TGM10 - based on NAIM by Julian Vereker

Status
Not open for further replies.
I seem to remember JLH (or was it Cordell?) who remarked that most amps sound very much the same until they reach a fault, such as a transient oscillation, or sticky clip, or loss of fb control at some point.

But these are technical, engineering issues. The elephant in the room is the way psychoacoustics intrude; what does the ear 'hear', to what extent is harmonic profile, how does the ear pick up 'time shearing' and why do spatial cues disappear on some amps and not in others..... in my work these are the dominant questions, because I see the engineering as relatively straightforward, not that it's easy.

I have enjoyed the NAIM approach for years, and first saw it close up in the Avondale circuit which is commonly available on the net. I applaud and thank Bigun's interest in it here.

Hugh
 

Attachments

  • AvondaleNCCamp.jpg
    AvondaleNCCamp.jpg
    191.4 KB · Views: 437
Simple amplifier with "sophisticated" compensation

I'm not so far with listening and understanding the different ways of compensation. One designer here has a complex compensation too and I disagree with his choices. Bigun also uses more than simple compensation (I think I will study his TGM1 soon).

[Even when using fast devices I always look for no-compensation possibility]

So what are your goals when you select for certain compensation scheme and value?

Answers to questions like this might answer the untold story about what make a good amplifier beside the already known criteria like THD. If you like high slew rate, you might have your own threshold for this variable (And if you choose too high a number, many will disagree, btw)
 
So what are your goals when you select for certain compensation scheme and value?

My goal is lowest THD at 20kHz as possible but still good harmonic profile (monotonic) and high slew rate. Of course it must have enough gain margin and phase margin.

How about the sound? I don't know 😀 I will know when I build it and I will compare to my other amplifiers that I built it.

I will learn how amplifier specifications affect the sound or how do I interpret the specifications.
 
My goal is lowest THD at 20kHz as possible but still good harmonic profile (monotonic) and high slew rate. Of course it must have enough gain margin and phase margin.

Thanx Bimo. Is PM/GM the only stability measure/test that you use (and what are the thresholds here)?

How about the sound? I don't know 😀 I will know when I build it and I will compare to my other amplifiers that I built it.

I will learn how amplifier specifications affect the sound or how do I interpret the specifications.

😀 The accuracy of the conclusion is low then (and you'll never know precisely whether your assumption/guess is correct or wrong), because there are too many variables involved.

To limit the variable we should compare options/schemes/scenarios in that same amplifier circuit.
 
Thanx Bimo. Is PM/GM the only stability measure/test that you use (and what are the thresholds here)?
.

I was able make a stable amplifier with PM 60 and GM 9, but I likely have PM 70 and GM 13 minimum because I can use cheap parts for prototype.

😀 The accuracy of the conclusion is low then (and you'll never know precisely whether your assumption/guess is correct or wrong), because there are too many variables involved.

To limit the variable we should compare options/schemes/scenarios in that same amplifier circuit.

Yes, I did those several years ago.
 
Had you ever thought of trying lets say Exicon 10N/P20 or BUZ900/905 Mosfet's? I talked with JV when these devices first were talked about ( He asked me and I knew nothing about them then ). He was very interested. As the Naim sound is about rhythm maybe it wasn't his cup of tea. I noticed after this Naim became strongly associated with all NPN output stages. To maintaing the rhythm one might try 2 sets of FET's. Gate stoppers at about 220R. The reason this might offer something better is bias can be set by ear up to 100mA per set or even a little above that, 20 mA might be the lower limit. The VAS should drive them directly. Bias runaway should not happen with FET's of this type, Often a 470 R pot is all that required.

I doubt very much anyone will try this. All the same I am almost certain Naim did.
 
Nigel, as I recall, Naim (JV) did try, but did not like the prospects for using Mosfets in their amplifiers. As Hitachi Lateral Mosfets were at their most popular phase in the 1980s, I assume that's what he also referred to.

I don't think there is much similarity in the crossover harmonics spread of a BJT quasi compared to a hybrid so its likely you are considering something that has quite a different sound character - certainly not in the realms of clone build anyway. Maybe better, maybe worse according to taste but I suspect you would struggle to get comparable sound.
 
As the Naim sound is about rhythm maybe it wasn't his cup of tea. I noticed after this Naim became strongly associated with all NPN output stages. To maintaing the rhythm one might try 2 sets of FET's. Gate stoppers at about 220R. The reason this might offer something better is bias can be set by ear up to 100mA per set or even a little above that, 20 mA might be the lower limit. The VAS should drive them directly.

Nice. This is what I thought/guess: rhythm.

I doubt very much anyone will try this. All the same I am almost certain Naim did.

If Naim did try but didn't release the result, it could be because it is expensive or it is simply not better.

I have tried quasi LATFET and the result is in general good (better than non-LATFET) but my objections have been: (1) Is it good enough compared to complementary design? (2) I need to design a quasi P-Channel to make use of the leftovers. This is fine as I will use it in my active system. BUT this is IF both designs can reach at least 80% of the performance of my complementary design, which I assume is unlikely 🙁

I don't think there is much similarity in the crossover harmonics spread of a BJT quasi compared to a hybrid so its likely you are considering something that has quite a different sound character - certainly not in the realms of clone build anyway. Maybe better, maybe worse according to taste but I suspect you would struggle to get comparable sound.

