How low can you go?

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
How about collecting in one single thread, the specs (and features) + description + link, of all amplifier projects, finished or ongoing on this forum, instead of having them buried in thousands of pages and hundreds of threads?


Benefits:

1. A general Reference to easily find, compare and chose an amp for builders,

2. Motivation to design or build a better amp,

3. Competition and reward for designers,

4. Motivation to actually measure the performance of an amp objectively (virtually or physically), instead of guessing and having no clue about its actual performance, playing a Russian roulette, or relying on subjective colorful poetic words like "the best" ;)

5. Unburial of forgotten projects.



In my view, faithful reproduction is the ultimate goal for an amp (provided it is also highly stable and reliable) that means: the least altering of the input signal.
Adding 2nd harmonics for example (tubes), while might sound good to some people, it is NOT a faithful reproduction, it was NOT the artist's intention (he can take care of all that in the mastering stage, no need for the user to add more) and it can even be added artificially as a sound effect that you can switch on or off if you want, instead of having it permanently and always coloring all kinds of music you listen to.

The level of distortion and quality one can hear is unknown, it varies depending on the content, the conditions, the person, etc. so we just aim for the least distortion (=highest linearity) possible. Of course, important is also the signal to noise (SNR), power bandwidth, etc. In my experience, one great benefit when aiming for perfection is that you end up discovering new methodologies and topologies, that aside from contributing to the amp technology, they can lead to an increased overall quality (not just THD).



Suggested specs (the more, the better):

- THD % at both 1Khz and 20Khz at full output and load and at a given input voltage,

- Power bandwidth range in khz at full load,

- Bandwidth -3db,

- Max load,

- SNR (signal to noise ratio) in db (can be easily calculated from rms output noise and peak output signal): 20*log10(VPEAK/sqrt(2)/rmsnoise) -even Google can calculate that)
measured at 0 input resistance and at an average real-life value.

- Amplifier's input and output impedance (when applicable),

- Phase, Group Delay graphs for the audio spectrum,

- Gain,

- Gain flatness for the full audio spectrum in db (16hz to 20Khz),

- Damping factor,

- Power consumption, efficiency, Class,

- Operating temperature range (for the claimed specs),

- Other specs, pros, cons.


Also I'd suggest that all kinds of amplifiers should be welcome for this general specs-reference, despite the amp's application eg for headphone amps too, if there is no objection from the forum's admins, as the common factor here is comparison of output sound quality. For example, while I'm designing a headphone amp, I'm very interested in all kinds of amps, to find out what has/can be achieved in terms of output sound quality.


So, is there interest about a one-thread reference for-all amp specs?
If so, how about posting the specs you are proud of, either as a builder or designer, virtual, or physical, finished or in development?
 
Last edited:
A searchable database of DiyAudio amplifier projects. I like this suggestion. Very much.

Yes, you can say that, since if there is no blah-blah, but only essential information, it would be far easier to search a large number of projects, even by visually scanning the pages.

I think designers and builders should NOT be afraid of comparison.
Comparisons make us all work harder for something better.

Posting here your project's specs will increase exposure, as all projects cannot be listed in the first page, especially the older ones, unless they are all mentioned in a single thread. For the history alone, it should be worth to include them here. Personally, I don't have a clue how many and what projects have been suggested in this forum.

If all users could access all projects in a single thread (specs + brief description + schematic/photo + link), many more would decide to build your projects even with less than "perfect" specs (one might prefer project A over project B due to wattage, cost, simplicity for example).
 
Last edited:
It would be irrelevant to me as most amps sound the same to me.
I have designed class A, AB and D amplifiers and cant tell between them.
Maybe someone else can, has better speakers, better ...ears (no offense, I see you play music too, several musicians especially drummers, have abused their ears -I have friends with problems), or prefers the best he/she can get, or simply wants to know what is offered. Post your specs/description/link!
 
I miss the gain flatness from 16 Hz to 20 kHz on your list.
Good idea.
And I would like to see damping factor in the spec list.
Done.

