Stan Curtis' ETI System A class A amplifier
Stan has recently created a website which covers and has links to PDFs relating to this design a link to which is provided here.
The last Lecson power amplifier, the AP4, was a massive Class A monobloc with an output of 70 watts and a much higher output in Class AB. Curtis at one time had six of them in a tri-amped active crossover system. Unfortunately Lecson failed just as the first batch was about to be delivered. He bought the assets of Lecson and sold the completed amplifiers to enthusiasts. The design was then published in "Electronics Today International" as the "System A amplifier" and the remaining parts were sold as kits. The publishers told Curtis that it was their most popular article ever and it was re-printed in different digests four or five times.
The high current slewing conditions when transient signals are fed into difficult loads can produce higher order harmonics as some output stages 'catch-up' with the demands made. These tend to 'harden' the overall sound. The gradual non-linearities in class A designs where the output devices continuously operate in the linear collector region are held to produce lower order (2nd and 3rd) harmonics commonly found in musical instruments which are easier on the ear. Because of the high quiescent currents involved, quite substantial hardware arrangements are required for outputs of 10W or more, compared to class B operation.
ETI 1981 & Electronics Digest Winter 1985 (pdf)
"Hello, I built two of these 60 W Class A amplifiers... Were very tricky to settle fine! For the best of my knowledge, they were also shown at an audio convention in Paris under the "Jeremiah Braithwaite" brand (never heard from them later !), and also was turned into a powerful Class AB under the Mission brand (was the 772 model, I think...). Hope this serves! Best regards, Jean-Bernard Lemaire"
When quiescent, a dissipation of 240W per channel was normal. The transformer might show a >30°C temperature rise whilst the bridge rectifier and smoothers would certainly be warm. Ventilation of the case could then be considered given that half of the heatsink fins were radiating inside the amplifier's sealed enclosure. A number of constructors using this type of heat-sink - Redpoint NV115-1 - appear to ignore correct mounting methods, another example can be found here. Some may see this as odd, given the financial outlay made compared to the rest of the amplifier. Further drift was noticeable when the top panel was fitted, thus effectively sealing the enclosure.
Another approach could be used to reduce the enclosure's internal temperature (normally 40-50°C), thus increasing the longevity of the amplifier's components, like the large computer-grade electrolytics, for example. As with so many class A designs seen, there is almost an assumption that everything must run hot.
An alternative enclosure design could employ a sub-frame to support the major components, whilst the upper and lower panels could be replaced with perforated sheet-metal thus allowing air-flow through the unit. One design considered mounting the heatsinks vertically in two columns, bridged by the PCB with the power supply below. Even a bridge rectifier heat-sink bolted to the outside of a rear panel would reduce the heat input by that of a small soldering iron. Thought has been given to a re-design using TO-264 MJL3281A / MJL1302A output devices with large area comb-type heat-sinks forming the side and front panels, eg;
Using output devices with a greater gain bandwidth product, say with an fT of 30MHz instead of 2MHz, it may be possible to reduce the quiescent current without affecting performance (the MJL3281A achieves a typical 2MHz GBP with an IC of 250mA). Additionally, a PCB redesign would enable direct connections between it and the bias and output transistors reducing problems associated with the comparatively long lead lengths found between them in the original. The separate feeds to each output device from the main smoothers could be retained, but each could be decoupled by it's own electrolytic and ceramic capacitors.
At the same time, mounting the PCB as shown in the original article subjects the small-signal transistors to the heat rising from the drivers and output transistor's emitter resistors. Inverting the PCB, or mounting horizontally if room permits, helps to reduce drift. A PCB redesign could use perforations, much wider tracks and put some transistors in closer proximity to others, eg; input pairs x-x and y-y (see below). The thermal compensation (z) could have been improved. This transistor (MPSA06) has a TO92 case, not a TO220 as some thought from the supplied wiring diagram.
Problems with a 'noisy' (but inaudibly so) offset appeared to centre on the quality of the offset pot itself. The prototype used open cermet for the quiescent adjustment and an open carbon for the offset. It should be noted that a carbon skeleton preset pot may exhibit 10 times the thermal drift of a comparable cermet type. Enclosed multi-turn wirewound types are far better. One always felt that the lack of local smoothing and RF de-coupling on the supply lines was worthy of attention, perhaps a ceramic tying the output device's collectors to their respective heatsinks could be considered.
