Lynx amplifier

[Kibi]

After six years on the shelf, I finally found some time to do some more work on this project.

I started out with preparing the heatsinks. I drilled all of the holes on the milling machine for accuracy and followed each hole with a thread forming tap. Rather than cutting the threads, I favour thread forming for small diameter holes in aluminium. These are all M3 holes some of which will need some decent torque on them, so they are at least 20mm deep.









The next thing to do was to develop a jig for bending the leads on the output transistors accurately and consistently.

Using a scrap piece of 25.4mm square aluminium I cut a pocket out to accommodate the body of the transistor with a step at the end at the correct length for the bend. The height of the step also needs to be sufficient to support the leads at the base of the package as they are bent over.



With the transistor mounted face down onto the jig, all that's needed is to tap the protruding leads with a soft rubber mallet to bend them over the edge to form a lovely consistent 90º bend for each lead of each transistor.







Now, with all of the devices mounted, all the holes appear (as if by magic) to align properly with the holes that I drilled in the heatsink.



With that done, we can start to get on with some testing. I set up a bunch of multimeters, an oscilloscope and a signal generator in order to perform the initial configuration, setup and testing.
I have configured this circuit to operate from a ±50V supply, however, my bench supply only goes up to ±30V. This will be OK for what I need to do today until I build the proper power supplies. Also, the bench supply has current limiting which can prove handy in case this amplifier decides to enter into some sort of vicious oscillation mode.
My cautiousness was well founded, actually, because the very first thing this amplifier did was to oscillate horribly at around 2.8MHz with a very expensive OPA627 no less. A TL071 did not exhibit any oscillations and a NE5532 caused oscillations at around 1.3MHz.

Following the advise of other members who have contributed to this thread especially Stephan sch, I replaced C109 with a 22pF capacitor, soldered 22pF capacitors from the collectors of Q103 and Q104 to GND. Reluctantly I also had to solder another 22pF capacitor over pins 2 and 6 of the lovely OPA627.

Various leads and probes hooked up to perform tests of the VAS. Don't worry about the test clips across one of the emitter resistors, look upper left, the wires destined for the CAS are flapping about in the breeze.



With the modifications, the VAS stopped oscillating and the bias voltage could be set easily.

Now it's time to connect up the CAS along with a 6.8? resistive load on the output.



I got some of my multimeters out so that I could monitor various aspects of the setup simultaneously.
From left to right, the Fluke 287 is just sitting in standby with the leads ready to take random measurements at will. The Fluke 89 is measuring the bias voltage across TPA and TPC. The Gossen is measuring the quiescent current through one of the drivers. Finally, the Fluke 25 is measuring the AC line voltage. Here I have set it to 230VAC. My workspace is purely solar powered, so when I perform high power tests such as these, I like to keep an eye on how the inverter is performing. It's  a high quality Victron 2KVA inverter, so there is no reason to doubt it's performance, I am just interested in looking at these sort of things.
Interestingly, the Solartron bench multimeter in the background was initially measuring the quiescent current over the emitter resistors but it caused oscillations, so the battery powered Gossen was employed for the task at which point the oscillations went away.



Here is a helicopter view of the whole lash up.



Now we can actually take a look at the performance of this thing. After the modifications, I am reasonably impressed. The only thing I am not so happy about is having to slow down a very expensive Op-Amp. Having said that, in my case, the rise time is still quicker with the modified OPA627 than with the respectable NE5532 (which also needs a 22pF across it).

So, with the DC bias and quiescent current set correctly, let's get on with some AC measurements.
I should note that I am not overly fanatical about a DC coupled front end such as this design, so the eagle eye'd may have noticed a great big 4.7uF capacitor which provides AC coupling between the amplifier and my signal generator.

In all of the below screen shots from my oscilloscope, the green trace is the input signal, the yellow is the output signal.

