UCD400 or ZAPPulse?

[Pasi P]

Re: Re: coupling caps?

Of course you can do this - no problem. This is electrically equivalent to simply replacing the onboard coupling electrolytics by the film caps.

Sorry about my not so good understanding of electronics but..
If i understand correct, input caps are in op.amp_output_(UCD180). Is it really so that these can be shorted and moved before input of UCD (between input RCA and input amplifier)?

Other (stupid?) question. If unbalanced (RCA) input is used, is -input capacitor "in use" at all?

 

[matjans]

mods, could you guys split this thread into a ucd400 tweaking thread? I for one am still very interested in how to squeeze the last few drops of quality out of these modules and Hypex's Jan Peter and Bruno, amongst more diyers around here, have proven to be very helpful in the process.

I think moving this stuff to the ucd400 q&a thread would be useless since it's gotten WAAAY too long already.

 

[Pasi P]

mods, could you guys split this thread into a ucd400 tweaking thread? I for one am still very interested in how to squeeze the last few drops of quality out of these modules and Hypex's Jan Peter and Bruno, amongst more diyers around here, have proven to be very helpful in the process.

I think moving this stuff to the ucd400 q&a thread would be useless since it's gotten WAAAY too long already.

Good idea. Maybe it is time to start "Hotrodding UCD" thread?

 

[GregD]

Using a WIMA MKS2 cap of 2.2 or 4.7 uF is a much better solution. Before you jump to other similar brands, check if the leads are iron or copper. Those with copper leads sound much better than those with iron leads (most brands).

Even better yet is the Vishay SAL 123 RPM series. They just sound GREAT as a DC blocking cap or analog decoupling cap.

I took a look on Vishay's web pages, and they show a SAL 128 RPM and a SAL 123-A series. It looks like the main difference between the two is radial versus axial packages. It also shows that they're aluminum solid electrolytic caps. I thought film caps were supposed to sound better than electrolytics in the signal path.

 

[simon dart]

I've been following the rise of ClassD for a little while now, my triode strapped EL34 power amp sounds so fine but struggles with my current speaker's load and the UCD's have got me curious, electrically they appear so elegant.

I've got an old-ish Tripath TA2022 development board which in my opinion sounds great apart from the gritty opaque HF almost all solid state amps I've heard posses to some degree.

Are these new generation any better? Will those delicious micro dynamics and natural sibilance dissapear in the hunt for more grunt? Will I be disappointed if I take the plunge?

Simon

 

[lucpes]

Are these new generation any better? Will those delicious micro dynamics and natural sibilance dissapear in the hunt for more grunt? Will I be disappointed if I take the plunge?

Simon

I guess this post may be relevant for your question: http://www.diyaudio.com/forums/showthread.php?postid=640188#post640188

 

[simon dart]

Thanks Lucpes,

I'm definately tempted!

Simon

 

[Lars Clausen]

The main difference between ZAPpulse and UcD is that ZAPpulse uses pre filter NFB, while UcD uses post filter NFB. (As discussed in another thread).

With the new feedback tweak as proposed by phase_accurate, and optimized to fit the ZAPpulse, you have the option to run ZAPpulse (any newer version) in post filter NFB mode. or what could be called 'UCD mode'.

So maybe it's now UCD and ZAPpulse

The tweak will bring about the sonic properties associated with post filter NFB amplifiers (like UCD), whether you prefer these or not is a matter of taste. But anyway some readers of this website might like to check it out, and so can now tweak their ZAPpulse module to this mode.
Since ZAPpulse's gate driver system has very low time delay, a ZAPpulse in post filter NFB mode will operate with a switching frequency of 700 kHz.

Will post tech details in 'the easy tweak for NFB' thread later.

BTW: UCD400 is not the closest match for ZAPpulse 2.3SE. As you might have noticed ZAPpulse 2.3SE is spec'ed to +/- 80V and produce 700 Watts RMS in 4 Ohms. (Like UcD 700). So it would be more correct to match:

UCD700 and ZAPpulse

 

[Bruno Putzeys]

I'd like to clarify a few things.

The tweaked Zap is not a UcD. The UcD control loop has no feedback taking off before the output filter. Done correctly, this provides the most neutral sound possible, a feature that mixed feedback loops do not share (ie. Lars' criticism of the sound of his mixed feedback modified modules is correct).

700kHz switching frequency does not sound like a good idea to me. The switching frequency in UcD is determined by several parameters of which the propagation delay is only one. One can get any switching frequency from any prop delay.

