Amplifier startup problem

I've implemented the circuit on page 2 of

formatting link

with the addition of 1K resistors across the 100uF capacitors to define ground as being midway between the power rails.

Most of the time, it works fine, but occasionally when it starts, it pulls the ground down to 1.8 volts or so above the negative power rail, and holds it there. I have multiple incarnations of this amplifier on the same power rails, and sharing a ground (by virtue of a MiniDSP board driving them), so all of them end up sharing this skewed ground. They don't work in that state.

I have not yet determined whether it's a specific amplifier that's doing this, or whether any of them can spontaneously start up in this state - it's a time consuming process, since I cannot create this fault state at will - I have to wait for it to happen.

Meanwhile, I would welcome any insights - is the design flawed? Have I got a dodgy LM1875 IC?

Sylvia.

Reply to
Sylvia Else
Loading thread data ...

I've had problems with long-tailed-pair based discrete class-A/B amplifiers pulling down a fold-back limited power supply at turn-on. The fix there wa s to replace the long-tail resistor with a constant current diode, which wo n't work for you and is probably implemented inside the LM1875 anyway.

I'd look carefully at the current limiting on the power rails before I did anything else.

My situation was that I'd just fixed the power supplies so that they didn't blow up on turn-on, and was unwilling to soften the fold-back. You may hav e more flexibility.

--
Bill Sloman, Sydney
Reply to
bill.sloman

Case like that I would sat try another IC first. If that doesn't work, look for HF oscillation. That print does not specify a coil between the IC and the actual output. I could be oscillating and if the capacitance of load, s uck as if the speaker has peizoelectric tweeter causes instability and burn s the thing up, you. as the engineer are holding the bag.

It is not just peizo tweeters, sometimes it is long runs. It has X amount o f capacitance peer foot, So when you lay those foots on you lay the pFs on, which then become nFs, and then... Got it ?

Anyway, if it not a bad IC that circuit is turning into a monostable multiv ibrator. If you build the circuit as shown you should not have that problem . Even if you changed the feedback resistors.

So, if it proves to not be the IC (and I mean by replacement) you probably have a ground problem. They can be creepy. This shit is a bit complex and of course there are some tough problems, but I had a 1976 Toronado and it h ad a constantly running off power clock. Every time I hit the brakes the cl ock stopped. It was not the fuse to the clock. It was not the fuse to the b rake lights. I had to take the voltmeter to it, I am not sure but I think t he problem was the fuse to the courtesy light or some shit. Seemed totall u nrelated, in a 1976 car.

Anyway, you are properly handling the mute pin on that chip right ? You can 't just leave them floating. Or maybe a bad connection causing it somehow.

Reply to
jurb6006

I probably don't have a clear picture of your resistor mod, but have you tried another hefty capacitor at the resistor tap?

Reply to
John S

I'd put zeners across the 1k resistors to make sure the 'ground' can't get too far out if there's a switch-on transient.

For example, if your total supply is 40V, then, say, a 24V zener across each 1k. Fairly big ones.

Cheers

--
Syd
Reply to
Syd Rumpo

No - I can't see what that would achieve.

The original circuit has Vcc and -Vee power inputs. It's not stated, but it may have been assumed that these are stabilised voltages relative to ground - that is, the power supply consists of Vcc, ground and -Vee.

I've just connected a power supply across Vcc and -Vee, the difference between them being 27 volts, with no power supply ground. So I've used the two resistors to define ground to be mid way between the power rails. The rationale for this was to avoid ground loops.

However, I've realised now that this means that if any amplifier output sinks a smallish current (about 80 millamps, given that there are five amplifiers) for short time, it can drag my defined ground down to near the negative rail through a loudspeaker. Since that places the amplifier inputs outside their permitted range, all bets are off as to what the amplifier will do, and I suppose a continuing sinking of current is not implausible.

Sylvia.

Reply to
Sylvia Else

The LF performance of your circuit will be poor, if that matters, as the output goes up, so the 'ground' follows as allowed by the 100uF capacitors. Slugging the 'ground' with extra capacitance will help this.

Cheers

--
Syd
Reply to
Syd Rumpo

It will, but not enough. 1k resistors are way too high to be returning speakers to, no matter how big your C.

