As a hobbyist, I strive only for the best amplifier, a perfect amplifier. An amplifier that gathers all the best traits from all the best amplifiers. And I think THD (it's only a convenient way to say distortion) at Nap250 level is audible in a negative way.
As I'm searching for the best class B, I'm a bit worry about the fact that many class-A amplifiers show certain favorable trait compared to class-B. Of course it is not the high level second order distortion. I suspect it is something related to the non-existence of very high order distortions in class-A amplifiers. I have not compared my class-B amp(s) side by side with class-A (may be soon, as I'm running out of worthy competitors
)
hi huggygood. I think every DIYer and Ebay type kit assemblers in particular, should see why we need heavy duty switches, relays and connections for speaker outputs. As you've gone to the trouble of taking pics, why not attach them to a post here with a # reference.@Ian....I took one or two pictures to show you this wonder.
When you have a multimeter you measure impedance, maximum current, dc offset, power.
When you have an oscilloscope you measure frequency response and noise and rise time.
When you have a spectrum analyzer you measure THD and spectral shape.
I don't have spectrum analyzer. I use ears to hear if I could hear distortion (increase/reduction), I use spice to guide my ears and the direction where to advance the circuit.
I have oscilloscope but I don't even use it. Just a waste of time when ears and spice (with some methodology) and multi-meter can do it faster.
From your list, I just measure the DC offset, along with some DC operating points around the circuit. I use calculator to measure current (I=V/R). Heck, LTspice model for BC546 is so awful, it doesn't behave like the one I have in the circuit.
I think you have an advantage in some ways by not relying on test equipment.
Controversial?
It is an easy trap to fall in to: measuring what is convenient to measure and fixating on it. In a way the inventors of affordable test equipment have defined the standard measurements for audio.
To avoid this trap, trust your ears. Fixate on what you hear.
So long as you can establish and maintain the stability of a circuit and don't ignore it in pursuing the fixation. That's not easy to do unless you have ways of monitoring responses beyond the audible bandwidth - particularly as we miss most everything beyond about 14kHz by using just our ears. Ears alone implies the need for a lot of experience too. Not everyone can hear and interpret what is occurring without training or at least studying old texts that cover practical issues of amplifier design and construction.
I think the issue is cost. Software is cheap, hardware is expensive and I couldn't afford a decent scope for many years but I can't live without one now that diyAudio presents so many different circuits and so many interesting and challenging ones worthy of investigation.
Over many years, I have found great old TEK oscilloscopes big enough to need trolleys and 100MHz models good enough for laboratory work at affordable prices. Good products stand up well over time and I occasionally watch for governmental auctions, when labs sell off obsolete equipment - some of it is not very old and is of better quality than you can buy new.
It is very easy to do using LTspice. Once you know the concept well (Nyquist), i.e. the Math behind it, it is just a piece of cake!
When I was a System/Business Process Analyst, I probably knew Accounting more than most Accountants, because I knew how to build Accounting software. In short, you cannot build a software without precise understanding of the processes behind it.
Hardware is more like a tool, a ready-to-use tool where you don't need to know the process to use it. While software can be more powerful especially when you know the Math behind it or when you know how it is built/programmed. LTspice is hard to use because it is not 'ready'-to-use. Needs a little programming skill and a lot of concept to use it properly. I don't even use the models for the transistors I'm using because I feel that the importance of accurate transistor models is highly overrated.
I think the way we think about how things work is influenced by the methods we use to measure them. Ultimately the human auditory system is the correct measurement by definition. In the absence of understanding how this system works we make indirect measures and these tend to be driven by equipment designed for other fields of study. Inevitably this can lead to a fixation on measures that are not entirely appropriate, not just by designers, but by marketeers.
THD is one such measure. Made widespread by reducing its cost through cheaper and more convenient equipment from the likes of AP. Now things have to have low THD or they are of questionable musical capability. So people will go on the pursuit of <0.01% THD in an amplifier even though it will be used with a speaker whose THD is >1%. Which is why speaker manufacturers never quote THD! Speaker companies stay old-school and quote frequency response.
