in Elektor 2008 That amp appears to be quite close to the Alexander amp.

Garza circuit

pushpullquasi

Audiopro circuit
pushpullvas

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: