There's a lot to be said for using a transformer. I haven't yet tried it myself, but the ground-breaking ability is important.
Indeed. As is the ease of getting excellent balance over the entire audio band.
Isolating the grounds at the input of my power amp was the single greatest quality difference in electronics that I've observed in my system in quite a few years; reduction of even subliminal hum seems to be a great positive for clarity. And, of course, putting the transformer at the input means that the signals are handled at a low level and that feedback may be more easily applied than is the case for interstage transformers.
(NB: This is not OT- my tube amp's input is a pair of pFETs.
Re: Re: Re: Re: Where do you draw the line between tube and sand?
Give the man a cigar!
You're spot on the money. Let's look at an 8 bit system, for example. With 2's complement math, which is almost universally used, 45 decimal = 00101101 binary, and -45 decimal = 11010011. Note that adding up the two numbers gets you 100000000, which is 256 decimal.
Now having to make sure that addition comes out correctly might become a real pain in the neck, but the error caused by merely inverting the data turns out to bias the output by 1 LSB in the positive direction. If you're passing 24 bit or even 20 bit data, that's literally lost in the noise. In other words go right ahead - let the math purists bitch, fewer moving parts mean less stuff to break.
Francois.
EC8010 said:
Do the phase splitting in the digital domain. It's a while since I've looked at digits, but I vaguely recall that to get a true analogue inversion of 2s complement notation you need to add or substract 1LSB. Or something. I could be wrong, but it might be worth checking.
Give the man a cigar!
You're spot on the money. Let's look at an 8 bit system, for example. With 2's complement math, which is almost universally used, 45 decimal = 00101101 binary, and -45 decimal = 11010011. Note that adding up the two numbers gets you 100000000, which is 256 decimal.
Now having to make sure that addition comes out correctly might become a real pain in the neck, but the error caused by merely inverting the data turns out to bias the output by 1 LSB in the positive direction. If you're passing 24 bit or even 20 bit data, that's literally lost in the noise. In other words go right ahead - let the math purists bitch, fewer moving parts mean less stuff to break.
Francois.
True that not adding an LSB 1 after inverting the bits will create a slight error, but the error is small and would be a DC offset anyway. Of no significance in audio. If you're using a serial bit stream DAC chip all you'd need is a single inverter. Some DAC chips have differential outputs already. Or you could construct a pair of SRPP circuits with both bottom triode cathodes tied together and fed by a constant current or a big resistor connected to a very negative supply. One grid of the bottom tubes goes to ground, the other grid of the other bottom tube goes to the DAC output IV resistor. Both top tube cathodes will produce a differential pair of signals.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.