Triangular waveform

[dead rat]

Hey homies

Im building a class d amplifier, and i need to generate a triangular wave for the PWM/comparator circuit.

Can anyone recommend an excellent IC? ...or a circuit?
A low THD is very important

Im considering the XR2206

Cheers
Me

 

[tiki]

Hello,
the THD of a triangle generator cannot be very low.
look here, it's based on a proposal of the good JohnW.

 

[dead rat]

But then how are people supposed to achieve good results with class d amplifiers?

Is it because of the high frequency of the triangular waveform?

 

[tiki]

The linearity(!) should be excellent.

 

[bwaslo]

But then how are people supposed to achieve good results with class d amplifiers?

The better ones are self-oscillating and don't use triangle waves. See, for instance, IcePower or UcD.

Check the "Paper 6690, Presented at the AES 120th convention, Paris 2006" in the Applications section at http://www.hypex.nl/

 

[dead rat]

Improving triangular waveform

hellooooo

I have made a triangular wave using the XR2206.

It works great at 1 kHz.

But when i increase the frequency to 400-500 kHz the triangular wave flattenes out at the top,
this is really not helping my PWM signal.

So first, can i fix this and how?

Recommended replacement waveform generator?

 

[Corax]

Well, if you don't mind have a look in various application/design notes from i.e. Linear Technology, Analog Devices, Texas Instruments, ...
In particular check the opamp and comparator sites. At least Linear Technology has some "Circuit Collections" with various stuff in it. The chances are good to find something.

I know there was one particular schematic with an ultra-precise triangle generator going all the way up to 1 or even 2 MHz. But unfortunately I cannot remember from whom it was. 🙁

Use the search function on there homepages and you should come up with a few or maybe a bunch of example schematics. 😀
If I find some time I'll look out for it too.

 

[ChocoHolic]

The better ones are self-oscillating and don't use triangle waves. See, for instance, IcePower or UcD.
[/URL]

I definitely like Bruno and his designs.
But saying that self oscillating is generally better.. - that's sounding like you never tried both in design and also never tried both in listening.
Couldn't it be that self oscillating can give equivalent good results at lower costs? 😉
And how about the issues with beating in multi chanel systems. Already in stereo systems this effect is existing and if you then add the fact that self oscillating amps change the switching frequency depending on the music signal, then you unavoidingly get a beating interference, which is depending on the music signal.

You can extract this effect for demonstration with the following set up.
Two signal generators , fast stereo amplifier, two piezo tweeters (please put 10 Ohms in series to each piezo tweeter, in order to avoid trouble with the fast amp).
Connect one side to signal generator 1 and the side to signal generator 2. Now adjust gen1 to 49kHz and gen2 to 51kHz. You will hear a 2kHz sound.
Now start modulating the frequency of gen2 between 47kHz and 51kHz (manually is fine).
==> You might feel like sitting in a park and listening to the birds.

Fortunately this effect is not loud at all and the higher you choose the frequencies, the less you might notice it. But it is hard to avoid in self oscillating designs and it is nothing else than what thousands of teachers tell to 16 year old pupils.

Both approaches have their advantages and disadvantages.
I.e. post filter feedback ist definitely an advantage of self osciallting designs.
But this does not mean that post filter feedback is impossible in clocked designs. In clocked designs it is more difficult to implement and is bound to more narrow limitations.
I.e. lock up effects is definitely a struggle that you will not have during designing a clocked amp, but it might 'sweeten' you one or the other evening during designing self oscillating amps.

 

[fredos]

I have already implement clocked design with post filter feedback, take a look at my old 1200 amplifier. I agree with chocoHolic, beating is alway there. The stereo amplifier are good exemple, but imagine a pro audio rig multi way, multi amp of self oscilating stuff...I wont be there!

In my opinion, the best way to get clean triangle is to use a simple RC network to get triangle wave, as close as possible to triangle input of comparator, BUT you should design it to be in the linear region of the RC network...Again look at my 1200 schematics! Just dont forget to drive the RC with low impedance source!

 

[bwaslo]

My tweeters don't respond at all at 300kHz+ and I doubt that yours do either!

So mixing would have to occur (maybe) in power supply rails. Has anyone heard or measured this happening with UcD or ICEpower? Bridged UcDs should be doing this problem all the time, with the two amps involved running in opposite phase and sweeping past each other in frequency constantly. ICEpowers often run several off a common supply. Has anyone heard birds chirping in their speakers when running multiple self-oscillating Class D? Anyone report listening to a triangle wave derived Class D and thought it sounded better than ICEpowers or UcD?

