Hello,
I'm working on a home made amplifier, with a full differential architecture and a single rail voltage.
I use a 24V power supply, so that each speaker terminals are bias to 12V.
I use a kind of counter capacitance load composed of self with a small resistors in parallel (R78//L3 and R77//L4):
My amps presents some instabilities and I took a look on Zobel output network, but I don't know how to design it (values of components) and how to arrange it.
In the example above I connect the traditional 10R (in my schema it's 5 R59) in series with the 100nF (C16) in // with the load (R28).
R78/L3 and R77/L4 are parts of the amp and the speaker is represented by R28.
The other way I can connect is the following:
really I don't know how to choose, even if this will help me with instabilities issue (I'm unstable even with no load).
I understand well the benefice of self with R in // (R78//L3 and R77//L4): when I plot phase with spice I could see the phase and gain margin added by those components but for the other branch (R&C in series) I don't see anything when plotting.. (only a very very few difference@frequency above Mhz).
Also my design is almost full smd, I use for R78 and R77 smd 0805 resistors, from my point of view (power dissipation) that's good but I see some large resistors on a lots of design, don't know why is anybody can light me about that 🙂
I'm working on a home made amplifier, with a full differential architecture and a single rail voltage.
I use a 24V power supply, so that each speaker terminals are bias to 12V.
I use a kind of counter capacitance load composed of self with a small resistors in parallel (R78//L3 and R77//L4):
My amps presents some instabilities and I took a look on Zobel output network, but I don't know how to design it (values of components) and how to arrange it.
In the example above I connect the traditional 10R (in my schema it's 5 R59) in series with the 100nF (C16) in // with the load (R28).
R78/L3 and R77/L4 are parts of the amp and the speaker is represented by R28.
The other way I can connect is the following:
really I don't know how to choose, even if this will help me with instabilities issue (I'm unstable even with no load).
I understand well the benefice of self with R in // (R78//L3 and R77//L4): when I plot phase with spice I could see the phase and gain margin added by those components but for the other branch (R&C in series) I don't see anything when plotting.. (only a very very few difference@frequency above Mhz).
Also my design is almost full smd, I use for R78 and R77 smd 0805 resistors, from my point of view (power dissipation) that's good but I see some large resistors on a lots of design, don't know why is anybody can light me about that 🙂
Can it oscillate only in differential mode or also in common mode? Do you see any impact of the RC series networks when the load is disconnected or inductive?
The dissipation of the resistors depends on frequency. At high frequencies, they dissipate much more power than at low frequencies.
The dissipation of the resistors depends on frequency. At high frequencies, they dissipate much more power than at low frequencies.
full circuit is a little bit complex:
Attachment removed at OP's request.
I use a kind of 'push pull' scheme with only nmos mosfet but with single ended drive. That's the purpose of opa who process current image to drive foot nmos.
Purpose of this scheme is to test a strange way to stay in a class, I can't explain here in public because I think about a patent.
I already have a first prototype of this amp which works good, I build a second prototype with bug fix which oscillate, then I think I was lucky with the first run and my design is certainly borderline in terms of stability despite the good results in spice simulation. Especially the layout around opamp is certainly not very good ('long' tracks for inverted input).
That's why I interrogate about zobel network which looks like a black magic box for me and would like to understand.
I don't know how to test if oscillations are cm or diff mode ?
Oscillate for me: noise about +/- 5-7V p-p in output when no load, and the same with a pure resistive load.
Initial prototype show very good results in terms of noise about +/-10µV p-p (approx the min my QA404 can measure)
Attachment removed at OP's request.I use a kind of 'push pull' scheme with only nmos mosfet but with single ended drive. That's the purpose of opa who process current image to drive foot nmos.
Purpose of this scheme is to test a strange way to stay in a class, I can't explain here in public because I think about a patent.
I already have a first prototype of this amp which works good, I build a second prototype with bug fix which oscillate, then I think I was lucky with the first run and my design is certainly borderline in terms of stability despite the good results in spice simulation. Especially the layout around opamp is certainly not very good ('long' tracks for inverted input).
That's why I interrogate about zobel network which looks like a black magic box for me and would like to understand.
I don't know how to test if oscillations are cm or diff mode ?
