When designing an all digital system (i.e. PCM in to PWM) how can feedback be deployed? There's been a lot of discussion within this forum regarding feedback - take it prior to, or post LC filter. But how do you do it for a PCM design? I imagine using some DSP, perhaps. To take into account the post LC filter requires an A/D. Adjusting the PWM is going to be cycle by cycle, I suppose. This would take considerable thinking on my part, to come up with a circuit for this. But, the commercial world has attempted to help out.
Looking at the commercial offerings, here's what I've found:
1. TriPath TCD6000 - Claims to have pretty low THD, and SNR. Uses digital feedback from the MOSFET output. It is supposed to compensate for MOSFET differences, and power supply affects. Tripath scares me though - it looks they are not doing very well as a company, stock price is in the toilet, I can't get a reply via email or voice mail, it's difficult to get a live person on the phone, and they've had some turn over of key personnel.
2. TI TAS5518 - nice feature set (besides the pwm), claims to low THD. No feedback. Digikey shows the part in-stock. I haven't tried tech support yet, but it's TI and they are a stable company.
3. Freescale Symphony FSA95601 - This is a new part on the scene, it's just sampling now and literature is scant at the moment. It does use feedback, presumably from the MOSFET output. Not much else to say.
4. D2Audio - I don't know much about these parts, anybody want to chime in?
5. Cirrus Logic CS44800 - somewhat interesting part, but I haven't looked all that closely. It uses some proprietary PSRR logic, which requires an aditional A/D for the power rails. Other than that, no feedback. Dynamic range is down a tad from other vendors offerings, and THD spec is a tad worse 0.03% vs 0.01 or better from others. Cirrus logic is also stable company.
6. Wolfson - another company that I haven't scrutinized. No feedback on these parts.
BTW, what is so bad about open loop on class-D? Is it the power supply problem? What if it's regulated really well?
Looking at the commercial offerings, here's what I've found:
1. TriPath TCD6000 - Claims to have pretty low THD, and SNR. Uses digital feedback from the MOSFET output. It is supposed to compensate for MOSFET differences, and power supply affects. Tripath scares me though - it looks they are not doing very well as a company, stock price is in the toilet, I can't get a reply via email or voice mail, it's difficult to get a live person on the phone, and they've had some turn over of key personnel.
2. TI TAS5518 - nice feature set (besides the pwm), claims to low THD. No feedback. Digikey shows the part in-stock. I haven't tried tech support yet, but it's TI and they are a stable company.
3. Freescale Symphony FSA95601 - This is a new part on the scene, it's just sampling now and literature is scant at the moment. It does use feedback, presumably from the MOSFET output. Not much else to say.
4. D2Audio - I don't know much about these parts, anybody want to chime in?
5. Cirrus Logic CS44800 - somewhat interesting part, but I haven't looked all that closely. It uses some proprietary PSRR logic, which requires an aditional A/D for the power rails. Other than that, no feedback. Dynamic range is down a tad from other vendors offerings, and THD spec is a tad worse 0.03% vs 0.01 or better from others. Cirrus logic is also stable company.
6. Wolfson - another company that I haven't scrutinized. No feedback on these parts.
BTW, what is so bad about open loop on class-D? Is it the power supply problem? What if it's regulated really well?
Hi,
It's party the power supply problem. Regulating the psu could work but not without great cost, complication, efficiency, reliability.. and possibly performance.
Another factor is that you can't really correct for output stage errors strictly in the digital domaine... because really, it's analog, so at most you'd be taking delay into account and nothing else if you tried digital feedback only. That forces the use of an ADC as you've seen... is it still true digital?
I think to get away with open loop you'd need to switch hella fast... one amp pulled it off thus far, said to sound inferior to a much simpler, more robust, fully analog, self oscillating design
It's party the power supply problem. Regulating the psu could work but not without great cost, complication, efficiency, reliability.. and possibly performance.
Another factor is that you can't really correct for output stage errors strictly in the digital domaine... because really, it's analog, so at most you'd be taking delay into account and nothing else if you tried digital feedback only. That forces the use of an ADC as you've seen... is it still true digital?
I think to get away with open loop you'd need to switch hella fast... one amp pulled it off thus far, said to sound inferior to a much simpler, more robust, fully analog, self oscillating design
classd4sure said:. one amp pulled it off thus far, said to sound inferior to a much simpler, more robust, fully analog, self oscillating design
Is TACT the only company that has pulled off the no feedback digital amp?
What would be the best design to take a digital signal input with digital Xover and room compensation, and drive a 0.5 ohm resistive ribbon with good fidelity?? Is the TI chipset the best path? What would be better? Does TACT use TI chips?
" TacT True Digital Amplificiation: The Sound of the Future
Tact True Digital Amplification, the sound of the future. The Tact Millennium is recognized as the first true digital audio amplifier. Other so- called Digital or Class D amplifiers have been various combinations of Analog and Digital technology, in most cases using analog feedback. TacT's True Digital Amplification is conceptually different from any other amplification in almost every area of design. Once you understand the fundamental differences you will realize that Tact's complete line of True Digital Amplifiers offer the potential of much higher fidelity than any other digital or analog design.
Having said all of this it is worth pointing out, that the Tact's True Digital Amplifiers are not amplifiers at all. They are D/A converters that just happen to put out enough current and voltage to drive speakers directly, without the need of any amplification. This way it will perform the functions of: D/A converter, Preamplifier and Power Amplifier, without having any analog circuitry after D/A conversion other than one coil and one capacitor performing a 60 kHz, 2nd order low-pass filter. "
Ah I could have been in error saying only one has done it, tact maybe true digital and it's possible they do use TI chipsets, it's been discussed here before by people who've taken a closer look at it than myself. However the mere fact it has "room compensation" I think goes to show its' not without feedback.
