You only have put down comments, you never comment on the calculation. Those are solid numbers and if you use one output pairs, you have a problem.
It's actually to do with long voice coils jumping out of the gap.
Your short voice coil / long gap avoids this EVIL
Kindhornman, I'm still sorting out my junk from the trip eg visit laundromat. I need to go to your DSP speaker thread but probably not till tomorrow.
_____________________
Bob, I've posted my suggestions for your 2nd ed. before but I hope you won't be offended if I summarize them here.
- label sims & measurements clearly
- list DBLTs that HAVE been carried out on amps and their implications for design. Jan_Didden has been involved in some this Millenium and they confirm stuff that I was involved with in the previous century. WARNING : results are heretical to both audiophools & objectivists
- do a bit more on CFAs then damn them with faint praise
Last edited:
Kgrlee,
It seems my thread is dying of a thousand cuts. I didn't want to make something that would need to cost enormous dollars and that seems to be a requirement with some on this site otherwise you can't be serious. I'll send you a PM and see if I can interest you in something beyond just picking your brain. I have made a request of Brad Wood, he seems to be interested, a team is always better than an individual.
Now go clean those damp clothes from you trip.
It seems my thread is dying of a thousand cuts. I didn't want to make something that would need to cost enormous dollars and that seems to be a requirement with some on this site otherwise you can't be serious. I'll send you a PM and see if I can interest you in something beyond just picking your brain. I have made a request of Brad Wood, he seems to be interested, a team is always better than an individual.
Now go clean those damp clothes from you trip.
No such luck, I just ordered the resistors and some other stuffs, they charged me sales tax. But the good thing is is they take Paypal, I don't have to type my credit card like when I order from Digikey and newark.
I have you to thank for, introducing to the resistor and ordered the first time from Mouser.
Now, we'll see how fast I receive them. Digikey is FAST, never over 3 days.
Hi David,
It's all over for the driver when that happens. But the end of the VC is usually bashed over from hitting the back plate first. Woofers normally cost more than the parts to fix an output stage. I would also expect some shorts in the coil, burning for certain once the coil leaves the big heat sink we call a magnet structure.
At this point some protection should be activated as the current is way high now. How high? No idea, but it's only of academic interest as you view the destruction. I would expect either the output stage could handle this load, or protection has acted to save the rest of the world. If not, it is the fault of the amp designer that the amplifier failed.
-Chris
It's all over for the driver when that happens. But the end of the VC is usually bashed over from hitting the back plate first. Woofers normally cost more than the parts to fix an output stage. I would also expect some shorts in the coil, burning for certain once the coil leaves the big heat sink we call a magnet structure.
At this point some protection should be activated as the current is way high now. How high? No idea, but it's only of academic interest as you view the destruction. I would expect either the output stage could handle this load, or protection has acted to save the rest of the world. If not, it is the fault of the amp designer that the amplifier failed.
-Chris
It's actually to do with long voice coils jumping out of the gap.
Your short voice coil / long gap avoids this EVIL
Kindhornman, I'm still sorting out my junk from the trip eg visit laundromat. I need to go to your DSP speaker thread but probably not till tomorrow.
_____________________
Bob, I've posted my suggestions for your 2nd ed. before but I hope you won't be offended if I summarize them here.
I look forward to your publication
- label sims & measurements clearly
- list DBLTs that HAVE been carried out on amps and their implications for design. Jan_Didden has been involved in some this Millenium and they confirm stuff that I was involved with in the previous century. WARNING : results are heretical to both audiophools & objectivists
- do a bit more on CFAs then damn them with faint praise
I appreciate your input and am certainly not offended. However, I'll probably stay away from the audiopolitical double blind testing issue
.Cheers,
Bob
Presumably you have references for these tests ...
... rather than "I met this guy, Otala in a pub who told me the latest fashion in amps will be incredibly high currents demanded from EVIL speakers". [1]
In more than half my life as a true speaker guru, I have encountered only one real life situation where a speaker will demand more from the amp than what you'd expect from the small signal impedance curve.
