you could buy the book, but Distortion In Power Amplifiers is a start
"Blameless" means identifying, addressing specific weaknesses/distortion in a evolution of the Lin/Thompson diff pair-TIS/VAS-unity buffer power amp
"Blameless" means identifying, addressing specific weaknesses/distortion in a evolution of the Lin/Thompson diff pair-TIS/VAS-unity buffer power amp
You make a good point. That is why I advocate the use of the 19kHz + 20kHz CCIF IM test with spectral analysis. The IM products that lie below 20kHz tell pretty much the whole story. However, the percentages or dB numbers are not directly comparable with THD-20 numbers, so caution is required in quoting these numbers. This test was not as viable many years ago due to the expense and limited dynamic range of spectrum analyzers of the time. Modern digital/PC-based spectrum analysis based on FFT largely eliminates this issue.
This type of testing is essential for honest testing of class D amplifiers.
I applaud John Atkinson for showing these testing results in most of his amplifier reviews in Stereophile.
Cheers,
Bob
I like it when manufacturers are game to put it all on the line: my Perreaux 2150B, a classic version of this circuit, in the owner's manual claims:
"200 watts per channel continuous, both channels driven into 8 ohms 20Hz to 20,000Hz at no more than 0.009% THD from 0.25 watts to rated power"
The big weakness in this unit is the power supply implementation, the more you do to refine that area the better the subjective performance.
I think they're awesome, 2SK135/2SJ50I've had those running in an amp for years and so far are still my favourites to listen to. Biassed well enough they make for hours on end listening pleasure (ofcourse it doesn't all depend on the outputs, but they do make a definite impact).
That's good to hear, so to speak
.Cheers,
Bob
No. Amplifiers with a single LTP input stage, a common-emitter VAS, and class AB push-pull EF output stage have been around since the late 1960's. Take a look at the Harmon Kardon Citation 12. "Blameless" is a name given to logically evolved (straightforward good engineering) of this basic design by Doug Self. Much of this evolution occurred in IC op amps of the 1970s.
Cheers,
Bob
http://www.angelfire.com/sd/paulkemble/sound7g.html
This famous schematic in the Hitachi databook actually hides important secrets that are in the Hitachi commercial amp. The databook circuit, as shown, isn't unconditionally stable with load. Also the commercial circuit regulates the PS to the driver & earlier stages and there are other subtleties. Yes. The excellent performance IS possible with this simple circuit ... but ONLY if you add the secret tweaks ... but then the complete amp is no longer simple
What about 0.25 watts and below. That is where most listening will take place. The critical zone.
Mooly, I read somewhere that under normal (whatever that is) home conditions, approximately 95% of our listening is up to the 1W continuous power dissipation, with only the transients surpassing that.
This depends on how loud we like to listen, what type of music dominates, the acoustics of our rooms, and of coure, the effciency of our speakers. A 3 dB difference will mean half or double the power.
Your concern is what is, I believe, called distortion comeback. It happens when one has an unusually low bias amp, so low that it crosses from class A to class B operation at power levels of say 100 mW or even less, so that even very low room power levels are already in class B. This causes distortion to rise, but if memory serves, this is usually not too much of a problem (up from say 0.05% to say 0.08%).
Actually, I don't think it's too much of a problem these days in general, it's a well known and well understood problem.
This depends on how loud we like to listen, what type of music dominates, the acoustics of our rooms, and of coure, the effciency of our speakers. A 3 dB difference will mean half or double the power.
Your concern is what is, I believe, called distortion comeback. It happens when one has an unusually low bias amp, so low that it crosses from class A to class B operation at power levels of say 100 mW or even less, so that even very low room power levels are already in class B. This causes distortion to rise, but if memory serves, this is usually not too much of a problem (up from say 0.05% to say 0.08%).
Actually, I don't think it's too much of a problem these days in general, it's a well known and well understood problem.
That's true and Panos "how much voltage (power) do your speakers need" thread Thread highlighted that.
An underbiased amp will perform better at very low loading and voltage swing. It seems curious to specify it at 0.25 watts and above. As if something untoward is dominating the readings below that.
An underbiased amp will perform better at very low loading and voltage swing. It seems curious to specify it at 0.25 watts and above. As if something untoward is dominating the readings below that.
Thanks guys. It is a mystery. The output inductor is intriguing and maybe just maybe that is part of it. As Bob says no way should it work. I have measured it and it seems true. When me it was 50 kHz. Surly that is already valid. What I can't understand is why the VAS doesn't object the the gate capacitance.
