Absolutely nothing to do with audio, but a great read for the techies.
This performance report was pulled together in the late 1960’s by Walter Haeussermann , one of the German rocket scientists who went to the US after WWII, where he settled and went on to have a spectacularly successful career with NASA. He was a key member of Werner von Braun’s team that made the 1969 US moon landing a reality. Just take a look at the section on the control computer they used to guide the beast . . .
There is a persistent assertion by small group of amplifier design practitioners that the diamond buffer input stage of CFA audio amplifier operates in class B (or AB) mode wherein there is a hard ‘handover’ between each half of the diamond buffer input stage as the input signal passes through 0V. The most recent claim of this kind is in the June 2017 issue of AudioXpress by M. Kiwanuka .
The short article below explores this subject and finds that the claim is wholly incorrect: the front end of a CFA diamond buffer input stage operates well within the class A region for all known audio signals
This is the famous analysis of class B amplifier cross-over distortion by the then head of HP Reasearch Labs, Dr. Barney Oliver, published in the February 1971 edition of the HP Journal. The bias current Iq for a class B emitter follower amplifier is shown to be approximately Iq = .026/(Re+re+(rb/hFE)). In practice, – the ideal value ending up somewhere between a 13 to 26mV drop across the output transistor emitter degeneration resistors. This paper provides the theoretical underpinings for that relationship. So, when you hear about the ‘Oliver’ voltage, you know where it came from. You can read more about Barney Oliver here and here
I recently (2017) had a recurrence of the problem on another high power design – some pictures are shown below. When I went back and looked at the notes above, I realised I had not followed my original advice, and the problem had returned to plague me – clearly a case of ‘those that fail to learn from their mistakes are condemned to repeat them’.
This is what you get when you place the scope probe on the emitter of the cascode transistor. The probe capacitance is probably contributing to the problem and causing it to break into oscillation, or it may be increasing the level of existing oscillation. Either way, its not acceptable, and especially so if you are trying to design a circuit to deliver single digit ppm distortion performance.
Here it is with the time scale expanded:-
If you want to prevent or limit the probe from affecting the circuits behaviour, one trick is to look at it with a 10x probe – the input capacitance is much lower. Another option on 1x, is to place a 50-100 Ohm resistor in series with the probe – this helps to isolate the probe capacitance although you will still have some attenuation because of the scope probe and scope input capacitance. Use a 1206 surface mount device and solder it upright on the node you want to probe. Note that a 1x scope probe input capacitance is about 50 pF//1MEG Ohm and a 10x probe is 15 pF//10 MEG Ohm.
Its very important to note that you can form Colpitts oscillator structures in the base, emitter and collector circuits of transistors. Small signal audio transistors have fT’s of 100 ~ 300 MHz so all you need is some inductance (on a PCB this is often in the 40-60 nH range corresponding to 4-6 cm trace lengths) and then capacitive coupling (10-30 pF – layout dependant) from each end of the inductance to a non-inverting terminal on your amplifying device along with some gain. Below is a screen shot of three circuit examples. They all show HF instability and oscillation to some degree with 10’s to 100’s of mV at 20 to 100 MHz frequencies, but with some value combinations, it is quite easy to get volt level HF oscillation. As you can see, the LC networks that lead to problems can arise across any two terminals.
Although the PCB traces in a conventional analog amplifier are unlikely to be long enough to qualify as antenna’s at the frequencies mentioned, you will still couple a lot of this garbage capacitively into other small signal parts of your circuit. As noted above, if you are working on high performance audio circuits, problems like this will quickly put paid to any ppm or sub-ppm distortion aspirations you may harbour.
The high voltage PNP MMBT5401 and its NPN counterpart the MMBT5551 are often used for level shifters and feature an fT of 100 MHz to 300 MHz and a Cob of 6pF – they are fast and in the cascode configuration will easily oscillate if the conditions are right.
The following preventative measures (not an exhaustive list) provide a good starting point:-
Place a 470 to 1k SMD (1206 or 0805) resistor as close as possible to the base of the cascode transistor. This lowers the Q of any inductance (i.e. ‘dampens’ it) in the base circuit and swamps any -ve resistance reflected into the emitter.