It is a fault condition. When it happens, you will be glad your amp DOESN'T give 300% more than you'd expect.
They quote Otala.
Perhaps one day, someone will record some 'music' (??) which sorta resembles these completely unrepresentative waveforms. You may like to simulate them in your favourite DAW and decide if anything close is worth listening to.
[1] This happened to me at AES Hamburg 1981 when I tried to drink the good professor under the table .. an endeavour doomed to failure. His revelations 'under the influence' have coloured my views of his work ever since.
Yes. He used the word 'fashion'
Aside from the drinking spree, should we not make a distinction between a speaker driver proper and a system with an often complex xover system with lots of L's and C's? Isn't that where possible excessive load current may come from? I never measured it, so you may well be 100% correct,but still.
Jan
There's no reason for a crossover and speakers to behave differently from what their impedance predicts, unless a nonlinear phenomenon comes into play. For instance, and inductor saturating or a voicecoil going too far. As I understand when the voicecoil leaves the gap or hits something and halts motion, the back-EMF is suddenly removed and so there is a surge of current. But not more than you would expect from the resistance of the voicecoil unless there was an insulation failure.
I think electrical disconnections and amplifier antics are more likely to drive the voicecoil like a DC converter and maybe cause large momentary currents/voltages. If protection kicks in and causes the amp output impedance to go high, then it's like a flyback converter. So maybe those reverse protection diodes on the amp output really do matter sometimes.
Hi Alan,
I have also not seen a lot out there that applies to CFA audio amplifiers, except, of course the thread on this site. Unfortunately, that thread got so mangled at points it was sometimes difficult to follow. This was definitely not the fault of the person who started the thread. Sometimes it seemed like 10 blind people feeling of an elephant. I have never designed or built a CFA power amplifier, so I would really have to brush up on it.
Anyway, I have a lot of new material to put in the book and I'll just have to see how much I'll be able to say about CFAs. It will definitely not get a chapter of its own. I'm guessing it might be a section in the IPS/VAS chapter, since that is mainly where a CFA is different.
I also was not strongly convinced in the thread that the CFA architecture had any advantages at the end of the day, since the biggest limiter of amplifier performance is usually the output stage, which is the same for VFA and CFA.
I can get all the slew rate I want in a VFA if I use Miller Input Compensation (MIC). My little 50-watt MOSFET power amplifier with error correction had a 300 V/us slew rate and 0.0006% THD at 20kHz.
Cheers,
Bob
I would also be glad if you could / able to write a chapter with CFA
Do you also explain in your book how you get so fine data for a 50W amplifier?
As I wrote earlier my expirence with CFA is very mixed, first I saw them as a step forward (or backwards in time if you look at Hiragas designs) later when I built I always felt them to be slim and dry in the lower octaves, especially when compared to leach type VFA"s. I believe that the main reason is that the feedback in a CFA is NOT pure as you inject both Voltage and current into the feedback node. Or rightly you inject voltage that is then transformed to current in a somewhat. Compromised V to I conversion
In Selfs book he claims that the patent is held by Audio Partnership PLC which seems to be a division of Cambridge Audio.
Biasing a Class B output stage heavily into Class A suffers from gm-doubling distortion at the crossover points where the output stage runs out of bias current and reverts back into Class B. Self calls this type of biasing "Class AB" to differentiate it from what he refers to as "Class B" which is an optimally biased Class B/AB amplifier. His books go into great detail as to why an optimally biased output stage is better, i'd recommend reading them - the short answer is that it avoids gm-doubling effects as much as possible.
Class XD is not the same as a heavily biased AB stage. It avoids gm doubling effects as much as possible because it is based on an optimally biased Class B and allows the amplifier to run in true Class A for small signals since the crossover region is displaced away from 0v.
Last edited:
Hi Bob,
I have built my first CFA and simulated many. Main difference I have found between CFA and VFA is that is much easy to compensate CFA. In my built CFA 200W HEXFET I use combination of the TPC and OIC, with some effect of a shunt compesation. It is very stable and good sounding amp, with very good bass, not differnce with the VFA as some have found. I use my tablet to send this and dont know haw to copy link here. The thread is called 200W MOSFET CFA and the post with latest schematic is #799.