Nope, I haven't heard every amplifier in the workd, and who knows maybe there's magic in some triangle wave fixed frequency derived Class D some where. No reason to ignore that technique, if some advantage can be found. I haven't heard every Victrola cylinder phonograph, so I can't say that CDs generally sound better than those either...

 

[Anthony C Smith]

A simple high quality triangle generator can be made with a crystal oscillator (11.2896MHz crystal and 74HCT4060 oscillator divider) feeding into an integrator (fet opamp, +in grounded 1M and 47p in parallel from out to in- and 6K8 from oscillator to in -)
depending on the rail-rail operation of the opam the 6k8 be trimmed to get maximum swing with lowest top distortion.
regards

 

[Pafi]

Has anyone heard birds chirping in their speakers when running multiple self-oscillating Class D?

Yes, I have, at about -80 dB, but only as long as the onboard supply decoupling capacitors were 1 uF, and sw. freq was only 200 kHz, and 2 channels were on the same PCB, and supply rails were directly paralelled.

Official UcD must be free from this problem.

 

[fredos]

I heard a lot of Icepower multi channel amplifier beating. Most of the time people did'nt notice, they think it was the noise floor or some white noise...To be honnest, I never hear more than 2 channel of ice power that didnt do beating..Maybe you just didnt realize that is that.

Fredos

 

[classdphile]

I heard a lot of Icepower multi channel amplifier beating. Most of the time people did'nt notice, they think it was the noise floor or some white noise...To be honnest, I never hear more than 2 channel of ice power that didnt do beating..Maybe you just didnt realize that is that.

Fredos

As with many class d caveats you have to know to listen for it, it can sneak by at lower levels of intrusion and particularly for certain types of music, or while using lower efficiency speakers.

I likely coined the term "birds chirping" as it's quite descriptive over beat frequencies, whistling or heterodyning. Non the less it's largely a question of EMI, and I think if you can't handle a simpler, more robust self oscillating amp without it occuring at problematic levels, a similar clocked based amp you'd handle no better.

Synching them doensn't alleviate EMI. You have to fix it at the root of the cause and not through trickery as synched clocks that might only half work.

Higher levels of EMI will also show up in ways that might make you think sounds good but really doesn't if you know what you're hearing and how to listen. IcePower is not the most advanced or best executed example of self oscillating technology. I could name others that also completely don't pass the grade, far worse than IcePower, or I could name a few that do. UcD is of course one worth mentioning that has no such issues, and it's both by design and implementation.

In terms of free lunch you don't get the same level of performance with a clocked amp as you can with self oscillating without resorting to additional complexity that can amount to reduced reliability, and more board space to pollute your control signals with, but I guess it's OK for PA 🙂

You've still got jitter with your clock, any induced noise on the line or circuitry that powers it will add to it. Let's not pretend that the best measuring and sounding class d amps in the world are all self oscillating by design, some of which are in fact used in multi channel home theatre or cinema setups and go beyond UcD.

Putting down proven technology that's on the cutting edge is a poor business model if you ignore the facts, but fear it if you must.

 

[ChocoHolic]

My tweeters don't respond at all at 300kHz+ and I doubt that yours do either!

I would say, if your tweeters roll off at 16 kHz, then they might respond at 300kHz with approx. -50db.
And this -50db are referencing to the HF-ripple of the amp output, not to the max power output.
That's why I say, this effect is not loud.
Also our personal perception regarding such effects seems to be very individual.
For some people, this is making magic.
For others it is messing up the entire sound stage.
I was listening to UCD (UCD400 if I remember correctly) only once and we compared it to a shameless priced Mark Levinson and to my both class AB Rookies and one R&D prototype of a friend. The speakers were ESL.
My Ranking:
1. Both Rookies and the other R&D proto ( I feel quite guilty... subjective ignorant idiot I am... )
The Rookies and the other R&D proto all had their own signature, but for my ears they were playing on the same level.
2. UcD. For my ears the UcD sound did NOT appear as obviously birding. Even if I was aware of the theoretical backgroung before the listening test. For me the UcD was not overwhelming magic, but basically OK and not tiring.
3. Mark Levinson. Some call it neutral. And if this is really correct, then I do not like correct neutral amps.

 

[dead rat]

A simple high quality triangle generator can be made with a crystal oscillator (11.2896MHz crystal and 74HCT4060 oscillator divider) feeding into an integrator (fet opamp, +in grounded 1M and 47p in parallel from out to in- and 6K8 from oscillator to in -)

Why a fet opamp?

 

[fredos]

As with many class d caveats you have to know to listen for it, it can sneak by at lower levels of intrusion and particularly for certain types of music, or while using lower efficiency speakers.

