Hacking the feedback loop

[cloth-ears]

Started messing with an old Class AB amp, opened the feedback loop and fed it back from the speakers after the x-over. Now the amp is correcting the x-over.

OK, so it's not exactly Phillips MFB but the results are interesting and worthy of some further effort on my part. Anybody else tried this fairly simple hack?

[lurcher]

Started messing with an old Class AB amp, opened the feedback loop and fed it back from the speakers after the x-over. Now the amp is correcting the x-over.

OK, so it's not exactly Phillips MFB but the results are interesting and worthy of some further effort on my part. Anybody else tried this fairly simple hack?

Why not just remove the xover then, thats what the amp is trying to do now.

Not sure what "after the xover" means, is there not two or more "afters"?

[cloth-ears]

HaHa,

(Not sure what"after the xover" means, is there not two or more "afters"?)

I mean after the x-over, I'm using OPA134's to decouple and remix

[Light Dependant Resistor]

Why not just remove the xover then, thats what the amp is trying to do now.

Not sure what "after the xover" means, is there not two or more "afters"?

But having the amp knowing what it is driving is a very good idea.

The concept of current drive vs voltage drive
is a logical extension and is explored in considerable depth here:
http://www.diyaudio.com/forums/solid-state/250272-current-drive-loudspeakers.html

There is also need for 2 Zobel networks possibly 3 and explored here at post 17
http://www.diyaudio.com/forums/multi-way/190300-best-place-put-zobel-network-2.html

Cheers / Chris

[cloth-ears]

Thanks guys, I understand the concept. My question was has anybody else tried adding correction post x-over. Apparently not, still I’ll persevere :scratch:

[Reffc]

What is it you're trying to correct? You mention that the amp is trying to "correct" the crossover. My understanding is that current (fashionable) theories on current drive systems applied to loudspeaker design are meant to iron out the issues with voltage driven passive speakers (lower distortion and kinder loads at resonance points). This pre-supposes a hell of a lot though such as measured distortion from bass drivers; phase inaccuracies and issues and a host of things connected with loudspeaker design. Sounds like it could be a case of the tail wagging the dog unless one starts from the stand point of knowing exactly where the "problem" is in the first place, and more especially, the actual magnitude of that problem. Passive speakers and crossover networks (internal and external....ie mechanical/electrical/acoustic of the drive unit and any modifying circuits needed to achieve the summed acoustic result) can be made that dont present silly phase shifts at low frequency/higher current loads and also the impedance profile can be smoothed very effectively passively to help limit current swings. Seems a little like a solution looking for a problem when going active might be the next logical step?

[cloth-ears]

Thank you Reffc, all very good points well made. I don’t need to tell you about the effects of negative feedback on gain stability, linearity, frequency response, step response. This is all well and good within the confines of the amplifier itself. I was just experimenting with applying this correction after the crossover (a more holistic approach) to see what would happen. I was more than a little surprised to hear and measure the results.

So far I am encouraged enough to crack on, I feel a hint of serendipity coming round the corner :rolleyes:

[StanleyB]

You mentioned that you fed the signal back from a point after the x-over. But you don't say if it is from the bass, midrange, or treble.
Anyhow, the principle of what you have done is not unknown. I remember that there were amps on the market once that used a three wire speaker cable. One wire was part of the feedback loop from speaker to amp. It is so long ago, I just can't remember which amps they were. But I shall do a search.

[cloth-ears]

Hi Stanley, I’m using a pair of Tannoy DC1000 with their own crossovers.

I’m taking samples of both bass/mid and treble via a few OPA134's to decouple and remix.

Amp is an old but pretty good HK 730. The 730 only uses a small amount of feedback so it’s a simple task to get the correct balance. The damping factor is seriously low by today’s standards at about 30, other than that it performs quite well

[Reffc]

Thank you Reffc, all very good points well made. I don’t need to tell you about the effects of negative feedback on gain stability, linearity, frequency response, step response. This is all well and good within the confines of the amplifier itself. I was just experimenting with applying this correction after the crossover (a more holistic approach) to see what would happen. I was more than a little surprised to hear and measure the results.

So far I am encouraged enough to crack on, I feel a hint of serendipity coming round the corner :rolleyes:

That makes sense if by experiment you noticed improvement and just wanted to press on to see where it leads. The DC's (all of them) have an impedance profile like the Alps, but I'd say a DF of 30 is way more than adequate to control them. A third of that would likely be adequate with domestic cable run lengths and a reasonable efficient crossover circuit.