Oscillate for me: noise about +/- 5-7V p-p in output when no load, and the same with a pure resistive load.
Initial prototype show very good results in terms of noise about +/-10µV p-p (approx the min my QA404 can measure)
Last edited by a moderator:
Don't see any kind of compensation in your circuit.
This is necessary to control the location of the dominant pole.
Oscillation in a real amplifier is inevitable without it.
The Zobel networks are a separate issue from this.
This is necessary to control the location of the dominant pole.
Oscillation in a real amplifier is inevitable without it.
The Zobel networks are a separate issue from this.
full circuit is a little bit complex:
I use a kind of 'push pull' scheme with only nmos mosfet but with single ended drive. That's the purpose of opa who process current image to drive foot nmos.
Purpose of this scheme is to test a strange way to stay in a class, I can't explain here in public because I think about a patent.
I think you have already published way too much on a public forum to have the slightest chance of getting a patent, or at least a valid one.
In real life (as opposed to simulation): connect a two-channel scope to the outputs and see if they move in phase or in antiphase.
In simulation: run a transient analysis, preferably with the trapezium integration method, push small and equal current steps into both outputs and see what happens with the two output voltages. Repeat it with one of the current steps in the opposite direction.
Common mode means equal signals on both sides, differential mode means opposite signals. The attenuation of the feedback signal by R20 is not effective in common mode, and the load impedance also doesn't load the output for common mode. The loop with the op-amps also doesn't do anything in common mode. Hence, the stability can be different for common and differential mode.
Last edited by a moderator:
Ok, so will remove this piece of schemas, but there is nothing extraordinary in, I remove the 'magic' part.
First pole of this architecture is fairly high (in the 1 Mhz), then stability is acquired with only BW limitation, thanks to diamond buffer input and specific I to V transducer made with current miror (a technique described in some doc relative to diamond buffer).
@marcel if you can edit your post to remove the quote of my post with schema, so that I can edit it and remove picture 🙂
First pole of this architecture is fairly high (in the 1 Mhz), then stability is acquired with only BW limitation, thanks to diamond buffer input and specific I to V transducer made with current miror (a technique described in some doc relative to diamond buffer).
@marcel if you can edit your post to remove the quote of my post with schema, so that I can edit it and remove picture 🙂
Amplifier needs to be stable as mentioned.
More likely the issue.
Stabilization networks on amplifier output is good practice
but is added to amplifier which is already stable.
They reduce further risk of difficult inductive loads or long cable runs
Causing further issues. Not a crutch to stabilize a amplifier.
More a safety precaution if already stable amplifier becomes unstable from
speaker load. Mainly multi way designs with passive crossover, or installations
like live audio where long cables are used for speaker location.
Large resistors are needed because because high frequency bursts of oscillation in
the Zobel or bucherot will dissipate, high burst of energy. Depending on amplifier
output power. In general any amplifier in continuous or short burst of oscillation
will dissipate lots of energy in output devices and any filter on output behaving as a load.
Basic idea is stabilization filters on output not conductive unless circuit is oscillating
then when conducting reduce the ringing to keep amplifier stable. The energy goes somewhere
so resistors and inductors need to be able to handle power levels seen.
Far as what is needed to stabilize amplifier design.
Didnt see circuit before removed.
Judging from hints of push pull design and diamond buffer.
Can only imagine a unseen current war which needs to be tamed.
May require additional capacitors as mentioned and even diodes
to make current behave. Another generalized assumption is maybe
more current is needed somewhere in circuit. Specially diamond buffer
since current mainly is traded over to current source transistors not
buffer transistors like typical gain stage.
Model is likely stable.
I test models with non ideal voltage sources.
And prefer model be unstable by adding internal resistance to voltage sources.
Or series resistance depending on what your software model allows.
I like model to be unstable intentionally, so power supply decoupling capacitors
are needed in model like real life to be stable.
Then find hidden ringing . likely high frequency issue.
Observe model with squarewave from 10k to 20k.
To find ringing that may have been missed observing a seemingly stable model
again assumptions with a mosfet design,
device transition frequency very high and can difficult to tame.
Also might consider high frequency transaction point changes dramatically
might require different better matched mosfet if high frequency cross conduction is issue.
Again assuming capacitance stabilization needed
and also a current war or lack of enough current.
