T Network resistors for Gainclone

[Puffin]

I have just implemented the T Network configuration for my inverted LM3875's. I had previously used a 10k input resistor and 22k feedback resistor.

The T network sounds great, but with my high efficiency speakers I am getting a lot of hiss. Is this right? or usual?

[Puffin]

I am told that the hiss is to be expected. I think I can live with it for the added clarity I am getting. A worthwhile tweak IMO

[danielwritesbac]

Can you post a schematic? A "scribble" is just fine.

[Nuuk]

Originally posted by danielwritesbac
Can you post a schematic? A "scribble" is just fine.

Come on Daniel - you should know where to look by now!

What about this ?

[danielwritesbac]

Originally posted by Nuuk


Come on Daniel - you should know where to look by now!

What about this ?

WOW! That link is very interesting. Its just cool! Thanks! This should be a nice adventure.

Assuming that its still negative feedback, I think I've spotted a noise. The carbon nfb poses a problem because the other half of the divider is metal film. That's backwards, and causes limitless gain on RF. You'll need a very small value capacitor (do I remember correctly 20pF?) between pins 3 and 8, if that's a negative feedback loop there.

Or maybe it just looks backwards to me because its an inverting amp? lol! I'm dizzy now.

[Nuuk]

Early to bed, early to rise
makes a chip amp builder
Not wealthy but wise!

[danielwritesbac]

Well, the hiss output is an AC signal, so we know that it came from somewhere.

What about drop RF again?
Right at the speaker terminals, you could connect a 5w5R (or similar) and a 10nF (3nF to 10nF range) for speaker zobel. That's not directly on the audio band, but it ought to load down some RF.

Howabout a 220pF load at output end of the decoupling cap?

And, why not try a LM1875? Unlike LM3875, the LM1875's noise figure doesn't rise with frequency during the audio band and its useful output power can be brought "near" the LM3875 if using 8 ohm speakers. And, you'll no longer require the midbass blur of big caps at chip pins. Instead you can move those to the rectifier and use smaller caps at the chip pins.

Or there's "wrong way Class A". . . using an resistor, range 2k to 5k, from the V+, you get the effect of a very ineffective noise gate. This does almost nothing whatsoever, but if you're already close, it may be enough to reduce the noise. Trying to reproduce only the "silence" in class A, might get you exactly that, because that transistor will "stick" open during very low output while its sister cannot output at all.

Headphone test! Perhaps the hiss is coming from the source? If its digital, some 1nF mylar (as a load upon its line level output) might cut off a digital hetrodyne pretty easily.

OH, I'm out of ideas!

EDIT: Not quite. There's a dirty trick. . . You can parallel your decoupling cap with a 100nF or 47nF 250v "hi-K" ceramic, and this will throughly "etch" the treble, and then you can brute force that away with a heavy-duty speaker zobel. Its like "Dolby B" or phono, in first boosting the treble and then reducing it. Sure that's a terrible way to do it, but a slight amount of that will remove a slight amount of hiss.

EDIT: Howabout replace that 100R metal film with a 100R Caddock or any other 1/2w carbon resistor (econo carbon film)? That might filter just a bit of HF away. Or, it might increase it. Well, the attempt costs about 3 cents.

[BWRX]

T-network: the better feedback solution?