Fullscreen HTML5Med Res (??MB)HTML4

Ultra-LD Mk.4 200W R Power Amplifier: Previ We have been working on a revised version of the very popular Ultra-LD Mk.3 amplifier module. While the circuitry will be very similar, it will be on a smaller PCB employing SMDs for the lowpower circuitry of the front-end, with new small-signal transistors to substitute for those that are now hard to get or no longer made. By NICHOLAS VINEN T HIS IS THE fourth power amplifier module in our Ultra-LD series and the third based on ON Semiconductor’s ThermalTrak power bipolar transistors. While the specifications for this module will be similar to the last, it has been considerably re-designed and there are a number of advantages compared to the Mk.3 module. Like its predecessors, this amplifier module has extremely low levels of distortion (including at higher frequencies), along with a substantial output power capability of 135W RMS into an 8Ω load, 200W RMS into a 4Ω load and substantially higher music power figures. We are using the same output transistors; they’re still state-of-the-art. It’s hard to fault the existing UltraLD Mk.3 module on its noise or distortion performance so while we aim to provide an incremental performance improvement, one of the the main reasons for this new design is to substitute more modern components for those which are quite dated. Specifically, the Toshiba 2SA970 low-noise input transistors used for the input pair are increasingly hard to find and the BF469/BF470 high-voltage transistors are now obsolete. As is the case with so many parts these days, modern signal transistors are available mostly in SMD packages; through-hole components, especially new devices are becoming less common. So being virtually forced to use at 80  Silicon Chip least a few SMDs in the new design, we decided to change the entire front end and some of the output stage to SMDs. Besides better availability and lower prices, there are several other advantages to using surface-mounting parts. Firstly, this allows the small signal section to be much more compact which means both a smaller PCB and less chance of RF and hum pick-up due to shorter tracks. In theory, there may also be a small improvement in performance due to lower parasitic inductance. Also, because the parts no longer have leads which must pass through the board, we can employ a ground plane on the underside. This makes a very effective shield for the input stage so it is far more immune to any stray magnetic fields, whether they are from the output stage tracks, output filter inductor or anything else in the chassis (eg, a mains transformer). Using SMD transistors for the voltage amplification stage (VAS) and its associated constant-current source also means that we can use the copper on the PCB for heatsinking, eliminating the bulky flag heatsinks which we used on those transistors in the earlier designs. Extra features & changes While updating the module, we’ve taken the opportunity to add some features that we’ve been asked for in the past and change some design deci- sions that we felt were not optimal. Firstly, we have added an offset adjustment trimpot to the design. This allows the input transistor offset voltage to be adjusted down to around ±0.1mV. This makes the amplifier much more suitable for driving a transformer with a low-resistance primary winding. The board will have provision for the required output voltage clamping diodes as well. Secondly, the extra diode featured in the January 2013 issue (Performance Tweak For The Ultra-LD Mk.3 Amplifier) is now present on the board. This makes the unit’s performance much better when it is driven into hard clipping; or should we say, less bad. It effectively makes recovery from negative-voltage clipping as clean and fast as that from positive-voltage clipping and thus improves signal symmetry and reduces ringing under these conditions. For this role, we are using an MMBD1401A SMD diode which has a low base capacitance of 2pF at 1MHz. We have also changed the relatively hard-to-get Molex power and output connectors to the commonly available pluggable terminal block type. However we have yet to confirm whether these will give the best possible performance for the speaker terminal connections as we’ve previously encountered issues with dissimilar metal junctions in connectors affecting linearity (see the panel on page 65 of siliconchip.com.au RMS iew the April 2012 issue, in the Ultra-LD Mk.3 Amplifier Pt.2 construction article). However, our testing so far shows that these connectors are certainly sufficient for the power input and are more convenient to wire up than the Molex types. New transistors Of the seven small signal transistors in the Ultra-LD Mk.2/Mk.3 design, six were arranged in pairs: two PNP input transistors, two NPN current mirror transistors and two PNP constant current source control transistors. The new SMD transistors (HN3A51F [PNP], HN3C51F [NPN]) we have specified are two to a package and have virtually identical performance to the 2SA970 low-noise transistors used in the earlier designs. This 6-pin dual package has much better thermal tracking between the siliconchip.com.au The new Ultra-LD Mk.4 power amplifier uses SMDs for the frontend circuitry, resulting in a more compact PCB design. This view shows a prototype version; the final version will have a few minor changes. two transistors. This is especially useful for the input pair: any differential heating will cause a shift in the differential base-emitter voltage between them and thus affect the output offset voltage. With both transistors in a single package, this should be essentially eliminated. It also means that any interference picked up by the two transistors should virtually cancel due to their close proximity. The benefit to the current mirror is smaller but its operation does depend on good base-emitter voltage matching which is a