The point that AKN makes I'm sure is the reason - the almost "perfect" Soulution 710 amplifier demonstrates this behaviour, Soulution 710 power amplifier Measurements | Stereophile.com, fig. 4. The key thing is that the slope of the distortion vs. power curve is straight, indicating that the crossover distortion is well in hand.
Also, for a conventional sensitivity speaker these performance figures mean that in-room THD(+N) is very close to 0dB, well below ambient sound.
Also, for a conventional sensitivity speaker these performance figures mean that in-room THD(+N) is very close to 0dB, well below ambient sound.
As distortion is so low, it is noise that dominates the THD+N vs. power plot.
To see harmonic components of the distortion, we need spectral analysis and probably notch filter to remove main harmonic, to improve resolution - speaking about lower power like 1W or 10W. I do not understand why John Atkinson does not show spectral analysis at lower power as well, I can see only a plot at 101W.
Last edited:
Is it possible that MOS FET's because they don't switch on abruptly never create the problems? It is startling to set up a MOS FET amp ( Hitachi devices ) with zero bias. That is about 3 mA standing current. The sound is like a not so good DAC. The sort most people have. By 20 mA the sound is more like a standard bipolar amp. At 100 mA there is a sound few bipolar amps have except some bass tightness is lost. I gave my ex employee James a Hitachi amp with a Allen key bias pot. He seems to set it for every piece of music. Fair enough. 40 mA should be OK. What is very nice is the bias voltage is so low. 1 to 1.2 V between devices seems enough to get the current moving. These are not industrial FET's if think more like 5 V. I suspect Exicon FET's ideal to replace Germanium for this reason. Some old amps are worth the trouble to do that. If you do get rid of the drivers.
I notice in some graphs Hitachi started at 2 watts. I never found the sub 1 watt to be a problem on the scope. Daft they should do that as it looks deadly suspicious.
I notice in some graphs Hitachi started at 2 watts. I never found the sub 1 watt to be a problem on the scope. Daft they should do that as it looks deadly suspicious.
http://www.angelfire.com/sd/paulkemble/sound7g.html
This famous schematic in the Hitachi databook actually hides important secrets that are in the Hitachi commercial amp. The databook circuit, as shown, isn't unconditionally stable with load. Also the commercial circuit regulates the PS to the driver & earlier stages and there are other subtleties. Yes. The excellent performance IS possible with this simple circuit ... but ONLY if you add the secret tweaks ... but then the complete amp is no longer simple
Is the amplifier in the link the commercial one? If not, do you have a schematic of the commercial one?
Cheers,
Bob
That's right.
Also when testing Hypex I had to build a 12 pole filter as my analyzer couldn't give me useful information. The 12 poles all the same gave a - 3dB point at 50 kHz. That was fine until I went from toroidal PSU to Hypex SMPS. I gave up at that point and said Hypex know what they are doing. The SMPS sounded brighter and very detailed and no fatigue. I slightly distrusted it and said my filter was telling the truth. I had to drastically redesign the toroidal PSU to be as good. I used a a few tricks to do it. I got there in the end.
When someone said the Hitachi needs the output choke I thought it could be true.
I met a Sony engineer in 1979 who showed me a class D amp. He said " Very interesting as it sounds like a tube amp. Maybe we always liked the transformer? " . The Sony choke was quite large. They never made that amp. What they did make was the Sony TA5650 FET amp. Lovely and I guess similar devices as in the D ? I think it was me talking to that guy that started the Esprit range. Just too many things I said were in that concept. It was a clone of the Linn Naim Isobarik system. The Sony was a totally better concept and no clone. I told him to buy the Linn system and I heard later they did that. It was bought from Grahams in London and shipped back to Japan. The Sony flat cone speakers really were something. The Isobar principle didn't float their boat, the KEF B139 did. FAL in Japan do a better version.
There is a good chance that they are specifying THD+N. At low levels, noise will be a greater contributor to THD+N than a low distortion value. So when we see a THD+N rise at low levels, we don't necessarily know whether what we are seeing is distortion or noise. 1 watt into 8 ohms is 2.83V rms, so 0.25 watt is about half that voltage, or about 1.4V rms. An amp with an un-weighted S/N of 86dB with respect to 2.83V rms will have the noise 80dB down from 0.25W, so it would be on the same order of magnitude as 0.01% THD.
It is thus easy to mistake a rise in THD+N at low levels for crossover distortion.
An amp with an output stage biased at 100mA will be able to deliver about 200mA peak into an 8 ohm load before it fully exits its class A region. This will develop 1.6V peak into the load, corresponding to 0.32W peak or 0.16W average power.
Cheers,
Bob
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.