I'm not going to publish a schematic of how to get the best performance from a Zap module using a proper UcD loop. Besides, while porting the UcD loop to other power stages will obviously work, most power stages will produce too high distortion figures, due to UcD's modest (but constant) loop gain. Having frequency-constant THD is a prerequisite for good sonics, but so is low THD.
Much of the secret in the UcD power stage lies in getting low THD at medium powers without resorting to overlapping FET drive. IC gate drivers generally result in very high THD at low powers when a similar dead time is chosen.

The UcD700 is called so because you can set it to deliver 700W continuously into a 4 ohm load and come back a year later with the module still playing happily. The semiconductors are not an issue here, the capacitors are. The ripple current in the local buffer caps is 0.2 times the RMS output current. This is 2.7A. You will need several caps in parallel to handle this at elevated temperatures with good lifetime. I've seen elcaps blow up on "high power" modules with insufficient local storage. Admittedly the current incarnation of UcD700 is seriously overrated with respect to this. 2 or 3 electrolytics per side (or 5 much smaller ones) would do quite nicely in all but "night&day" applications.

 

[Lars Clausen]

The tweaked Zap is not a UcD.

Nobody claimed that, and definitely not what we are trying to achieve. We just want DIY'ers to have a chance to play with post filter NFB.
In this implementation prefilter NFB is limited to above 150 kHz, (like also suggested by charles - the filter cutoff frequency) and therefore have no significant effect on the sound.

700kHz switching frequency does not sound like a good idea to me.

As can be seen from the datasheet of ZAPpulse it runs very stable even in this high switching frequency.

IC gate drivers generally result in very high THD at low powers when a similar dead time is chosen.

I agree with you. However our IC gate drivers have no built in dead-time, for this we have the proprietary 'pulse field' technology, that adjusts the dead time according to the level of power. This keeps low level THD to a minimum without adjustment. We have used this principle for years.

The UcD700 is called so because you can set it to deliver 700W continuously into a 4 ohm load and come back a year later with the module still playing happily.

As can ZAPpulse 2.3SE.

We are not using 'standard grade' electrolytics for this purpose, but the very low impedance RVI 136 types from Vishay (former BC Components).
With 150 millioms Zmax they will only dissipate around 1W at 2.7A RMS. So this transforms to a 40 degree elevation, making safe long life performance (7000 hours spec'ed) up to 65 C ambient temperature. In real life this lifespan is actually longer because the electrolytic caps are supported by lowZ Plastic pulse capacitors, and a low impedance link to the main PSU capacitors.
So the comparation with UCD700 is accurate

Below you can see our post filter NFB setup for the ZAPpulse. The values shown are the ones that work best for ZAPpulse, and will give a constant switching frequency independent of the load. For other Class D amplifiers other values might work better. The sound is best with plastic capacitors, compared with ceramic NP0 caps.

 

[Bruno Putzeys]

Hi Lars,

As can be seen from the datasheet of ZAPpulse it runs very stable even in this high switching frequency.

It's not about stability, but about dissipation and about a higher relative contribution of switch timing error to distortion.

I agree with you. However our IC gate drivers have no built in dead-time, for this we have the proprietary 'pulse field' technology, that adjusts the dead time according to the level of power. This keeps low level THD to a minimum without adjustment. We have used this principle for years.

I've had difficulty finding full THD data on your website. Do you have measurements (e.g. THD vs power at 1kHz and 6kHz or THD vs freq at various power levels)? It would be nice to see the added expense of the dynamic timing control reflected in the performance data.

We are not using 'standard grade' electrolytics for this purpose, but the very low impedance RVI 136 types from Vishay (former BC Components).
With 150 millioms Zmax they will only dissipate around 1W at 2.7A RMS. So this transforms to a 40 degree elevation, making safe long life performance (7000 hours spec'ed) up to 65 C ambient temperature.

The data sheet "ripple current" rating of elcaps is usually based on an internal temperature rise of 10ºC. While this does in no way preclude operation at higher ripple currents if traded for lower ambient temperatures, 40ºC is really high and liable to create even higher localised heating. It is precisely because of this that manufacturers are reluctant to specify operation at higher internal temperature rises - it becomes difficult to guarantee that all elcaps will have the same lifetime.

In real life this lifespan is actually longer because the electrolytic caps are supported by lowZ Plastic pulse capacitors, and a low impedance link to the main PSU capacitors.