Like I wrote to you in another thread, worrying about spec points sounds more like work than like hobby to me.
Perfection can only be distinguished objectively by specs, we have nothing else.
Perfection of course requires hard work, but in this case, the more perfect, the more creative it requires from you to be, which is a fun process and a rewarding end result ;)
 
To give you an example of how blindly optimizing spec points sometimes messes up the real-life performance:

When you design a DAC, the signal to noise ratio and dynamic range are the key specification items. The higher the numbers, the better they sell. Anything over 80 dB(A) is dead quiet under normal domestic listening conditions, but still, 130 dB(A) sells better than 129 dB(A).

As these specs are always measured with sine waves, this tempts designers to design their DACs such that they are driven as hard as possible with a full-scale sine wave. Hence, they have no headroom at all for interpolation filter overshoot and clip when you use them to play music rather than sine waves. If you back off by a few decibels from the clipping limit, practically all music will be reproduced perfectly, but the numbers measured with full-scale sine waves will get worse by the same number of decibels.

Admittedly, I don't know any such examples for amplifiers.
 
@MarcelvdG

First, a DAC is orders of magnitude more complex (but not more difficult) than an amp, with digital filters than can even "cheat".

Second, this is a case where the designer optimizes one parameter at the expense of other parameters. That's NOT proper development and has nothing to do with optimizing an amp, or any other device, or even software, to perfection. To reveal what is hidden by the maker, more specs are required, not less.

You can have the same effect with amps, if for example you optimize THD at 1Khz to have "ultra low distortion" ignoring the performance at 20Khz which might skyrockets 100 times or more, and which is why I suggested THD at both 1Khz and 20Khz.

Or make an amp with a poor power supply, etc.
All this can be expressed or revealed in specs, and perfection definitely requires a holistic approach. The more specs and the more detailed, the better, not the opposite.
 
Last edited:
What you need to design a good hobby circuit is knowledge of electronics, common sense and an idea about what the circuit will be used for in real life.

Specification lists are a necessary evil in professional organizations where marketers, system designers and circuit designers work on the same product, but at different levels of abstraction. Those lists are almost always incomplete and can then lead to problems such as the one I mentioned. On top of that, spec points are often meant to impress potential customers rather than to be useful in real life.

Your list already shows that it is difficult to make a complete spec. You extended it twice and still have no spec on quiescent current variations due to output device temperature fluctuations in class (A)B amplifiers and their impact on distortion. I also wouldn't have a clue how to spec that.

Of course there is nothing wrong with measuring a couple of performance metrics, if you have the equipment, to see if things work as intended, nor with putting them in this thread.

For what it's worth, my non-switching class-AB amp that was published in Electronics World February 1996 has an output power of 2 times 20 W in 8 ohm (measured), an output inductance of 1.6 uH (theoretical value), -3 dB points at 1.1 Hz and 140 kHz (measured), second-order Butterworth response at both low and high frequencies (theoretical), quiescent current by design independent of temperature changes of the output devices, 0.006 % distortion at 16 W into 8 ohm at 10 kHz (measured) and 0.0025 % at 10 W into 8 ohm at 10 kHz, mainly second harmonic (measured).
 
Last edited:
What you need to design a good hobby circuit is knowledge of electronics, common sense and an idea about what the circuit will be used for in real life.

Specification lists are a necessary evil in professional organizations where marketers, system designers and circuit designers work on the same product, but at different levels of abstraction. Those lists are almost always incomplete and can then lead to problems such as the one I mentioned. On top of that, spec points are often meant to impress potential customers rather than to be useful in real life.
So you're suggesting that we should buy a car without looking at the specs, a TV, a kitchen appliance,
an airconditioner, a computer, loudspeakers and finally amplifiers... :rolleyes:

Well, the philosophy of "if it works and I (think I) can't see/hear a flaw, then it's good enough for me",
is interesting, but it has its drawbacks:


Without specs (and specs awareness):

- you cannot measure and compare performance, quality, efficiency, reliability.

- you cannot optimize any such parameters in any design.

- you cannot feel and use the competition spirit to motivate and surpass yourself, expand your knowledge, experience and inventiveness, and have more fun.

- you cannot pursue perfection, as a guarantee of maximum end-user satisfaction (yours, or anyone else's).

- you cannot have the same amount of reward, like knowing you have achieved "the best", not because you think so, but because you can prove it with actual measurements.

- you cannot push technology beyond its limits, which requires from you to be more creative and innovative, and finally contribute, because no radical solutions are necessary to achieve the average.

Aiming for something better, is what pushes evolution in all sectors, and specs is the compass that guides development.


Your list already shows that it is difficult to make a complete spec. You extended it twice and still have no spec on quiescent current variations due to output device temperature fluctuations in class (A)B amplifiers and their impact on distortion. I also wouldn't have a clue how to spec that.
My intention wasn't to make a complete list, I merely gave some suggestions, which is why I wrote "other tests" in the end.
Operating temperature range (for the claimed specs), is enough and it's what the end-user wants to know, not what happens internally (added).


For what it's worth, my non-switching class-AB amp that was published in Electronics World February 1996 has an output power of 2 times 20 W in 8 ohm (measured), an output inductance of 1.6 uH (theoretical value), -3 dB points at 1.1 Hz and 140 kHz (measured), second-order Butterworth response at both low and high frequencies (theoretical), quiescent current by design independent of temperature changes of the output devices, 0.006 % distortion at 16 W into 8 ohm at 10 kHz (measured) and 0.0025 % at 10 W into 8 ohm at 10 kHz, mainly second harmonic (measured).
Congratulations, you are the first one daring to post your amp's specs! :D
 
Last edited:
Operating temperature range (for the claimed specs), is enough and it's what the end-user wants to know, not what happens internally (added).

I'm not just an end user, neither is anyone else on this forum.

In any case, I'm not referring to operating temperature range, but to transient crossover distortion: temporary increase in distortion after a change in volume. Solid-state class-(A)B amplifiers that rely on thermal compensation to keep the quiescent current constant all suffer from that to some extent, but none of the typical audio tests cover it at all.

Congratulations, you are the first one daring to post your amp's specs! :D

By the way, I forgot to mention that my amplifier can process a full-scale square wave without slewing (confirmed by measurement) and that it has a LED that warns the user when it is driven into clipping or when anything else happens that causes a large difference between the actual and the intended output voltage. When clipping severely, the amplifier automatically disconnects the loudspeakers.
 
Without specs (and specs awareness):

- you cannot measure and compare performance, quality, efficiency, reliability.

- you cannot optimize any such parameters in any design.

- you cannot feel and use the competition spirit to motivate and surpass yourself, expand your knowledge, experience and inventiveness, and have more fun.

- you cannot pursue perfection, as a guarantee of maximum end-user satisfaction (yours, or anyone else's).

- you cannot have the same amount of reward, like knowing you have achieved "the best", not because you think so, but because you can prove it with actual measurements.

- you cannot push technology beyond its limits, which requires from you to be more creative and innovative, and finally contribute, because no radical solutions are necessary to achieve the average.

Aiming for something better, is what pushes evolution in all sectors, and specs is the compass that guides development.

I don't agree with your claim that tough specs necessarily lead to innovative designs. Sometimes they do, sometimes they don't, and sometimes undue emphasis on some "key" specs leads to an unbalanced design that really doesn't make much sense.

To give you an example of the second case, when you compare my amplifier from the mid-1990's with Douglas Self's blameless amplifiers that were published around the same time, you'll see that his amplifiers are far more conventional, yet have lower distortion. I couldn't care less, because my amplifier's distortion is low enough for all practical purposes and due to the fact that it has no transient crossover distortion and can not even be driven into slewing by square waves, I'm pretty sure it also has low distortion on any musical waveform (in fact I've done some subtractive tests with music from Tracy Chapman's CD Matters of the heart).

