UCD180 questions

[Jan-Peter]

Peranders,

We are working on two version of the UcD400 in parallel one with faston connectors and one with JST (eVH) connectors. The UcD400 with JST connectors is earlier ready for delivery.

Regards,

Jan-Peter

www.hypex.nl

[tiki]

For o-shaped transformers, available in Germany, look here.
Is that
clock something for John?.
Just found it, did not try.

I like Bruno's idea of a (secondary side?) PFC. To reduce possible spikes from the supply and ringing between combined low-ESR-Cs, one could insert (lossy!) ferrite beads into the wiring between the bigger electrolyte cap and the smaller (and lower ESR one) cap. Is it a good idea to use Schottkies for the rectifiers instead of FREDs?
Regards, Timo

[Konrad]

Kind of show to bring clock and jitter into the tread.

Jan-Peter as i see on the prev pictures the connection's are kind of close when using kable connectors (it seems), it may be an idea for the mkII.

abt switching amp:
I find it kind of hard too supress the switching using only 2.order filtering. As i see the photos and the output filtering using one core coil(s?) (i belive) and coils at the same core and one cap ( i belive). The only filtering of that kind i remember seeng is 5.order ( to coils at same core and one cap, the coils magneticaly opposit copled or wond) remember it as sub bass filtering values different ofc.

If it really measures as it does, is it with ONLY 2.order filtering at the output?

Abt crossover distortion: it depeds on the linearity of the 'linear' transfers involved, in some designs they are present at other stages than output stages, if so they contribute as well.

[Jan-Peter]

Konrad,

Faston are one of the best connectors because of the high current capability and the low resistance. But multipole connectors are easier for manufactures.

Indeed we use only one L and one C as outputfilter. For instance by the UcD400 the voltage is 60VDC and the Fsw is 450kHz, we have only 300-400mV of HF outputvoltage. Thereby it is a sinewave (1e harmonics of the switching frequency)

I don't understand;

Abt crossover distortion: it depeds on the linearity of the 'linear' transfers involved, in some designs they are present at other stages than output stages, if so they contribute as well.

Regards,

Jan-Peter

www.hypex.nl

[Bruno Putzeys]

Originally posted by Konrad
abt switching amp:
I find it kind of hard too supress the switching using only 2.order filtering. As i see the photos and the output filtering using one core coil(s?) (i belive) and coils at the same core and one cap ( i belive). The only filtering of that kind i remember seeng is 5.order ( to coils at same core and one cap, the coils magneticaly opposit copled or wond) remember it as sub bass filtering values different ofc.

If it really measures as it does, is it with ONLY 2.order filtering at the output?

Abt crossover distortion: it depeds on the linearity of the 'linear' transfers involved, in some designs they are present at other stages than output stages, if so they contribute as well.

Hello Konrad,

Nobody in their right mind uses higher order filters than 2nd these days.

There used to be a time when people wanted to have a squeaky clean signal from a class D, just because they were used to seeing such signals from linear amplifiers. Seeing remnants of the carrier gave them goose-bumps.
Distortion analysers too weren't exactly thrilled to be greeted with a carrier residual. Because of that, some went to great length at the output filter. The famous Brian Attwood 1983 paper showed a 6th order filter with parallel notches!

Apart from the optical (oscilloscope) quality of the signal, such an amount of filtering is completely unnecessary, because speakers do not react in any way to the presence of carrier related components. A second order filter leaves about 100mV worth of HF ripple on the output and that is low enough.
Some people are afraid that there might be some sort of demodulation but it simply doesn't happen! Speaker nonlinearities only manifest themselves when the cone starts moving, and at 400kHz this is clearly not the case. A tweeter acts like a fairly steep filter right after 20kHz. At 400kHz the acoustic response is likely to be 100dB down. Add to that 50dB from the output filter.
You might wonder about dissipation in the voice coil then? Well, suppose a 400kHz, 100mV signal hits a 6ohm tweeter. At 400kHz, this tweeter's impedance is more likely to be around 30 ohms. 100mV into 30 ohms translates into 3.3mA. Now, most of that 30 ohms is reactive ie. dissipation only happens in the 6 ohm part: 6ohms*(3.3mA)^2=66uW! This is not going to warm up any voice coils!

Higher-than-two orders of filtering are also undesirable. You get the nonlinearity of multiple coils and higher/more complex output impedance. The output impedance of 4th or even 6th order filters makes it impossible to get anywhere near a flat frequency response in a realistic speaker. Using feedback to control the output impedance becomes increasingly ineffective as filter order goes up. A second order is the highest you can go while retaining full control over the output voltage. At higher orders you can only reduce the problem, but not eliminate it.

What the output filter should do is simply attenuate the carrier sufficiently, including the harmonics. A well designed output filter will leave only a sinusoid-like residue standing, without further switching hash. A second order filter is more than sufficient to do this, provided the coil has low parasitic capacitance and the capacitor has low self-inductance.

