Which chip and whatever happened to Mueta and UCD?

Tiki,

Thanks for your nice comments.

For the moment we sell this module direct to everybody for the same price. It will be available within a few weeks.

The UcD is excellent for a more way system, we already build for a big loudspeaker manufacture a 2-way and a 3-way active studio monitor. Not heat at all and sound is much improved in comparing with the previous used TDA7294.

To Charles:
Upper frequency is -3dB at 50kHz, 1st order, see graph.
With a powersupply of 2 x 45VDC, the outputcarrier is around 300mV. I don't know yet if this was Vtt or Veff, I will measure this next week and post the answer.

Jan-Peter

www.hypex.nl

P.S. Graph is from the original document from Philips and measured by Bruno.
 

Attachments

  • freq.jpg
    freq.jpg
    8.1 KB · Views: 602
Thank you Bruno for your blessings🙂, actually i've nothing problem with coil calculations generally, i just saw same core in 100w-180w UcD (maybe 400w too?), which pretty small, and i thought that this is gapped EP17 with ~15 turn of 1mm wire or a something else.. I use unclosed core (or air) usually- bigger size & worse EMI (it's ok, i'm just DIYer), but 50w or 1000w it works well.
Other experiment is much dramatic for me now, i've plan to add current feedback loop from filter cap like MUETA to the UcD_like amp..simulations is ok at least. Don't know why, but i wanna do it. Really interesting, how can affect a sound the inverse dependence of THD from frequency?
 
IVX,

Some info what I would like to share with you:
I would advise you to make a good outputcoil where also the HF-emission is as low as possible. This will definatly also improve the performance of the Class-D amplifier because you have less inteference and keep al signals more clean.

Good luck with implementing a UcD Class-D look a like together with a capacitor feedback just like the Mueta. I am looking forward to your testexperience!

Perhaps we can start a new topic? ;-)

Regards,

Jan-Peter

www.hypex.nl
 
Hi Bruno,

I just voted, but I think it would be more noticed, if you would open a new thread with that proposal. Maybe this forum could also include switch mode power supplies, since it deals with similar techniques. As it stands now class d amplifiers are discussed in Solid State and Digital forums and there are more posts and threads about class d than there are all the posts in Car Audio forum.

I hope moderators would consider your proposal, although I am not sure they know you are one of the best class d designers in the world.

Best regards,

Jaka Racman
 
Whatever happen to.......

Hi Bruno,

A quick couple of things....

What ever happened to the Phased Power Analogue concept?

Has it been put back in the drawer for a while whilst you devlop the UcD?

Lastly, if i were in your position, i would have at least played with the mueta style of feedback on a UcD amplifier to satisfy my own curiousity.... Are you willing to let us know the effects of doing such a thing (if indeed you have).

Just curious..

Cheers

Sherrif
 
Re: Whatever happen to.......

Sheriff said:
What ever happened to the Phased Power Analogue concept?
It's pretty much up in the air still. The problem is that a discrete implementation is too complicated to be practically useful and that the interest from semiconductor folks in really high-end (but more expensive) solutions is zip. So there isn't any funding to set it rolling again.

Sheriff said:
Lastly, if i were in your position, i would have at least played with the mueta style of feedback on a UcD amplifier to satisfy my own curiousity.... Are you willing to let us know the effects of doing such a thing (if indeed you have).
I'm not sure what you mean by "applying Mueta style feedback on a UcD". That would automatically result in a Mueta amp...

The phase lead of UcD serves a purpose similar to that of Mueta's current sense (in practice they use the derivative of the output signal too). The two concepts are closer than they look. The difference between hysteresis switching and phase-controlled switching is not instrumental. Hysteresis switching has the advantage of 3dB better loop gain, but produces greater frequency modulation of the carrier (Mueta uses quite a complicated circuit to counter this by modulating the hysteresis). Phase-controlled switching has better carrier frequency stability with modulation and in terms of parts count is frighteningly simple.

Then there's the control loop affair.
The Mueta concept used a second integrating loop around the self-oscillating portion from the onset. The UcD schematic publicly known from the patent didn't have that, but the one used in products does. The chief difference is that UcD's second loop is gain-limited such that loop gain does not increase below 20kHz. You may remember that I'm a firm believer in frequency independent THD, hence. It also allows for the amp to be unconditionally stable, even when clipping with no load attached.
Mueta's amp has an op amp based integrator that does continue upward below 20kHz. This explains their lower THD numbers.

