I think I saw exactly that today when doing a 47 kHz test . Well above 34 Vrms when 30 Vrms at 1 kHz and clipping ( Hypex Ucd 180 ) . The idle PSU voltage +/- 47 V . That is not to say the PSU being switch mode wasn't doing similar stuff . The inductance of the resistor must be part of it . However it is the same set I use all the time and never saw this before . The amplifier is about 3 db down at this frequency so doubly odd .
The stored energy is in the output inductor, not the speaker. The side switching at a longer duty cycle puts energy into the inductor, which is then transferred into the opposite supply rail through the opposite side switches. Look at the diagram posted above. You're not "creating" energy, you are simply transferring it from one rail to another.
Measured the Hypex SMPS 400/180 . Not the quietest of devices . Made an RF probe and found the fuse end to be quietest . 3 cm is the optimum minimum spacing . Good news is the RF didn't carry up the mains cable .
As I understand it the Hypex PSU is not fixed voltage . It tracks the mains voltage . Rather a shame because I want to try an optimum class H design . I do see 48 V switch-modes now . That would be ideal . My brief look at this says to me that the Hypex 180's deserve a conventional PSU . Mine I have been trying is 500 VA only as it was what I had handy . I suspect 160 VA would power a pair of UCD 180's if the Hypex 1/8 crest factor and 92 % efficiency rating is factored in ( less if space requires it ) . From my measurements that would even allow continuous 70 Wrms if heat-sinked on a biscuit tin or better . I would suggest 0 -25 0 - 25 transformer as a excellent choice if so . I used 2 x 22 000 uF . I was testing on a piece of 2.6 mm aluminum about the size of A4 paper . Slightly more as it has 70 mm right-angle folds at each end with no side pieces . Even after 15 minutes at 70 W rms the heat build up was fine .
Anyone know of a simple cheap temperature switch ( over temperature ) . The ones Naim audio used need high current , I don't trust a toaster switch to work at uA . If the toaster switch ran a 5 V relay at 70 mA I guess that might be OK ? It could have a shunt resistor to help . I guess a thermistor and comparator might be OK ? It would switch the 12 V logic of the modules . 0V = ON .
As I understand it the Hypex PSU is not fixed voltage . It tracks the mains voltage . Rather a shame because I want to try an optimum class H design . I do see 48 V switch-modes now . That would be ideal . My brief look at this says to me that the Hypex 180's deserve a conventional PSU . Mine I have been trying is 500 VA only as it was what I had handy . I suspect 160 VA would power a pair of UCD 180's if the Hypex 1/8 crest factor and 92 % efficiency rating is factored in ( less if space requires it ) . From my measurements that would even allow continuous 70 Wrms if heat-sinked on a biscuit tin or better . I would suggest 0 -25 0 - 25 transformer as a excellent choice if so . I used 2 x 22 000 uF . I was testing on a piece of 2.6 mm aluminum about the size of A4 paper . Slightly more as it has 70 mm right-angle folds at each end with no side pieces . Even after 15 minutes at 70 W rms the heat build up was fine .
Anyone know of a simple cheap temperature switch ( over temperature ) . The ones Naim audio used need high current , I don't trust a toaster switch to work at uA . If the toaster switch ran a 5 V relay at 70 mA I guess that might be OK ? It could have a shunt resistor to help . I guess a thermistor and comparator might be OK ? It would switch the 12 V logic of the modules . 0V = ON .
Exactly right. And it is usually true that in a bridged configuration bus pumping is not a problem.
Cheers,
Bob
compensation of the feed-trough capacitance of CCS
Referring to APAD5, page 262, Fig 8.49 and APAD6, page 394, Fig. 15.7
The feed-trough effect can also be reduced by the use of the variable capacitance of a reverse biased diode or the BC-junction of a transistor of the same type than the one used in the CCS, connected in lieu of the speed-up capacitor Cs. By doing so, the compensation capacitance would vary approximately in the same way than Cbc of the CCS transistor (Fig. 8.48 and 15.6 respectively).
Referring to APAD5, page 262, Fig 8.49 and APAD6, page 394, Fig. 15.7
The feed-trough effect can also be reduced by the use of the variable capacitance of a reverse biased diode or the BC-junction of a transistor of the same type than the one used in the CCS, connected in lieu of the speed-up capacitor Cs. By doing so, the compensation capacitance would vary approximately in the same way than Cbc of the CCS transistor (Fig. 8.48 and 15.6 respectively).
Nigel, I've used the Airpax thermal switches like this one: http://www.farnell.com/datasheets/1561970.pdf They are rated for switching logic signals.
I believe mechanical energy stored in a woofer would cause bus pumping just the same as energy stored in the output inductor. That is the great thing about Class-D amps. A linear amp has to dissipate all the power returned to it by a reactive load, but a Class-D amp can recycle it to the DC bus.
I believe mechanical energy stored in a woofer would cause bus pumping just the same as energy stored in the output inductor. That is the great thing about Class-D amps. A linear amp has to dissipate all the power returned to it by a reactive load, but a Class-D amp can recycle it to the DC bus.
