Intermission -- news flash --> Hey JC !! check it out... MSN News has proof that where I spent many Mondays flying to and Fridays flying back home actualy exists... CIA finally approved a single photo of Area 51 (North of Mercury, Nevada). It only shows the admin buildings near the run way strip, though. Those were the good old days of the truely wild, wild west. The short art. isnt sure what was done there.... it was live nuclear device tests for decades. I was there during such event. Guess CIA thought everyone was dead by now. hah! Ok John, now you finally know. Thats why I couldnt leave the USA afterwards for years without gov approval. Cold war stuff. Better than any movie... you just cant make up the stuff that goes on.
Ooops. back to the quiet and peacefulness of music... .. never mind.
-Richard Marsh
Ooops. back to the quiet and peacefulness of music... .. never mind.
-Richard Marsh
Last edited:
yes, underground. Worked on the whole thing electronic from the device/experiement with all the control, dagnostic and singal cables as it was lowered down hole to the trailers full of equipment. The remote control detonation was back at a bunker miles away (edit) Watching the whole thing on CCTV... I can tell you it was felt way past me into downtown l.Vegas. Then we waited for the crater to drop... earth fell into the void created below. Its as clear as it was yesterday and everything in that area.
Thanks-- not NM. The two labs competed on weapons designs only. The best presentation got the job. Eventually the tests went above ground in the Laaser lab to do simulations and tests there.... no more need to do live tests. The models were then good enough and so we could do a test ban. The tests were always at Area 51. If you would be allowed over that area... it looks like all the pot-marked craters on the moon's surface.
-Richard Marsh
Last edited:
Richard,
I asked about White Sands as that is one of the places my ex-partner went for months at a time while working on the B2. Not sure exactly what they were testing down there and I knew not to ask about that. My guess was aerodynamic testing of some models but didn't want to get any confirmations on that. Don't need the NSA asking me or him any question that I don't need to know about.
I asked about White Sands as that is one of the places my ex-partner went for months at a time while working on the B2. Not sure exactly what they were testing down there and I knew not to ask about that. My guess was aerodynamic testing of some models but didn't want to get any confirmations on that. Don't need the NSA asking me or him any question that I don't need to know about.
Regarding audible effects of changing the AC Mains frequency, I would guess that for some amplifiers, a minimum PSU reservoir capacitance was used, but it was calculated incorrectly, and was a bit too low to handle the very low audio frequencies, presumably even at well-below the rated maximum output power.
A while back, I looked at ways to calculate the minimum reservoir capacitance for a linear PSU for a power amplifier that was driving a single low-audio-frequency sine wave into a resistive load.
That "beat tone" problem, i.e. the difference between the mains frequency and the signal frequency, was prevalent.
There is a ripple-voltage minimum that occurs at the difference frequency of the mains and the signal. So if someone didn't calculate the minimum required capacitance while considering the worst-case phase relationship between the mains and the sine signal, they would miss the true lowest dip in the rail voltage and would use a capacitance that would allow clipping, every so-many cycles, for the lowest-frequency sine they designed for.
Of course, if they just assumed that the caps had to be able to handle constant DC at the sine's RMS level, that automatically limits the lowest sine frequency that won't cause clipping. BUT, if they instead assumed that the PSU caps had to be able to handle constant DC at the PEAK (rather than RMS) level of a sine, then they'd be golden.
The latter case can handle ANY signal (even constant DC at the max rated peak output level), whereas the former case (using RMS-level DC as the requirement/assumption), can easily be shown to clip for an infinite number of signal shapes that do not exceed that level, but are not a single simple sine.
I was able to find a closed-form solution for the minimum required PSU capacitance for the case where fsignal >= fmains. I also found a solution for fsignal < fmains but I convinced myself that it was based on a wrong assumption:
Considering the time period between charging pulses, I had assumed that the worst case would be when the mains and signal phases aligned so that the sine signal peak was centered between the charging pulses. Alas, simulations did not agree. So my equations for that case predict a minimum capacitance that is a little bit too low.
