That would depend on how clipping is avoided. Running near clipping for most of the time I take to mean that the average level is near clipping, as for most of the time music is near its average level. Maybe that is not what he meant, but that is what he said. If he meant to say that the music is at a much lower level, nowhere near clipping, then this tells us nothing about the PSU because it is not having to deliver very much current and there will be very little droop.
Precisely!
The amp power supply may be modulated pretty severely but that does not necessarily imply that amp is driven into clipping.
Also music and pure sine waves would behave quite differently and that is what Frank is referring to. If it is a sine wave, the the powers supply modulation would be in sync with the signal and may affect the way it sounds but it would not be horrible.
However, if it is a complex signal the power supply will be modulated most by the highest power signal which would "mix" with all other signals and besides smearing due to inter-modulation artifacts these artifacts could also contain dissonance which is most fatiguing because it is sounds our brains do not like.
Exactly what this thread is trying to address.
Last edited:
There has been a misunderstanding of what I meant, I most certainly am not talking about average levels of the signal being near clipping, rather whether the peak level ever clips. I'm very much aware that music has dynamics
, so when I say close to clipping I am referring to the piece of music as a whole; in the same way that a recording engineer will master a non-compressed CD so that the maximum level ever reached, at probably one point in the whole disk, is perhaps zero, one or two bits from the absolute maximum - the term used is to normalise the levels.I have got my chip amps to overload, but this was thermal overload, the sound starts to chatter as the protection cuts in and out; interestingly, this is never on pop but rather operatic -- the soprano hits a big note and sustains it, a very pure, high level tone over a decent period was enough to start overcooking the chips ...
As regards drooping, it is, or was, endemic to amplifiers. Back when I was really into trying to understand what caused sound quality problems I made a point of experimenting with every system in a dealer's showroom that I came across, and they all displayed this characteristic. Irrespective of how massive the metalwork was, there would be a certain volume where the sound would start to collapse, not clip, which I interpreted as the power supply being overtaxed ...
Frank
Last edited:
It would seem to me that what we need here is an absolute definition of clipping in this instance. I can understand when a power supply reaches the point where the sag in the rails is enough to change, or cut a waveform that the amplifier is trying to produce. A square wave that has rounded corners or a sine wave that has the tops cut off would make me think of clipping. On the other hand you could have a power supply so over-sized that you will never get into that condition but the amplifier is driven so hard it is out of the SOA and also is clipping in that the waveforms are distorted by thermal conditions. So who is going to define the actual term clipping as we are trying to use in this instance? I have always assumed that if you never reach the point where the loudest or sustained peaks do not become changed in any way that you are under the clipping level of the system. You have enough headroom to stay out of that condition.
Perhaps as good as any -- from the Rane dictionary: http://www.rane.com/par-c.html:
Frank
I am talking otherwise of the power supply voltage rails being sufficiently modulated by the current demands to interfere with the correct functioning of the amplifier circuitry: the PSRR not being good enough, or the regulation or smoothing being insufficient, causing the SQ to deteriorate ...
Frank
Thank you Frank for the definition. It appears that I had the correct idea about the term. Now I have a separate question. On an amplifier with a clipping light what exactly is the light indicating when it is triggered? I know it is called a clipping light but where in the circuit is the clipping indicator located and what is it actually measuring?
At least one take on an indicator, by that ever prolific chap, Rod Elliot, is here: Power Amp Clipping Indicator. It doesn't cover how it's often implemented in commercial units, but explains precisely how such a circuit should work ...
Frank
Frank
If you have an inverting (shunt feedback) amp, there's a slightly simpler way to detect clipping - put a window comparator on the inverting input. Normally the voltage here will be very low - the output voltage divided by the OL gain at any particular frequency. When the amp clips, this input is no longer a virtual earth. This method does away with the need for the clip detect circuit to look at the PSU rail voltage - that's now taken care of by the f/b network and amp itself. This method would also respond to slew limiting which can be thought of as a kind of temporal clipping.
I once built a compressor using this method to detect clipping. The basic premise was that I didn't want to adjust the gain until the amp was very near to clip. So I used a CMOS 'model' opamp (which swings between the rails) and ran this from a scaled down (using a resistor network) version of the main supply rails. The gain was adjusted so the CMOS opamp clipped around 80-90% of the output of the main amp and the inverting input was then amplified and fed into a diode pump to reduce the gain via a JFET element.
Bi-Amp/HT setup with the mids-n-treble section, "TDA chip" powered up to 20w on clean regulated power? Looks doable for hi-fi. Which chip is it?
