This is exactly the reason why it is used - the push pull VAS can be driven into class B when a lot of current is demanded from it. At low impedances, the beta of the output and driver devices can drop significantly requiring a lot of current from the VAS. If the VAS can cope with this by running into class B instead of just clipping when you hit the limit imposed by the CCS load, you are better off. Higher distortion caused by leaving class A is better than clipping.
These circuits also don’t require critical matching to work. Yes, it will sound better, but just using nominally complementary devices, and pulling pairs from the same batch of parts gets it close enough to be listenable and produce low enough DC offset not to be dangerous.
Good luck with the design. I followed edbarx on the development of his amp. It was a brave undertaking and he managed to pull it off.
wg_ski, helpful follow up. It is more important to match the NPN and PNP pairs. The NPN PNP differences can be taken care of by DC correction. Differential pairs are capable of low distortion with good matching. I have a recollection of a Doug Self analysis indicating that 5% current match gets most of the way.
I have posted before that devices from Diodes Inc. make implementation easier with BCM847BS, BCM857BS, DMMT5401 and DMMT5551.
wg_ski, helpful follow up. It is more important to match the NPN and PNP pairs. The NPN PNP differences can be taken care of by DC correction. Differential pairs are capable of low distortion with good matching. I have a recollection of a Doug Self analysis indicating that 5% current match gets most of the way.
I have posted before that devices from Diodes Inc. make implementation easier with BCM847BS, BCM857BS, DMMT5401 and DMMT5551.
If amplifiers must be tested with loads that are completely different from a loudspeaker, why not just test with a short circuit too?
You misunderstand the purpose of the test. It was devised to emulate a difficult load. Calculate the impedance of the 2uF with frequency.
It has been employed as a standard test for amplifiers from at least 1974 to the present day.
It has been employed as a standard test for amplifiers from at least 1974 to the present day.
The fact this test has been used by some people does not make it a standard test procedure. And this load is indeed far from simulating worst case realistic loads.
why not? here is my one:- before i start a pcb, any comments?
also checkout VSSA type amps, diamond buffer amps.
This is one reason devices with sustained beta at high current are preferred as output devices, for instance the 3281/1302 pair have very flat gain curves upto 6 to 8A or so. Older output devices were truly awful by comparison, gains falling from 1A or less.
Well I believe that's a required safety test for commercial equipment, but only to prove it won't catch fire!
A method I've used (only with simulation so far) is to connect a different amp to the other side of the load, at a different frequency. This gives very reactive currents into each amp - and would be easy to do with a stereo amp, just bridge the outputs through a 15 ohm dummy load (for 8 ohm amp) and play non-harmonically related tones to each channel.
You can look for intermodulation products or just see if the output devices undergo secondary breakdown due to over-optimistic design criteria.
Or measure the damping factor.
I built an amp based on this one out of an old AudioAmeture article. It is a beast with about 200 watts into 8 ohms. Great design!
I would like to thank all those who took some of their time to answer my original question. From the various replies, it is clear emitter resistors in input stages play a very important role to (i) linearise the EB junction behaviour and (ii) make transistor behaviour more stable and predictable.
My personal preference is for one single input stage, albeit elaborate, to minimise any dependencies upon rail voltage variations and modulation of the constant current by the signal itself. I think, my amplifier would have benefitted with a cascode to enshroud the constant current source in the input stage.
My personal preference is for one single input stage, albeit elaborate, to minimise any dependencies upon rail voltage variations and modulation of the constant current by the signal itself. I think, my amplifier would have benefitted with a cascode to enshroud the constant current source in the input stage.
With Re added to the input transistors, the current of each input transistor is well defined. I don't think cascode will make anything better (except it allows you to use low Vce, high beta transistors)
The power rail ripple has the same effect on the + input and - input, they will cancel each other.
Last edited:
Yes, this is important.
I've caught Fairchild 2SC5200/A1943 on this. The PNP loses hFe with current, so they cheated by increasing its hFe so it still has decent current gain. However this results in higher hFe at low current, which means a large mismatch in hFe with the NPN, which increases crossover distortion. Womp womp.
Yeah, IMO tripping the protection and shutting down the amp is a reasonable response to sticking a capacitor in the output.
That's basically how my output stage test works, and it gives really useful results! And since there is a resistor between the two, you can test out of phase current without wrecking phase margin by putting a cap on the output.
If you make one amp play a small signal at say 6kHz, and the other play a large signal at 40Hz, connected with a resistor... the 6kHz AC voltage at the output depends on output impedance (or damping factor or gm), and therefore you get the wingspan of the output stage.
Nothing to do with safety. The capacitor introduces a load phase angle which should not trip the protection. A simple test with a scope allows comparative stability behavior. In this example the asymmetry is clearly apparent.
Attachments
Last edited:
The two diff amps are usually referred to as a complementary balanced input or a complementary diff input.
The advantages they offer are symmetrical slewing, high symmetrical drive from the VAS (but see later remarks), cancellation of most of the 2nd Harmonic in a well balanced design, lower input bias currents and offsets, and an additional 6 dB open loop gain.
On a big power amp, the additional cost is marginal.
Re the VAS output current, it’s unlikely that it will ever go into class AB. Most designs run the VAS current at between 5 and 15 mA and if you use an EF3, even into a 2 Ohm load, the variation will be only 20-30% of the VAS standing current value.
BTW a competently designed feedback amplifier will always include an output inductor of between 0.5 and 1 uH. This is very effective in isolating capacitive loads from the amplifier.
The advantages they offer are symmetrical slewing, high symmetrical drive from the VAS (but see later remarks), cancellation of most of the 2nd Harmonic in a well balanced design, lower input bias currents and offsets, and an additional 6 dB open loop gain.
On a big power amp, the additional cost is marginal.
Re the VAS output current, it’s unlikely that it will ever go into class AB. Most designs run the VAS current at between 5 and 15 mA and if you use an EF3, even into a 2 Ohm load, the variation will be only 20-30% of the VAS standing current value.
BTW a competently designed feedback amplifier will always include an output inductor of between 0.5 and 1 uH. This is very effective in isolating capacitive loads from the amplifier.
Last edited: