Basic 50W hifi amp -for beginners?

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
This amp is designed to provide high quality sound yet be fairly simple to set up and build. Unfortunately, perhaps, it uses quite a few transistors, so isn't the simplest possible amp, but does have some good features.

The overall architecture is based on the one which Doug Self expounds. It differs in a number of respects. First, the input stage is degenerated with quite high value resistors - 1k. This reduces the gain to a measly 1 mA/V (actually half this but is multiplied by two in the current mirror). The benefit is that the input stage won't "clip" or cut-off for any input signal up to 1V input, so should keep transient distortion to a minimum.

The low gain makes it essential to use a Darlington VAS pair, in this case two PNPs, driven by a PNP mirror. The reason for the mirror is that the input impedance of the Darlington VAS is so high even a 10k resistor degrades the gain significantly, but with a mirror the dynamic impedance is around 1M.

The VAS transistor, a BD140, operates at a fairly high current- 15 mA compared with the usual 6 mA or so- for the same reason as the input degen resistors - so that the VAS does not cut off in a transient. The emitter resistor is split into two so that the base resistor is connected to the tap point. This allows us to keep the base-to-emitter resistor ratio at 1k to 10 ohms, cutting the gain by about half, but reducing Early voltage distortion by raising the effective Early voltage, otherwise a ratio of 1k:20 would have given an unnecessarily large gain reduction. The additional 10 ohms, however, is included in the short circuit protection with a simple current limiting transistor on the VAS.

The compensation capacitor need only be a small value at 33 pF because of the large input degen. To aid stability, 100pF sprog stopper capacitors are used across the driver collector-base junctions. The compensation capacitor is connected between the output rail and the feedback compensation point (Darlington VAS base) which reduces crossover distortion remarkably. However, it is not quite as stable as connecting the capacitor to the collector of the VAS, but several of these circuits have been built and at worst may need 150 or 220 pF stopper capacitors. AS a result the THD figures are around 0.02% at 20 kHz and .005% at 1 kHz. This is not the headline THD performance which the so-called "blameless" amplifier might achieve, but it offers this performance with transient intermodulation distortion effects minimised while using a fairly safe compensation approach.

Short circuit protection is passive current limiting to about 5A in the output transistors. The amplifier is intended to operate with 8 ohm loads and a 2A fuse in each rail is a good idea.

Quiescent current should be about 50-100 mA in the output transistors. It could be set with just two fixed resistors in the bias stabiliser, but if the current is significantly outside these limits a parallel resistor can be used to adjust the current as needed. Alternatively the base to emitter resistor in the bias stabiliser could be a 680 ohm plus 470 pot to allow the current to be tweaked.

The output transistors are the very popular 2N3055/MJ2955 pair. These should be the epi types (all PNPs will be; the NPNs should be) as these have a higher frequency response than the original hometaxial RCA devices. The latest from ON semi also have an SOA almost as good as the original.

John
 

Attachments

  • basic50.png
    basic50.png
    29.3 KB · Views: 992
Hi Lineup.
Thanks for your comments. Indeed the performance of such a circuit is to be heard rather than seen. I have found that its quality is surprisingly good. I would go further and say that it still sounds OK even if the sprog-stopper capacitors are eliminated and the Miller capacitor moved to the normal position (collector from the VAS) but the THD at 20 kHz rises to about 0.2%. It still has a low THD at 1 kHz though. This version would be recommended for anyone needing an unconditionally stable amplifier - such as when using transformers for line matching, other loads (motors) etc for non-hifi use.

I would also ask contributors to diyaudio for their listening impressions of dual-slope Miller capacitors (the ones with the split miller and resistor to ground). I have found these to be highly strange. It's hard to describe but a sort of "artifically crisp". However, evaluating simulations reveal that the leading edge transients can be completely wrong: the overall amplifier response is delayed, but when the edge "cuts in" it over-compensates and goes faster than the actual signal. Ears were not lying. My version of the blameless sounds better!


John
 
And your point is?
.. to understand your design approach and where it differs from what I've heard or seen before. In my designs I've found the amount of compensation to have a noticeable impact on the sound of the amplifier quite unconnected to stability concerns. I've only ever 'played' with the usual Cdom type compensation. Everyone seems to hate Cdom but they always end up using it, you've gone a different route. I'd be particularly interested to hear what your impressions are on any changes in the sound of the amplifier as you adjust compensation with the scheme you have employed.
 
Last edited:
Hi Bigun
I first came across this scheme in an article by Cherry. I have used it ever since.
While the overall capacitive burden on the VAS is about the same as a conventional Miller (150-200 pF) this is not translated into base current needed by the input stage.
However, the high degeneration in the input stage means that in practice (as Self pointed out) the overall effects between a 330 pF Miller (say) and 100 ohm degen is the same as a 33pF and 1k.
What the scheme does have though is reduced crossover distortion, because the current needed in the input stage is only that needed by the VAS darlington pair (capacitor current divided by the gain) rather than the input stage having to provide the current as well.
And because the transistor stages remain conducting there should be no transient distortion.
I have not tried different capacitors. I generally use the smallest sprog stoppers necessary for stability.
The scheme seems to work best with the slower transistors (2N3055/MJ21194 etc) - the newer ones (MJL3802/1893 etc) show oscillations in the crossover region (internally) unless really large capacitors are used, in which case something else has to be done.
John
 
There is some mistake Ellis, in your tittle, the amplifier

is basic, but it is not really for beginners..more for seniors or lucky beginners.....it is the end design i have made and have listened too..anything beat these Blameless style design...no chance for others.

Not a try for beginners...they will find the ONE in their first try....lucky guys!

If beginners decide to build..well...they are going the the best possible sonics and they will feel very deceptive in further tries.

As you know, John, despite i respect your knowledge a lot, i have build a lot of units and no chance to anyone of them in these last fifty years of activity..absolutelly no chance..it is my final design, my final amplifier..i can put power, or increase bass, or different sound stage..but globally it is unbeatable, and unbeatable design.

My own reaches 0.018 of THD in 20 Kilohertz just tuning the miller..anything more is needed...and yours had that because low speed output transistors...give a try in others more modern to see what will happens.

The harmonic distribution is lovely, less than 0.002% is very easy with these circuits.

Lucky your beginners...i am in the near the end and i found that after half century.... beginners believing in you, in me, in Doctor Self, Ostripper and others that supports this kind of schematic, will be very lucky and will jump to the happy ending...starting with the best possible unit they would ever listen too.

All these Blameless...the sons of blameless, are awsome...do not matter how much mistakes you make, bad decisions taken...the one will sound great.

regards,

Carlos
 

Attachments

  • They are all this way.jpg
    They are all this way.jpg
    414.3 KB · Views: 618
Last edited:
The overall architecture is based on the one which Doug Self expounds. It differs in a number of respects. First, the input stage is degenerated with quite high value resistors - 1k.

True that gain is reduced, but the % difference between 're' of the transistor, around 25 Ohms, and the degeneration resistor is awlfully large. Current gain is important for the input stage, this is why we use small signal devices for input. Adding degeneration does raise the BW but going overboard on degeneration brings less return at the expense of current gain. Lowering the open loop gain also results in the feedback factor being reduced, reducing the ability of the feedback loop to correct for distortions created by the VAS and OPS within the closed loop BW. It is true that a non-degenerated BJT input stage may produce PIM distortion in the amplifier becuase the CL BW becomes more of a function of output current, but I believe much above 10X re, ~250 Ohms, goes beyond the point of diminishing returns.

The low gain makes it essential to use a Darlington VAS pair, in this case two PNPs, driven by a PNP mirror. The reason for the mirror is that the input impedance of the Darlington VAS is so high even a 10k resistor degrades the gain significantly, but with a mirror the dynamic impedance is around 1M.

The VAS buffer has nothing to do with signal voltage gain, but with 15mA bias and Hfe of ~100, the BD140 VAS requires 150uA of base current, just for bias. AC current gain depreciates as frequency increases especially for those slow BD transistors. The VAS buffer helps the input stage with this significantly but I might lower R29 so that ~2 - 3mA bias is in Q5. The next stage after the input is technically not a Darlington but rather an emitter follower cascade buffer. Darlington would imply the two collectors are connected together. I would not recommend a true Darlington VAS, what you have done with the buffer is correct. Although if the amp is hard clipped, Q4 current would saturate Q5 and perhaps effect and corrupt the clipping recovery, or result in smoke from Q5.:eek: A clamp for the input signal limit might not be a bad idea with this amp.