If one figures the time constant between ESR of the Elcap and the capacitance of the local film caps, it quickly becomes obvious the film caps do not shunt an appreciable portion of the ripple current. At 150 milliohms and 500kHz switching, you'd need 2.2uF of very high quality film capacitance to make a dent in the ripple going into the local elcaps.
Counting on a direct link to the storage caps is not something I would do. If one wants the EMC behaviour of a module to be independent of the application, it is imperative that all HF currents remain on the module, with only the LF (audio) currents flowing between the module and the storage capacitors.
The UcD modules all measure invariant, EMC-wise and performance-wise, even if connected using 1 metre long -or longer- flying leads. This is how I measure EMC anyway.
Like with the film capacitors, I would doubt the effectiveness of short wiring between the modules and the storage caps. For the same 150mOhms and 500kHz switching, the total wiring inductance should remain below 47nH. This is a loop of around one square cm. This should including the leads of the storage capacitors of course.

So the comparation with UCD700 is accurate

Not exactly. Following this reasoning is only that UcD700 will live even longer. All things equal, the design that stresses its parts least will have greatest longevity.

 

[Lars Clausen]

Hey Bruno

I've had difficulty finding full THD data on your website.

I had also difficulty finding the very same data on your website

We make THD measurements, and an AP system 2 is always just a phone call away. However as you probably know my opinions about THD as a measure of sound quality, i would be reluctant to use THD measurements to support my product.

Standard electrolytic caps are spec'ed to 2000 hours of lifespan at the specified ripple current. The ones we use from Vishay are spec'ed to 7000 Hours.

I think the comparison UCD700 / ZAPpulse 2.3SE is accurate enough. Anyway our warrenty extends to the use of ZAPpulse for outputting 700 Watts RMS in 4 Ohms. So........

 

[ghemink]

BTW: UCD400 is not the closest match for ZAPpulse 2.3SE. As you might have noticed ZAPpulse 2.3SE is spec'ed to +/- 80V and produce 700 Watts RMS in 4 Ohms. (Like UcD 700). So it would be more correct to match:

UCD700 and ZAPpulse

Why this focus on a few watts more or less when comparing these amps??? This is for sure not the only important performance parameter for all users.

Many users built active systems with these modules. For a tweeter, an 180W amp is more than enough to blow it away. In an active system, probably 180W is more than enough for all drivers as each amp has to cover only a smal part of the freq response. It is not only power that counts. As it is not only THD that determines the sound of an amp.

In my case where I use the amps in an active setup where I use many amps, low noise and EMI/RFI are for more important parameters as the drivers are connected directly to the amps without any attenuation from passive x-over networks. The UcD's were extremely good at that (low noise and EMI).

Best regards

Gertjan

 

[Bruno Putzeys]

I had also difficulty finding the very same data on your website

Right. We've got some plots floating around in several documents on the web site but it's not awfully well structured. Our web master has been alerted.

However as you probably know my opinions about THD as a measure of sound quality, i would be reluctant to use THD measurements to support my product.

THD plots aren't a panacea but they do tell a lot. After the audio community first attached too much weight to THD, the current state (where measurements don't count) is equally incorrect.

Standard electrolytic caps are spec'ed to 2000 hours of lifespan at the specified ripple current. The ones we use from Vishay are spec'ed to 7000 Hours.

Lifetime specs are always given for a specified ambient temperature and a specified ripple current. From this 1 data point, the user should calculate life time based on the temperature and ripple current in his application. What you'll often find is that the cap manufs massage the numbers a bit by specifying lifetime at a lower ripple current for "long-life" types. This is not to say that long-life caps aren't different from standard ones, but the difference is quite small. Using derating is usually a cheaper way of getting long lifetimes from caps. For professional applications, 20000 hours is really an absolute minimum, but with derating standard caps will cut this.

 

[Lars Clausen]

Gertjan: Of course i agree with you on this. You can compare products in many ways, dependent on your needs and situation.

About EMI / RFI we made huge reductions 2 years ago when we introduced ver 2.1. Of course improving step by step since then. And we are at comparable levels with the other Class D amplifiers on the market, not only from Bruno and Jan-Peter, but in general.

 

[Bruno Putzeys]

Edit: The post to which this one is a reply seems to have disappeared in the meantime.

Can you show us how you made this calculation?

Following arrhenius' law and presuming that most suppliers spec their rated ripple current to produce an internal temperature rise of 10ºC produces the most conservative estimates.