But note that as the sine frequency gets lower, and the sine gets "wider", and much wider than the mains' period, then near the peak of the sine, it will appear to almost be a constant DC at the peak level. So for VERY low frequencies, relative to the mains frequency, it seems clear that we should not try to design the minimum capacitance for anything other than constant DC at the PEAK sine level that corresponds to the rated maximum output power, and should never use the RMS-level assumption.
I also have a good LT-Spice simulation of a transformer/rectifier/capacitor power supply and a power amplifier with a resistive load, with a sine wave input signal. It's important to include at least the transformer windings' leakage inductances and resistances, because then the model/simulation will more-correctly produce the power rail voltage peaks, which can be higher than the secondary's ideal peak output voltage.
Anyway, I have attached the paper with the equations. Maybe there's a clue, in there, about the audible effects of changing the mains frequency that have been described in this thread.
I have since obtained the Mathcad software. So I hope to re-visit the case where fsignal < fmains, at some point.
Although the solutions are only for a single sine, they should be able to be combined with Fourier theory to derive equations for the actual minimum required PSU reservoir capacitance for any signal shape, if that ever happened to be required.
But, again, if you really don't want to worry about any of that and just want it to be "bulletproof", at the cost of possible slight overkill in the capacitance value, then simply assume that the signal could be constant DC at the PEAK sine level that is implied by the rated maximum output power.
A while back, I looked at ways to calculate the minimum reservoir capacitance for a linear PSU for a power amplifier that was driving a single low-audio-frequency sine wave into a resistive load.
That "beat tone" problem, i.e. the difference between the mains frequency and the signal frequency, was prevalent.
There is a ripple-voltage minimum that occurs at the difference frequency of the mains and the signal. So if someone didn't calculate the minimum required capacitance while considering the worst-case phase relationship between the mains and the sine signal, they would miss the true lowest dip in the rail voltage and would use a capacitance that would allow clipping, every so-many cycles, for the lowest-frequency sine they designed for.
Of course, if they just assumed that the caps had to be able to handle constant DC at the sine's RMS level, that automatically limits the lowest sine frequency that won't cause clipping. BUT, if they instead assumed that the PSU caps had to be able to handle constant DC at the PEAK (rather than RMS) level of a sine, then they'd be golden.
The latter case can handle ANY signal (even constant DC at the max rated peak output level), whereas the former case (using RMS-level DC as the requirement/assumption), can easily be shown to clip for an infinite number of signal shapes that do not exceed that level, but are not a single simple sine.
I was able to find a closed-form solution for the minimum required PSU capacitance for the case where fsignal >= fmains. I also found a solution for fsignal < fmains but I convinced myself that it was based on a wrong assumption:
Considering the time period between charging pulses, I had assumed that the worst case would be when the mains and signal phases aligned so that the sine signal peak was centered between the charging pulses. Alas, simulations did not agree. So my equations for that case predict a minimum capacitance that is a little bit too low.
But note that as the sine frequency gets lower, and the sine gets "wider", and much wider than the mains' period, then near the peak of the sine, it will appear to almost be a constant DC at the peak level. So for VERY low frequencies, relative to the mains frequency, it seems clear that we should not try to design the minimum capacitance for anything other than constant DC at the PEAK sine level that corresponds to the rated maximum output power, and should never use the RMS-level assumption.
I also have a good LT-Spice simulation of a transformer/rectifier/capacitor power supply and a power amplifier with a resistive load, with a sine wave input signal. It's important to include at least the transformer windings' leakage inductances and resistances, because then the model/simulation will more-correctly produce the power rail voltage peaks, which can be higher than the secondary's ideal peak output voltage.
Anyway, I have attached the paper with the equations. Maybe there's a clue, in there, about the audible effects of changing the mains frequency that have been described in this thread.
I have since obtained the Mathcad software. So I hope to re-visit the case where fsignal < fmains, at some point.
Although the solutions are only for a single sine, they should be able to be combined with Fourier theory to derive equations for the actual minimum required PSU reservoir capacitance for any signal shape, if that ever happened to be required.
But, again, if you really don't want to worry about any of that and just want it to be "bulletproof", at the cost of possible slight overkill in the capacitance value, then simply assume that the signal could be constant DC at the PEAK sine level that is implied by the rated maximum output power.
Last edited:
Every year the Army at White Sands Missile Range has a tour of trinity site where the first atomic bomb was tested.
