UCD180 questions

[paradigm]

So, why can't it cost a Dollar in audio implementation?

Everyone knows that medical gear, always is heavily over priced.

Hmmmmmm.


/Roland

[UrSv]

Ahem, what was that about sticking to the UcD topic earlier in the thread?

[Swedish Chef]

Everyone knows that medical gear, always is heavily over priced.

This is a real sidenote but Roland, since medical instrumentation is one of my majors I have to both agree and disagree. Medical instruments for sure are expensive - mostly for good reasons. I would be glad to discuss why with you in personal!

But most audio gear (especially in the "high end" segment) are even more overpriced. While the price you pay for medical gear goes into high performance quality parts, rigorous testing and scientific research, the money you pay for most audio stuff goes into a pretty box, marketing hype and just general cans of snake oil. JBL Pro components is one exception to this and I think we both agree on that!

Now, back to the topic...

/M

[xman737]

Hi Bruno,
The spec. of UCD180 said it can output 105W in 8 ohm load. Does the output power change with the input voltage? The max. input voltage is +- 50V. What is the output power when supply +- 40V into UCD180? Is that 84W?

[Bruno Putzeys]

No news there... Output power scales with the square of supply voltage.

At 50V the theoretical power output into 8 ohms is about 150W.
In fact, if you hook up the amplifier to a +/-50V regulated lab supply, over 140W/8ohms will indeed be delivered.

The reduction to 105W is because a realistic power supply (transformer-rectifier-elcaps) has a lower output voltage at full load.
A normal-size unregulated DC supply that puts out 50V at no-load is likely to drop to around 40V when the amplifier is playing at max power.

[FuriousD]

FuriousD,
Hereby a FFT of 7.2kHz at 20Vrms in 8 Ohm load, of the UcD180.
I am sorry I can't give info about OEM customers they will not appreciate this. But when the amps are ready we can show you some details about the product.

Jan-Peter

IVX, OK here it comes ...

Sorry for the delay, I've been away travelling for a week or so, and have only just got back.

Firstly, I'd like to congratulate John W on the performance of his amplifier. >-140dB noise floor, All Harmonics below -108dB, and falling with order. I think that kicks just about every class AB amplifier into touch and proves well and truly that class-D is the way to go.

On the UCD; Once you take the scaling into account (Jan-Peter, the F4 button or your audio-P will scale your plot reference to 0dB, it just makes the whole plot more readable), you have second harmonic at -89dB and third at -78dB with a noise floor of -135dB (approx). The THD would most definately still give a good class-AB amplifier a run for its money at 100W, but this plot is not at 100W, it's at 20V RMS into 8Ohms, only a mere 50W.

Also, a-135dB noise floor on the UCD equates to about 18 Bits-ish resoultion (I'd have to get my calculator out to check exactly)

It's pretty hard to compare apples with pears after all, and I have to wonder why Jan-Peter has only done a 50W plot to compare against a 120W plot from John W, and at the end of the day, this is only one measurement, and as we all know with measurements...

Once again, over to you Jan-P.

F.

[Bruno Putzeys]

Originally posted by FuriousD
but this plot is not at 100W, it's at 20V RMS into 8Ohms, only a mere 50W.

It's pretty hard to compare apples with pears after all, and I have to wonder why Jan-Peter has only done a 50W plot to compare against a 120W plot from John W, and at the end of the day, this is only one measurement, and as we all know with measurements...

Well as far as apples with pears are concerned...

JohnW's amp consists of a very highly refined PWM DAC followed by a synchronised power stage with <2 nanosecond switch timing precision and with a high order control loop around it.

The UcD module uses a gain-limited 2nd order modulator with a simple power stage that has (as a consequence of its simplicity) a >20ns switch timing error.

I could duplicate JohnW's amp and the results produced by it. This would be a totally futile exercise: John already has it all-singing-all-dancing. I could only confirm that his distortion results are exactly what is to be expected from the selected circuit topology (thus also confirming the quality of his hardwere implementation).

For UcD the choice was firmly made to have low output impedance and use minimal circuitry. Above Prated-3dB, THD due purely (=mathematically) to the modulation technique starts to dominate over power stage THD. There you have it from the designer, no need for JP to measure at 100W/8ohms and confirm this fact. Besides, a THD+N plot is already published in his data sheet and shows this effect as well.

Both JohnW's amp and the UcD represent "optimal design points" in their own right ie. a set of requirements was put forward and a minimal amount of circuitry produced to get to this result. This is what apples to apples is all about. You're invited to make a similar expenses/returns comparison on other class D amplifiers.

I see no reason why anyone (such as JP) should waste his time trying to defend the THD figures coming out of the UcD180 module, just like I would not advocate harrassing JohnW to "show his output impedance plot" and prodding him to defend it in comparison to UcD's performance there.

If you want lower THD, it can be done - at a cost in terms of other performance specs or in price or both, depending on what you want precisely. This is what engineering is all about.

Originally posted by FuriousD

Also, a-135dB noise floor on the UCD equates to about 18 Bits-ish resoultion (I'd have to get my calculator out to check exactly)

It may be a good idea first to obtain information concerning the number of points and the type of window used in the FFT before getting out your calculator.

Best,

Bruno

PS & FYI: the unweighted 20Hz-20kHz noise output of the bare UcD circuit is somewhere below 20uV.

