3 stage LIN topology - NFB tappings?

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hi,
Doug Self makes a big play on where to tap off the upper leg of the NFB from the output. But I cannot recall anyone suggesting where best to tap off both the upper leg of the NFB and the lower leg of the NFB.

When the Amplifier is drawn in the same format as a balanced input to unbalanced output, it becomes obvious that the output leg could tap off the speaker terminal (but before the stabilising Thiele network) or as Dr Cherry suggests between the Zobel RC, but where should the lower leg go to?

All the way back to the source? then NFB compensates for the interconnecting cable?
or locate the lower leg resistor at the source component (and the source resistor controls the amp gain, =input 3),
or tapping into the RCA ground at the input to the amplfier casing (= input 1),
or as conventionally done, just to the PCB signal ground point? (=input 2)

Which is likely to give best results and which may be adversely affected by long wires?
 

Attachments

  • nfb options 1000.png
    nfb options 1000.png
    27.8 KB · Views: 2,633
Well, I didn't figure that part out. ;)

The point of my (above) post is that any resistor sturdy enough to last more than a month on the zobel is also the same sort that would put infinite gain on HF noise if used for the hi leg of the NFB. So, I thought that also using one of those (exact same model) for the lo leg of the NFB might cancel that potential problem.

Sorry that the response took a bit. I model via soldering iron, because "applied" is the only thing that makes sense to me.

And, here's a guess on the input area. Once again that sturdy power resistor was a problem for me as it couldn't provide a sufficient protection against HF noise if used for the input ground, so here's (photo below) the homely bus ground system that results from chasing this little issue about.

This (pictured below) is obviously wrong. The return line from the load is pack full of AC (audio that we just amplified).
If that 8R load involves a speaker with a crossover involved (an even worse noise on the speaker return line), I'd like to show load/speaker return line connecting to an offboard power supply (which is, hopefully, the only location of caps large enough to mash the unwelcome signal into DC, so we don't mistakenly amplify it)--but I didn't know how to show that.

Is there not some way to employ a diode to protect the amplifier against the return signal from the speaker? I didn't know how to do that either.
But, if our power supply is either regulated or switcher, then it probably won't have the 6000uF (minimum) resource to neutralize the speaker's very powerful return signals. Without that resource (or "something" to do that same job), the bass response is awful. I wouldn't put large capacitance at the amplifier board because that's yet another way to make a noise. Howabout a diode to neutralize the load/speaker return signal? Help?
 

Attachments

  • alternativetap.jpg
    alternativetap.jpg
    17 KB · Views: 1,999
Hi T,
it is not a balanced input. It is only drawn to look like that for the sake of clarity so formumites can see the circuits are identical (except Cherry a/b)
What I am asking is which if any are better for an NFB topology and whether any of them are bad.
 
Hi Daniel,
the gobblede gook has lost me again.
But back to your last posted diagram.
Show the input ground attached to the left side of C2+R2.
Move the signal ground connection from the far end of R6+C1 to between R3 & R6.
You now have a working circuit that looks like input 2 but with the power ground connected to signal ground with a 12r resistor just like Leach and many others (except you show the Cherry/b NFB point).
 
Administrator
Joined 2007
Paid Member
Hi Andrew,
Going back to post#1 I followed Dougs work at the time but there were aspects I couldn't "reconcile" and I think your asking the same question.
I found it helped to just think of it as a DC problem, but the real issue was "what happens when the input two grounds of a stereo pair" are connected together, wherever that may be.
Does working it as DC problem help if you put in nominal resistance values for all the wiring. I found it did :)
 
AndrewT said:
Hi T,
it is not a balanced input. It is only drawn to look like that for the sake of clarity so formumites can see the circuits are identical (except Cherry a/b)
What I am asking is which if any are better for an NFB topology and whether any of them are bad.

I may be missing the point (it wouldn't be my first time!). Ignoring the Cherry switch for the moment...I sort of look at your circuits like the attached. So the output is going to be approximately proportional to (Va - Vb). You seem to be asking where the best place is to attach Vb to signal ground, and where to locate R3.

With NFB you have to worry about at least two things: loop stability and noise. Provided the op-amp is unity gain stable, then the circuit stands a good chance of being stable regardless of where Vb is grounded, noise notwithstanding. Noise is important and will affect both terminals of the op-amp. To minimize it usually requires the track areas of those pins to be minimized. So R1, R2 & C2 should be as close as possible to the + pin and R3, R4 & C3 as close as possible to the - pin. So I would be inclined to rule out your "input 3" option.