feature of these dual transistor packages. We’ve replaced the BF469 (main VAS transistor) and BF470 (its constant current source) with FZT696B and FZT796A transistors respectively. These are in SOT-223 packages which are capable of up to 2W dissipation with suitable PCB heatsinking. In operation, they normally dissipate well under half a watt, so this is not an issue. Still, it’s desirable to keep them at a stable temperature to avoid changes in performance as they warm up or cool down. Compared to the BF types, the FZT transistors have a slightly higher transition frequency (70MHz vs 60MHz), much higher peak collector current rating (1A vs 100mA), slightly lower but still sufficient voltage rating (180V vs 250V) and dramatically higher current gain (150-500x compared to ~50x). This means that the open loop gain and open loop bandwidth of the amplifier should be higher and in an ideal world, this will result in greater distortion cancellation. We’ve also improved the open-loop bandwidth by replacing the BC639 in the first stage of the VAS Darlington with a BC846, the surface-mount equivalent of a BC546. The BC639 was originally chosen for its voltage July 2015  81 Q10 NJL3281D Q11 MJE15030 BD139 MJE15031 NJL3281D + CURRENT FLOW DURING POSITIVE EXCURSIONS Q7 Q9 Q12 NJL1302D Q13 NJL1302D Q8 CURRENT FLOW DURING NEGATIVE EXCURSIONS + FROM POWER SUPPLY L1 TO SPEAKER rating of 80V; its relatively high collector current rating is not important since the collector current is limited by a series resistor. The BC846 has an identical collector-base voltage rating and only a slightly lower collector-emitter voltage rating of 65V but it has better linearity and a much higher typical hFE of 200-450, compared to just 40-160 for the BC639. Preliminary testing shows that this new amplifier is capable of producing very low distortion figures (well below the limits of our analysis equipment at some frequencies and power levels) but we have not finished tweaking it yet. At this stage, we are simply not able to quantify how good it is. Stability & compensation While greater open loop gain is desirable as it can result in better distortion cancellation via global feedback, it does come with challenges. We’ve had to go to greater lengths to stabilise this amplifier compared to previous revisions. Due to the very high open loop gain, we’ve had to add a capacitor across the VAS current-limiting resistor (in series with the collector) to reduce local feedback due to the Early Effect (where gain changes to some extent with collector voltage). We’ve also had to use a slightly more complex VAS compensation scheme, similar to the 2-pole version used in 82  Silicon Chip the Mk.3 amplifier but with an extra capacitor across the ground resistor. We’ve also incorporated components to allow for high-frequency roll-off within the feedback loop. Specifically, this consists of a step circuit, ie, a series combination of resistor and capacitor across the main feedback resistor. We’re also looking into tweaking the values used in the output RLC filter. This filter has a dual purpose; it acts as a Zobel network, which is a type of snubber at the output that helps stabilise the amplifier and it also isolates any extra capacitance in the speaker and its wiring from the amplifier, which could otherwise cause enough phase shift in the feedback loop to trigger oscillation. But you may recall from our articles on the Ultra-LD Mk.3 design that we discovered that the magnetic field generated by the filter inductor also interacted with the magnetic field caused by currents flowing in the PCB itself and thus its value and orientation affected performance. With this new design, the magnetic loops are tighter and so this should be less critical. We’re hoping that this means we can reduce some of the filter component values (keeping them sufficiently high for stability) and in the course of doing so, also reduce the inductor resistance and thus the amplifier’s output impedance. This Fig.1: the current prototype board with the high-current flow paths shown for lowfrequency signals (ie, at frequencies where onboard bypassing capacitors do not supply much current). Since many of the current paths overlap and flow in opposite directions, this provides a high degree of magnetic field cancellation thus minimising inductive coupling between the output and input stages. Note that the output transistor emitter resistors are directly under the fuseholders (ie, mounted on the bottom of the board). should improve its damping ratio and possibly also reduce the possibility of the inductor’s magnetic field interacting with anything else in the amplifier. At the time of writing, this is still being investigated. Magnetic cancellation As you may be aware, all of our lowdistortion amplifier PCBs have been laid out carefully in order to avoid the magnetic fields caused by high ClassB currents from interacting with the rest of the components on the board and injecting distortion. This is an especially difficult problem because of the fact that the Class-B currents are essentially half-wave rectified versions of the output waveform. In theory, the Ultra-LD Mk.4 has the best magnetic cancellation of any of our designs, as the main Class-B current paths are directly on top of each other. In other words, when current is flowing into the board along one layer of the PCB, the same current flows along the other side of the board in the opposite direction and thus the magnetic loop is only as wide as the PCB is thick (~1.5mm). This arrangement is shown in Fig.1. Current flowing from the positive supply to the loudspeaker via the upper pair of emitter-follower output transistors is shown with red and magenta arrows, while the equivalent flows for continued on page 87 siliconchip.com.au How Far Do You Go With Restoration? Old valve radios present many wellknown problems for restorers. These include