BR Damir
I would also be glad if you could / able to write a chapter with CFA
Do you also explain in your book how you get so fine data for a 50W amplifier?
I'll count your vote for CFA
.The full JAES paper for that amplifier is on my web site. It is titled "A MOSFET power Amplifier with Error Correction". There you will see the complete description. It uses an N-channel JFET input VAS that is cascoded. The load is a differential current source (like a current mirror, but with a differential output and common mode rejection). It uses a cascoded differential VAS with a Darlington cascode current mirror below. The IPS and VAS are driven from boosted power rails. The output stage uses vertical MOSFETs with Hawsford Error Correction (HEC). The IPS/VAS is covered in my book, as well as the HEC. Extensive measurements were made on the amplifier and shown in the paper.
Cheers,
Bob
I agree. However, some might argue that the gm-doubling that does occur is sufficiently (and symmetrically) displaced from the crossover that it may not be as audible at the somewhat higher volume levels when a large class A current is used. In any case, one should run THD or 19+20kHz CCIF as a function of power to see the effects of crossover distortion in both cases.
There is another effect that is ignored by many. That is thermal bias shifting in the output stage as the power level of the program material goes up and down. In many cases that precious 26mV or so is not maintained at all times during dynamic program material. If one chooses a large class a region bias, the temperature is always quite warm and thermal variations with program material are quite small. Doing this incurs the gm doubling and keeps the amplifier quite warm all of the time, so it is not a very "Green" solution and will increase your electric bill somewhat.
Cheers,
Bob
A picture speaks a thousand words:
From 'Audio power amplifier design handbook' fifth ed. "XD Const" is basically a Class-B amplifier with a constant current source hooked up to the output. "XD PP" is the addition that the CCS is voltage controlled by the output. I won't post any more than that because I don't want to infringe on Doug's copyright.
A picture speaks a thousand words:
From 'Audio power amplifier design handbook' fifth ed. "XD Const" is basically a Class-B amplifier with a constant current source hooked up to the output. "XD PP" is the addition that the CCS is voltage controlled by the output. I won't post any more than that because I don't want to infringe on Doug's copyright.
I read Self's XD patent quite some time ago and also noted his XD-PP version, where the current source is dependent upon output voltage. It would seem that the XD-PP approach would only be optimal into a particular value of resistive load. So if these curves were measured with a resistive 8 ohms load and the circuit was optimized for 8 ohms, this would be the best case. Not sure how much its advantage over straight XD degrades into, say, a 4 ohm load.
Since the use of a constant pull-down current for op amp outputs is as old as dirt, I don't think Doug would have gotten this patent without the XD-PP tweak.
I'll have to go into his book and see what the bias current was for the class AB curve in the figure.
Perhaps the best apples-apples comparison is where the class A total dissipation is the same as the class XD total dissipation. Next best is if the class AB bias current is the same as the class XD bias current. Note that the idle current drawn from the positive and negative rails is not equal.
Notice also that the NPN power transistor with the pull-down current runs much hotter than the PNP power transistor. Actually, at no-signal I believe that the PNP transistor carries no current. Of course the fact that they are on the same heatsink mitigates this a bit (but certainly not with respect to junction temperature rise and case temperature rise over heat sink temperature).
I think the XD-PP circuit would work better and be less dependent on load reasistance and phase angle if the pull-down current source were made dependent on output current. Output current can be measured with a very small resistor in series with the return leg of the loudspeaker. I don't think Self patented this approach.
Of course, if you are going to the trouble of generating an output current signal, you can put in a nice damping factor control that can even create a negative output impedance to increase the damping on some loudspeakers with sloppy bass, or cancel some of the speaker cable resistance in a long run of modest gauge wire.
There are many other neat tricks you can do if you have a measure of the output current, including power and impedance monitoring and some protection schemes. Some pro-sound amplifiers do this. Indeed, some feed the output voltage and current signals through A/D converters and into a DSP.
Cheers,
Bob