Class D glitching

Having a problem with a class D variant design in Spice. This is the output, the hf on af is correct but something's going very wron g at the top of the waveform. The bad news is I can't show the circuit, whi ch is more than a little limiting, but any ideas of the sort of issue that might cause that with a class D would be welcome. I realise it may well be not much can be said without the cct diag.

formatting link

NT

Reply to
tabbypurr
Loading thread data ...

ong at the top of the waveform. The bad news is I can't show the circuit, w hich is more than a little limiting, but any ideas of the sort of issue tha t might cause that with a class D would be welcome. I realise it may well b e not much can be said without the cct diag.

The switched element that pulls the output up runs out of puff close to the top of the waveform, and isn't supplying enough current to pull the output up to the desired voltage.

As the desired voltage falls away on the other side of the peak it eventual ly gets to a region where it can deliver more current than is necessary and starts switching off from time to time.

You should have been able to articulate that for yourself. Without the circ uit diagram there's no way of working out whether this was worth saying.

--
Bill Sloman, Sydney
Reply to
bill.sloman

ong at the top of the waveform. The bad news is I can't show the circuit, w hich is more than a little limiting, but any ideas of the sort of issue tha t might cause that with a class D would be welcome. I realise it may well b e not much can be said without the cct diag.

Right, there are only 3 places the problem could occur. Comparator, driver pair, output pair.

According to Spice P_diss is 3.5nW in each output tr, which does not equate with an op tr going linear. The lower rate of fall of V is due to the load .

Disconnecting the inductive load reveals the problem: it's only switching h alf the waveform. Problem turns out to be the comparator output, it's the w rong type of comparator! Its input is only specced down to -0.3v, and it's seeing -0.6v.

I figured after sleeping on it it had to be the comparator, but looked firs t at the output drive rather than input. Not posting the cct is a pita.

NT

Reply to
tabbypurr

wrong at the top of the waveform. The bad news is I can't show the circuit, which is more than a little limiting, but any ideas of the sort of issue t hat might cause that with a class D would be welcome. I realise it may well be not much can be said without the cct diag.

r pair, output pair.

te with an op tr going linear. The lower rate of fall of V is due to the lo ad.

half the waveform. Problem turns out to be the comparator output, it's the wrong type of comparator! Its input is only specced down to -0.3v, and it' s seeing -0.6v.

rst at the output drive rather than input. Not posting the cct is a pita.

Taking comparator inputs below the negative rail isn't a good idea.

You are lucky that the Spice model caught it.

If you do it in real life, the current drawn out of the substrate by the ca tching diode can go anywhere, and the comparator outputs can behave very od dly.

I did something like that back in 1979 with a quad-slope A/D converter circ uit, and was lucky enough to be able to create a -0.7V negative supply rail to stop the odd antics.

--
Bill Sloman, Sydney 

The story gets recycled here from time to time.
Reply to
bill.sloman

Tabby, Bill, what were your signals, that went below the rail?

--
 Thanks, 
    - Win
Reply to
Winfield Hill

It's an amp run on +&-6v, a variant of Class D that I'm playing with. I did n't pay much attention to the comparator used for the sim, real life ones w on't use the same part. Comparator inputs can go more than 0.3v below the 0 v rail, the Spice model doesn't like that though.

NT

Reply to
tabbypurr

I had to deal with an input signal to be digitised that could get below the rails - I didn't have to produce a digital code to represent it, but I did have to stop the system doing anything silly when that happened.

I used a germanium diode to clamp the input at -0.3V and adding the -0.7V negative rail to the vulnerable part left me well protected.

--
Bill Sloman, Sydney
Reply to
bill.sloman

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.