In some cases, a SMD ceramic capacitor from the base of the cascode transistor to ground may help. I’ve found values between 10nF and 100 nF work well. Do not use film or anything else exotic – XR7 dielectric rated at 3-4 times the voltage on the base is about right. The capacitor ESR is also part of the fix.
Make sure the overall loop area from the cascode base reference voltage to ground and the driver transistor in the emitter is small. If not, you will simply be adding inductance in the base circuit and will exacerbate the problem – loop areas have to be kept small to minimize inductance.
Following on from (3) above, recall that the output at the cascode transistor collector is a current, so you can run fairly long traces from the cascode collector to the next part of the circuit – typically a common emitter stage referenced to the supply rails. However, you must minimize any capacitive coupling from the cascade collector circuit to its emitter – the best way to do this is through attention to layout.
If the signal currents are low (1~10 mA), the propensity for the cascode circuit to break into oscillation can be further reduced by inserting a resistor of 100 Ohms to 1k in the collector of the cascode, located as close as possible to the device. This technique also helps by the way in emitter followers or beta helpers.
If your circuit currents are low enough to allow, insert a low value resistor (100~200 Ohms) in the trace close to the cascode transistor emitter – this will help reduce the Q of the trace inductance.
Pay attention to layout during the design stage – keep the cascode, driver transistor and associate circuit compact and with short traces. Keep loop areas small.
One final point about using Zener diodes as the reference voltage to the base of the cascode transistor. Zeners above about 7 V generate a lot of broadband noise right up into 100’s of MHz. Without filtering, damping and careful layout as described above, this noise can promote instability in cascode circuits.
This article was published in October of 1952, when Baxandall was 31 years old, the design having already won him a prize 2 years earlier in a competition. His design basically relegated other tone control circuits to the scrap heap, although cheap passive, or non-inverting circuits still persisted well into the 1980’s, especially in low cost mass market products from Japan. Of course, now tone controls can be implemented digitally, but for the analog type, this remains the gold standard.
Stan Curtis was involved in both engineering and management roles in companies like Quad, Cambridge, Rotel (a family owned Japanese brand) Mission and Lecson to name a few. He is primarily an audio business consultant and remains highly respected and a noted product designer within the sector in his own right.
This is his 60 Watt DIY class A amplifier published in ETI in 1985 which has been referenced on numerous web sites and publications over the last 25 or so years. It’s a rather complex design, but, 60W of class A power translates to a BIG amplifier with significant standing dissipation.
A Design Idea of mine that appeared in EDN around 2000. The article appeared with the title ‘Lost Cost . . . ‘ – seems someone got a bit mixed up. The idea has been referenced in a few papers on the web – always good to see!
Here’s another one using a delta modulator that also works very well. I submitted this to EDN but it was rebuffed. I guess whoever reviewed the idea wasn’t used to seeing an opto operated in photo-diode mode.
A while back, one of the diyAudio forum members who was building the nx-Amp (here’s the thread on DIYaudio.com) enquired about either reducing the amplifier’s overall gain, or providing a volume control facility. Since both the 15 W class A sx-Amplifier and its bigger 100W sibling the nx-Amplifier, are Current Feedback Amplifiers, performance is quite carefully optimized for a specific gain – they are not as tolerant as voltage feedback amplifiers in this regard. Although not specifically raised in the thread, a consensus seem to develop that what was actually needed was a general purpose buffer that would accommodate the input from a pot, and be able to comfortably drive a power amplifier with typical input impedance of 10 k Ω.
No originality is claimed – Nelson Pass’s B1 buffer has been around for years and is well regarded – the UBx ‘Universal Buffer’, as this will I hope come to be known, also uses this classic configuration (but without the cascode), which I believe was first published by either Siliconix or National Semiconductor way back in the late 1960’s.
This design showcases a discrete opamp for audio applications. It features very low noise, low distortion and a class A output stage that will deliver 1ppm distortion into a 600 Ohm load at 12 V peak.
A complete relay that is smaller than a Tyco RT series EMR Simple_solid_state_relay_Updated . This design uses a small double sided PCB and some SMD components. Here are the Gerber files SSLR. The recomended mosfets for this design are the PSMN4R3-100PS[1] for supply rails of up to +-50 V absolute maximum, and for supply rails of up to +-75 V, the Fairchild FDB075N15A[1]