[cloth-ears]

Sorry Paul, I have to disagree with you there. A DF of 30 is insufficient to stop any cone with a high loss surround. A stiff paper cone perhaps. A DF of 200 or more would act like a shunt, and that is what is needed to keep the buggers under control.

[Reffc]

Sorry Paul, I have to disagree with you there. A DF of 30 is insufficient to stop any cone with a high loss surround. A stiff paper cone perhaps. A DF of 200 or more would act like a shunt, and that is what is needed to keep the buggers under control.

We may have to agree to disagree there ;). In reality you can get away with less. I am referring to net damping factor, accounting for all chain losses, not just the amp output impedance which most people use without considering crossover and speaker cable losses. So, for an amp with an output impedance of 0.03 Ohms, the 200 DF becomes more like 15 accounting for all losses prior to the drive unit (more in some cases) resulting in a true DF into say a 6 Ohm load of 15 with crossover losses and cable losses totalling say 0.35 to 0.4 Ohms which is realistic.

[cloth-ears]

Not quite sure how you came to those exact figures but I do agree that the DF at the end of the chain is lower than at the start, inversely proportional as it were. It is therefore important that you start with the biggest possible number in order to minimise the effects of these losses. Many manufactures realise this and aim high.

[Light Dependant Resistor]

For readers of this thread DF is Damping Factor and explained here:
http://www.crownaudio.com/media/pdf/amps/damping_factor.pdf

Cheers / Chris

[cloth-ears]

Thanks Chris, I sometimes forget that I'm a scientist and babble on a bit. :)

[Reffc]

Not quite sure how you came to those exact figures but I do agree that the DF at the end of the chain is lower than at the start, inversely proportional as it were. It is therefore important that you start with the biggest possible number in order to minimise the effects of these losses. Many manufactures realise this and aim high.

Just an example using realistic figures based upon what you might expect from a SS amp. Crossover losses are straightforward to explain and a typical air cored LF inductor might range from .25 up to .7 ish of an ohm depending on value, number of turns and gauge and cored inductors much less, so I took a value of about 0.3 Ohms as a typical average and added on a fraction more for speaker cables/connections (connection losses are usually minuscule). 6 Ohms is a very typical minimum impedance for an 8 Ohm nominal speaker load, and lowest impedance usually (but not always) occurs in the bass frequencies, so is a reasonable figure.

DF = speaker impedance divided by total output circuit (ie amp output impedance plus speaker cable losses plus crossover losses)

So we have: 6 / (.03 + say 0.1 Ohm total cable/connection losses + 0.3) = 13.95 (I just rounded up to 15 as a handy figure before). If speaker crossover and cable losses are ignored, we would have had DF= 6/0.03 = 200.

That's how misleading quoting straight DF losses as a function of speaker to amp impedance is if the crossover and circuit losses are ignored, which of course they cannot be. That's why I stated that in reality a net DF of 30 would actually be reasonable.

[anthonyTD]

I did a thread on Damping Factor a while back, [Cant for the life of me remember where it is now] which outlined much of what Paul is stating, if you start off with an amp that has a high damping factor, it is only relevent if it connects directly to the drive unit, once you put an inductor, or any form of resistance in the way [even speaker cables] you have to re-calculate with everything in the path to get an actual DF figure, which in most cases turns out to be a fraction of what one started off with at the amp output!
A...

[cloth-ears]

I can’t believe that you are agreeing with me without actually agreeing with me (confused face).

YES, DF does reduce down the path to the coil; therefore unless I’m a complete idiot big numbers at the start are still bigger at the end than small numbers at the start.

If we go mad and say it reduces by 75% then 25% of 35 is 7.5 whereas 25% of 400 is 100

Big numbers at the start are better than small numbers at the start. What’s not to understand about that?:scratch:

[Reffc]

No one is arguing about that at all. Anthony and I are merely highlighting that it's the end (net) figure that matters.

It does make a difference because quoting DF without quoting net DF can be meaningless.

EG, lets say two amps, one a single ended valve amp with a =3 Ohm output impedance is driving 4 Ohm speakers via a simple first order network with an air cored inductor of 0.4 Ohms DCR and a SS amp with a 0.1 Ohm output impedance drivers the same load. Allowing some minor circuit losses, DF using the SE amp would be around 1.2 (woeful) and the SS amp would manage closer to 11, so an order of magnitude more. If the SE amp were replaced by a PP design with an output impedance of 0.5 Ohms, things would be closer.

There's a wide variation, but having measured frequency response from various loudspeakers (including my own designs) being driven by both SS and valve amps, a net figure on a demanding load of 20 to 30 is plenty enough for many loudspeakers. There are a few designs that require more, but I've yet to come across a speaker demanding a net DF of several hundred.