Another constructor/user (Lindsay Haworth) writes;
"... the issues were as follows:
1. We found that the ripple on the Output from the AMP was about 10mvolts with a strange sine wave (Amost sawtooth) which we think was the onsett of instability. It would make rinigng noises I had one do that during set up and it scared me, if I went near it with a test probe it would start, it was waiting to take off and become an Oscillator.
2. The wiring to each output device was way to long and we think it contributed towards the instability.
3. The versions we made shared the same transformer output secondaries so maybe there was some comprimize to the sound quality but larger or multiple smoothing caps up to 100,000 made a huge differnce in producing that V8 sound.
4. The 3 amp quiesent current is essential, it degrades below 2.75amps, I ran one at 1 amp and it killed the sound stage and dynamics of the sound.
5. I agree that the PCB needs laying out to reduce point to point wiring to allow side mounted Heat sinks, but that may mean reducing quiesent current which is a negative affect to get rid of that heat.
6. Dedicated toroids with seperate wiings for the front end perhsp with decent regulators weel decoupled would really help.
7. Actually the Meridian 105 100watt AB amps are virtually the same design except being biased into class A but there were known to be fire crakers and unstable, the 103, 35 watts PA was not. Its all in the layout.
Unfortunatly I let all my notes go many years back on these amps but I remeber some of the best Jazz I ever played via an LP12 was epic on this PA."
A mono-block format eliminates cross-talk and intermodulation problems, apart from the weight/bulk considerations that would be involved in a 'stereo' unit (12 heat-sinks, 2 500VA transformers, 4 Sprague 36D 15mF 50V elects, etc). The input filter's bandwidth is below that of the open-loop gain so TID is non-existent. To avoid 'colouration', protection circuits were omitted, although the inclusion of a soft-start is recommended, as is the use of a robust speaker load (5-10A output fuse), the only other protection in this design, apart from a thermal cut-out in the primary's live feed, being the mains fuse. The author's preference was to remove the output fuse and instead fit two in the supply rails between the output devices, smoothers and PCB. The transformer was not screened and no mains filtering was provided, the mains and signal inputs lying side by side.
Notwithstanding perceived sound quality considerations, it can be seen how easily class AB operation (for outputs of say >10W), in terms of expenditure on energy and hardware, was viewed more favourably in commercial terms, some class A designs offering efficiencies of 10% or less. The high output current capability being intended for 'awkward' loads, experienced constructors can assess class AB operation by reducing the quiescent current and, if necessary, by increasing the gain (which reduces the power bandwidth) or the amplitude of the input.
Kits for both the PA and preamp were offered by Jelgate Ltd (Offord Cluny, Cambs) in the 'Buylines' of the original ETI articles. Versions labelled 'Jeremiah Braithwaite Esq' from Braithwaite Kits, at the same address as Jelgate Ltd, appeared briefly later (Sept,'81). This company was owned by Curtis and Mike Harris and produced the RA-14 (mains-powered) and ST-17 (battery) MC preamps, Lecson versions being produced under licence.
A photograph in the Braithwaite ads showing two power amps, one above the other, with the preamps on top led some owners to assume that this was the recommended operating position, whereas the original intention was for the amplifiers to sit close to their respective (low-impedance) speakers connected with short runs of thick (30A) cable. The very compact preamps could then sit (reasonably) anywhere, making for a very unobtrusive, but powerful, arrangement thus aesthetically contrasting with Curtis' earlier Lecson AP3. Some users sat their speakers on top of each amplifier. With this on thick carpet the case was effectively insulated thus 'cooking' the interior giving rise to premature failures, especially of the electrolytics.
The matching minimalist preamp (copies of scans of original article available on request) used a modular approach to meet individual needs and 'discrete component operational amplifier's using selected devices. A shunt feedback arrangement was used for the RIAA stage. In the given PCB layout C109 (left RIAA eq) is shown connected incorrectly to the MC supply rail and the right MC supply decoupling cap is missing. Other constructors reported problems including oscillation with the volume control set at zero (return earth sides via screen to PCB and not as shown) and another that a PCB supplied had two tracks joined which needed cutting and a link added, so caution is advised. A tape output selector switch can offer flexibility and a conventional balance control with fixed output stage gains will reduce problems with a noisy pot. Any signal path switching should be via hard gold-plated contacts and not silver-plated brass. A separate power supply was intended to power other modules including a parametric equaliser and 'two other blocks still under development'. Does anyone have information on these?
Contact me at firstname.lastname@example.org
especially if you want additional content to this page
or if you find any links that don't work. Don't forget
to add the page title or URL. Take care!
Back to index, sound, tips or home.