leaky or shorted capacitors, high or open-circuit resistors, dead or lowemission valves, open-circuit transformer windings, battery corrosion and noisy volume control pots. My own experience with all kinds of radios shows that while a set may appear to “work”, a thorough examination often reveals defects that detract from its intended performance. Now add a novel type of deterioration for early transistor sets: leakage in (mostly) germanium transistors and capacitors that allow a set to work “pretty well” but not up to its original specifica- tion. Both the 78T11 and the Pye Jetliner that I recently restored suffered AGC faults due to leakage (in a transistor and a capacitor, respectively). Often, a restorer won’t bother to troubleshoot further if it works OK on local stations. Indeed, it’s up to the individual to decide just how far to go in the restoration process and whether they want the set to perform to its maximum potential. Some things to consider include: nostation current drain, distortion and current drain at full output, sensitivity, freedom from oscillation (or “howling”), the AGC action and the audio frequency response. level “wip-wip-wip” oscillation on all volume settings. An oscilloscope check showed a trace much like the parasitic oscillation that’s sometimes seen in high-gain audio and HF/VHF RF power amplifiers. The culprit was C17, the main audio bypass capacitor. A faulty AGC bypass capacitor (C9 in this set) can cause audio oscillation. It certainly did on the TR-1 set that I restored (see SILICON CHIP, September 2012). pressively, with a frequency response from the volume pot onwards of about 45Hz to 7kHz (-3dB points). By contrast, the response from the aerial terminal to the output is about 40Hz to 2kHz. The distortion (THD) was well-controlled: 1.7% at 10mW, 3.5% at 50mW and 5.2% at the onset of clipping (160mW). At full output (about 200mW), the THD rises to some 13%. Performance The supply voltage for the set is nominally 6V (4 x 1.5V cells). When the supply is down to just 3V, the maximum output is around 40mW for a THD of 5%, falling to about 2.6% at 10mW. All in all, the Stromberg-Carlson 78T11 is a solid performer and is an important example of early Australian transistor radio design. If you have one, get it out and restore it to full working order. Describing a set as being “very good for its age” can be a cheap shot but this set really is a good performer. In fact, it matches the excellent Philips 198 – it’s pretty much the same design but with better audio response according to my test results. Getting down to actual figures, at maximum gain, it needed field strengths of 30µV/m and 35µV/m for 50mW output at 600kHz and 1400kHz respectively – but with corresponding signal-to-noise (S/N) ratios of just 7dB and 5dB. For a 20dB S/N ratio, the sensitivity at 600kHz is about 100µV/m and at 1400kHz about 150µV/m. This set’s AGC action has a very early onset, so delayed AGC would have given an even better figure than my test results. As for selectivity, this measured ±1.5kHz at -3dB and ±11.5kHz at -60dB. The AGC held the output to a 6dB increase for a signal increase of 34dB and the set needed some 40mV/m in order to go into overload. Distortion measurements The audio stage also performs imsiliconchip.com.au Supply voltage Further Reading For schematics, see Kevin Chant’s website: www.kevinchant.com/uploads/7/1/ 0/8/7108231/78t11.pdf www.kevinchant.com/uploads/7/1/ 0/8/7108231/79t11.pdf For Stromberg-Carlson’s Australian history: www.radiomuseum.org/dsp_ hersteller_detail.cfm?company_ id=7578 Many references also exist for the US parent. Among them, see: www.radiomuseum.org/dsp_ hersteller_detail.cfm?company_ SC id=751 Ultra-LD Mk.4 Power Amplifier Preview . . . continued from p82 negative output excursions via the other pair of output transistors are shown in blue and cyan. During positive output excursions, current flows from the positive supply input connector to Q10 and Q11 (the NPN output transistors) and then to the output filter (L1, etc) and the positive speaker lead, via paths that overlap almost completely. Return current from the black speaker lead to the power supply ground connection completes the loop. The part of the loop where the current paths diverge is the section around the RLC output filter and this is difficult to avoid. The negative path through Q12 and Q13 is shorter but otherwise similar; again, the only real loop area is through the output filter. In fact, since the positive and negative paths converge at the top end of L1, the current in this section of the loop is not half-wave rectified (ie, it is effectively just the output current) and so it’s far less of an issue in terms of radiation and distortion as it lacks the sharp transitions of the Class-B current. Note that the 0.1Ω emitter resistors for the power transistors are 3W SMD types mounted directly under the respective positive and negative supply fuses. Besides being far more compact than the previously specified 5W wirewound resistors, the SMD types are non-inductive and their positioning gives much better field cancellation. L1 will generate its own magnetic field due to this current flow and this is why its winding direction and the number of turns are quite critical; if orientated correctly, the field generated by the output current flowing through L1 will at least partially cancel with the field generated by current flowing through the loop formed by the PCB tracks that was explained above. This does not consider current supplied to the output from any of the on-board bypass capacitors, however their paths have been designed to be relatively tight loops as well. Acknowledgement Thanks to reader Alan Wilson for suggesting many of the part substitutions that we are using in the new design and prompting us to investigate some of the other changes we were considering for our next amplifier. SC July 2015  87