Sounds very impressing. What legal limit are you using for this?

In absence of a proper anechoic chamber and all that, I do a pre-test using a current clamp. It has been shown that the following test is a good indication of compliance:
Wire the module using sufficiently long (3m or so) wires to the load and source. If you're testing the module on its own using a lab PSU, have a decently long cable between the PSU and the module (otherwise you test the mains cable). Set the spectrum analyser to peak hold mode and run an RF current clamp along the full length of all cables.
Spectrum analyser settings are:
150kHz to 30MHz: 10kHz RBW, limit=30uA.
30MHz to 1GHz: 100kHz RBW, limit=3uA.
Detection= peak.

This test is conservative in that, because it measures conducted emission (instead of radiated) and because it uses a flat limit(the standard is more complicated), the limit is set quite low. For most frequencies that means this pre-test is 10dB too stringent.

On the 400W module, the plot only shows the analyser noise floor and radio transmitters. In a faraday cage and using a low-noise preamp for the current probe, you can see a small hump at 0.6uA around 150MHz but the rest is noise.
The 180W module is about 6dB worse around the same frequency.

 

[Lars Clausen]

The data sheet "ripple current" rating of elcaps is usually based on an internal temperature rise of 10ºC. While this does in no way preclude operation at higher ripple currents if traded for lower ambient temperatures, 40ºC is really high and liable to create even higher localised heating. It is precisely because of this that manufacturers are reluctant to specify operation at higher internal temperature rises - it becomes difficult to guarantee that all elcaps will have the same lifetime.

Well not if you are using high grade quality caps like the Vishay / BC Components 136 RVI series.

Look at the datasheet page 171: http://www.vishay.com/docs/28321/136rvi.pdf

You will see nice linear curves for higher ripple currents. In our case it shows that for the 2.7A RMS you have a 1.5 - 2 times life expectancy (meaning 10500 - 14000 Hours) from 43 - 48 C ambient temperature. And as we all know this is only when playing sine waves. With music, the average ripple current is about 1/3 of that, increasing the life expectancy about 3 times the 10500. (If the juices have not naturally evaporated before this time anyway )

I don't think there is any issue here!

In absence of a proper anechoic chamber and all that, I do a pre-test using a current clamp. It has been shown that the following test is a good indication of compliance:

Ohh ok, this kind of measurements.... Yes, well me too i found most noise emission from the ZAPpulse was right in the same area as FM and lower TV bands, making it difficult to see on the spectrum analyzer when you turn the amplifier on and off

Our emissions at 10 cm distance from the module were comparable to those of a normal harddisk drive.

 

[Bruno Putzeys]

Well not if you are using high grade quality caps like the Vishay / BC Components 136 RVI series.

I am using those caps in some of our OEM products.

You will see nice linear curves for higher ripple currents. In our case it shows that for the 2.7A RMS you have a 1.5 - 2 times life expectancy (meaning 10500 - 14000 Hours) from 43 - 48 C ambient temperature.

Glad you discovered that a blunt "7000 hours" spec is quite a simplification of matters.

I don't think there is any issue here!

Nah, not for normal use. It's only a problem when you're designing for really tough use. And even then. I prefer to err on the safe side.

Yes, well me too i found most noise emission from the ZAPpulse was right in the same area as FM and lower TV bands, making it difficult to see on the spectrum analyzer when you turn the amplifier on and off

Well, if you see *any* difference at all in this band outside a faraday cage, an FM tuner will, too.

Our emissions at 10 cm distance from the module were comparable to those of a normal harddisk drive.

That does not sound like a recommendation. Hard disks are notoriously problematic in agency tests.

 

[IVX]

Where i can see THD/freq plot of the ZAP (20-20khz, or in the rated frequency range)?

seems ZAPpulse haven't chance for the flat THD/freq plot above 5khz (in the best case even -lm6172 real pole around 5khz), isn't it?

BTW, what about IMD figures for the both, UcD & Zap amps?

 

[Pierre]

I don't want to offend anyone but... don't you guys think this forum is becoming an endless discussion between LC Audio and Hypex?
I think that's not the purpose of this site, but to offer and take ideas.
I am not meaning that a slight advertisement of both product lines (as long as they offer some help and advice) is harmful, but I think this is too much, are you with me?
Lars, Jan Peter, Bruno: don't discredit yourselves this way because this can only cause harm to both parts and of course no benefit to us, users of the forum.

Best regards,
Pierre