Still, if I were to design a discrete amplifier at work and the main spec points were low harmonic distortion and low production costs, I'd probably be forced to design something conventional, similar to Self's amplifiers.
 
I'm not just an end user, neither is anyone else on this forum.
In my view, only the resulting performance and reliability is what counts, that's the ultimate goal, not what happens inside or how you got there. More output tests for performance and reliability will reveal internal weaknesses, but internal specs won't necessarily guarantee an excellent output performance, let alone the huge list that would be required to cover all cases (rather impossible) and most parts of it would be irrelevant to irrelevant topologies.


I'm not referring to operating temperature range, but to transient crossover distortion: temporary increase in distortion after a change in volume. Solid-state class-(A)B amplifiers that rely on thermal compensation to keep the quiescent current constant all suffer from that to some extent, but none of the typical audio tests cover it at all.
That's interesting, it made me look for more info about that and any other weird behaviors I'm missing, thanks. My amp definitely won't suffer from that, and I'll suggest a test when I'll get there.
That said, if some specs are missing, that doesn't mean we should abandon specs altogether.


To give you an example of the second case, when you compare my amplifier from the mid-1990's with Douglas Self's blameless amplifiers that were published around the same time, you'll see that his amplifiers are far more conventional, yet have lower distortion. I couldn't care less, because my amplifier's distortion is low enough for all practical purposes and due to the fact that it has no transient crossover distortion and can not even be driven into slewing by square waves, I'm pretty sure it also has low distortion on any musical waveform (in fact I've done some subtractive tests with music from Tracy Chapman's CD Matters of the heart).

Still, if I were to design a discrete amplifier at work and the main spec points were low harmonic distortion and low production costs, I'd probably be forced to design something conventional, similar to Self's amplifiers.

You said it yourself: "I couldn't care less", that's the only cause your amp had more distortion, you just quit pushing for lower distortion. You could have all the other specs, in combination with low distortion, if you kept optimizing, even if you had to change your topology -most likely not to a conventional one, as probably those didn't have the other advantages you had achieved.


sometimes undue emphasis on some "key" specs leads to an unbalanced design that really doesn't make much sense.
I have addressed that previously: optimizing one parameter at the expense of others is NOT proper development, it is BAD development, and therefore irrelevant.


I don't agree with your claim that tough specs necessarily lead to innovative designs. Sometimes they do, sometimes they don't
I'm talking about really tough specs or tough combinations of specs, that no one else has achieved so far.
Then, obviously to achieve something extraordinary, you need extraordinary solutions. There is some threshold, beyond that, you need to think "out of the box".

Cost-optimization can also be a very tough spec, if combined with other demanding specs.

For example, once (mid-90's) a friend brought me a simple schematic from a magazine where a mosfet was used in class D as an RF amplifier (instead of linear) in the AM broadcasting range, with a power output of 10 watts. He asked me if I could make it more powerful.
I took the challenge and started developing it from scratch. Day-by-day I was increasing the RF output power: 25, 50, 75, 100... Then I set some initial tough desired specs.

In the end, it was a design with a unique set of specs and features (and I bet it still is, I haven't heard of anything like that).
it could output 500 Watts RF while operating from a 12V car battery(!), it consumed only 40% vs a conventional device in AM mode and it had 94% efficiency (operating "cold") in CW mode (also covering the Morse amateur band too with full morse support (hand & Iambic) and automations for worldwide communication via the ionosphere), it was also highly optimized for cost, it used cheap-mosfets instead of extremely-expensive RF ones, it had a custom dual-modulus PLL that costed 10 times less with 2792 channels with fast lock and low noise, the led display could show freq, amperes (0-50), volts, temp, UV meter, it auto controlled the fan, it had soft-on-off (a simple push button), alarm and auto shut-down, battery protection, output protection, temp protection, it could self adjust if driven from a poorly regulated power supply by detecting the useful voltage bellow ripple to avoid hum, it had a "hi-fi" high-level transformerless AM modulation, input mixer, adjustable audio compressor for perceived loudness, etc etc a long spec and feature list of >40 items.