I think "input 1" is attempting to sense gnd at the source end of a cable so as to reduce cable effects. I think that will only work if any unwanted noise/voltage drop on the signal part of the cable is equal to that on the gnd part. This is why I pointed out that the impedances are different.

Any help?
 

Attachments

  • at2.jpg
    at2.jpg
    13.5 KB · Views: 1,909
Noise at great force.

Any effort we do to send a lovely signal to the load (speaker) then comes back as a fiercely strong loud noise via the load (speaker) ground line!
Gobbledegook + amperage, is indeed, a description of the problem.
So, please don't ground the NFB at that point. :D

Input ground versus NFB ground:
At high output usage (or transients), the load return line (speaker ground) will lift the Input ground, with terrible results if Input ground goes same or higher than NFB ground (unbalanced non-inverting amplifier, as shown).
Not traces, nor cables, nor rectifier, nor transformer is sturdy enough to prevent this.
A 1/2w or larger carbon resistor (groundlift) added to the NFB ground leg is daffy-looking but works in practice (retards capacitance / low capacitance). A Caddock price is unnecessary when a half penny, half watt, will retard capacitance just as well. That's an exploitation of component variances, just like the Cherry switch.

Ditch the noise?
Instead of thinking like the year is 1920 and sending the unwanted AC signal (from speaker ground line!!) into the power supply board neutral, where it could be rectified (flattened) by big capacitors (if using a linear supply so equipped), lets please figure out how to do that with a diode?
If my imagination is working correctly, that's two series pairs of diodes required to ground Either the speaker OR NFB to V+ and V-, with a creative virtural ground--And thus the point is keeping load return (speaker ground) noise permanently away from the NFB ground.
Otherwise. . .

However, if you really want to restrict this to linear supply, then this might do (photo below). What do you think?
At this point, I'm a little lost. See photo. That's as far as I got.
 

Attachments

  • alternativetap.jpg
    alternativetap.jpg
    25.8 KB · Views: 1,863
AndrewT said:
. . .or locate the lower leg resistor at the source component (and the source resistor controls the amp gain, =input 3). . .

Hi man!
So sorry that it took such a long time.
I don't have much tools for this.

Each physical model that I've built operates best with the NFB ground farthest from the speaker ground (a noise). All attempt some of your #3 above, except that I always had the additional series resistor on the -input. Its all similar to a balanced amplifier that just happens to have an RCA jack.

In my opinion, the "long wire for - input" concern is a smaller concern. Its unknown if there will be a noise on that line, but it certain that its farther away from a known source of noise.

And, that's the best I can do, until someday there will be a method devised with diodes to make sure that the 0v line is never polluted with AC from the speaker ground.

Cheers!
 
My suggestion:

RCA ground wired to central ground, then to amplifier module as signal and feedback ground. This prevents any parasitic current to disturb the signal ground reference on the module. No DC should flow through this wire and it shouldn't be used for decoupling either.

Speaker ground wired to central ground, then to amplifier module as power and decoupling ground. This prevents signal ground to be disturbed by the currents flowing through decoupling capacitors and other elements.

Both grounds joined only for RF at the amplifier module through a RC network (like 10n and 15 ohms which becomes effective above 1Mhz).
 
Administrator
Joined 2007
Paid Member
Eva, Wavebourne, AndrewT, Daniel and all,
What's your take on this ?
This follows D Selfs work on his blameless amps and is the arrangement Doug recommends. Each amp is discrete, the grounding line on the PCB follows that in my drawing. The speaker ground is as shown on the PCB.
What I can't grasp is this. In isolation ( as a Mono amp )each amp is perfect.
What happens when the the two input grounds connect together at the source.
If the top amp is supplying say 5 amps positive into the load then the ground lead to the star will have a volt drop along it with reference to the star. No problem in itself. The lower amp will be at a different ground potential to the upper, so what happens when those input grounds connect. To me an unwanted volt drop is now developed along the ground track of both amps.
Hope you can follow that. This is something that really bugs me, I can't grasp this.
To me the speaker return should not be on the PCB for a stereo pair on a single PSU.
Help :)
 

