kgrlee and bcarso,
I am so on-board with both of your understanding of the need to keep out of the clipping stage of a power amplifier when listening at high spl levels. I have been working on long excursion small cone loudspeakers that though they are very low distortion the power requirements are very high. I think the last time I checked the efficiency was in the 84 to 86db range, definitely not in the high 90's for a pro-audio large format driver or a compression driver territory. So power output from the amplifier to get up to let's say 110db spl is going to take some real power to say the least. Definitely not happening with a 50 watt @ 8 ohm power amp! Steady state average power is not the problem but the peaks for something like a kick drum or a piano can tax any low power amplifier. It is those peak moments that make all the difference in the sound, as soon as you hit those limits all that hard work is for not without the instantaneous power to carry the load. Clipping is not your friend in those instances and I surely wouldn't be happy if that was the case up to 30% of the time under normal conditions.
I am so on-board with both of your understanding of the need to keep out of the clipping stage of a power amplifier when listening at high spl levels. I have been working on long excursion small cone loudspeakers that though they are very low distortion the power requirements are very high. I think the last time I checked the efficiency was in the 84 to 86db range, definitely not in the high 90's for a pro-audio large format driver or a compression driver territory. So power output from the amplifier to get up to let's say 110db spl is going to take some real power to say the least. Definitely not happening with a 50 watt @ 8 ohm power amp! Steady state average power is not the problem but the peaks for something like a kick drum or a piano can tax any low power amplifier. It is those peak moments that make all the difference in the sound, as soon as you hit those limits all that hard work is for not without the instantaneous power to carry the load. Clipping is not your friend in those instances and I surely wouldn't be happy if that was the case up to 30% of the time under normal conditions.
A couple of comments.
I like the base current recapture a la Aldridge for Q20 and Q22. This works fairly well since the impedances at the emitters of Q19 and Q21 are a good deal less than 10 ohms. Since Q20 and Q22 bear the brunt of the dissipation shifts with signal swing and consequent alpha modulation, this will be very helpful, and as well their variable output capacitances' displacement currents are also pumped back, mostly eliminating that as a distortion mechanism.
One concern is bias stability. The BC parts are fairly good-sized (625mW for 25C ambient) and hence have a fairly small delta Vbe with dissipation-induced temperature change, maybe about 440uV/mW. But that's going to affect quiescent currents for at least reasonable ambient and average self-heating temperature shifts. To some extent the output voltage won't shift much since the circuit has decent symmetry, but there are cumulative changes in the quiescent currents that get pretty large. For the assumption of 25C junction temperatures for example, the dissipation of Q20 is about 280mW. If however all four BC parts have a Vbe change of |50mV| as they heat up (associated with about a 23C increase for each over ambient), Q20 dissipation goes to about 717mW.
With the relatively long thermal time constants involved, the thermal distortion due to Q15 and Q16 should be minor, except maybe at lowest frequencies, but I can't very easily model that at this point.
Otherwise, given a good output stage, it looks promising.
kgrlee
With respect forr, if you don't think the overload performance of an amplifier is an important part of its sound ...
[...]
I can't let forr get away with claiming overload performance isn't important.
The overload behaviour of an amp is unimportant because it should never happen.
As suggested, adjustable "klippers" ahead of the power unit are the solution if well controlled clipping is really wanted.
The notion of FSD of digital formats allows to correctly handle the highest voltages at the input of an amplifier (using an attenuator) such as its output voltage limits are never reached.
If an amplifier clips in an audio system, it means that its power specification is insufficient in the intended application, it needs to be changed for a more powerful one.
I open a new thread on the demands on power amps as this is a bit out of topic of CFAs :
http://www.diyaudio.com/forums/soli...s-dynamics-sound-levels-home.html#post3620198
With respect forr, if you don't think the overload performance of an amplifier is an important part of its sound ...
[...]
I can't let forr get away with claiming overload performance isn't important.
The overload behaviour of an amp is unimportant because it should never happen.
As suggested, adjustable "klippers" ahead of the power unit are the solution if well controlled clipping is really wanted.
The notion of FSD of digital formats allows to correctly handle the highest voltages at the input of an amplifier (using an attenuator) such as its output voltage limits are never reached.