The deep optimization of the (unheard) combination of the initial specs (power, voltage, efficiency, cost) required innovative solutions, and after achieving them, those specs made it worthwhile to add more specs and features. The full development lasted 18 months incl. the PCB design, but I finished the output stage at full power + the driving in the first 2.5 weeks due to previous relevant experience.

Many more projects were like that in both the digital and analog domain, in software and mechanics too.

This also is the case in my current project BTW that made me realize the benefits of deep spec optimization and how fun it can be.


By the way, I forgot to mention that my amplifier can process a full-scale square wave without slewing (confirmed by measurement) and that it has a LED that warns the user when it is driven into clipping or when anything else happens that causes a large difference between the actual and the intended output voltage. When clipping severely, the amplifier automatically disconnects the loudspeakers
That's very nice!

Let's give this thread a chance. Let's stop philosophical blah-blah and start posting specs.
MarcelvdG, why not re-post all the details in a more readable format and also add a link / photo / or schematic if you want -png or other?

I will also post my current project's specs (simulated for the time being) as an example.

Amp-designers and builders, you are welcome to post your specs/description/links, even simulated ones.
Any bold members out there? :)
 
Last edited:
Project: SS DC DD Electrostatic Headphones Amplifier.

Status: Under development.
(to be slowed down to a minimum for a few months, as I have to finish another project first (software), that will help earn a living.
That said, I just received the first batch of components, in order to have them when I'll need them, as the availability for some of them is very limited. I'll also use them to verify some parts of the simulated design in practice.

Description: Solid state, Direct-drive, DC, high voltage output, extremely-low distortion and noise, capable to drive two+ ES headphones simultaneously at full specs via two connectors.

Specifications: (preliminary, simulated for the time-being in LTspice, just a few specs for now)

THD at 20Khz: 0.000068 % (0.68ppm, -123.3 db)

THD at 1Khz: 0.000056% (0.56ppm, -125 db)

SNR (signal-to-noise ratio): -132 db

DC-20KHZ freq. response flatness: 0.0000428 db (42.8 micro-db)

DC-20KHZ phase difference: 0.044 degrees (44 milli-degrees)

DC coupled,

Direct drive

Conditions for the above specs:

Input voltage: 1V pp.

Max output voltage: 1300V pp. (gain = 1300).

Max load: 400pf (equivalent to more than two ES headphones driven simultaneously (eg Stax SR-009 = 150pf).


TODO: (I hardly have covered 30% of the total development).
DC servo, output protection, regulated supplies, addition of xlr inputs, digital inputs, a high performance DAC chip to be embedded, system health monitor, self-test, user safety, an attempt to solve a couple of essential unsolved amplifier issues, actual testing and development incl. reliability and stress-tests, a custom futuristic case and interface, PCBs, etc.

THD @20KHZ:
38671da6c11d326350ee7ee0ee6c30b1.png



THD @1KHZ:
f5f785d5a7ed4eb21cda2eb7c88ff40b.png


SNR:
ea4376de3c989aa87e2e3e3e550e0b92.png
 

Attachments

  • Fourier20khz_5.png
    Fourier20khz_5.png
    24.4 KB · Views: 294
  • Fourier1Khz_5.png
    Fourier1Khz_5.png
    21 KB · Views: 294
  • SNR_5.png
    SNR_5.png
    4.8 KB · Views: 29
Last edited:
Fair enough, let's agree to disagree.

Description:
Non-switching class-AB power amplifier using a non-linear common-mode loop to control the AB behaviour as published in Electronics World February 1996

Features:
Quiescent current does not depend on temperature tracking between the output devices and a temperature sensing device, hence there is no transient crossover distortion (and no quiescent current trimmer).
Designed to process a full-scale square wave without slewing (confirmed by measurement).
LED that warns the user when it is driven into clipping or when anything else happens that causes a large difference between the actual and the intended output voltage. When clipping severely, the amplifier automatically disconnects the loudspeakers.

