I have experience with the TAS5630 and TAS5706B from TI. The former never put a foot wrong with regards to offending the ears but was coloured. I have since rebuilt and slightly improved the design but have yet to listen to it properly. My main setup is 4 way active and currently it powers the subs.
The TAS5706B is an interesting device, I used two for four channels, paired them with a WM8804 S/PDIF receiver and then used the internal biquads to create a 2 way active loudspeaker. About as direct a signal path as one can get if you are concerned with such matters. I auditioned the combination driving a small two way comprised of a W15CY001 from SEAS with a Peerless HDS tweeter in a wave-guide. I was very impressed with how good it sounded, again the class D amps didn't do anything to offend the ears and certainly did nothing to embarrass themselves.
The below image shows the distortion of the TAS5706 driving my 9.4ohm test load to 1 watt, as you can see the distortion performance isn't exactly stellar. The second image however is the distortion that the loudspeaker produced with the W15CYs driven to 2.83Vrms by said amplifiers. Done as such you can see that the high order harmonics fall away as the mid/bass falls away and the tweeter kicks in. This is no doubt because the tweeter needs a tiny amount of power compared to the woofer and the TAS5706 probably gives out much lower distortion at very low power levels. I have yet to test this out as my earlier 1 watt measurement was made only to confirm that my build of the amplifier was performing to spec so I didn't do a measurement at 0.1 watts! As can be seen however the main short coming of the TAS5706 (rising distortion with frequency) isn't an issue with this design and it also helps to show why I liked the way it sounded so much.
The TAS5706B is an interesting device, I used two for four channels, paired them with a WM8804 S/PDIF receiver and then used the internal biquads to create a 2 way active loudspeaker. About as direct a signal path as one can get if you are concerned with such matters. I auditioned the combination driving a small two way comprised of a W15CY001 from SEAS with a Peerless HDS tweeter in a wave-guide. I was very impressed with how good it sounded, again the class D amps didn't do anything to offend the ears and certainly did nothing to embarrass themselves.
The below image shows the distortion of the TAS5706 driving my 9.4ohm test load to 1 watt, as you can see the distortion performance isn't exactly stellar. The second image however is the distortion that the loudspeaker produced with the W15CYs driven to 2.83Vrms by said amplifiers. Done as such you can see that the high order harmonics fall away as the mid/bass falls away and the tweeter kicks in. This is no doubt because the tweeter needs a tiny amount of power compared to the woofer and the TAS5706 probably gives out much lower distortion at very low power levels. I have yet to test this out as my earlier 1 watt measurement was made only to confirm that my build of the amplifier was performing to spec so I didn't do a measurement at 0.1 watts! As can be seen however the main short coming of the TAS5706 (rising distortion with frequency) isn't an issue with this design and it also helps to show why I liked the way it sounded so much.
Attachments
I don't need more then 16 watts for the midrange and tweeter in my main loudspeakers. Most people for that matter don't need more then about that much either if Pano's thread was anything to go by.
http://www.diyaudio.com/forums/mult...much-voltage-power-do-your-speakers-need.html
yes, for how this amp is intended to be used ie. in active or digitally crossed multiamped speakers, I would go out on a limb and say that for even midwoofer duties, without any XO parts to interfere, many people do not need more. Its a matter of application, i'm using opc's higher power amp for where needs it, but i'm betting I could power my 2 ways with 4 of them no problem at all.
hypex claims that including the filter in the feedback loop the frequency response is not load dependent.
this stands true for both ncore and ucd.
check the response at 8/4 ohm and open circuit at 8.2 chapter.
http://www.hypex.nl/docs/UcD700HG_datasheet.pdf
Well it isn't just Hypex, including the output filter in the feedback loop is something I consider (from my fairly uneducated stand point) to be vital for giving you 'state of the art' class D performance. Not only would it help remove variations in the frequency response, caused by a varying load, but it should also help counter any linearity issues with the output filter too.
So far Zetex are the only manufacturer to have an IC solution class D amp that includes the output filter in the feedback loop and this isn't exactly readily available to DIYers.
I for one have been waiting a very long time for TI to release chip amps that include the output filter in the feedback path but I don't see this happening any time soon.
So far Zetex are the only manufacturer to have an IC solution class D amp that includes the output filter in the feedback loop and this isn't exactly readily available to DIYers.
I for one have been waiting a very long time for TI to release chip amps that include the output filter in the feedback path but I don't see this happening any time soon.
NPX (former Philips) TDA8920 class D chip amps is what I have built years ago and compared with Hypex UCD.
Tweeking the output filter with quality parts (air core inductor) resulted in a preference for the TDA8920 soundwise.
I am not so sure that including the output filter in the feedback loop is a must as long as the filter is optimized with respect to the load, but yes it is more practical to include it in the feedback loop.
That depends, I don't know if performance like that of the ncore is achievable if the filter is outside of the loop. I am talking of distortion here rather then anything else. The ncore is rather special in how low it's measured distortion is and no other class D amp with the filter outside the loop even comes remotely close to matching it.
Class I can be temptating, but is basically just multiphase - which of course deserves a new letter....
Take care, aside multiple obviously temptating properties it has less obvious down sides:
- The two paralleled stages must have excellent DC-properties, otherwise you will get unpleasant DC balancing currents.