I like a CCS loaded VAS, it has larger BW, due to the larger Zout, and sounds better than a resistor load, IMO. Bootstrap works quite well also, just ask Carlos.:)
There might be some inprovement by using more suitable transistors, particularly for the VAS and CCS. KSC2690/KSA1220 are far superior to the BDs. I think you would be surprised at the results if you added a pre-driver to the output stage making it a tripple, then using a small signal device for the VAS, bias only a few mA. This means you could get rid of the VAS buffer. But this would add 1 more transistor to the circuit.:mischiev: I bet you would get more BW, being able to use less compensation, even with the 2N3055/MJ2955 outputs.

However, it is not quite as stable as connecting the capacitor to the collector of the VAS, but several of these circuits have been built and at worst may need 150 or 220 pF stopper capacitors.

The lag compensation cap would be best placed at the VAS collector. Encompasing the phase margin of the output stage within the Miller loop is unnecessarry. I would think such a circuit could be tamed by using just lag (conventional Miller Cdom) and lead compensation. Using a cap with no series resistor to load the VAS to GND (rail is same a GND in AC realm) would increase distortion as frequency increases, I would think.


I would also ask contributors to diyaudio for their listening impressions of dual-slope Miller capacitors (the ones with the split miller and resistor to ground). I have found these to be highly strange. It's hard to describe but a sort of "artifically crisp". However, evaluating simulations reveal that the leading edge transients can be completely wrong: the overall amplifier response is delayed, but when the edge "cuts in" it over-compensates and goes faster than the actual signal.

Have you experimented with TMC (Transistional Miller Compensation)? It is like two pole but the resistor connects to the output node instead of GND. There is significant difference between it and two pole. There is much info on the forum to this subject, simply search.;)

Unfortunately, perhaps, it uses quite a few transistors, so isn't the simplest possible amp, but does have some good features.

Heck thats only a few transistors... its not that bad. The amp I'm working on, well a new revision of utilizing more specialized devices this time:rolleyes:, has 205 transistors.:D All descrete except for a LM556 timer and a duel comparator. However, this includes: pre-amp, DC servos, +/- series shunt voltage regulators, VAS, two seperate output stages (full bridge output), clipping detection, and DC detection/protection.:cheeky:
 
Disabled Account
Joined 2006
CBS240 , the bd140 transistor is actually much better than you think. According to phillips its a bc639 in a TO126 package for higher dissapation. It has a ft of 100Mhz at 10ma and a Cob figure of around 7pf. This is the reason diyers find them to perform quite well especially when used as drivers in lower power amps. Used in a Vas position they are certain to outperform ksa1220 /ksc2690 when it comes to high frequency distortion above 1khz because they have much lower Cob compaired to the ksa/ksc. Both of these transistors are suboptimal for vas duty because of their high Cob figures.
 
Hi Carlos,

Hope you ll be not subject to what is called the HOLLANDESE SYNDROM
and wich is prevalent in all oil rich countries...
In a few words, it owes its name to discoveries of large amount
of gas in Nederlands in the early 50s.
The following years , industrial production did decline as
the reasonning was : why work more since there s
large amounts of such ressource to live with without
having to tear down the slightest sweat...


w
 
CBS240 , the bd140 transistor is actually much better than you think. According to phillips its a bc639 in a TO126 package for higher dissapation. It has a ft of 100Mhz at 10ma and a Cob figure of around 7pf. This is the reason diyers find them to perform quite well especially when used as drivers in lower power amps. Used in a Vas position they are certain to outperform ksa1220 /ksc2690 when it comes to high frequency distortion above 1khz because they have much lower Cob compaired to the ksa/ksc. Both of these transistors are suboptimal for vas duty because of their high Cob figures.

Yes, you're right about Cob. The 2690/1220 would make better drivers for the output stage as they are more rubust, considering the beta droop of the 2N3055/2955. KSC6282/KSA1142 would be much better VAS. Cob ~5pf.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.