Is there anybody built a non feedback amplifier??

Status
Not open for further replies.
janneman said:

This is utter nonsense. Read the patent (or better, don't bother).
Either the load on the lower amp is not equal to the speaker load, and then the feedback signal has no bearing on the output signal and isn't doing anything for the oputput signal.

Or the load on the lower amp EXACTLY mimics the speaker load, and in that case you might as well take the feedback from the speaker and save yourself an output stage. Really...

Jan Didden

Hi Jan,,

of course the feedback signal is not doing anything for the output signal. We are talking about zero feedback.

The load does not mimic the speaker, again, we are talking about ZERO FEEDBACK. How about waking up plus thinking several octaves higher before you try to diss my amp topology.

Thanks,

Charles 😉
 
Altmann's SPLIF technology was criticised by some forum members. Altmann wrote that speakers do not like be watched by NFB because they become a little nervous, as you dear forum member, when your chief is standing behind you and looking over your shoulders. You get your balance when chief (NFB) will be next room to yours; a kind of voodoo, nes pa?
 
mandat said:
Altmann's SPLIF technology was criticised by some forum members. Altmann wrote that speakers do not like be watched by NFB because they become a little nervous, as you dear forum member, when your chief is standing behind you and looking over your shoulders. You get your balance when chief (NFB) will be next room to yours; a kind of voodoo, nes pa?


To put it in other words:

A worker that needs to be watched all the time by his boss in order to do his job. Would he be considered a good worker ?

If you go to the bathroom, do you need control ? Would you need somebody standing behind you, in order to let it flow ?

Charles 😉
 
I am a bit suspicious with the "great benefit" of Altman SPLIF topology.

For example if there is a reactive load, which loudspeakers use to be, then the crossoverdistortion can be as far as 60 degrees or more out at the loudspeaker-output pole regarding to the NFB-output pole, because it's not the zero voltage crossing, but the zero current crossing which will give the crossover distortion because it's around the zero current crossing point the outputstage is working in class-A (if we have a class-AB amplifier)and not the zero voltage crossing point (am I thinking right here?).

...so it must be very good amplifier even without feedback, so why not skipping FB, or take the feedback from the previous driving stage driving the output stage if we want to "isolate" the output stage.

Michael 😉
 
Ultima Thule said:
I am a bit suspicious with the "great benefit" of Altman SPLIF topology.

For example if there is a reactive load, which loudspeakers use to be, then the crossoverdistortion can be as far as 60 degrees or more out at the loudspeaker-output pole regarding to the NFB-output pole, because it's not the zero voltage crossing, but the zero current crossing which will give the crossover distortion because it's around the zero current crossing point the outputstage is working in class-A (if we have a class-AB amplifier)and not the zero voltage crossing point (am I thinking right here?).

Hi Michael,,

the Splif amp works only 100% isolating with Mosfet power transistors. With bipolars the base current would manifest as feedback.

But with Mosfets crossover distortion is not as prominent as with the harder switch-off behaviour of bipolars. The Mos transistors has a comparably very large cutoff region similar to a valve. So I would not give that an issue. You can also adjust bias to compensate this, plus you can adjust the linear dummy load.

QUOTE]
...so it must be very good amplifier even without feedback, so why not skipping FB, or take the feedback from the previous driving stage driving the output stage if we want to "isolate" the output stage.

Michael 😉
[/QUOTE]

If you take FB from the driver stage only, then you leave the output stage out of the FB-loop, so you will get much more sonic signature and coloration from your output transistors.

With the SPLIF design, output transistor distortion (nonlinearity, coloration, etc) is handled by the feedback loop of the dummy-load, that includes a copy of the speaker driving output transistors.

This way, I would consider the SPLIF concept the least distorting zero-feedback amplifier topology available today.

Charles 🙂
 
Charles said:
Hi Jan,,

of course the feedback signal is not doing anything for the output signal. We are talking about zero feedback.

The load does not mimic the speaker, again, we are talking about ZERO FEEDBACK. How about waking up plus thinking several octaves higher before you try to diss my amp topology.

Thanks,

Charles 😉

Hi Charles,

I think we miscommunicated here. In reaching my conclusion, I skipped a couple of what I thought were obviousl steps, but it seems I took too larges steps.

Right. Using the feedback, you create a "distortion free" signal at the output of block 10, which drives the resistive load. Since block 10 and 4 are (ideally) identical, the assumption is that the output of block 4, which drives the speakerload, is smilarliy distortion free.

Let us suppose this is true, that would mean we have the same signal at the resistive load and the speakerload. We can then connect the two outputs and nothing will change. This will demonstrate that the speaker is controlled by the feedback, therefor we have a conventional global feedback amplifier and we can save ourselfs an output stage.