Here is a 1KHz square wave with a 1V P-P input. The amplifier has some slight ringing but nothing to worry about. The specified (sine wave) gain of 25 at this stage has been superseded. Although I suspect that the oscilloscope is measuring the peaks of the ringing, it looks good to me.
What I am looking out for in particular with a square wave input is the rise time. Here we can see a rise time of 740nS which roughly suggests a bandwidth of 500KHz. This, to me, is an acceptable sort of bandwidth to expect from an audio amplifier.
The bandwidth of my generator is 20MHz, so expect it's representation of a 1KHz square wave signal to be pretty faithful.



Here is the sine wave. Here we have a gain of 25.05, what more can you expect?



How far can the input signal go? Let's investigate. Once agin, with a square wave, I wound up the wick from the signal generator. The negative half of the output signal starts oscillating with an input voltage of about 1.9V P-P. At nearly twice the regular input voltage, I am not one to complain.



Back down to 1V P-P, let's zoom into that ringing a bit........



.....and a bit more.



In order to get a better view of the ringing, let's wind up the input frequency to 100KHz. Note, the rise time is what we are interested in from a bandwidth point of view, this is still 740nS. We are winding up the input frequency in order to get a closer look at the ringing. The generator is still producing a reasonably faithful signal.



At 100KHz, sine looks good too. Gain has dropped to 24.7, who's complaining, we are way out of the design bandwidth.



Let's double the input frequency, 200KHz. The gain is now down to 23.5, wow, not bad. Although you can now see some distortion at the peaks.



So, that was fun. Let's look at the band that matters, the audio one.

Sine at 10KHz, input 1V P-P:



Square at 10KHz, input 1V P-P:



Sine at 20KHz, input 1V P-P:



Square at 20KHz, input 1V P-P:



Goodness knows how Jan managed to get acceptable performance from his original design. I could only achieve the performance that I have demonstrated with the modifications that have been suggested by other members who have contributed to this thread, thank you.

[fivefish]

I like the transistor jig you made.

On the square wave, it looks more like a tiny bit of oscillation (based on the overall shape), rather than a square wave ringing. Maybe try a 33pf?

Nice job.

[Kibi]

I like the transistor jig you made.

On the square wave, it looks more like a tiny bit of oscillation (based on the overall shape), rather than a square wave ringing. Maybe try a 33pf?

Nice job.

I like your thinking, I should try and slow my super Op-Amp down?

[fivefish]

This is what a square wave ringing looks like. The first bump is the peak, then it dies down.



But in your square wave, you don't have that first peak, instead it's more like a tear-drop or oval shape. It's like on the very edge of oscillating, but then dies down... probably due to that 22pf you added.

[T3sl4co1l]

Links (ha?) to schematic?

Slowing down an op-amp by throwing capacitance right across it is usually a poor solution: it directly trades stability for lower loop gain, and not very evenly at that.  Better is to add a resistor in series with the capacitor, so you can add a zero near the crossover frequency (whatever it is in the loop that happens to be causing that problem), extending phase margin and GBW.

I would guess the ringing is a need for ferrite beads on the output transistors or drivers.  Or maybe just crunchiness in your load resistor (okay, so it's whatever resistance it is at DC... how about at 1MHz+?).

BTW, regarding bandwidth: it's often a bad idea to use more bandwidth than you need.  Fortunately, audio isn't at all phase-critical (ha, depending on which audiophools you listen to.. but we have science to better them all), so you can push that cutoff down as low as needed, no problem.

The main hazard of excessive bandwidth in audio amps is usually EMC: you'll get AM radio, or random CB, or dit-dit-bee-dit from GSM cell phones, piping in via wiring and parasitic rectification.  If nothing else, some input LC filtering helps with this.

Tim

[MagicSmoker]

Agree with others that this is oscillation/incipient instability. Sometimes this occurs when driving a purely resistive load (or, at least, one with very little inductance, especially without a Zobel network or one that is incorrectly designed), but the fact that the oscillation gets worse with higher fT op-amps suggests that it isn't the op-amp causing the problem, oddly enough, rather, it is likely you need a bit (more) of damping in the base connection to each output BJT (for a quick kludge of the existing board insert a lossy ferrite bead on each base resistor).