Reinventing IcePower topology

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hi all

Lastly it stroke me that B&O IcePower do some tricks on their amps and I think I start to understand why, need some sanity check here.

First they advertise the tricks which are to my understanding:

1. Class BD full bridge - with some well known advantages and disadvatages compared to AD or half bridge

2. Nested feedback with local differentialling self-oscillating pre-filter loop and global post-filter voltage feedback. The differential characteristics of self oscillator together with output LC give a single dominant pole global loop (hopefully), which is something feedback loops do like.

BUT, assume one uses such a nested feedback arrangenent with half-bridge.
It won't work!!!

The global loop can and will take over the self-oscillation and switch into UcD mode at somewhere close to filter resonance frequency. The inner loop will get permanently overdriven in either direction so that it does not self-oscillate any more.
My point is that the nested feedback arrangement like this can ONLY work with class BD and balanced feedback (or withvBCA, which is patented by someone else), due to the fact, that the output does not see the switching ripple of the first harmonic, same for global feedback loop.
Any base-frequency switching ripple would immediately turn the whole effort in a global voltage ripple-based modulator (like UcD).

Need opinions on that, thanks in advance.
 
Icepower is plain class AD, I have just removed a 1000ASP PCB from the heatsink and checked to be sure.

The inner loop produces the phase shifted self oscillation, but there are two op-amps on the way of the pseudo "triangle wave", which is not a good thing at all because it contains frequencies that two MC33078 can't handle properly. The outer loop drives the inner loop with gentle negative feedback for further linearization.

I think they use two feedback loops (one pre-filter and other post-filter) because they feared using the carrier residual on the output directly as a "triangle wave" as done in UcD. This requires a high quality carrier residual, which implies an high quality output filter with negligible parasitics, and ICEpower output filters exhibit high parasitic inductance on the (big) capacitors, not to mention the complete absence of ground planes and the compromised layout to reach such a small size.

Then Bruno demonstrated that it was possible to use the carrier residual on the output directly as a "triangle wave", and he did it routing the carrier residual directly to the comparator without any op-amps on the way, and using a differential approach. Bruno layouts use ground planes, and the modulator is in a spatial plane perpendicular to the power plane, so they are much better.

ICEpower comparator (LM319) is not wired in differential mode like in UcD, they rely on the MC33078 to add up both carrier residuals (pseudo "triangle waves") and then the signal goes to both sections of the LM319, one wired as inverting and the other as non inverting, to generate complimentary logic outputs and drive the two IR2010. This is not very clever.

I derived my own self oscillating modulator scheme from these two, which also uses two feedback loops, but both are post-filter...
My output filter has high C, low L and negligible series inductance on the capacitor. My carrier residual is sensed fully differential and goes to a TL3016 high quality comparator with only passives on its way. Only audio frequencies go through op-amps in my modulator, and all their inputs (from "outer world") have capacitors to ground, so RF rejection is very high. I use two poles, one passive and one active. ICEpower also seems to use two poles.

The resulting amplifier was just presented at NAMM last weekend... It's about to be released.

In general, you can safely and completely forget about ICEpower (and all their marketing hype) if you understand UcD, because UcD is ICEpower and a lot more, but made simpler :D:D:D (although it's useful to know the weaknesses of ICEpower to avoid repeating them)

EDIT: Another major difference from ICEpower to UcD is that ICEpower modulator does not damp output filter properly and relies on a 5 ohm resistor mounted on the heatsink for some damping (which imposes limitations on sustained trebble output), while UcD (and my approach) result in load-independent frequency response.
 
Last edited:
Yes, I meant you with pseudo-science. Sound is a loudspeaker thing :)

Concening the other question I can't tell much for the moment. I was just hired to design the amp, and my customer told me that they were at NAMM and they presented the product (preliminary prototypes) and many customers got interested due to the high power and compact size. I wasn't there, but I think there were another thousand companies presenting several thousand other products, and my customer only sells plate amps to high volume speaker manufacturers, so even if I told more, you have little chances to find more information or references on the web (or get the product once it's on sale). Sorry.


