Parallel chips - output resistor questions

[sangram]

Tried searching for an obvious question, couldn't find it, so here it goes:

1. Influence of output resistor on sound quality? I was using locally available 5% wirewounds, and switched one channel to industrial metal oxide resistors. It seems the sonic influence of the output resistor is extremely marginal, if at all - which surprises me as it is the signal path.

2. Why do smaller output resistors cause less heat? I was using 2 x 0.47 ohm in parallel for output resistance of 0.23 ohm, now dropping it to half seems to lead to a marginal reduction in heat output, where intuition shows a higher idle current would flow through the smaller resistors. Both times, I am using unmatched 5% parts.

Thanks for any hints.

[abraxalito]

1. I concur - the construction of the resistor makes relatively small differences to the sound.

2. Assuming that your amplifiers are reasonably well matched for gain and offset, then smaller valued resistors will dissipate less power. More energy ends up in the speaker for lower values. Using 0.1ohm sharing resistors does mandate reasonably good matching of gain - I'd aim for no worse than +/-0.1% myself.

[tiltedhalo]

First of all if you are getting heating in the output resistors you need to look at your circuit. You have a overly large amount of offset or oscillation. To be getting noticeable heating at idle or full output there is something going on there that should not be. Disconnect your resistor output legs and check DC offset at the chip, remember if you have three chips, two show marginal -DC offset and one shows small +DC offset the + offset will dissipate the combined power of all chips, the same applies to oscillations, a oscillating chip will push its power into the output matching resistors of the non-oscillating chips.
Also check your balanced line driver or circuit if you are using a BPA design, you should show 0 offset period, if working properly, if not you will globally multiply the offset by the gain factor of the amp. If you have +DC at one leg and -DC at the other you could push a lot of current across the resistors. Again this also applies to oscillation in the driver circuit.
Allot of people build BPA amps and never see there full potential because of problems like this, they seem to work fine but there are underlying issues. 30mw multiplied across circuit can lead to 60W of wasted power.

[Minion]

Your Paralelling is not going to work properly with 5% resistors , You need to match the all the resistors in Both Chipamps to 0.1% , I would also use oversized feedback caps and input caps ......

[abraxalito]

Originally Posted by Minion

Your Paralelling is not going to work properly with 5% resistors , You need to match the all the resistors in Both Chipamps to 0.1%

Not necessary for the output sharing resistors - I've Spiced it. But sure, the gain resistors really should be 0.1% or better. 0.1% for a high power, low value resistor really is a waste of money, those kinds of parts don't come cheap.

[sangram]

@Tiltedhalo - no, I was not talking about the resistors heating up (they are not). I'm only using a PA100 style design, and I get about 2mV output offset.

In a parallel chipamp, imperfect matching of the gain setting resistors causes some increase in quiescent current due to additional current flowing between the chips. This increases the standing power dissipation (idling) and warms up the sinks. I noted that with smaller resistors, the amps out out less heat than with the larger ones, that's what I was asking about, actually .

@minion: I use no input cap, the feedback cap is 100uF, feedback resistors are 22K/1K - lower end rolloff is 1.6Hz, plenty low enough I see the point about oversizing caps - I had 47uF caps and not much bass. Once I moved up to 100uF that problem was tackled.

I handmatch the resistors but a 3 1/2 digit meter does not allow me to see a variation of 20 ohms in the 20K ohm range setting (and I have to use 200K, where it gets even more interesting - plus meters are usually accurate to only 1% and are meaningless for higher-spec parts). Ideally I should be buying 0.1% parts for feedback, but those are simply not available in my country for love or money, so I make do with what I got.

As it is I'm using PRP and Rodersteins, of which I bought a small batch. I've seen abraxalito's sims and I know I may be pushing my luck a bit, but I'll take my chances.

[abraxalito]

Originally Posted by sangram

I handmatch the resistors but a 3 1/2 digit meter does not allow me to see a variation of 20 ohms in the 20K ohm range setting (and I have to use 200K, where it gets even more interesting - plus meters are usually accurate to only 1% and are meaningless for higher-spec parts). Ideally I should be buying 0.1% parts for feedback, but those are simply not available in my country for love or money, so I make do with what I got.

As it is I'm using PRP and Rodersteins, of which I bought a small batch. I've seen abraxalito's sims and I know I may be pushing my luck a bit, but I'll take my chances.

Even a 1% accurate DMM will do better on repeatability than on accuracy. You don't need accurate values, you need close matching. To improve the sensitivity of selecting resistors, you could try using a bridge arrangement and the most sensitive voltage range on your meter (usually 200mV) across the arms of the bridge. This will allow you to match your 22k/1k resistors more closely than 0.1%. Of course, you'll need an external power source too to energise the bridge. With 20V then a 1ohm change in a 20k resistor will correspond to 0.5mV so your 3.5 digit multimeter will in theory be able to see a change of 0.2ohms. Way better than you need

[sangram]

fair point. I do not plan to do this kind of a circuit again, the one up and running is doing absolutely fine - but I'll surely look at a bridge the next time I need that close a match

[AndrewT]

we home experimenters cannot cheaply measure anything to ~0.1%.

We can compare DUT against REF to better than 0.01% with care and still do that cheaply. It justs takes time and thought.

Do not try to do absolute accuracy measurements unless you are prepared to spend a lot of money on equipment and even more money getting all your measuring equipment recalibrated/certified at regular intervals and learned how to check that your calibrated equipment is still in spec.

[tlmadsen]

I think we are heading for a sligt form of overkill. As stated you don't need absolute accuracy, but close matching. Actually, you only really need close matching for the gain resistors.

Just for the record. I only went trought 100 pices of 1% metal resistor from the same batch on a claibrated 6½ Digit DMM once. No one was higher than 0,3% from absolute value.
If you have your avg. 3½ DMM, and you measure say ten times as many resistors as you need, if should be farily easy to get close matching on what you need.

It is not 100% clear to me, why you would need 0,1% output resistors. (.......need to make a calculation).

As you might know it is not always easy to get 0R1-5W resistors and one way of getting around this (and possibel need for accuracy) is to parallel 10x 1R0-1% metal resitors.
Have fun.