Alternatively, let us suppose that it doesn't work that way, and that the output signal at the resistive load is different than at the speaker load. In fact, I think this will be the case because the open-loop output of block 4 will surely be influenced by the varying speaker load, without the feedback in the other loop seeing it. That would mean that the feedback does nothing for the speaker signal and therefore can be deleted. In fact, I feel that the feedback makes things worse because in correcting block 10 it will generate a "correction" signal that will appear at the output of block 3. This is not the same as the correction required by the different load (speaker) on block 4. Therefore, what the feedback does is:
- nothing to correct block 4, and
- inject an distorted correction signal for block 10 into block 4 and the speaker.

In reading Ultima Thule's post above, I feel he sees it similarly.

Jan Didden
 
janneman said:
Using the feedback, you create a "distortion free" signal at the output of block 10, which drives the resistive load. Since block 10 and 4 are (ideally) identical, the assumption is that the output of block 4, which drives the speakerload, is smilarliy distortion free.

Let us suppose this is true, that would mean we have the same signal at the resistive load and the speakerload. We can then connect the two outputs and nothing will change. This will demonstrate that the speaker is controlled by the feedback, therefor we have a conventional global feedback amplifier and we can save ourselfs an output stage.

Alternatively, let us suppose that it doesn't work that way, and that the output signal at the resistive load is different than at the speaker load.

Hi Jan,,

I guess you are looking at the signal in terms of voltage.

Of course the output voltage at 10 is different, as we have different loads. One pure resistance, and one reactive (speaker) load.

In fact, I think this will be the case because the open-loop output of block 4 will surely be influenced by the varying speaker load, without the feedback in the other loop seeing it. That would mean that the feedback does nothing for the speaker signal and therefore can be deleted.

Of course, as I said, we are talking about a zero feedback amplifier.

In fact, I feel that the feedback makes things worse because in correcting block 10 it will generate a "correction" signal that will appear at the output of block 3. This is not the same as the correction required by the different load (speaker) on block 4. Therefore, what the feedback does is:
- nothing to correct block 4, and
- inject an distorted correction signal for block 10 into block 4 and the speaker.

You are now in near proximity of understanding my idea.

At 3 you find an output drive signal with the ability to null the output stage's distortion, without being reactive to speaker behaviour.

That is the key feature that the SPLIF topology can achieve where the other no-feedback approaches fail.


Charles 🙂
 
Charles said:
[snip]At 3 you find an output drive signal with the ability to null the output stage's distortion, without being reactive to speaker behaviour.

That is the key feature that the SPLIF topology can achieve where the other no-feedback approaches fail.


Charles 🙂

No, sorry. I will find at the output of 3 a signal that can null the output stage 10's distortion, which is not at all the same as the speaker driving block 4 distortion. Because, if it was, then the output signals of 10 and 4 would be the same, and you would have a conventional global feedback amp as I noted in my previous post.

I also feel you make the underlying assumption (but please correct me if I am wrong) that the speaker-back EMF injection into the global feedback loop in a conventional case is somehow worse than the influence of the speaker impedance varying all over the place (modulus and phase-angle) on the output linearity, in combination with an openloop output Z which also varies all over the place as well.

Jan Didden
 
Hi Charles,

first I must admit I didn't realize that you are the designer behind the SPLIF invention, sorry if my post might been interpreted as offensive.

Anyhow, I have a question if you do know what is the damping factor of your amplifier at the speaker output node?

I still think if it's not a really rugged design with high OP stage biasing etc. that the output signal will be sacrificed by the reactive load of a loudspeaker.

If the speaker output node has a current out of phase it must be obvious that the output signal will be distorted.

Lets look at the point the voltage is crossing zero, at the FB output node the current will be zero because of the resistive load, but at the speaker output with a lagging or leading current the output FET's will surely have a diffrent DS voltage than the FB output FET with ZERO current.

You could easily do an X/Y-ing with an oscilloscope between the both output (eg. the speaker output and the FB output), I would be curious to know what the result will be.

Whatever the kind of FB or not, the output will be distorted by a reactive load, and that's why I said it must be a verry good amplifier design which should have an inherent low distortion even without any kind of FB among other things in a way allready mentioned here.
(But, observe, I wont let it to be a semantic question about what FB is, some have allready on this forum claimed the EF to be a FB and let it be so for him or her.)


Do we have Charles Hansen somwhere here around!? :wave2:


Michael
 
Yes Pavel,

that's my thought too.

Susan Parker & Co. are also very pleased with high output impedance amplifier which gives "special effects" in combination with reactive speaker loads! 😎

No offense Susan! 🙂


Michael
 
PMA said:
Michael,

even if it was very very good output stage (class A very linear), it would be impossible to completely eliminate speaker impedance influence. You know that I have done similar listening tests, and the results depended on listener's taste to great extent.

Pavel,

I understand Charles point was that it surely eliminated back-EMF injection in the feedback loop, and that I can agree to. But I don't think that that is worth the price paid in terms of forward linearity (or lack thereof). I think it is indeed a very clever idea, but it concentrates too much on ONE issue thereby letting (IMHO) worse things in through the back door.

Jan Didden
 
Status
Not open for further replies.