Last edited:
Excellent . Thanks . That's the company that made motors for the LP12 I suspect ? Part of Philips if so .
Could you post a pic for those of us beach bums who only have older versions and are too indigent to buy the latest?
Try the TI/National LM35 Centigrade temperature sensor in the TO220 package, which is convenient for mounting on a heat sink.
Cheers,
Bob
While on the subject of class D, it is important to recognize that class D amplifiers fall into the category of a sampled-data system. Whether the class D is PWM or PDM (e.g., Delta-Sigma), the analog input signal is sampled at only discrete times. As such, in principle, class D amplifiers should include an anti-alias filter. This filter must have high attenuation at one-half the lowest sample rate and higher frequencies. Some PWM modulators can have varying carrier frequencies (sample rate) with signal, and sometimes the sample rate can swing to a lower-than-expected frequency. Many SACD players can produce significant levels at ultrasonic frequencies. Moreover, unwanted EMI sources can fall in a frequency range where aliasing may occur as a result of the class D sampling process.
Bear in mind that the D in class D does not stand for digital, but is just the next letter in the alphabet after class C. Even though class D is not digital per se, it is a sampled-data system.
Cheers,
Bob
Bear in mind that the D in class D does not stand for digital, but is just the next letter in the alphabet after class C. Even though class D is not digital per se, it is a sampled-data system.
Cheers,
Bob
Fig. 7.44 and 7.45 in APAD3&4, pag.230-235. No pics. They're copyrighted.
Last edited:
I don't understand.
What is the difference in nature between an analog tape recorder, where the signal is a polarisation frequency modulated in amplitude by the analog audio signal, and a class D amp, where the signal is a square wave modulated in step duration, analog way - means continuous, no step- by the audio signal ?
Of course, both need to filter the analog signal to avoid beats with the switching/polarization frequency.
Tim de Paravacini say they are mostly the same . The magnetization taking place at the top and bottom of the bias wave . He went on to say if people had thought about this the 200 kHz bias frequency would have suggested the 192 kHz sampling that eventually became the true hi fi standard . The big advantage of class D and tape recorders is they don't need decoding . They are ready to eat so to speak . The only digital I strongly dislike is CD . Others systems claim less or give more . CD I suspect is dogged by the method of storage and playback and not 16 bit linear quantization . Early digital LP's seem better ( no , not colouartion or my ears , just better from the same master ) . Donald Fagen The Nightfly for example made on the 3M machine . Very glassy , not horrible because of it . In fact not unlike the Hypex UCD 180 . It is said to deliberately have 44.1 kHz sampling running from a 48 kHz master doesn't help . The LP will be one stage better in that . Also the timing errors are one off errors . The cutting of the LP is not without problems . At least errors are of a different origin . Some overlap .
From a suggestion made to me earlier in this thread the need for oscillator stability in class D is perhaps the Elephant in the room . I doubt it is a problem if COG caps ( 30 ppm typical ) are part of the RC devices . One thing I do not understand about Hypex is why the Hypex SMPS is not locked together with the amp . This must be a free lunch that has not been accepted ?
Agree on all this, Nigel 
About CD, i believe the mistake is that the standard use "flux" and not pure datas (that you could sample and re-clock) with a poor error correction system.
And optical disk very fragile. 44.1Khz is too low, too.
Well, i quite enjoy the Ry Cooder's Bop t'ill you drop CD (one of the first digital recordings), despite all the evils ;-).
About CD, i believe the mistake is that the standard use "flux" and not pure datas (that you could sample and re-clock) with a poor error correction system.
And optical disk very fragile. 44.1Khz is too low, too.
Well, i quite enjoy the Ry Cooder's Bop t'ill you drop CD (one of the first digital recordings), despite all the evils ;-).
Hi Esperado,
Are you saying that an analog tape recorder is a sampled-data system?
If so, are you saying that it is due to the recording bias signal and that the effective "sampling rate" is the frequency of the bias oscillator?
Cheers,
Bob
I wonder ? The frequency doesn't seem to vary much . I seem to remember about 390 kHz to about 405 kHz . I didn't really look that carefully . As I was measuring spot frequencies maybe I didn't see what I should have if using music .
The point made to me earlier in this thread was that sub harmonics can not exist where the oscillator is stable . That seems reasonable to me . I am reasonably happy to believe the inherent stability of Hypex is good ( espeailly in a small time window ) . If it varies it is to accommodated it's way of working . Thus it might produce sub harmonics . However at a very high frequency . I still see locking the PSU to the output stage as helpful . If it is identical it can not beat or hetrodyne ? That stimulates a though . Sub harmonics always seem to plague radio so why not class D ? I did see stuff . Alas it was very hard to say if it was generated in my home brew output measuring filter (12 pole - 3dB 50 kHz ) ? It got much worse with SMPS attached ( other was 500 VA + 2 x 22 000 uF at 42 V ) .
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.