Either provide a suitable supply or painfully bodge a pair of shunt regs across the rails.

NT

Reply to
tabbypurr

When you say ground I assume you mean pin 1 (+ve input).

I'm wondering if there is some kind of latch up as the circuit doesn't imply a current path from pin 1 to -Vee.

It is possible that the choice of resistor values and the circuit, ie

22k were to ensure any holding latch current was below some critical value.

BICBW

--
Mike Perkins 
Video Solutions Ltd 
 Click to see the full signature
Reply to
Mike Perkins

** So you created an artificial centre tap on the supply rail with a couple of 1k resistors and then directed the SPEAKER return current there?

An you claim this scheme actually worked ?

On what planet ??

FYI:

the sane method is to use coupling caps ( ie 1000uF electros )from the ICs to the speaker loads, returning to the 0V rail.

But don't that that bother you.

.... Phil

Reply to
Phil Allison

A big C is a low Z, much lower than the 1k at sensible LF. The 1k just allows any DC to settle.

In the single supply example given in the datasheet, there's 2200uF in series with the speaker. That would suggest using two 1100uF supply capacitors to maintain LF response.

It's an easy thing to simulate to see what happens, just use an op-amp with appropriately scaled R's and C's and load.

Cheers

--
Syd
Reply to
Syd Rumpo

I think the 100uF supply capacitors take the speaker return current, albeit not very well at LF.

Yes, the datasheet gives an example of this, it's the best way.

Cheers

--
Syd
Reply to
Syd Rumpo

Oh dear. Pin 1 is intended to maintained at a point mid-supply. It is a virtual ground as shown in Figure 2.

The speaker output is coupled through a 2,200uF capacitor with one end of the speaker grounded as shown in Figure 2.

I have no idea why any further slugging of any ground is required, virtual or otherwise?

--
Mike Perkins 
Video Solutions Ltd 
 Click to see the full signature
Reply to
Mike Perkins

That depends on the claim that figure 2 was implemented and the speaker was connected through a 2,200uF capacitor and to ground/-Vee.

--
Mike Perkins 
Video Solutions Ltd 
 Click to see the full signature
Reply to
Mike Perkins

It's perfectly able to work as is, albeit within constraints and as a horrid bodge. Keep the output low enough and the 100u caps could keep the rails within functioning voltages. But as Sylvia found, it's very far from a satisfactory circuit.

NT

Reply to
tabbypurr

of course. But it isn't good enough.

no it doesn't. Just 10mA output offset would skew the rail voltages by 10v! 10mA on 8 ohms is just 0.08V. You might or might not get away with it as a 1 off, but it's definitely not reliable or sensible.

2200uF 20Hz = 3.6 ohms, so your 2200uF decoupling caps would take close to 1/3 of the output voltage swing across them at lf, throwing the supply rails to the IC all over the place. Not sensible.

Does your sim replicate the latchup behaviour you get in reality? Don't you need to add dc output offset? This is just the sort of situation sims aren't designed to replicate.

NT

Reply to
tabbypurr

They're bypassed by the 100uF capacitors.

Yup. And SPICE says it works too.

What you're probably overlooking is the 90dB open-loop gain, and considerable negative feedback reducing the closed-loop gain to just a factor of 6.

Sylvia.

Reply to
Sylvia Else

You're mixing up the two circuits. Sylvia is using the first, without the speaker coupling capacitor.

Cheers

--
Syd
Reply to
Syd Rumpo

That's a thought worth examining.

However, my original reason for using this symmetrical power supply arrangement was in the hope of eliminating start-up/shut-down clicks. Unfortunately, this hasn't been successful, so I'm going to include some relays to delay connecting the speakers. If the speakers aren't connected at start-up, then there's no path by which the ground can be dragged down, so it may be that the relays will solve both problems. If not I'll look at the zener idea further.

Sylvia.

Reply to
Sylvia Else

** Hopelessly inadequate.

** A obvious lie.

** What you are overlooking is that the input zero reference has a massive amount of LF audio imposed on it all the time when it should have none.

A recipe for horrible instability.

Wot a fool.

.... Phil

Reply to
Phil Allison

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.