My understanding is that JV did not buy-in to the THD race to the bottom. He had to insist that magazine reviews of Naim did not quote THD as this would make Naim look worse even though it sounded better.
THD is one such measure. Made widespread by reducing its cost through cheaper and more convenient equipment from the likes of AP. Now things have to have low THD or they are of questionable musical capability. So people will go on the pursuit of <0.01% THD in an amplifier even though it will be used with a speaker whose THD is >1%. Which is why speaker manufacturers never quote THD! Speaker companies stay old-school and quote frequency response.
My understanding is that JV did not buy-in to the THD race to the bottom. He had to insist that magazine reviews of Naim did not quote THD as this would make Naim look worse even though it sounded better.
The purpose of the low distortion THD at all costs, is a bit like the maid who hides the garbage under the carpet!!!!
One of the ways to get for a low THD distortion is, for example, to use high rates of negative feedback, but unfortunately the counterpart is to pay the price of a high Dynamic Intermodulation Distortion, terribly unwelcome to our ear even in very small quantities.
One of the ways to get for a low THD distortion is, for example, to use high rates of negative feedback, but unfortunately the counterpart is to pay the price of a high Dynamic Intermodulation Distortion, terribly unwelcome to our ear even in very small quantities.
The purpose of the low distortion THD at all costs, is a bit like the maid who hides the garbage under the carpet!!!!
In England we say "Sweeping the problem under the rug/carpet".
@johnego
I think we understand that THD is not the only or best metric for indicating whether an amplifier is going to sound good.
When I raised the matter of oscilloscopes earlier, it was not to support making measurements such as THD but specifically to monitor stability. If you don't realise your amp is on the edge of taking off, you won't be listening to anything for long. As I explained, predictive analysis or simulation only works with sufficiently accurate parts models and many, dare I say most clone builders, will be using generic semis or "fakes" that are unlikely to fit the models in use. That's probably the reason for concluding that models are overrated when logically, accurate simulation largely depends on them. Bonsai's recent KSC2690A/KSA1220A thread discusses many of the issues. KSC2690A and KSA1220A
In some cases, it may not matter that we are using something like 2SC945 when the label shows the part as 2SC2240 for example. In others it will result in unexplained failures, noisy operation or poor sound quality. Bear in mind that unless you buy semis from authorised distributors, you are probably using generic junk which renders the models dubious anyway.
I say this with the experience of watching lots of kids and adults build and test their Naim clones over several years here. I know what assumptions people make that lead to mistakes and sometimes failures, mostly because of the uncertainty of substituted semis. People are amazed when they realise the kits and bags of other semis are full of substitutes and so-called "fakes" which often prevent their projects from sounding like the original. You may be aware of how poor the Naim sound becomes when you substitute the VAS transistors in a NAP clone for example. Very few kits have been supplied with correct or at least closely similar parts so very few clones will ever sound right, no matter how much simulating and thinking we do about it.
I would agree that we can over-rate models when the parameters in question don't have much bearing on our concerns. This could just be DC circuit conditions or AC linearity within a narrow range of operation. I disagree regarding the VAS transistors, drivers or output transistors though. These require higher accuracy or they aren't much use to you in simulating more than DC and maybe Nyquist conditions for which you may not need the complexity of LT Spice.
I think we understand that THD is not the only or best metric for indicating whether an amplifier is going to sound good.
When I raised the matter of oscilloscopes earlier, it was not to support making measurements such as THD but specifically to monitor stability. If you don't realise your amp is on the edge of taking off, you won't be listening to anything for long. As I explained, predictive analysis or simulation only works with sufficiently accurate parts models and many, dare I say most clone builders, will be using generic semis or "fakes" that are unlikely to fit the models in use. That's probably the reason for concluding that models are overrated when logically, accurate simulation largely depends on them. Bonsai's recent KSC2690A/KSA1220A thread discusses many of the issues. KSC2690A and KSA1220A
In some cases, it may not matter that we are using something like 2SC945 when the label shows the part as 2SC2240 for example. In others it will result in unexplained failures, noisy operation or poor sound quality. Bear in mind that unless you buy semis from authorised distributors, you are probably using generic junk which renders the models dubious anyway.