- The switching residuals that are being injected back into the modulator through the feedback have double frequency of the modulator carrier and the resulting distortion mechanism appeared difficult to cure, while I found a simple but extremely effective method for traditional designs.
Both topics are not necessarily show stoppers. It is more a question of your target applications.
For very high power the extra efforts for the DC-balancing won't bother in costs and complexity. Also nobody will panic, because of slightly higher THD - and of course there might already exist appropriate cures of these specific double-fs-feedback/modulator distortions.
The two stages need not have excellent DC-properties, since DC bias has to be actively controlled anyway to avoid runaway, which is inevitable w/o active control. This topology basically needs 2 control loops: 1st for output voltage, 2nd for bias current.
Yup, the 2nd control loop is what I adressed by the 'extra efforts for the DC-balancing'. I also agree that the inherent DC-accuracy , or better inaccuracy, without bias/balancing loop, will almost guarantee trouble except we use lossy resistors for paralleling - or air coils with high Rdc in the filter.
==> which is leading us again to the discussion, which solution might sound better.
The one and only question. In fact it is easy to reach clarity about technical advantages and disadvantages of different solutions, but the impact on sound is not always easy to foresee.
Many sonic myths might have their origin from certain set ups, where the sonic effect was real, but the given explanation was incomplete or sometimes just rubbish.
Back to class I / dual phase:
What is the benefit of biasing the two stages with a certain DC instead of biasing them to zero differential bias (==>balance)?
Basically I would just expect that the two typical areas of non linearity (in the range were load current and peak filter ripple equal) would be shifted.
In any case I would expect that the output nonlinearity of the dual phase is less than in single phase, because of the time shift between both dead times at the critical load current ranges.
I did not dig into full depth of this circuitry. So I am curious about Tom's findings (in case you intend to dig deeper).
"So I am curious about Tom's findings (in case you intend to dig deeper).[/QUOTE]"
As usual, my simulations focus 1st on getting the topology to work stable with as high loop gain as possible. Since I am simulating in the time domain with simetrix, I can only see a sine wave, square plus Overshoot, etc. at the output with the switching ripple superimposed. I can also try to find effective output impedance vs. frequency by using different load impedances and comparing the voltage. But: For me, there is no way to determine if the distortion is 0.02 or 0.01 % THD, as long as the distortion is not visible from the output waveform.
As usual, my simulations focus 1st on getting the topology to work stable with as high loop gain as possible. Since I am simulating in the time domain with simetrix, I can only see a sine wave, square plus Overshoot, etc. at the output with the switching ripple superimposed. I can also try to find effective output impedance vs. frequency by using different load impedances and comparing the voltage. But: For me, there is no way to determine if the distortion is 0.02 or 0.01 % THD, as long as the distortion is not visible from the output waveform.
Well, 0.01% is already pretty nice. In fact in my simulations the distortions caused by switching residuals that are feed into the modulator by the feedback were more in the range of >0.1% - especially at higher modulation levels, say above 50% of max output voltage.
I am using LT spice and let it calculate the FFT from the output wave form.
But of course you are 100% right, that simulation of low distortions is sort of gambling. ...have to build a proto, in order to countercheck theory and simulation with reality.
Last edited:
What would you think of these measurements?
Frequency response 0.3 Hz – 33 kHz -3 dB points, 8 Ohms load.
Frequency response -0dB/+0.2dB 20 Hz - 20 kHz, 8 Ohms load
Frequency response -0.2dB/+0dB 20 Hz – 20 kHz, 4 Ohms load
THD+N, 1 W /8 Ohms 0.004% A-wgt.
THD+N, 1 W/4 Ohms 0.006% A-wgt.
THD+N, 100 W/8 Ohms 0.01% A-wgt.
THD+N, 180 W/8 Ohms 0.07% A-wgt.
THD+N, 275 W/4 Ohms 0.07% A-wgt.
S/N ratio 117 dB A-wgt. ref 200 W/8 Ohms.
Channel separation 84 dB 1 kHz, 200 W/8 Ohms.
I am Class A man myself but I have just purchased my first Class T pcb.
When I have put it together I will let you know what I think.
Edit: On paper at least what I like about the Class T amps, is that the distortion does not increase with reducing impedance (according to the graphs I have seen), as opposed to Class AB amps which distort more as speaker impedance falls.
When I have put it together I will let you know what I think.
Edit: On paper at least what I like about the Class T amps, is that the distortion does not increase with reducing impedance (according to the graphs I have seen), as opposed to Class AB amps which distort more as speaker impedance falls.
Last edited:
Congrats, Markus, you've therefore achieved what not many came close to.
For the record, all my intergate&reset, sample&hold and similar efforts have been a plain failure
"The switching residuals that are being injected back into the modulator through the feedback have double frequency of the modulator carrier and the resulting distortion mechanism appeared difficult to cure, while I found a simple but extremely effective method for traditional designs."
Since in class I, no adverse body diode / dead time issues exist, I would suggest to use a high performance schottky diode and ramp up the frequency as high as possible. Why not go 2 x 500kHz or greater? Then the residuals will get way smaller.
Since in class I, no adverse body diode / dead time issues exist, I would suggest to use a high performance schottky diode and ramp up the frequency as high as possible. Why not go 2 x 500kHz or greater? Then the residuals will get way smaller.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.