Sorry to be so late in chiming in.
I would have to assume they meant it was not clipping after taking into account the dynamic range of the music.
On a related note, the point of the Cmin analysis is to find combinations of C, Vrail, VA, and Pmax that make it impossible for a square wave to clip at Pmax, even by a little bit, which would mean that no other type of signal could ever clip. And it can easily be done. Then it is guaranteed that the lowest trough of the ripple waveform cannot intrude into the headroom reserved for the signal (load) and amplifier voltages, as long as the output signal voltage peak stays at or below its specified maximum level.
I have a new spreadsheet that calculates it much more accurately, taking into account the transformer droops, and even the cap's ESR. I will try to upload it today or tomorrow.
I have been comparing its predictions with my simulation results, to validate it, and it seems to be able to account for everything, within about one percent. One of the last things that I still need to add is the variation of the rectifier voltage drop versus the average rectifier current.
In order to do the calculations needed, I had to resort to an iterative method, since some of the equations have no closed-form solution, and some are approximate until you work backwards to the transformer primary winding and then forward again through the rectifier, enough times. But it's quite compact and only needs three iterations, although I provide four because it's nice to see the smallness of the final changes in the values.
Cmin is good to know, for any combination of transformer size, output voltage, and Pmax. But I realize it's not what is needed when designing a power supply, except when it points out that a particular combination is marginal, or impossible.
So I will be using the information I have found, while researching for this, to create a more-design-oriented spreadsheet.
Cheers,
Tom
Last edited:
Sorry for the late reply, Daniel, bits of life happening ...
The HT setup uses a real basement unit, TDA7269A, specs look pretty dreadful! These fellows actually shut down completely, I has to do a power off/on reset, from memory, to get them going again. Otherwise, they do remarkably well because they directly drive a widerange speaker, no crossover to place extra current stress on them; with something like Foo Fighters I have no trouble getting my ears to start ringing within a few minutes ...
An earlier effort was my own LM3875 gainclone design, these chips start chattering when they thermally overload. These were driving decent sized 2 ways, so had to contend with crossover demands. The heatsinks were oversized for the job, but even so the transient power dissipation was too much for the heatsink contact area to get rid of fast enough, so the protection cut in ...
Frank
Considering the performance you got, it is mostly likely to be a singleton input amp or similar audio delight with dreadful specs. Those work best with clean power, low gain and a light load. . . and you did that.
So, when you added the regulators, what power caps did you put on the amp board, the 1000u in the datasheet or something more like 330u or maybe the 10u~47u range suggested by regulator datasheets?
Normal regulators don't help much at HF unless they're set a lot lower than the transformer voltage. What's the voltage difference between the regulated versus unregulated? You could add RC's across the transformer's primary and secondary. Is the bass amp powered by the same transformer as the mid-n-treble amp?
Do you have additional HF filtering at the amp, like a picofareds cap from in+ to in- (the biggest size that didn't change the audio) and/or at the RCA jack? I was thinking that maybe some additional filtering could prevent the amp from getting stuck in protection mode.
Last edited:
Don't confuse 2 different projects, Daniel, I'm confused enough by life as it is,
!! The HT used tweaking essentially with the supply as is, everything could be better in myriads of ways but I've done it a certain way to try some other ideas. The regulator technique was done on a LM3875 gainclone, which was all my own design ...The HF issue with the regulator approach was addressed by having very low ESR local decoupling, power and earthing planes, minimal lead lengths -- I was aiming to get absolutely the lowest supply impedance at HF without going really silly. One of the key qualities the power supply had was effectively high reserves of energy, as I've mentioned before: the amp could keep working for a few minutes at moderate volumes with no audible change if you pulled the mains plug ...
For the HT on the other hand, no such fanciness: why it can do the job is that, as you guessed, the subwoofer is completely separate, electrically and physically, it has its own mains cable. So part of the tweaking was to isolate the glitching of the mains caused by the current draw of each, from the other to some degree. It all could be done better, but it performs well enough to prove some points, for me ...
Frank
Last edited:
Aaah.. have just located it http://www.diyaudio.com/forums/power-supplies/216409-power-supply-resevoir-size-79.html#post3137770
Later there was a science fiction piece that went un answered http://www.diyaudio.com/forums/power-supplies/216409-power-supply-resevoir-size-114.html#post3164323
Later there was a science fiction piece that went un answered http://www.diyaudio.com/forums/power-supplies/216409-power-supply-resevoir-size-114.html#post3164323
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.