More likely the issue.
Stabilization networks on amplifier output is good practice
but is added to amplifier which is already stable.
They reduce further risk of difficult inductive loads or long cable runs
Causing further issues. Not a crutch to stabilize a amplifier.
More a safety precaution if already stable amplifier becomes unstable from
speaker load. Mainly multi way designs with passive crossover, or installations
like live audio where long cables are used for speaker location.
Large resistors are needed because because high frequency bursts of oscillation in
the Zobel or bucherot will dissipate, high burst of energy. Depending on amplifier
output power. In general any amplifier in continuous or short burst of oscillation
will dissipate lots of energy in output devices and any filter on output behaving as a load.
Basic idea is stabilization filters on output not conductive unless circuit is oscillating
then when conducting reduce the ringing to keep amplifier stable. The energy goes somewhere
so resistors and inductors need to be able to handle power levels seen.
Far as what is needed to stabilize amplifier design.
Didnt see circuit before removed.
Judging from hints of push pull design and diamond buffer.
Can only imagine a unseen current war which needs to be tamed.
May require additional capacitors as mentioned and even diodes
to make current behave. Another generalized assumption is maybe
more current is needed somewhere in circuit. Specially diamond buffer
since current mainly is traded over to current source transistors not
buffer transistors like typical gain stage.
Model is likely stable.
I test models with non ideal voltage sources.
And prefer model be unstable by adding internal resistance to voltage sources.
Or series resistance depending on what your software model allows.
I like model to be unstable intentionally, so power supply decoupling capacitors
are needed in model like real life to be stable.
Then find hidden ringing . likely high frequency issue.
Observe model with squarewave from 10k to 20k.
To find ringing that may have been missed observing a seemingly stable model
again assumptions with a mosfet design,
device transition frequency very high and can difficult to tame.
Also might consider high frequency transaction point changes dramatically
might require different better matched mosfet if high frequency cross conduction is issue.
Again assuming capacitance stabilization needed
and also a current war or lack of enough current.
Hmm interesting tips WhiteDragon 😎
Ok, Zobel network is only to help when load become too much capacitive, it's a way to dissipate energy generated by oscillations due to a resonance made by load impedance and amplifier output impedance 'mismatch'.
And as you say a proper design is a stable design before speaking about zobel nw.
I understand that instability on classical negative fb amplifier is due to classic I/V stage traditionally made with a common emitter structure, common emitter have to be bw limited to obtain stability because of it's positive fb natural behavior, that's I understand about amplifier stability. Nor the current gain output stage or diff amplifier input stage are more involved in global stability than the i/v transceiver ?
The usage of current mirror structure embedded in a diamond buffer to made I/V transceiver is for my point of view naturally stable, just like the UGS design:
I/V stage is made with current mirror with ZTX450 and 550, generating voltage with amplified current through Routx resistors.
I don't use Pass topology, I use a less complex with diamond buffer as input stage. As you can see there is no stability capacitor to limit Bw in such design and it's the beauty of such design.
Ok, Zobel network is only to help when load become too much capacitive, it's a way to dissipate energy generated by oscillations due to a resonance made by load impedance and amplifier output impedance 'mismatch'.
And as you say a proper design is a stable design before speaking about zobel nw.
I understand that instability on classical negative fb amplifier is due to classic I/V stage traditionally made with a common emitter structure, common emitter have to be bw limited to obtain stability because of it's positive fb natural behavior, that's I understand about amplifier stability. Nor the current gain output stage or diff amplifier input stage are more involved in global stability than the i/v transceiver ?
The usage of current mirror structure embedded in a diamond buffer to made I/V transceiver is for my point of view naturally stable, just like the UGS design:
I/V stage is made with current mirror with ZTX450 and 550, generating voltage with amplified current through Routx resistors.
I don't use Pass topology, I use a less complex with diamond buffer as input stage. As you can see there is no stability capacitor to limit Bw in such design and it's the beauty of such design.