I likely coined the term "birds chirping" as it's quite descriptive over beat frequencies, whistling or heterodyning. Non the less it's largely a question of EMI, and I think if you can't handle a simpler, more robust self oscillating amp without it occuring at problematic levels, a similar clocked based amp you'd handle no better.

Synching them doensn't alleviate EMI. You have to fix it at the root of the cause and not through trickery as synched clocks that might only half work.

Higher levels of EMI will also show up in ways that might make you think sounds good but really doesn't if you know what you're hearing and how to listen. IcePower is not the most advanced or best executed example of self oscillating technology. I could name others that also completely don't pass the grade, far worse than IcePower, or I could name a few that do. UcD is of course one worth mentioning that has no such issues, and it's both by design and implementation.

In terms of free lunch you don't get the same level of performance with a clocked amp as you can with self oscillating without resorting to additional complexity that can amount to reduced reliability, and more board space to pollute your control signals with, but I guess it's OK for PA 🙂

You've still got jitter with your clock, any induced noise on the line or circuitry that powers it will add to it. Let's not pretend that the best measuring and sounding class d amps in the world are all self oscillating by design, some of which are in fact used in multi channel home theatre or cinema setups and go beyond UcD.

Putting down proven technology that's on the cutting edge is a poor business model if you ignore the facts, but fear it if you must.

I did'nt say thatself oscilating are bad...I just a matter of thinking...I dont like ''uncontroled'' design. If a parts change value (you heard about tolerance of parts?), switching performance\frequency can change...In a clocked design, with proper feedback this should not happen. I agree that clocked design need more parts, but anyway, in pro audio you need lot more for stupid guy. Exemple: short protection, thermal proteck, clipping limiter, under\over voltage protection, turn-on\turn-off muting, HF oscilation protection for drivers, DC protection, etc...

Other hand, you can optomize dead time, implement variable dead time in function of power to increase effiency, use zobel for power limiting, etc...All this need predictable performance, not easily achievable with self oscilating...

EMI by the way cause same probleme in clocked and self oscilating design...So dont say that that's the cause of beating of self osc amp...A badly designed clocked design will do same if EMI are not properly managed. Just use a 2'' compression drivers with 110db on any self osc and you will know what I talk about...

Fredos

 

[classdphile]

Hi,

So you're saying you prefer clocked as you fear the uncontrolled aspect of controlled chaos? So long as we agree that doesn't make it inferior.

You had stated your preferred method to form the clock was with a simple RC. Perhaps you can explain to me how that's immune to this phenomena of tolerancing and stability you speak of, I'm pretty sure it isn't.

In self oscillating the frequency may change wrt to such things, but it's as an intrinsic corrective measure since they're already a part of the control loop. In a simple clocked case a global loop ignores it, hence a requirement for additional measures.

The protections mentioned can be seen as transparent to the actual methodology for the most part, falling outside of the control structure. They're used because things aren't always predictable either. I don't view that as being any more involved for one variety than the other.

Other hand, you can optomize dead time, implement variable dead time in function of power to increase effiency, use zobel for power limiting, etc...All this need predictable performance, not easily achievable with self oscilating...

Much of that too can be seen as transparent to methodology, mostly, they switch the same in either case. Variable deadtime is better sensed and not predicted per se. The use of a zobel is something of a bandaid with drawbacks of its own, and some amps don't need it. In fact it's proven itself a common part to burn up and may require additional protection just for itself, you know, for mr. stupid of which there's no shortage.

EMI by the way cause same probleme in clocked and self oscilating design...So dont say that that's the cause of beating of self osc amp...A badly designed clocked design will do same if EMI are not properly managed. Just use a 2'' compression drivers with 110db on any self osc and you will know what I talk about...

EMI wreaks havoc in a variety of ways. I am saying heterodyning is caused by unoptimal EMI since even non fixed frequency self oscillating amps that don't suffer heterodyning solved it through optimal EMI alone.

Try running multiple clocked amps off the same supply and don't synchronize anything. If you hear the birds it's because EMI could be optimized further. Synchronizing them is a common trick to make it robust against less than perfect EMI performance in some respect or other. If lowest EMI is always superior, which amp is then best, one that relies on synchronizing, or one that gets away without 99% of the time?

I think your amp remains one of the best examples of a clocked amp around here anyway.

 

[Anthony C Smith]

why fet opamp?

Fet op amps have a much higher input impedance and from experience give much less distortion than the 5532 bipolar types
part I use in this case is TS272 from SGS
http://www.st.com/stonline/products/literature/ds/2303/ts272.pdf

 

[fredos]

They are more sensitive to noise, EMI and glitch too...Just dont forget to drive them with low impedance source...