[JohnW]

Recently I had the chance to meet with Bruno in Belgium,

I guess you could gauge the depth of discussions we had, that while we talked in the Hotel bar that evening – we completely missed the fact that the Hotel reception was being raided by a masked Gunman less then 10 meters from us. Who then apparently screeched away in a Black BMW waiting outside – and then having to explain to the Police, with all honesty, that we hadn’t noticed anything.

During the visit I had a chance to listen to Bruno’s UCD technology, and also directly compare 44.1KHz CD / SACD.

The most immediately obvious signature of Bruno’s system was the complete lack of “Digital sibilance”, and the very tight & controlled base (this does not mean to say “bass light”). The fact that we had the system playing while we spoke – gives you an idea that was no annoying sibilance (I find it impossible to hold a conversation while sibilant systems play in the back ground). However, I also failed to notice Hotel being raided later…

The Bass performance was typical of what I have become accustomed to from high performance Class D amps (by this I mean closed loop multi order designs) – incredibly tight, deep and controlled – no hint of “Bloom” – and in no means Bass light!

Sound stage was very good – but I’m used to electrostatic speakers – and just missed that extra dimension – that said, I can honestly say Bruno’s “Box” speakers where probably the best I can remember hearing of there breed (yes Wilson Pup’s / Puppies are also very good – but for the $$$$ they had better be)!

There’s absolutely no question that UCD has the potential to perform – you simply could not have an issue with any aspect of its audio performance.

I also had my first chance to compare Red Book CD (44.1KHz 16Bits) verses SACD – played though the SAME DAC this means 2.8MHz PDM output stage. In this DIRECT comparison, I was VERY under whelmed by SACD. Sure, the Base was deeper (but strangely on the recordings I listened too, it seemed slightly distorted in the low end), and the sound stage opened JUST a tiny amount. Apart from the “Bass” I found it VERY hard to tell the difference between the 2 formats – certainly less then say changing cables – as an engineer and I cannot believe I find myself saying this….

This was a DIRECT comparison – not like in convectional units where the DSD stream and PCM formats are handled completely differently.

I would like to thank Bruno for his time and hospitality, and am gland to have finally met - in person, an undeniably competent Class D / digital audio designer.

John

[FuriousD]

Bruno,

Thank you for taking the time to respond on this topic. I have not meant to cause any offence by my comments. Do not take things so personally.

I am fully aware of the engineering & costs involved in full and half bridge switch implementations, having spent many years designing in this field. I like everyone else on this forum, like to have a level playing field. John W's amplifier plots have been of great interest to me, as I am sure that they have been to everyone else. No one is asking for a defence of anything, merely more information to compare against other implementations, and besides which JP did ask!

FuriosD,
Tomorrow I will do some FFT measurements on the UcD180, do have some more specific measurments what you like to see?
Best regards,

Jan-Peter

As a technology still seeking real uptake and approval in the higher end of the audio market, it is necessary for class-D designs to not only meet the standard of current class AB designs, but to more than exceed them! The audio market place is far more forgiving of a bad Class AB than it is of similar Class-D designs.

In the end, the technical performance is of no real interest if the product does not sound fantastic, and from all comments on this forum the UCD more than aquit's itself here.

Regarding the -135dB noise floor, I was assuming the best possible set up on the AudioP (Usually woeful admittedly given the AP input stage), but like I said, thats not exact.

Also, I'm not really sure what the fascination is with THD+N plots. It's like hiding the information under a blanket when you really want to knows whats under the bed.

F.

[Bruno Putzeys]

Originally posted by FuriousD
Thank you for taking the time to respond on this topic. I have not meant to cause any offence by my comments. Do not take things so personally.

No offense was taken. If you feel I took it personally it could be projection
Your plea for a level playing field is well taken, hence my surprise at output power apparently being the only matter that needed clearing out in order to level it.

Originally posted by FuriousD
As a technology still seeking real uptake and approval in the higher end of the audio market, it is necessary for class-D designs to not only meet the standard of current class AB designs, but to more than exceed them!

This is somewhat problematic. A good class D designer can always design a class AB amplifier of better performance. If someone asks (pays) me I can build a linear amp with -140dB THD in a week's time. It's a moving target.

Originally posted by FuriousD
The audio market place is far more forgiving of a bad Class AB than it is of similar Class-D designs.

My own experience with the hifi mags (rags) is quite the opposite. I continue to be amazed at how far reviewers are willing to push back their performance standards in order to be able to endorse S-Master, or at least not to break it down to the last morsel. Good analogue class D's get the boot because they're analogue and therefore passé.

Originally posted by FuriousD
Regarding the -135dB noise floor, I was assuming the best possible set up on the AudioP (Usually woeful admittedly given the AP input stage), but like I said, thats not exact.

Especially if you're comparing to plots made on a Rohde&Schwarz.
I personally do not join the choir of AP bashers. If you have problems with the AP input stage, you've got an EMI problem, or are very near to one. That said, the first system two's were indeed a disaster. Luckily I don't have one of those.

Originally posted by FuriousD
Also, I'm not really sure what the fascination is with THD+N plots. It's like hiding the information under a blanket when you really want to knows whats under the bed.
F.

In this case, a mere cursory glance at the THD+N graph would have told you not to expect <-80dB harmonics at 100W. It would certainly not have required a fascination for it.