Now, there's a chap called Jan-Peter who is more competitively minded in THD matters. He's constantly prodding me for extended control loops to get "really low" numbers. I still owe him the optimised values for a second-order integrator, but tests with estimated values were very encouraging. The last thing I saw was on the order of 0.0005% at 100W (but not stable above that). I'm leaving it to him to post graphs when the time is ripe.
 
That was quick...

Hi Bruno,

Thanks for the reply.

Is the PP-A in a position to be put on to an FPGA?


I am expecting a couple of UcD400 boards from Jan Peter early next week and very much look forward to testing them.

Now a high powered UcD with THD 0.0005% across the entire audio band should keep everyone happy. What filter did yo use on the AP to get these results?

The Mueta website quotes 0.003% THD and noise measured without any filters).

Best Regards

Sheriff
 
Re: That was quick...

Sheriff said:
Is the PP-A in a position to be put on to an FPGA?
Hmmm. Not sure if you're not confusing PPA and PPD(SD). The latter is open loop digital, so it does all processing on an FPGA. That thing exists and works fine. The PPA uses the same power stage arrangement but it has a control loop (ie is analogue). The output from a 6th order integrating loop filter is fed into a fast 4-bit ADC. This signal is then processed by an FPGA to set the right switches.
So, part of PPA goes into an FPGA but there is still a significant amount of analogue processing going on.


Sheriff said:
What filter did yo use on the AP to get these results?
The Mueta website quotes 0.003% THD and noise measured without any filters).
Not too fast 🙂
I think Mueta already regrets having made that claim, for it suggests that the AP was set to its full 500kHz bandwidth. That's not the case. It was set to 22kHz. However, what they meant is that they were using neither the AES17 filter nor the passive fiilter box AP introduced later (opt 0025 or what was it).
Now, the 22kHz filter of the AP is a fourth order butterworth affair. The AES17 filter is a 11-pole elliptical filter. As far as THD and noise measurements on class D amplifiers are concerned, this means precisely the same: components outside the audio band are ignored.

I suppose everyone is entitled to a few less-scientific more-commercial claims in their sales talk. Mueta's is "no measurement filter used". Mine is "zero order loop behaviour". I'd strongly doubt that either they or I have sold one amplifier more because of that but it's part of the audio folklore.

Some background information on the AP:
The raison d'être for the AES17 and the 0025 is that the autorange circuitry in the AP bases its settings on the wideband signal content. The traditional bandlimiting filter operates after the THD analyser notch (in order to prevent the THD from the filter from dominating the measurement). It means that if there is HF noise around (like in DACs) or a carrier (like in class D), the analyser's gain is too low to properly resolve THD and noise for signals that are small compared to their outband noise.
The AES17 filter is placed right at the beginning of the measurement chain (and hence is required to have extremely good THD+N specs itself). It allows the AP to choose its gain settings according to the inband content alone.

In general I use the AES17 setting for THD vs power sweeps in order to get proper readings down to the noise floor. For THD vs frequency I set it to 30kHz, because the output signal is kept at a constant high level.

The 0025 was introduced to remove very high-frequency noise (EMI) to prevent it from causing demodulation in the AP. I'll tell you this: if you need the 0025 to get a proper measurement result, the amp under test has a serious EMI problem. If it performs within legal EMI limits, the 0025 is an unnecessary investment.

Needless to say, I don't have one.

Cheers,

Bruno
 
Apples with apples

Hi Bruno,

Yes i was refereing to PPA - but i now appreciate that the tricky stuff still resides in the analogue processing. Thanks for the info.

With the THD, It is good that we can compare apples with apples as it were.

We have in fact measured the Mueta hybrid amplifier on a sys 2 cascade, with no AES17 filter with various standard AP filters in place, 22, 30, 80 500 etc, but i did not make a note of the results (i listen rather than create audio circuits).

We have an AP system 1 dual domain - not as flash as Jan Peter's but is quite useful. Did not manage to catch the recent online seminar on class D that AP ran - and i am aware of the problems class D amps can cause them.

The proof of the pudding is in the listening....

P.S. Was that 0.0005% across the audioband? I have seen a graph where it is at that level to about 50Hz..?

Thanks for your time.

