High Performance Class A amplifier

[Aniket]

I started this project more than one month ago, when one of my friend requested me a 50W Class A amplifier for his DIY speakers he made using the worlds best drivers according to him.

So, i designed it with the foremost goal in mind: lowest possible distortion with a simplistic design. and a very stable and robust output stage, and high gain, high bandwidth voltage amplification stage and simple input stage with low noise very high gain discretes. the design goal was to achieve 50W rms into a 4ohm load in Class A.

Designing of the schematic and PCB is complete and very soon will receive the PCB's. posting some simulation results:

1. 53W into 4ohms, THD=0.003% @ 20kHz

2. 53W into 4ohms, THD=0.003% @ 1kHz

3. 53W into 4ohms, THD=0.003% @ 100Hz

4. 10W into 4ohms, THD=0%, yes 0% @ 1kHz

5 10W into 4ohms, THD=0.001% @ 20kHz

6. 20kHz square wave output

7. 1kHz square wave output

8. Clipping @ 1kHz

9. The PCB

The amplifier is actually a modification of a Class AB amp i designed some time ago, it had very natural sound stage and very good PSRR and Signal to noise ratio >120dB, which is critical for a Class A amp. The bias stage actively monitors the output stage idle current irrespective of the signal being amplified.
The distortion figures achieved for this Class A amp is 0.003% at rated power from 20Hz to 20kHz, and 0% at 1kHz @ 10W, and 0.001% at 20kHz @10W. Now Beat this. !!!!


Regards,
Aniket

 

[tcpip]

Your figures bring back one question I had about Class A.

In typical Class B amps, the GNFB helps to cut distortion. And since the effectiveness of GNFB reduces with frequency, plus some other factors also reduce their effectiveness with frequency, there is a rise in distortion with frequency. Correct me if I'm wrong.

But the primary source of distortion is crossover distortion. And Class A does not have any crossover distortion. Therefore, shouldn't a Class A amp have a flat distortion profile with frequency? Your amp seems to show this, at least in simulation, whereas most other Class A amps I've seen seem to have a rising distortion profile like a Class B amp. Why do they have this profile?

BTW, can you share the schematic of your amp? What OPS devices are you using?

 

[Aniket]

In typical Class B amps, the GNFB helps to cut distortion. And since the effectiveness of GNFB reduces with frequency, plus some other factors also reduce their effectiveness with frequency, there is a rise in distortion with frequency. Correct me if I'm wrong.

correct:thumbsup:

But the primary source of distortion is crossover distortion. And Class A does not have any crossover distortion. Therefore, shouldn't a Class A amp have a flat distortion profile with frequency? Your amp seems to show this, at least in simulation, whereas most other Class A amps I've seen seem to have a rising distortion profile like a Class B amp. Why do they have this profile?

the main difference between a Class A and Class AB amp is in the output stage. while most Class A amps can transition into Class AB and vice versa.
high frequency distortion is largely caused by the preceding stages, the voltage amplification stage being the main culprit. and is dependent on the devices and the compensation scheme.
For most devices the gain decreases with frequency, so the distortion increases. it also depends on Cdom, and/or any other compensation scheme in the VAS which also limits the bandwidth of the amp.

Now, here VAS device used is a very high gain, low Cob, fast transistor. so distortion is considerably low.

BTW, can you share the schematic of your amp? What OPS devices are you using?

i can't share here, i can send you through PM.
its a simple design with CCS for differential pair and VAS. emitter follower output stage. output devices are NJW0281/0302.


Regards,
Aniket

 

[HormusPeston]

There was a discussion on diyaudio about finding matched, fast, low Cob complementary pairs such as the C3503/A1381. The ones in the same gain-grade were difficult to match and one chap had to purchase 100 pairs before he found a 5% match. High-gain, low Cob transistors were used in CRT televisions. But if I remember correctly, they were all NPNs?

Cheers,

~HP

 

[Aniket]

There was a discussion on diyaudio about finding matched, fast, low Cob complementary pairs such as the C3503/A1381. The ones in the same gain-grade were difficult to match and one chap had to purchase 100 pairs before he found a 5% match. High-gain, low Cob transistors were used in CRT televisions. But if I remember correctly, they were all NPNs?

Cheers,

~HP

I am using the same transistors here, KSC3503E/A1381E, made by Fairchild(now On-Semi), E is the suffix for hFe classification. there are several other great VAS transistors made by Sanyo(now owned by On-semi) earlier.
2SC3503/2SA1381, 2SC3788/2SA1478, 2SA1407/2SC3601, 2SC2912/2SA1210 were originally made by Sanyo, during the CRT era. these all are complementary and make perfect VAS devices.

Regards,
Aniket

 

[Kannan]

Ankit, if it is ok with with you, how much does making one cost?

 

[Aniket]

Mine MJ's are also from ON-semiconductors. New production it seems.

MJ15003/15004, MJ21193/21194 or MJ15024/15025, these are really robust transistors. in fact the legendary 2N3773/6609 can also be used.

Ankit, if it is ok with with you, how much does making one cost?

it depends on the components and semiconductors used.