Attachments

  • grounding.jpg
    grounding.jpg
    37.2 KB · Views: 1,417
Mooly said:
Eva, Wavebourne, AndrewT, Daniel and all,
What's your take on this ?
This follows D Selfs work on his blameless amps and is the arrangement Doug recommends. Each amp is discrete, the grounding line on the PCB follows that in my drawing. The speaker ground is as shown on the PCB.
What I can't grasp is this. In isolation ( as a Mono amp )each amp is perfect.
What happens when the the two input grounds connect together at the source.
If the top amp is supplying say 5 amps positive into the load then the ground lead to the star will have a volt drop along it with reference to the star. No problem in itself. The lower amp will be at a different ground potential to the upper, so what happens when those input grounds connect. To me an unwanted volt drop is now developed along the ground track of both amps.
Hope you can follow that. This is something that really bugs me, I can't grasp this.
To me the speaker return should not be on the PCB for a stereo pair on a single PSU.
Help :)
Back in 2000 I modified my Crimsons to match the D.Self grounding layout. It reduced the buzzing I had all those years (Crimson grounding is plain wrong). It did not eliminate the buzzing.

After joining this Forum I discovered that the speaker return to star ground was an option, voila!
This is closer to the Leach layout with RCA terminal being the signal ground and the main star ground (Audio Ground) where all come together.

It was this experimentation (I have tried with literally hundreds) that prompted the original question in this thread.
Where best to take the signal ground?

I have not tried input3 since it means modifying my preamps/sources to similar to pseudo balanced. But if it works, universally, then it's something I could and would do.
 
Administrator
Joined 2007
Paid Member
Hi Andrew,
I tried everything I could think of on Dougs amp. I always got a small buzz when those inputs were all shorted together.
The only answer ( but is it the right one :) ) to me is too connect the speaker grounds back to the star.
Doug talks of distortion figures in the 0.00xx region. This problem must absolutely dwarf that. I don't get it !!
If you look at my drawing and imagine the top amp is belting out 100 watts into 8 ohms. That puts a very measurable signal between the two (as yet not connected ) grounds of the two amps. If just the input of the lower amp is connected to the ground of the upper amp a very recognisable signal will now be fed into the lower amp.
Question ? If you connect the grounds together does or does not a "circulating" current flow in the loop that has now been created and this current is related to the audio. If you isolate the grounds with lift resistors doesn't this in itself cause the input to the amp to be modified slightly.
This is the "Final Frontier" for me. I have got to understand this :)
 
back to the original question, the reason that nothing is said about where the feedback ground returns to is, it'd not as important as where the feedback taps off from. there are significant currents and voltage drops present in the output stage wiring, which can be sources of error in the feedback. the input ground does not have these large currents circulating through it and so does not present a significant source of error.
 
Mooly said:
Hi Andrew,
I tried everything I could think of on Dougs amp. I always got a small buzz when those inputs were all shorted together.
The only answer ( but is it the right one :) ) to me is too connect the speaker grounds back to the star.
Doug talks of distortion figures in the 0.00xx region. This problem must absolutely dwarf that. I don't get it !!
If you look at my drawing and imagine the top amp is belting out 100 watts into 8 ohms. That puts a very measurable signal between the two (as yet not connected ) grounds of the two amps. If just the input of the lower amp is connected to the ground of the upper amp a very recognisable signal will now be fed into the lower amp.
Question ? If you connect the grounds together does or does not a "circulating" current flow in the loop that has now been created and this current is related to the audio. If you isolate the grounds with lift resistors doesn't this in itself cause the input to the amp to be modified slightly.
This is the "Final Frontier" for me. I have got to understand this :)

You are most wise if this is your final frontier! Your questions show you have a brain and aren't afraid to use it. Good.
Let me remind folks of the basics: voltages are relative and currents flow in circuits (loops) - at least at the frequencies we are concerned with here. It is really very simple, but often overlooked on the erroneous assumption that wires have negligible resistance.

Of course, if the input gnds of the two amps are connected via wires of finite resistance carrying high currents, the input gnds will be at different voltages as a result. And if you connect the two inputs to the same source component, you'll cause ground currents to flow to and from the source component through the interconnect shield. This is not good. One solution is to connect the input gnds to the star gnd directly using their own, dedicated wires. With some pcb layouts you'll need to cut the connection between input gnd and spkr gnd (or other grounds). You also have to be careful how you connect things at the star ground - for it is not a zero-resistance point, it is usually a bolt thread - the sequence of eyelets on the bolt matters too.

Keep questioning everything and thinking for yourself. In the audio world, notoriety and amount of publishing is no guarantee of competence.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.