Cheers

Sheriff
 
Hi Bruno / Sheriff,

Just like to add my 10 cents worth on the topic of AP filters and Class D.

I have AP1, AP2 C+ and R&S UPD05, all of the units are affected by RF on their inputs to varying degrees – with the AP1 being the worst. Not only are the AP auto-ranging circuits affected, but also their (and R&S) ADC’s intermodulate with any carrier component that gets to them.

I’m guessing that due to the spread spectrum (correlated) nature of hysteresis type modulators you use, you might not notice noise floor artefacts – but with fixed frequency Class D modulators – the result of this RF IMD is obvious, with nasty spurie on the noise floor.

In my experience the AUX0025 filter is only useful for measuring class D units without a LPF on there outputs (some Sub-Amps designs which drive the speaker cone directly - etc), while the internal AES17 Pre-filter board is almost mandatory on AP2’s for high dynamic range / low THD measurements.

Many hours have been wasted tracking down spurie components on FFT noise floor that are the result of AP RF IMD “issues”. These designs have pass CE RF emission tests with a large headroom.

Unfortunately, the AES17 pre-filter cannot be fitted to AP1’s. The AP1 ADC is EXTREMELY sensitive to RF on its input – I find the unit practically useless for accurate low THD / DR / FFT measurements of Class D / Noise-shaped DAC’s without an external filter. Use the FFT mode on the AP1 to see the IMD products on the noise floor The AUX0025 makes little difference as it has a very little attenuation below 100KHz (about –7dB @ 100KHz), see graph below.

Any internal filter boards installed in the AP1 /2 will not help, as they are AFTER the auto range circuit – the AES17 Pre-filter needs to be install (on the AP2’s it can be retro fitted by the customer).

John

AUX0025 Frequency response, 0.002 Ohm input impedance, 100K output impedance
 

Attachments

  • aux0025 100khz.jpg
    aux0025 100khz.jpg
    24.4 KB · Views: 524
Below is an FFT of a TRUE digital amplifier, 118W 8Ohms 997Hz THD 0.0003% modulator and clock limited. Awtd dynamic range 114.5dB – again modulator limited.

The noise floor is analyser DR limited at low frequency (R&S UPD05) and noise – shaper limited (NS operating at 384KHz) at 20KHz. The R&S only has a ~30dB notch filter – I should use the AP, but I hate the AP’s user interface, the AP is not really a R&D tool – but its great for the production line….

118W 8Ohms 997Hz THD 0.0003% DR 114.5dB
 

Attachments

  • 118w 8r 8fs.jpg
    118w 8r 8fs.jpg
    36.8 KB · Views: 530
1W 8Ohms FFT plot of the above TRUE digital amplifier (about –30dB), now the “Poles & Zeros” of the noise shapers PWM “Linerizer” can be clearly seen. This is a 1st generation 384KHz PWMFs modulator. Later generation modulators achieve about –160dB noise floors (again @ 384PWMfs)

THD 0.0008% - but this result is noise only, as no harmonic (or non harmonics) can be seen above the noise floor – note the ~35dB notch of the fundamental by the R&S (0dB is referenced to the 118W 8ohm measurement).
 

Attachments

  • 1w 8r 8fs.jpg
    1w 8r 8fs.jpg
    36.3 KB · Views: 500
AP and production

Hi John,

I thought that the AP was a bit of a nightmare as far as production is concerned.

Take a look at:

http://www.milleraudioresearch.com/qcsuite/qcsuite.html

This thing is a doddle to set up and use - even i can test a bit of equipment to ensure that it is fit for shipping.... (and i am in sales...)

On a different topic to THD, Bruno mentions in an earlier post that UcD with the second integrator loop is unconditionally stable.

From this statement can i assume that it will drive any capacitive or inductive load, and give essentially loudspeaker independant performance?

'Then there's the control loop affair.
The Mueta concept used a second integrating loop around the self-oscillating portion from the onset. The UcD schematic publicly known from the patent didn't have that, but the one used in products does. The chief difference is that UcD's second loop is gain-limited such that loop gain does not increase below 20kHz. You may remember that I'm a firm believer in frequency independent THD, hence. It also allows for the amp to be unconditionally stable, even when clipping with no load attached.
Mueta's amp has an op amp based integrator that does continue upward below 20kHz. This explains their lower THD numbers.'


Cheers

Sheriff