Diff amp's what I had in mind, hence the mention of diff to single ended conversion in post 18. Buffering with, say, a 49720's probably not a bad idea. But I would expect filtering to degrade performance; should be better the more closely the front end's Vref and the 1592's Vref match. Also, referencing the input to Vref should decrease the requirements on Cref. My bet would be leaving Cref at the 100nF || 4.7uF used in the datasheet and eval board would be sufficient with further increases resulting in minimal improvement. Though, yeah, if you want 4.7+uF of actual capacitance using a 22+uF 6.3V part is probably wise. If other manufacturers spec at TDK's level of detail I've not found the docs, but TDK's 6.3V X7R parts are down 65% or so with 1.65V of bias and a few more percent should be taken off for aging.
So far as I know the 1592's not a spread spectrum part and, even if it does spread without Analog mentioning it, the spread should be from PVdd and---maybe---DVdd but not from AVdd.
Pull up to DVdd sounds fine on a quick look.
So far as I know the 1592's not a spread spectrum part and, even if it does spread without Analog mentioning it, the spread should be from PVdd and---maybe---DVdd but not from AVdd.
Pull up to DVdd sounds fine on a quick look.
Last edited:
I see your point, though, I'm skeptical of any circuit that tries to cancel noise by subtracting a 'copy' of the noise. Dealing with charge injection cancellation, I've found that most circuit that look to work great on textbook paper actually makes circuit performance worse. This may/may not apply here, but it definitely triggers a deja poo (haven't I stepped in this dung before?)
I think I'll leave the filter in place. A couple of 0603 parts take up little room and one can short it out with a 0 ohm resistor if one doesn't like the filter.
Pathetic I say. That's what it is... TDK makes good parts. I suspect the parts from other manufacturers are of similar or worse performance.
They do mention a 5th order delta-sigma modulator that shifts the noise outside of the audio band. I take that to mean 5th order MASH or something similar. Basically a pseudo-random dither around a fixed clock frequency. Maybe it doesn't spread enough to qualify as a spread spectrum clock... But it does create a bounce house for digital signals that can couple into the analog circuitry on the chip. Whether the cap on Vref is to mitigate this coupling or if it's just rolling off the 1/f noise of their reference is unknown.
~Tom
Hey Neil, do you have findings on load cap sizing for the 1592's oscillator? I notice the datasheet doesn't spec the input capacitance. Small C0Gs are cheap so it's not a big deal to step through different crystal load cap sizes. But I don't have hot tweezers and it'd suck to accidentally lift a pad.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.