Well, friday was the last day of this seminar,
He spoked a little between class A and class AB and his opinion on the topic, using class A in nominal operation that turns into class AB with high peaks and more talk on the clipping stuff, so, he designing the amplifiers to do class A for the optimal point but not burn a lot of power using class AB. Also he made a point on the sources of class AB distortion coming from improper PS decoupling because of current shape more than compensation.
Going to equalizers he took the approach of proportional Q so the eq is affecting the same freq bands in all the gains, instead of doing a filter and then add or subtract from the program. Also he pointed his preference about mirror curves rather than symmetric ones. In graphics he specifies that matching the bands to be as one at higher boost/cut instead of separating two peaks is a must because of phase shift.
From that he went to phase shift to group delay and the problems with that, we are not so sensitive as phase shift it self but group delay, making us confuse what we are hearing if the time of a part of the spectrums comes before or after from the rest of the same signal, and as we are more sensitive to time of the signal than amplitude. Also because we are moving in the space where the speakers are we may compensate the errors of those but not the ones from the material it's coming to them. Speakers are fixed sources of distortion but material is changing and we can't compensate that. A thing he told us to look for is how much time we can listen to something without getting tired even if it's sounding good or not. About the distortion generated in eq he marked about HF boosting for example are making a differentiation that may build spikes that weren't there before in the signal, to avoid that low Q makes low phase shift and no/less artifacts are created, "don't distort the image, just shadowing".
The shape of the filters in the 550 wouldn't be possible without the switches, not possible to do them with pots because too many things changing at the same time, but doing it with switches takes them to need caps with more precision so every 550 sound the same with the same configuration, with continuos varying parametric you would match it with the pot.
From the components used he spoke about tempco being a source of shift of the freq and amplitude in a slow way but the voltage coefficient (much smaller than temp) being a source of THD and that's why it's important. Real inductor Vs gyrator, Johnsons noise vs active noise, saturating the core vs clipping. High voltage from inductors when unloaded fast aren't done with gyrator but shouldn't be a problem in normal working situation. Real inductors has different resistance for each frec, not gyrators, temperature varying the values.
Butterworth filters being preferred to maintain time relationship for example at the zobel networks for transformers.
DC servos instead of caps or transformers aren't recommended in connection to the output world, they are ok inside the box where environment is controlled, if the outside environment is controlled is ok too. He recommended non polar caps for the signal coupling and tantalum too. For some big broadcasting companies only accept tantalum and not aluminum electrolytic are accepted.
When choosing opamps IMD is a problem, sometimes no data on datasheet, hard to measure, affects audio, transient created within the opamp not corrected by NFB. The fewer the active components and less open loop gain, the less likely to have IMD and use less feedback, this is why less NFB is better for him, and discrete sometimes better than IC. With more active devices, you need more feedback and more un-linearities to correct.
He spoked about some 2515 DOA or something like that, which had less open loop gain, it was better than 2520 but not sufficiently to go into production, if someone has more info please make it public!
In dynamics he mentioned his approaches and creation of the blackmer cell, he did a lot of talk about time constants, as sources of distortion and in lunch we were discussing the 525 approach of automatic release, depending with frequency content.
For summing amps he said switching should be between resistor and bus and preferred to go to ground for the end of the resistor to maintain the load on the source constant and reduce picking up noise and radiating to other resistors next to it. A cap to avoid DC shift when switching was recommended. An remote electronic switch was recommended instead of mechanical switch so the switching occurs as close to the bus as possible, also preferred an electronic switch rather than a relay because of speed and softens of switching. He showed P15A4684 datasheet as an example, he says it's ok because no voltage being applied there and should be less noise than a discrete one made from FETs.
About ground and shielding he prefer to connect all shield to chassis and no reference from the outside to local ground, use differential input even with unbalanced sources. For non differential outputs balanced take the chassis as reference.
That would be all the last day talk. I'm really proud and touched of having the chance of knowing such a designer, I'm really grateful of UnTreF to giving that opportunity. After he finished we went next to him to talk a little bit more and say goodbye, he is a really nice person, too bad this already finished.
JS