I say this with the experience of watching lots of kids and adults build and test their Naim clones over several years here. I know what assumptions people make that lead to mistakes and sometimes failures, mostly because of the uncertainty of substituted semis. People are amazed when they realise the kits and bags of other semis are full of substitutes and so-called "fakes" which often prevent their projects from sounding like the original. You may be aware of how poor the Naim sound becomes when you substitute the VAS transistors in a NAP clone for example. Very few kits have been supplied with correct or at least closely similar parts so very few clones will ever sound right, no matter how much simulating and thinking we do about it.
I would agree that we can over-rate models when the parameters in question don't have much bearing on our concerns. This could just be DC circuit conditions or AC linearity within a narrow range of operation. I disagree regarding the VAS transistors, drivers or output transistors though. These require higher accuracy or they aren't much use to you in simulating more than DC and maybe Nyquist conditions for which you may not need the complexity of LT Spice.
Last edited:
Oops, I was supposed to reply to this post not the other one. IM wondering of there are any populated boards available for purchase that are decent clones? How about the LJM one?
LJM's clone board has been around with revisions, for many years. It's a quality, full sized board but the semis supplied are are not able to deliver the sound quality you expect. The SQ is not good unless similar, or in the case of the VAS semis, the exact correct parts are fitted.
The best clone is presently the NAP200 AFAIK. The improved protection design performs and sounds better too. Beware the semi quality though. As time passes, there are new suppliers who send inappropriate or fake parts to be competitive. You may think that you can buy and fit correct parts yourself but time is running out for them - most are now obsolete and some don't have SMD or through-hole equivalents that can be easily bought and adapted. A favourite supplier I have used for clones is Along1986090. His kits have not been the cheapest but have been among the best quality for as long as I've been a DIY member.
The best clone is presently the NAP200 AFAIK. The improved protection design performs and sounds better too. Beware the semi quality though. As time passes, there are new suppliers who send inappropriate or fake parts to be competitive. You may think that you can buy and fit correct parts yourself but time is running out for them - most are now obsolete and some don't have SMD or through-hole equivalents that can be easily bought and adapted. A favourite supplier I have used for clones is Along1986090. His kits have not been the cheapest but have been among the best quality for as long as I've been a DIY member.
Over the years I have built for tests, tests and for friends, 5/6 NAIM clones and I also confirm that the kits offered on eBay by the dealer Along1986090, upgraded with minimal circuit modifications and equipped with components selected and bought separately, are actually very interesting, well made and above all very well sounding.
Ian Finch said:
In my experience, it is surprising that when we design an amp properly, the THD will automatically go low. But yes, we can make the THD low without making the amp proper or sounding good. Similarly, a circuit working well in real world will work also in simulation but circuit that work well in simulation does not necessarily work well in real life.
Yes, you are right. Simulating DC is the basic use (but is very important and underrated) and stability analysis, which in most/all situations do not require model precision. But what else do you need for designing good amps that require accurate models?
Garbage in garbage out. Asking for accurate models and asking simulators to give good circuit is a tall order. I don't think SPICE have that power yet. User intervention is so critical that without good 'engineering judgement', accurate models will be useless. And when you have very good engineering judgement, I'm afraid that you can even design a very good amp without simulator.
Having accurate models is equivalent with measuring all the transistors and then (with the help of the simulator) operate ALL these transistors in their sweet spots to build the circuit. This has been highly underestimated. If you don't believe it, ask members to measure DC along their amplifier schematics and post it here and only from that we may be able to know why the amp doesn't sound good enough.
Chapter 9, Hardware Design techniques by the Analog Devices gurus Kesler, Bryant, Jung, Garcia and McDonald
Prototyping and simulating analog circuits
Its just common sense. If you google and read the first few paragraphs, I don't think we'll have much left to discuss on simulators.
My practical experience is that simulated circuits don't match reality, unless the models are adjusted to the reality. It took about an year for an engineer in a lab where I was working to align his simulations to what we were measuring. On a HIL simulator. After he did, though, he could predict exactly the outcome.