If an amplifier clips in an audio system, it means that its power specification is insufficient in the intended application, it needs to be changed for a more powerful one.
I open a new thread on the demands on power amps as this is a bit out of topic of CFAs :
http://www.diyaudio.com/forums/soli...s-dynamics-sound-levels-home.html#post3620198
One idea I have is to use the BC847BPN device for the input pair. Seems to have good specs at any rate. Being in the same package would help thermal stability (though I would add a cascode). Anyone have any experience with this part?
Last edited:
Easy on the burgundy ol chap.
😀No, audiolab is a totally different company, audiolabor was a german company from the 1980 s which only produced CFA amps. Some interesting designs were seen from them. The mainstay was the topolgy seen here on the forum called VSSA or SSA or whatever.
PM your email, Ill send some schematics of these and others where this topology is used. Some from amps from 70s and some from a PA company.
Too bad I cant email a bottle or 2 burgandy .....
That's a Hawksford VAS you are showing there and if set up correctly, the linearity is exceedingly good.
You might want to consider using a mirror for the input stage level shifter, but that will necessitate some sort of balancing circuit however.
I have a suspicion that anything other than a mirror load for the front end level shifter introduces excess phase shift (mirrors are FAST and suffer little phase shift) and the whole thing degenerates (unfortunate choice of word here) into a VFA since we still have to meet the stability criteria and that means placing a pole down at a few kHz. Output stage phase shifts remain the biggest challenge, but We tackle that with TPC.
So, my suggestion is, we need to focus designing a very linear, wide band width mirror if we are to retain the essential features of the CFA.
Last edited:
I agree, mostly. However, hard clipping is also associated with power amplifier's appetite for (self) destruction. You really don't want an accidental overload (think cable handling) to blow the output stage and/or the tweeters. That's also why I think an over current protection is essential in a high power audio amp. Protections that don't interfere with the signal are not trivial to design, though.
Good mirrors are indeed essential. And the fastest ones tend to be quite noisy, being the primary noise contribution to operational transconductance amplifiers like the LM13700. Ballasting in the emitters of ones like the Wilson slows things down while improving the noise performance, although Groner has mentioned that he's had some success with bypassing them with capacitors for speedup while shifting the noise increase to out-of-band.
Hawksford analyzed and popularized for the audio community what was actually discovered by Aldridge in about 1963, it turns out. He called it a cascade circuit. I stumbled on a reference to it the same time I discovered that Frank Boxall invented (in ~1956!) another base-current-recovery one that gets primarily credited to Baxandall. Of course I do not accuse either Hawksford or Baxandall of plagiarism here, particularly as it's very unlikely that they saw the work, given that the apparent prior inventors' applications were to common-base amplifiers for telephony. More recently Larson, it turns out, (also surely no plagiarist!) patented the Boxall, applying for this a bit ahead of the Baxandall/Shallow article. John Addis told me that at Tektronix, Larson and others referred to it as the super-alpha. Boxall called it a "feedback compound transistor" and even enumerated some triple-device topologies. I've found them to be essentially hopeless to compensate.
Thanks for that info Brad.
I have not used this configuration in a practical amp - only sims - but I've heard about the stability issues.
I have not used this configuration in a practical amp - only sims - but I've heard about the stability issues.
I've not used it as yet, but bear in mind that the quiescent bias stability issue will not depend on input device matching or close thermal coupling. It's an effect arising from the resistive feedback dividers, which we suppose don't change much with temperature, and the magnitude of the device Vbe's which do. And then things get compounded further on, with the 300 ohm loads not changing but again the Vbe magnitudes doing so. And these two effects are not compensatory, but cumulative.
In the case of the input stage, at 25 C the model I have shows that most of the 10mA from the current generators goes to the 100 ohm resistors. As the Vbe magnitudes get smaller or larger the amount that goes to a given emitter changes. Although not that much, the current change produces a significant voltage change across ~300 ohms.
Bootstrapped cascoding for each input device will reduce thermal distortion, although there will be some negative input Z effects, usually easy to compensate with a little lumped C to common there. For ambient temperature changes one could consider making the 10mA gens themselves temperature-dependent.
Many are the ways.