Even if you get you patent through (and that alone costs a lot), you would need to make so much money out of it that it justifies to pay the per year fees for at least China, Japan, Korea, Europe and the US. And if anybody claims something against you, you need years to clear the case, and you will be blocked from selling your idea for that time. Big companies do this all the time: they know they will loose the case, but they prefer to block the smaller competitor from the market. So in many cases it's better to hide the idea instead of publishing a patent. A big company would copy your idea and claim that their solution is significantly different, and you would need years if not decades to prove them wrong which will be almost impossible. On top of that in many areas you win a patent case only if your product creates jobs in the economy of that area. That's the reason why Omega produces at least some parts in the US, to name just one example...
Interesting feedback, this is a high risk, especially for a tiny company, to go on patent.
Trade off between publicity / light that a patent can give versus risk (chat about you are speaking)...
An other point of view: hom about protect you if big competitor company reverse eng your product and patent the idea ?
In french we have 'enveloppe soleau' a kind of letter with state certified creation date which can prevent bad big company to block small competitor selling the idea (of course of the enveloppe soleau owner is the small competitor). Don't know is it can works for foreign country. Off course I have a enveloppe soleau with my idea.
Trade off between publicity / light that a patent can give versus risk (chat about you are speaking)...
An other point of view: hom about protect you if big competitor company reverse eng your product and patent the idea ?
In french we have 'enveloppe soleau' a kind of letter with state certified creation date which can prevent bad big company to block small competitor selling the idea (of course of the enveloppe soleau owner is the small competitor). Don't know is it can works for foreign country. Off course I have a enveloppe soleau with my idea.
If a big company uses your idea they can only patent it if there is enough difference from your solution: so they at least will not be able to block you from using your own idea. Actually to publish your idea somewhere would be a possible way to create some basic protection: and many do exactly this. Another possibility would be to write a draft of the patent and offer this to a big company: in that case you would need to make sure that they don't use your idea, so some kind of contract is needed - not easy, but still much simpler than a global patent protection.
The problem with all of that is: innovation is often not appreciated in the market. So patent systems are widely used to mark down business territory, not to protect innovation. There are agencies that make their lives by selling bundles of patents to big companies, and those big companies don't buy the bundles because they want to use the innovations but rather to expand their territory. The result: innovation is blocked.
Many try to hide their idea, in case of a circuit they would encapsulate it into epoxy (i.e. in the car industry), to make it impossible or extremely difficult for the competitor to reverse engineer it. But that's also a common way of using someone else's patent.
Hence you would need to make sure that nobody could claim a misuse of an existing patent. A combined solution would be to publish an article with the innovation, describing your actual solution only in general lines, and sell your product as an encapsulated module which is difficult to reverse engineer. That way you have your protection in your own hands.
The problem with all of that is: innovation is often not appreciated in the market. So patent systems are widely used to mark down business territory, not to protect innovation. There are agencies that make their lives by selling bundles of patents to big companies, and those big companies don't buy the bundles because they want to use the innovations but rather to expand their territory. The result: innovation is blocked.
Many try to hide their idea, in case of a circuit they would encapsulate it into epoxy (i.e. in the car industry), to make it impossible or extremely difficult for the competitor to reverse engineer it. But that's also a common way of using someone else's patent.
Hence you would need to make sure that nobody could claim a misuse of an existing patent. A combined solution would be to publish an article with the innovation, describing your actual solution only in general lines, and sell your product as an encapsulated module which is difficult to reverse engineer. That way you have your protection in your own hands.
Last edited:
Even integrated circuits that sell well get reverse engineered. They etch open the package, deprocess the chip (remove layer by layer) and end up with the layout and schematics automatically generated from the layout.
Indeed I'm aware of pugnacity of people to reverse...
jpk73 the other way you speak about are very interesting, offer to big company or publish.
Anyway I'm in the good mood today, only technic, found why a prototype works good and the new one didn't.
Objective for me is to leave my consulting job to be 100% on amplifier design, not easy but the hunger for working on his own device with his own design and own rules is a so beautiful challenge.
The prototype I speak about, off course not a game changer black magic inside but enough to go for a patent, or not 🙂
jpk73 the other way you speak about are very interesting, offer to big company or publish.
Anyway I'm in the good mood today, only technic, found why a prototype works good and the new one didn't.
Objective for me is to leave my consulting job to be 100% on amplifier design, not easy but the hunger for working on his own device with his own design and own rules is a so beautiful challenge.
The prototype I speak about, off course not a game changer black magic inside but enough to go for a patent, or not 🙂