So, simulators work, but don't work in the way most DIYers are using them. When they work, they save time as any other design tool. I don't know of any professional organisation that doesn't use some form of circuit simulation. But also none of them go in production just based on simulation
As hobbyists we sit wherever we decide. We can go crazy and spend months to adjust our simulation, or we can build and try several variations/prototypes of our circuit.
Prototyping and simulating analog circuits
Its just common sense. If you google and read the first few paragraphs, I don't think we'll have much left to discuss on simulators.
My practical experience is that simulated circuits don't match reality, unless the models are adjusted to the reality. It took about an year for an engineer in a lab where I was working to align his simulations to what we were measuring. On a HIL simulator. After he did, though, he could predict exactly the outcome.
So, simulators work, but don't work in the way most DIYers are using them. When they work, they save time as any other design tool. I don't know of any professional organisation that doesn't use some form of circuit simulation. But also none of them go in production just based on simulation
As hobbyists we sit wherever we decide. We can go crazy and spend months to adjust our simulation, or we can build and try several variations/prototypes of our circuit.
Manufacturers of qualified components (i.e. the genuine OEM components specified by the designer) generally publish SPICE models for their products. Where these are not so good, better ones are usually available free, from various on-line interest groups such as Components Library and Circuits - LTwiki-Wiki for LTspice, who have refined particular models for more demanding uses. Bob Cordell, who is an experienced design engineer and author, (ex-National Semiconductor) also provides some models online for the common audio semis used in his book(s) and by others here.
Spice and LT Spice in particular, is actually a very good circuit simulation tool for audio, as the many high-level designers and engineers who post here will testify. The results it predicts are often more useful than actual measurements though you need to know how to use it, which is the sticking point for most casual DIY users who probably don't know much about the software or whether their results are realistic. Perhaps you could check bimo's threads, posts and Indonesian website for the LTSpice simulations of his designs. He may have some practical advice to share.
It is pointless though, to talk about simulations if we are using generic parts or inaccurate models though. They may be functional with DC voltages as expected and be acceptable in many non-critical applications but will probably be different enough in AC performance to the real parts, to call them fakes. Unless they are functionally equal to the original part, you are wasting your time trying to make useful simulations for audio.
In regard to sound quality, note that Naim's early models do not conform to good design principles. They are modified (tweaked) to sound nice and to have that pace, rhythm, timing effect but not to minimise THD, though it measures well anyway. Looking at the Nait i5 design recently posted, you'll see that most of JV's modifications that were part of the Nait1, 2, 3 and also the common NAP250 design which covers the other NAP models, have been removed, making it a more conventional sounding amplifier.
We have agreed that this is a disappointing development but it's what their present management wants to market and it's for them and their retail customers to support, not DIYs like me who will never buy expensive new products if we can avoid it.
Spice and LT Spice in particular, is actually a very good circuit simulation tool for audio, as the many high-level designers and engineers who post here will testify. The results it predicts are often more useful than actual measurements though you need to know how to use it, which is the sticking point for most casual DIY users who probably don't know much about the software or whether their results are realistic. Perhaps you could check bimo's threads, posts and Indonesian website for the LTSpice simulations of his designs. He may have some practical advice to share.
It is pointless though, to talk about simulations if we are using generic parts or inaccurate models though. They may be functional with DC voltages as expected and be acceptable in many non-critical applications but will probably be different enough in AC performance to the real parts, to call them fakes. Unless they are functionally equal to the original part, you are wasting your time trying to make useful simulations for audio.
In regard to sound quality, note that Naim's early models do not conform to good design principles. They are modified (tweaked) to sound nice and to have that pace, rhythm, timing effect but not to minimise THD, though it measures well anyway. Looking at the Nait i5 design recently posted, you'll see that most of JV's modifications that were part of the Nait1, 2, 3 and also the common NAP250 design which covers the other NAP models, have been removed, making it a more conventional sounding amplifier.
We have agreed that this is a disappointing development but it's what their present management wants to market and it's for them and their retail customers to support, not DIYs like me who will never buy expensive new products if we can avoid it.