Ray
ChamberMaster | Alamogordo Chamber of Commerce - Trinity Site Tour
Ray
ChamberMaster | Alamogordo Chamber of Commerce - Trinity Site Tour
Yeah, there's that. I should have been simulating it, way back when, but never have.
Off the top of my head, I think that would at least not affect my mathematical derivations, very much, because it would mainly only affect the charging pulses, which I conveniently assumed were "sufficient" (noting that they might not be). I didn't reach farther upstream than the reservoir caps, for those calculations, because a) it's complicated, and b) there is no closed-form solution for the differential equations that result, unless approximations are used, because to find the rectifier turn-on and turn-off points (time or phase angle), it always comes down to trying to calculate when a sine intersects an exponential, which is a transcendental equation and thus has no closed-form solution.
But now that you mention it, if I ever get back to this stuff I think that maybe I should stop resisting using the approximations, so I can start with a general mains(t) waveform and include everything from there to the load.
Some of the approximations that others have resorted to using aren't all that bad. Usually it amounts to throwing away all but the first two or three terms of the Taylor Series expansion of the sine or the exponential or both. Using a portion of a parabola to model a portion of a sine turns out to be pretty accurate. Unfortunately, the current draw from the caps is sinusoidal instead of linear, so that part would be a little more complicated than what others have published, that I am aware of at least. Don't forget, too, that the rectifier resistance is related non-linearly to its current and voltage.
Back when I was looking at this stuff I did also come up with the differential equations for an "exact" numerical solution, with no approximations (except for a curve-fit for a particular rectifier diode's R vs I equation, if that counts), and implemented it as a VBA macro in Excel (with 4th-order Runge-Kutta for doing the time-stepping), along with a scalable transformer model. I uploaded the Excel sheet to the Power Supply Reservor Size thread. I could also re-visit that and see what it would take to use distorted AC Mains waveforms.
Of course, using a real simulator, like LT-Spice, would work great. But a lot of people aren't good at that, or don't even try. So I'd still like to be able to come up with something "simpler to apply"; probably either a spreadsheet or a simple equation. Just thinking out loud, maybe there's an equation that will approximate the "envelope" of a more-complicated solution, that will give minimum C values (e.g. as a function of minimum signal frequency and some mains parameters) that are sufficient, but still less than when using the DC-at-max-peak-level assumption (which, itself, also would need to be modified, if wanting to account for distorted mains).
Gee, that ALMOST motivated me to start working on that project again. Maybe someday...
For a while the Monster AVS2000 (motor driven variac ac regulator) regulated the Peak AC voltage as a proxy for the RMS voltage and it mapped much better to the actual voltage on the supply rails in amplifiers etc. However the constant complaints that the meter was not reading right and the inability of support staff to explain the details to customers forced us to change the unit. Measured better but significantly worse performance. Classic problem of measuring the wrong thing.
Am looking around a bit for DBT related experiences, and found this, relevant to the current 'flow' - nicely sums up the sort of thing that people 'bump into' when they become aware of such matters ... Listening for Electrical Interference - DIY HIFI - HIFICRITIC FORUM - HIFICRITIC FORUM : hi fi audio systems forum
Edit: my apologies, I just realised yours truly appears in the mix, wasn't intending to "point" to that ... one of those things ...
And further, the main protagonist makes excellent comments on what can easily 'pollute' a DBT - if such is done simply then the results will likewise be 'simple' ... it's a can of worms, in reality ...
Edit: my apologies, I just realised yours truly appears in the mix, wasn't intending to "point" to that ... one of those things ...
And further, the main protagonist makes excellent comments on what can easily 'pollute' a DBT - if such is done simply then the results will likewise be 'simple' ... it's a can of worms, in reality ...
Last edited:
You have done some good work there Tom.
Re the Behringer active speakers, yes it looks like 6800uF per rail is insufficient.
Add to that, simplistic regulation stage for the line level stages, and likely lousy earthing arrangements all adding up to the 'imd/veiling' that is present...subtle but present none the less.
So it all comes down to stout and clean power supplies, and good PSRR....qualities taken care of in good gear, but typically lacking in mid-fi consumer stuff.