Performance parameters (I won't call them specs because no-one demanded that I met them):
Output power of 2 times 20 W in 8 ohm (measured), 2 times 35 W in 4 ohm if I remember well
Output inductance of 1.6 uH (theoretical value)
-3 dB points at 1.1 Hz and 140 kHz (measured)
Second-order Butterworth response at both low and high frequencies (theoretical)
0.006 % harmonic distortion at 16 W into 8 ohm at 10 kHz (measured)
0.0025 % harmonic distortion at 10 W into 8 ohm at 10 kHz, mainly second harmonic (measured)

Further information:
See Marcel van de Gevel, "Audio power with a new loop", Electronics World February 1996, pages 140...143

The basic working principle is explained in this post: http://www.diyaudio.com/forums/solid-state/163231-class-ab-soft-switching-3.html#post5082095 (I can't post the entire schematics because of copyright issues.)

A similar design with more features can be found here Amplifier with variable bias that indicates whether it works in A or AB , but it is neither finished nor tested.
 
Last edited:
Hi
Mr.Magic: Just as a point of interest: ES headphones are typically driven differentially, that is, separate amplifiers driving the two stators in ooposite phase. So, is your 1300Vpp for one output or measured across both outputs?

I built a batch of ES headphone amps that are symmetric throughout and are 360ppb20 (20kHz) at 160Vpp output. Since the headphones have high efficiency of 100dB at 100Vrms across the stators (meaning 50Vrms per side, 70Vp per side), I decided that I was only going to design it for whatever output +/-100V rails would provide. So, the performance for the total system, including balanced input stage, EQ, width control and ES drivers:

140Vpp: 300ppb20 120ppb1, 100dB SPL
14Vpp : 10ppb20 <10ppb1 80dB SPL

I have no interest in average SPLs above 80dB, and even at that, it is very loud and takes a long time for your ears to recover.

I also made a low-z drive output stage for the above to drive dynamic head phones. At 20kHz and output levels corresponding to <1mW (100dB SPL), THD20 <10ppb. At very high drive corresponding to the "5W recommended" for Audeze at various impedances (20,70, 110) with balanced drive, THD is <300ppb20. At 1kHz THD is <100ppb.

For both head phones, at the typical levels of drive I ever use, THD is below waht LTspice can simulate and far below what modern test equipment can measure. SNR is >100dB unweighted.
 
@nauta,
-It is for one output, (it's a dual supply design in order to be capacitorless, all-dc) so diferentially the amp provides 2600Vpp between the two stators.

-I understand your point since you are making those amps for existing sensitive commercial headphones.
I decided to provide such a high voltage output, for two reasons:

1) After the amplifier, I am going to design and build a custom pair of headphones, that will take advantage of the high-voltage output to provide an extended and more powerful low frequency range (they will have a larger diaphragm area and a larger DS spacing).

2) I want the amp to be as powerful as the more powerful amps on the market (but with far better specs), capable to drive insensitive headphones.
For example, the Stax SRM-T8000 (in production) outputs about the same voltage (1330Vpp) but it's a hybrid one (tubes/solid-state).
An old discontinued hybrid Stax model SRM-T2 could output 1782Vpp! So there are insensitive old ES headphones out there.

140Vpp: 300ppb20 120ppb1, 100dB SPL
14Vpp : 10ppb20 <10ppb1 80dB SPL

You mean:
140Vpp: 0.3ppm at 20Khz, 0.12ppm at 1Khz, 100dB SPL
14Vpp: 0.01ppm at 20Khz, <0.01ppm at 1Khz, 80dB SPL

Nice!
I think ppm THD (if not THD %) is less confusing than parts per billion.

Thanks for posting the specs.
 
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.