You mean, apart from me, of course😉
JBL 2226G 100 dB/2.83 V @ 1 m. But this is a bit of a cheat since they are 4 ohm nominal, so equivalent to 97 dB for the 8 ohm version.
I think John Atkinson tested 260 Hi-Fi speakers and found only 2 at 96 dB or above, both professional monitors.
Now back to CFA.
Best wishes
David
Thank you Mr. Zan. My involvement in this CFA saga was cos your requirement. I think 50W 8R with good performance into 4R might be big enough 😀
I wanted to use a little known advantage of CFAs .. that they can be designed to have much lower input noise than an equivalent VFA 🙂
Very sensitive speakers will hiss at you even with nothing playing.
forr, I'd very much like to know your 96dB/2.83V @ 1m speakers.
For those not speaker people, there is a limit to the size vs efficiency vs bass response of a speaker. If you do it right, you might approach this limit. These days, many TS pseudo gurus don't realise this and fall far short.
http://www.aes.org/e-lib/browse.cfm?elib=2485
A loud speaker with good bass needs to be BIG. A small speaker may have good bass .. but then it needs the 1000W amp 😱
The room is not small. I sit between 3 and 4 metres from the speakers so 11 dB down from 1 m sensitivity.
That make 86 dB / W. So 106 dB peak means 100 W into 4 ohms. Spot on.
I plan 600 W so I should have plenty of headroom.
Yes, I know all about that! That was the start of the whole search for a better amp. I suspect "CFA" is not much lower noise. For a powerful amp it becomes limited by how much power one is prepared to burn in the feedback network to lower the feedback resistor. This is identical in both "CFA" and VFA. The rest of the circuit contribution in a VFA can be kept to about the same level so even if the rest of the "CFA" was silent it would only be about 3 dB better.
I have started to consider a VFA with inductor bypassed LTP, a la JE 990 discrete op-amp.
I haven't done detailed noise simulations yet but I expect this and a "CFA" to be quite similar in practice.
You looked at this?
Yes. About 16 cubic feet for each sub woofer + woofer box. JBL 2245 + 2226G. Plus horns on top.
The whole unit lines up with the tops of the doors.
Best wishes
David
Last edited:
As to be expected, if you consider parasitic capacitors from the output node, raising the input current has a big impact on the bandwidth. If scaled for 5mA and a suitable load resistor is used, the bandwidth extends out to 3 or 4 MHz. Practical considerations will mitigate against this, but this is far wider than you can get out of a classic VAS/TIS. The load resistor, within reason, defines the gain.
Now, if you set the OLG correctly, the output stage pole will fall above the ULGF, and that means you can get away with very wide bandwidths and high slew rates. The trade off of course is lower loop gain and higher distortion - but that is not the b all and end all of a good design.
This type of structure is therefore well suited to class A output stages, where the distortion is inherently lower, or HEC.
Further, since you could use the TIS load resistor to define the loop gain, the adjustment range could indeed be made from 0dB through to 50 or 60 dB so you could in theory go from ZGNF to High global feedback by simply adjusting a potentiometer.
Last edited:
Hmm then then the unavoidable question arises.. what would be a good opamp to drive 100R input impedance at low signal levels? (2Vpp max)? Too bad I don't have a model for the AD826, but from using it in Multisim, it did well for the function.
But, from what you derive the 1K feedback resistance? In unity gain the output would go straight to the - input without a feedback resistor.
Edit: I tried the LT1152 audio opamp but then I need to parallel at least 3 of them to keep the distortion lower than that of the amp itself.
Also, it wouldn't matter much if I increase the amp feedback network by 3.3 to get an input resistance of 330R.
Last edited:
It's a cinema system, rather than a home cinema system.
Also similar to some of the classic studio monitors, except improved of course😉
Professional components are not cheap, but still less expensive than many frou frou "boutique" systems where the price seems to be excessive just to make them seem "special".
Here the headroom should mean that power compression and excursion limits for the speakers and slew rate and clip limits for the amp, are so far away that they will never be even remotely of concern. I just need to keep the noise equally inaudible. Hence my interest in this thread.
Best wishes
David.
Why, yes, my partner does own her own home a mile away, why do you ask?😉