Lars Clausen said:
I agree - especially for audio the inductor should be designed to be linear and normally nowhere close to saturating, but I strongly disagree that feedback can't or won't reduce what distortion may arise from the inductor.
No, not at all, the arbiter of truth is the real world - the scientific method, repeatability in the lab, the double blind listening tests ... simulation is only a very useful preliminary step.
Regards -- analogspiceman
OK, i can respect you for saying that. But ..... did you ever do some real THD measurements to support your point?
I have, and must say the benefit of including the inductor in the feedback loop did very little to improve the THD. Much less than i had expected from my own calculations and the general perception of the matter.
Sorry if i stepped in on something .. 😀
OK let's forget the discussion if nobody wants it anyway 😀
Have a nice evening everyone...............
sovadk said:
Point taken. Getting a low cost, high fidelity sense signal for current is much more difficult than for voltage. One could supplement the dc coupled scheme with current transformer sensing for all but the low frequencies, or perhaps one could sum in and switch over to the (ac) derivative of output capacitor voltage a la Mueta.
One of the potential benefits of current based control that hasn't been touched upon is the easy way it would allow paralleling outputs for more power (with phase staggering, of course).
By the way, what "everything else" were you alluding to? For the sense output MOSFET dc bias currents would cancel and capacitive switching currents probably would too (on average within the audio range). I'm guessing most nonlinearities would stem from the bipolar junctions.
Regards -- analogspiceman
Perhaps your inductor was spewing flux that coupled distortion into critical circuit board traces. This can be especially troublesome with non homogeneous (i.e. gapped) cores that saturate or partially saturate. A good test for this is to physically remove the inductor from the immediate vicinity of the circuit board by extending its board connection via a tightly twisted pair of leads that are run perpendicularly away from the board (and then repeat the distortion measurement).
Regards -- analogspiceman
You're right, everything else is not much.analogspiceman said:
I wonder if a high side driver and a bootstrap circuit would mess things up. I'm definitely going to work some more with that current sense circuit. I's plain simple and seams to do it's job.
What's you view on the realative difference on the UCD FFT and the leapfrog FFT? Leapfrog seams to have more high frequency distorsion as what was somewhat predicted.
analogspiceman said:
Hi Analogspiceman,
In my setup I have placed the Aircore inductor Off-Board...it seems its a good thing from your point of view.....
K a n w a r
A coreless inductor does not necessarily leak magnetic field. Configure it as a multilayer solenoid and you will have built a nice electromagnet, but configure it as a toroid, with enough winding strands evenly arranged to distribute the current in homogeneous curved sheets, and the field will be confined to entirely within the interior of the toroid. No leaks.
By the way, an air core inductor has a terrible L to R ratio, usually leading to excessive resistive losses.
Regards -- analogspiceman
Hi Analogspiceman,
I think the Aircore Inductors are virtually Unsaturable and hence produce less distortion also...
K a n w a r
I think the Aircore Inductors are virtually Unsaturable and hence produce less distortion also...
K a n w a r
It is my strong suspicion that the extra distortion comes from the current sense circuit and not from the leapfrog configuration per se. - a.s.
snip:as
It is my strong suspicion that the extra distortion comes from the current sense circuit and not from the leapfrog configuration per se. - a.s.
Hello analogspiceman
Your suspicion can be proven easily in a simulation.
And later, if it will be necessary: With 2 small shunts You can get
I C8 and Iout, with both sum up You get "I L1"
Regards
Heinz!
It is my strong suspicion that the extra distortion comes from the current sense circuit and not from the leapfrog configuration per se. - a.s.
Hello analogspiceman
Your suspicion can be proven easily in a simulation.
And later, if it will be necessary: With 2 small shunts You can get
I C8 and Iout, with both sum up You get "I L1"
Regards
Heinz!
Hello 
and all,
I also do a comparison between leapfrog and as-UCD and find that the as-UCD has less harmonics. Further I try to do current-fb with a (ideal) B-source : with that the higher harmonics are less!
and all,I also do a comparison between leapfrog and as-UCD and find that the as-UCD has less harmonics. Further I try to do current-fb with a (ideal) B-source : with that the higher harmonics are less!
Final I take current-fb from shunts that senses Iout and Ic8 (outputfilter) but with this the 3. harmonic is worsen!
Regards
Heinz!
Regards
Heinz!
The shuted leapfrog does seam to have less distortion. The carrier frequency is lower however -> less distortion due from imperfect switching but also less feedback.
Immediately leapfrog shunted seams to outperform UCD in distortion.
The only problem is that you need to amplify the voltage across the current sense resistor, so an opamp with a very high GBW and low noise is needed.
As I said ealrier, I think that sensing the bootstrap current, comperator current .... in addition to the inductor current, is what causes the distortion in analogspicemans leapfrog implementation.
Immediately leapfrog shunted seams to outperform UCD in distortion.
The only problem is that you need to amplify the voltage across the current sense resistor, so an opamp with a very high GBW and low noise is needed.
As I said ealrier, I think that sensing the bootstrap current, comperator current .... in addition to the inductor current, is what causes the distortion in analogspicemans leapfrog implementation.
Attachments
I might have to take that one back. Although the attachment shows a difference in the inductor current and the sum of the currents in R21 and R24 (the current sense resistors), the FFT plots shows no alerting frequency components.