To darkfenriz:

Concerning ICEpower open loop gain, I don't know, but frequency response plots from datasheet tell that the outer loop is probably running out of OL gain around 20khz, and it's likely to increase at 12dB/oct below that frequency. The inner loop probably has some gain up to a higher frequency, but it's not very useful because it does not provide linearization or damping to output filter.

Have you seen this?
http://www.diyaudio.com/forums/clas...te-finished.html?highlight=icepower+schematic

This schematic seems to have several mistakes but it's still useful for learning.
 
Last edited:
Congratulations with the amp Eva ;) (a picture wouldn't hurt :))

Hadn't seen that (B&O schematic).
Seems thay are using 4. order output filtering ... didn't know that.
I have a hard time understanding why they use LM319 .... it's just so damn slow ....
I use TL1711 for my latest design. Rail to rail, damn fast and you can run the output as +-5V if desired. Only draw back is the size when hand soldering .... but it's doable ;)

Wonder why thay want to be able to switch between different resistors in the feedback path ... does it have to be set according to PSU voltage or something??
 
The 1000ASP that I have does not use 4th order output filtering, the output inductor has 4 sections but they are just connected as two series pairs. I think the schematic has some mistakes.

If you don't understand why they use LM319 (which is just one step ahead of LM393 and LM311, and is not that bad) compare prices, particularly for high volumes. They did a good cost reduction job. LT1711 seems crazy expensive.

btw: I use LM319 for current limiting for the same reason, it's dirt cheap and propagation delay with low overdrive voltages is not that bad.
 
Hi Eva

Yes of course they want the cheapest components, but as you have pointed out your self, the comperator is a critical component in the design .... so I just saying that it might impede the overall performance .... On the other hand I'm just doing DIY for the fun of it and don´t have to care much about whether a key component is 2 cent or 2 dollars ;)

Yes the LT1711 is quite expensive, but you should try it out ... think it will beat the TL3016 in performance :)

And LM319 for sensing purposes .... yes by all means ... one shouldn't overdesign (though a lot of that is seen in the HiFi world :D)
 
Re: LT1711

Worth the price of admission. Speed and I get a Q and a Qbar output, no charge and no addition inverter delay. And yes, I need both outputs for my design.

Over design?? On a site with nut jobs that think silver wire, teflon caps and ever higher and more expensive levels of unobtainium components sounds better than the pedestrian stuff?? All designs on this site have been value engineered to save the last penny! NOT!
 
You know, everybody wants to feel like doing great things.

In the audio context some use silver wire, some use teflon caps, some use >$1000 power cords, some are still trying to design the perfect class AB amplifier (with the usual layouts and single sided PCBs and even attaching the power transistors to the PCB with wires :dead:) while all that is about to become phased out, and a few even try to innovate doing things that few people have done before ;)

ICEpower was a big step ahead but it left a lot of room for improvement too.
 
Last edited:
If something is repeated enough times without somebody claiming the opposite, it becomes true... If it's said by someone with a big name, there are little chances to see someone daring to say the oppsite.

Figure out that they even made their way into the pro-audio market with such a product :D:D:D it's widely used in powered speakers, although manufacturers are now progressively switching to other better alternatives. On the other hand, B&O were among the first to dare to propose class D as a serious alternative to replace class AB.
 
Disabled Account
Joined 2014
The 1000ASP that I have does not use 4th order output filtering, the output inductor has 4 sections but they are just connected as two series pairs. I think the schematic has some mistakes.

If you don't understand why they use LM319 (which is just one step ahead of LM393 and LM311, and is not that bad) compare prices, particularly for high volumes. They did a good cost reduction job. LT1711 seems crazy expensive.

btw: I use LM319 for current limiting for the same reason, it's dirt cheap and propagation delay with low overdrive voltages is not that bad.

Eva,

I also was digging into the ICEPOWER 1000ASP module, I don't know how comes this amplifier works without any NAD/AND gates like other amplifiers, I only can see direct output of LM319 going to the IR2010. this of course after after the 33079 and 33078.

And there is two feedback loops from the output going back to the OPAMPS.

Never seen such design, and YES there are many mistakes in the schematics.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.