Regards,
Aniket

 

[Aniket]

Just thought of sharing this simple explanation of Class A amps:

Class A

As compared with the other amplifier classes well cover, Class A amps are relatively simple devices. The defining principle of Class A operation is that all of an amplifiers output devices must be conducting through the full 360 degree cycle of a waveform. Class A can also be broken down into single ended and push/pull amplifiers. Push/pull diverges from the basic explanation above by utilizing output devices in pairs. While both devices are conducting through the full 360 degree cycle, one device will shoulder more of the load during the positive portion of the cycle, while the other handles more of the negative cycle; the primary advantage of this arrangement is reduced distortion relative to single ended designs, as even order harmonics are cancelled out. In addition, push/pull Class A designs are less susceptible to hum; single ended designs tend to require special attention to the power supply to mitigate this issue.

Because of the positive attributes associated with Class A operation, it is considered the gold standard for audio quality in many audiophile circles. However, there is one important drawback to these designs: efficiency. The requirement of Class A designs to have all output devices conducting at all times results in significant amounts of wasted power, which is ultimately converted to heat. This is further exacerbated by the fact that Class A designs require relatively high levels of quiescent current, which is the amount of current flowing through the output devices when the amplifier is producing zero output. Real world efficiency rates can be on the order of 15-35%, with the potential to drop into the single digits using highly dynamic source material.

Very true.

add to this the most vital part of a Class A amp- The power supply. it should be capable of supplying steady 3-6 amperes(depends on the power output and idle current) continuous without sagging. More importantly it should have least ripple and noise. as it could creep into the audio chain.


Regards,
Aniket

 

[tcpip]

More importantly it should have least ripple and noise. as it could creep into the audio chain.

Yes, this is a big issue for a lot of Class A amps I see among DIY builders, because they combine Class A with minimalist. Nelson Pass is a big guru of the minimalist school. These amps often have poor PSRR.

If the topology is designed like a more complex amp (like a lot of Class B amps) then PSRR can often be quite high. Things like a 2-transistor CCS in input and CAS stages, current mirrors in the LTP of the input stage, etc, all increase the PSRR. G Randy Slone's classic book, "High Power Audio Amplifier Construction Manual" has this awesome 40W Class A amp which will reject supply rail noise very well. Its parts count however is quite high, and it's a fully symmetric topology.

In short, having very low-noise rails is a requirement only for minimalist topology Class A, I suspect. Not for all Class A.

 

[Aniket]

Yes,

correctly said.
I will add to this. All Class A amps require high continuous steady current which is mainly consumed by the output stage. any ripple or noise present in supply rails will somehow affect the output. it could be stability issues and could be oscillation as well.
To me the best and simplest way to cope with this is to place a 100-1000uF low esr electrolytic and a 100-470nF film cap closest to the power supply pin of the transistor. the performance gain through this is noticeable and there's sonic improvement also.

Regards,
Aniket

 

[Aniket]

Started collecting parts for the build. was Lucky to find some Wima film caps in my parts bin

 

[sadik]

Awesome Bro ....

 

[Aniket]

The PCB's arrived today for the amp modules and protection .

Regards,
Aniket

 

[omishra]

Interesting.
I would like to build one when I get back to work again. Thank you for sharing.

Sent from my ONEPLUS A3003 using Tapatalk

 

[SanjeevM]

Aniket, is there a significant difference between the LKG and LS series of Nichicon capacitors? Is the price difference justified.

Secondly, if one uses 10000uf caps, then there will be more caps in parallel for the same bank size, so the overall impedance may be lower.

 

[Aniket]

Interesting.
I would like to build one when I get back to work again. Thank you for sharing.

Sure

Aniket, is there a significant difference between the LKG and LS series of Nichicon capacitors? Is the price difference justified.

It depends on the application. As LKG series are designed for high grade audio equipment, I would prefer to use them in the signal path and decoupling.
For power supply and filtering, low esr, high ripple current rated caps will serve better.

Secondly, if one uses 10000uf caps, then there will be more caps in parallel for the same bank size, so the overall impedance may be lower.

Yes correct, the overall esr will be lower and you get much higher ripple current rating. For example, its better to use two 4700uF in parallel than a single 10000uF. you get a more sturdy power supply.


Regards,
Aniket

 

[tcpip]

For power supply and filtering, low esr, high ripple current rated caps will serve better.

Isn't temp rating important too? Doesn't a 105-deg-C cap last much longer than an 85-deg-C cap in these situations?

 

[Hari Iyer]

Couple of years ago, i had DIYed the Beta 22 class A amp. One of the channels got burnt due to DC in the input stage and after that i was never able to fix it. Also getting the matched jfets gets increasing difficult. Hence i settled for a 300B tube for my audio journey.

I may try and restore this Beta 22 one day but do not know now when.

 

[Aniket]

Isn't temp rating important too? Doesn't a 105-deg-C cap last much longer than an 85-deg-C cap in these situations?

Yes a 105 degree C rated cap will last longer in a hot environment, like SMPS or if there's any hot component nearby the cap. But in our linear power supplies they rarely bear heat.

Regards,
Aniket

 

[Aniket]

Huh, finally the amp is complete. was quite a lengthy project. some pics:

https://drive.google.com/open?id=0B83NgqcNMzasTmIzaDRBN2ZCYzA

Cheers !!!