The one I had in mind is an amp by Bruno Putzey, can't recall the name of it, but it's a DSD multi phase amp. Each phase handles a portion of the DSD signal so they can switch at a slower rate than the DSD frequency. The outputs are combined and there is no feedback at all.
Tact's method would have to be somewhat complex, and it may not sound bad at all, but is it worth it? Competing non "true digital" methods are so much simpler and more robust.
The one I had in mind is an amp by Bruno Putzey, can't recall the name of it, but it's a DSD multi phase amp. Each phase handles a portion of the DSD signal so they can switch at a slower rate than the DSD frequency. The outputs are combined and there is no feedback at all.
Tact's method would have to be somewhat complex, and it may not sound bad at all, but is it worth it? Competing non "true digital" methods are so much simpler and more robust.
LineSource said:
No! The panasonic digital amplifier SA-XR10 and his successors (50/70 ) all use the some equibit system , with no feedback .
See the thread :
http://www.diyaudio.com/forums/showthread.php?s=&postid=114080#post114080
classd4sure said:Right,
So, these are the no feedback amps with 0 PSRR..
Do you really think it's worth it?
I can only talk about my experience with my Panasonic XR10.
The power supply is variable and with a tight regulation.
As it have been designed for five channels and in my Panasonic is only feeding two , with a 8 Ohms load and the amp playing very loud, I can't see any ripple or other artifacts in the rail , with my scop...
Sounds like some worthy mods to it anyway, I like your idea of ridding yourself of the unused channels and freeing up power.
How do you like the effect of such a tightly regulated supply? Is it fast, responsive, dynamic??
What sort of regulation do they use? Must be fairly complex ?
Really my only beef with Equibit is you need to be an exclusive member, how hard is it to get samples I wonder.
Cheers
How do you like the effect of such a tightly regulated supply? Is it fast, responsive, dynamic??
What sort of regulation do they use? Must be fairly complex ?
Really my only beef with Equibit is you need to be an exclusive member, how hard is it to get samples I wonder.
Cheers
classd4sure said:
Yes, the sound is very fast and specially the bass is very "clean" and have lots of punch and speed.
They use a IC feeding a Half Bridge Fet drive , that feed a pair of Mosfets.Is similar to the audio channels output stages , but this time feeding the electrolytic power supply capacitors...
I dont know, but if you use TI chips, then I recommend you to couple via transformator.
ON:
Effect of varying supply can be cancelled out by dividing with its voltage, and this needs only relatively slow A/D. It's not feedback, I'd rather say feedforward. The other problem is crossover distortion caused by dead time. To decrease it one can measure delay between gate driver input and mosfet output cycle by cycle, and differences can be compensated in next cycle. This needs a fast comparator and a very very fast counter (some dosen gigahertz), or an analog time-multiplier (time-to-charge-charge-to-time conversion). This is not a real feedback either.
If you want to avoid the ultra tight power supply regulation, which - as Chris pointed out earlier - has some significant downsides, then you need some kind of feedback. In theory you can get away with measuring the V+ and V- and then predicting a correction to your signal from that, but i doubt you will ever get any useful THD or PSRR.
OK we need feedback. You can take feedback through an A/D converter, and use your DSP to make correction, which works fine of course. However unless you can find a 24 bit 1 Mss ultralow distortion A/D converter, you will end up having time delay in your feedback loop, which is never good. Since it is pure time delay, and not roll-off / phase difference, you then need to slow down your PID regulator loop just to keep the amplifier stable.
Better is to make an analog PID regulator feedback loop, with virtually no time delay (like a sigma-delta loop), and then use a normal (sorry - less fancy..) D/A converter chip, to connect to PCM. Inevitably you will get better THD, S/N and sound out of this solution, than you can with any digital loop. At least for the time being. In the future you can probably find better and faster A/D chips to narrow the gap to that solution.
About whether Equibit is the only non-feedback Class D system, i think i have heard Bruno Putzeys mention that also Sony S-Master is non feedback. I have no exact details on it though.
OK we need feedback. You can take feedback through an A/D converter, and use your DSP to make correction, which works fine of course. However unless you can find a 24 bit 1 Mss ultralow distortion A/D converter, you will end up having time delay in your feedback loop, which is never good. Since it is pure time delay, and not roll-off / phase difference, you then need to slow down your PID regulator loop just to keep the amplifier stable.
Better is to make an analog PID regulator feedback loop, with virtually no time delay (like a sigma-delta loop), and then use a normal (sorry - less fancy..) D/A converter chip, to connect to PCM. Inevitably you will get better THD, S/N and sound out of this solution, than you can with any digital loop. At least for the time being. In the future you can probably find better and faster A/D chips to narrow the gap to that solution.
About whether Equibit is the only non-feedback Class D system, i think i have heard Bruno Putzeys mention that also Sony S-Master is non feedback. I have no exact details on it though.
Yes as long as you are running at full modulation index ( = Volume on MAX). However when you turn down to say -40dB you lose the most significant 7 bits (because they are not in use at that level). However you still want 16 bit resolution at this level, to keep your S/N and THD at acceptable levels. Of course a 24 bit converter does not have 24 bit true performance, so you will get some deterioration of performance af -40 dB, and even more at -60 and 80dB. So i would say you need all the bits of resolution you can get.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.