I must be the current sense circuit that causes the additional distortion.
Simulation based on analogspicemans leapfrog curciut.
I must be the current sense circuit that causes the additional distortion.
Simulation based on analogspicemans leapfrog curciut.
Attachments
Hello Kaspar,
thank You for comparison.
Yes and also AS and think about , but...see Your next post...
thank You for comparison.
sovadk said:
Yes and also AS and think about , but...see Your next post...
Kaspar snip
I must be the current sense circuit that causes the additional distortion.
.....I try it separately and get at 10kHz only 0.02% distortion...!
I put in one leg of the current-fb a 1MHz lowpass ("equivalent to C7") : it seems that the harmonics get a bit smaller,
may be You try it also with a deeper frequency cutoff?
My problem with this whole pwm-stuff is
1. simulations take too long time, even with swcad and as´s leapfrog, and
2. I have still a lot problems to use swcad ! (not with my old EWB5.1 but this is for pwm-stuff too slowly!)
Please, can You explain how to put 3 ffts to one pic?
Regards
Heinz!
I must be the current sense circuit that causes the additional distortion.
.....I try it separately and get at 10kHz only 0.02% distortion...!
I put in one leg of the current-fb a 1MHz lowpass ("equivalent to C7") : it seems that the harmonics get a bit smaller,
may be You try it also with a deeper frequency cutoff?
My problem with this whole pwm-stuff is
1. simulations take too long time, even with swcad and as´s leapfrog, and
2. I have still a lot problems to use swcad ! (not with my old EWB5.1 but this is for pwm-stuff too slowly!)
Please, can You explain how to put 3 ffts to one pic?
Regards
Heinz!
It has to generate very low harmonic distortion as well. Distortion generated in the feedback path shows up proportionally in the output signal. It is better to use a differential forward path if differential input- and feedback- signals have to be processed.
Personnaly I don't like current feedback. It is a gainst the KISS principle. UcD on the other hand is the ultimate class-d amp from the KISS point-of-view. It's modulation scheme might not be IDEAL but as a whole it is OPTIMAL.
Regards
Charles
Have you ever heard about Paint Shop Pro? 🙂 LTspice lacks some options. The fact that you can't import data from an old simulation is one. I also bothers me that you can't copy from one schematic to another.powerbecker said:
I prefer Orcad PSpice for simulations.
You're right that simulation is slow and then it's not accurate even at small time steps. However you can do a relative comparison of the distortion, I think.
When I add additionally feedback arround a switching amplifier, I place a sinus generator at the output and mesure how well it's attenuated compared to before adding feedback. In that way I can mesure the relative improvement (you really need a small time step and a long simulation time to see the difference in THD). Often things turn out differently than I'd expected. It's not just about adding a lot of feedback. From observation it seams that it's the ratio between the openloop gain at the oscillation frequency and the openloop gain at a given audible frequency that determines the attenuation of the sinus disturbance. In other words carrier feedback messes things up.
Hello Charles,
snip
It has to generate very low harmonic distortion as well. Distortion generated in the feedback path shows up proportionally in the output signal. It is better to use a differential forward path if differential input- and feedback- signals have to be processed.
Yes indeed, but it was only a alternate idea to use it, for the case, that the original current-fb ads distortion....and a welcome opportunity to train a little bit the use of SwCad🙂
snip
Personnaly I don't like current feedback. It is a gainst the KISS principle.
What means KISS?
I agree with You in principle, if You had to design a voltage-amp the fb of (load depend) current is not good. On the other side You get a simple design which is currentlimited.
snip
(UCD)It's modulation scheme might not be IDEAL but as a whole it is OPTIMAL.
Also with this I agree.....BUT only in the case for low outputvoltages (and powers)!
If one need high outputvoltage or power, the use of min. a 2. filterstage is necessary to get a better rejection of the switching frequency.
Although I don´t know the limits I have strong doubts if not even
e.g. a UCD700 produce too much hf-pollution.
If You had to design a pwm-amp with 2stage outputfilter You cannot use the UCD-scheme....but You can use as´s Leapfrog-Design. (AND no patent, thank you as.!)
Regards
Heinz!
snip
It has to generate very low harmonic distortion as well. Distortion generated in the feedback path shows up proportionally in the output signal. It is better to use a differential forward path if differential input- and feedback- signals have to be processed.
Yes indeed, but it was only a alternate idea to use it, for the case, that the original current-fb ads distortion....and a welcome opportunity to train a little bit the use of SwCad🙂
snip
Personnaly I don't like current feedback. It is a gainst the KISS principle.
What means KISS?
I agree with You in principle, if You had to design a voltage-amp the fb of (load depend) current is not good. On the other side You get a simple design which is currentlimited.
snip
(UCD)It's modulation scheme might not be IDEAL but as a whole it is OPTIMAL.
Also with this I agree.....BUT only in the case for low outputvoltages (and powers)!
If one need high outputvoltage or power, the use of min. a 2. filterstage is necessary to get a better rejection of the switching frequency.
Although I don´t know the limits I have strong doubts if not even
e.g. a UCD700 produce too much hf-pollution.
If You had to design a pwm-amp with 2stage outputfilter You cannot use the UCD-scheme....but You can use as´s Leapfrog-Design. (AND no patent, thank you as.!)
Regards
Heinz!
- Status
- Not open for further replies.