So how about THD measurements then? Well, since the outband components don't get reproduced by the speaker, they don't matter. What we care about is what happens inside the audio band (up to 20k, maybe 40k at most). Instead of adding the filtering that's necessary to keep the carrier away from the analyser to the output of the amplifier, it's added to the input of the analyser (or past the notch filter). Remember, only the analyser needs that extra filtering, not the speakers and neither do our ears. It saves quite a lot of money by the way: only a low-power filter on each analyser instead of a high-power filter on each amplifier.

Analysers like the AP2 tend to come equipped with such filters anyway. DA converters too are no longer fitted with unnecessarily deep filters. A modern 120dB DA converter, measured without a pre-filter (or post-notch filter) at the analyser is unlikely to score much better than 80dB SNR. Especially when it's fed DSD signals - brace for 40dB SNR if you don't have an AES17 filter installed.

I hope I've been able to put your concerns about filtering to rest.

Now, concerning cross-over distortion - that class D amplifiers should not have cross-over distortion is an oversimplification. At larger output currents, a class D amplifier changes from so-called soft-switching to hard-switching. This régime change happens twice every audio cycle. This could be construed as a form of crossover distortion which is offset from the zero-current point. Power stages with short dead times are less clear-cut in their behaviour, so things get smeared out over the entire output current range. The distinction between crossover and ..uh.. non-crossover distortion becomes vague.

Of course (I believe this is what you are referring to), if someone manages to put an op-amp with crossover distortion in front of the amplifier, well, the audio quality will be marred by the distortion of the op amp. This is not exactly pertinent to class D...

Cheers,

Bruno

[Konrad]

Thanks info & fast reply.

Abt crossover distortion: it depeds on the linearity of the 'linear' transfers involved, in some designs they are present at other stages than output stages, if so they contribute as well.

The comment is kind if OT.
What i ment is even a simple diff stage do have their limmits, and transfers. and is comparable to crossover dist. If differential stage is used after the integrating stage it do have its part in the amp and is supressed by the openloop-feedback gain, trigging at the same level allways .... will remove crossover distortion it just came to me as a suprice !!

BTW 3-400 mV ac at high freq may show as dc even at some of todays digi instruments.

Regards

[Jaka Racman]

Hi Bruno,

At larger output currents, a class D amplifier changes from so-called soft-switching to hard-switching. This régime change happens twice every audio cycle.

One way to avoid this is to use very large ripple current in output inductor (more than twice the output current) and very large MOSFETS paralleled with some nF caps. Advantage is lossless ZVS and low on state losses and I also think parasitic diode does not come into conduction at low enough Rdson. Disadvantage is unsymetric transition time and requrement for adaptive gate drive. Is there any disadvantage of using this approach in non cost sensitive applications (besides that it works best in fixed frequency applications) ?

Best regards,

Jaka Racman

[JohnW]

BTW 3-400 mV ac at high freq may show as dc even at some of todays digi instruments

While this might be true, its simply a "measurement" issue, and does not have an effect on sound quality. More of a concern is to insure that the front end (buffer / gain stage) of the amplifier is not affected by the circulating RF on the PCB & through the feedback path – great care has to be taken with the choice of OPAMP / input stage design.

Hi Tiki – thanks for the Link, I’ve now design an ECL based clock / DAC.

John

[pburke]

Originally posted by Bruno Putzeys

The 20000uF per rail that I have on my "esoteric amps" is really as far as I dare take it. I'm planning on a PFC supply to address this.

So if I were running a monoblock UCD400 with a power supply design similar to the one posted a few posts back by OA51, I'd be in the ballpark?

To me it would seem more logical to me to pair the woofers together on one two-channel amp and the mid/highs on the other, so that heavy bass content isn't seen by the mid/high amps (in so far as this should be a problem). Usually the performance loss on 2-channel amps (or performance gain on mono blocks) lies in the inherent impossibility of a 100% correct grounding scheme combined with ground referenced inputs. The UcD modules have differential inputs, so crosstalk over the ground connections is nonexistent.

The reason for thinking high/low in one chassis was to a) maximize the power supply for the bass, since the higher frequencies won't draw as much juice, and more importantly b) to allow me to place the amps right behind the speakers. I don't believe in long speaker cables, while 6 foot ICs are just fine. So if I were to biamp from just two stereo amps, I'd be placing one behind each speaker. Meanwhile, I have decided to go with monoblocks right away, although the price of admission is a little higher.

Since the power dissipation in idle is low, actually you can afford to keep the amps on all the time.

I don't turn my current amps off - and they dissipate 60watts at idle each. 3watts are a welcome savings on the power bill.

You need a "soft start" circuit for charging the power supply caps without blowing the mains fuse.

So I'll be adding one. Not a huge deal.

You can use a plastic enclosure if you wish (except for cooling of course). Any EMI emanating from the amp is negligible

that's great - almost like an invitation for a "who can build the coolest and most original UCD amp"

Thanks for the replies. I'm looking forward to the UCD400 modules

Peter

[matjans]

Originally posted by pburke
I don't believe in long speaker cables, while 6 foot ICs are just fine.

why is that, if you don't mind me asking? I would think it's just the other way around...