The interesting thing, to me, is that choosing a musically related AC supply frequency substantially reduces these sonic effects in a sub-optimal system.
All to be expected really.
Next time I do some recording I will try 55Hz supply.....I expect it will reduce that zzzzzeddding characteristic present in many recordings.
Dan.
Re the Behringer active speakers, yes it looks like 6800uF per rail is insufficient.
Add to that, simplistic regulation stage for the line level stages, and likely lousy earthing arrangements all adding up to the 'imd/veiling' that is present...subtle but present none the less.
So it all comes down to stout and clean power supplies, and good PSRR....qualities taken care of in good gear, but typically lacking in mid-fi consumer stuff.
The interesting thing, to me, is that choosing a musically related AC supply frequency substantially reduces these sonic effects in a sub-optimal system.
All to be expected really.
Next time I do some recording I will try 55Hz supply.....I expect it will reduce that zzzzzeddding characteristic present in many recordings.
Dan.
Reasonably independent of level IME so far.
I have house mates so finding a decent length time window is difficult.
Results so far is that regenerated 50Hz sounds much cleaner than direct wall power...to be expected because of low THD and low noise output.
55Hz cleans it further....removes a veil of IMD junk, and vocals in particular sound much more natural, bass significantly also - please refer to my comments re ABC Classic FM and Helen Reddy in a post above.
I am about to shift digs this weekend and will have a dedicated workshop/listening room....haha, I will be able to play any kind of music, and/or repeat sections ad infinitum without strange looks or affecting others, and knuckle down and do some proper measuring.
Going from 55Hz to 50Hz or 60Hz, subjectively adds a 'modulation', not directly audible as 5Hz of course, but audible as an instability, or 'vibrato' that effects/modulates the whole audio spectrum.
I have read arguments/observations that low/very low frequency instability in DAC clocks has a similar effect.
Dan.
Ok here is the common mistake. A regulator does not have infinite bandwidth or even a close approximation. For small signal circuits or even power amplifier input stages a CRCRC may have the same voltage drop as a regulator but can be 60 db or more quieter even at frequencies as low as 20,000 CPS. Now a voltage regulator does of course provide a stable voltage but it may also introduce a resonance.
Just for those folks who believe no one today uses op amps as obsolete as a 741 the amplifier inside most voltage regulators is often worse.
So for today's fashionable results you have to use a regulator and have good filtering.
Just for those folks who believe no one today uses op amps as obsolete as a 741 the amplifier inside most voltage regulators is often worse.
So for today's fashionable results you have to use a regulator and have good filtering.
Thanks, Dan. But also, please keep in mind that it's not peer-reviewed, or anything like that.
It looks like I didn't explain the Vclip parameter very well, in that PDF file. And since it's something that seems to be missed by a lot of people, I'll just put it here:
Vclip is the minimum voltage between the amplifier's power rail connection and the amplifier's load connection. For one rail of a class AB amp, for example, it could be the minimum Vce of the output transistor plus the voltage across its small emitter resistor; so maybe around 3 Volts. But it will vary with rail voltage and load impedance.
The datasheets for the LM3886 and LM3875 chip amps, for example, provide plots of Vclip versus the rail voltage, for both the positive and negative power supply sides, and for both 4 Ohm and 8 Ohm loads in the case of the LM3886. For those chips, Vclip can vary from about 2 V to 6 V (but not for all cases), as supply voltage goes from +/-10V to +/- 40V. And it seems to always be higher for the negative rail's side of those amps.
Anyway, it's amazing how many times people seem to forget or not realize that the amplifier has to occupy some "voltage space", above the signal and below the ripple. It should be obvious, from looking at an amplifier schematic.
NOT accounting for the Vclip voltage across the amplifier can lead one to significantly over-estimate a power amplifier's rated max output power. i.e. Assuming that the max peak output voltage is the rail voltage minus the max p-p ripple voltage is very wrong. It's Vclip volts less than that.
I looks like the Vclip voltage is exactly like a regulator's dropout voltage. If the bottom of the ripple comes down too far and intrudes into the Vclip + Voutput range, it causes ripple-shaped chunks to be gouged out of the output signal waveform.
Cheers,
Tom
Last edited:
- Status
- Not open for further replies.