This is what I was talking about, that the "thing" with the Naim might have nothing to do with whether it is LTP or not, balanced LTP legs or not. If Nigel thinks this is "rhythm" then it might be achieved with any other topologies. Of course, I'm not 100% sure what Nigel meant with "rhythm". Then we need to know what this is technically...

My prediction is that this "rhythm" can be achieved with linear system, that's why LATFET could fit in. But I think I read somebody mention that Naim's output transistor is not linear?? But may be a transistor is linear only in region we will not operate it in.

I also found that to have a good "rhythm" the speed of the music (all instruments) shouldn't be too fast. Strangely, fast (and low damping) transistor often create a bass sound that is too fast. That's why I doubt 2SC5200 can make it. Sankens such as C2922 might be faster(?) but in my experience it can create slower and more defined bass than C5200.
 
dd, Naim used a number of different output transistor types in the same models over a period of more than 30 years - That's plenty of time for anyone to appraise any differences in the sound details like their quaintly titled "pace-rhythm-timing" sound qualities.

In the early models, Vereker used the fastest switching transistors available because in principle, speed is essential for the smooth crossover of class B amplifiers. That was until it was realised that you could reduce the stored base charge of some BJT output designs in other ways and even faster, more linear Japanese audio devices began to appear. He eventually changed transistors to quite different types of device in the excellent Sanken LAPTs. The cheaper Toshiba perforated emitter types also work but few single pair devices have the power dissipation and ruggedness required for Naim's range of models. Nait models BTW, used very ordinary power/switching transistors but users were still quite impressed - having all the elements of the desired sound character.

My point is that the BJT type can't be all that critical if 4 or more very different transistor types can be used in the one model without other changes and without losing key sound qualities. In other words, the desired sound quality or distortion components contributed by the output stage of NAP models, essentially derives from basic BJT characteristics.
 
Last edited:
My point is that the BJT type can't be all that critical if 4 or more very different transistor types can be used in the one model without other changes and without losing key sound qualities. In other words, the desired sound quality or distortion components contributed by the output stage of NAP models, essentially derives from basic BJT characteristics.

One would think so however as far as accurate simulations go that can depend on which .model is used. Bob Cordells .models differ from those on On-Semiconductor website.

According to the latter VAF (Early Voltage) for MJL21194 is 29.66 whereas Bob Cordell gives a phenomenal value of 500.

BF(Beta Factor) are close - values 62.4 and 70.

The equivalent .models for MJL3281 for VAF at 63.7 and 1000 are poles apart (BF respective values 63.7 and 158).

Can anyone with ppm simulation experience shed some light on this puzzle.
 
I also found that to have a good "rhythm" the speed of the music (all instruments) shouldn't be too fast.
That's my experience too. It is a symptom of circuit problems. The sound should be relaxed when it should be and then take you by surprise. When the emotional dynamics get compressed or confused the music can sound too fast and ill at ease. Soon you lose interest, even if the sound is pretty clear.

Years ago I heard an uncanny demonstration of a Linn Sondek turntable. After listening on a lesser but still very capable deck the Linn sounded more involving but also slower. It wss not slower at all of course...the track lasted the same time by a watch. But it sounded slower...like I had been listening for a longer time.
 
The Sinclair Z30 might be one origin point for the Naim design. Not the same. Just what was typical of the era. Nothing looks much like the Sinclair before it arrived. The RCA design is not really the same.

The way the Sincair was biased was to have a fixed resistor and use the current source as the heat sensor. That is not a lot worse than how Naim did it in terms of how accurately it tracked the output. Theory has it the Naim's needed to be left switched on to be in the same set up as when they were on the test bench at Naim as the drift was less than optimum. It is very likely one of Douglas Self's better Vbe bias devices would track better. Not sure if he did one for Quasi.

This might have been the other origin of the design. Page 439/440.

http://www.keith-snook.info/wireles...973/An approach to audio amplifier design.pdf
 
The Sinclair Z30 might be one origin point for the Naim design. Not the same. Just what was typical of the era. Nothing looks much like the Sinclair before it arrived. The RCA design is not really the same.

http://www.keith-snook.info/wireles...973/An approach to audio amplifier design.pdf

While the Naim circuit looks nothing like those in RCA application notes Sinclair produced amplifiers in module form.

As it happened RCA also produced amplifier modules.

These were H2000H and HC2500 both Quasi-Complementary types - described in RCA Power Device Handbooks as "Multi Purpose 7 Ampere Operational Amplifiers".

The internal circuits are given in the Handbooks - mine a 1978 issue.

If JV claimed to have been inspired by an RCA design, I suspect these modules would have been a strong if not compelling influence.
 
Alan Mornington West wrote me a slightly angry email playing down how important the Naim was as a design and his part in it. Sad really as he needs to take credit for a good design. Alan did say it was a completely original circuit without reference to any other. If you read the email he seems the last person in the world to lie.

I really like Bob Stewarts P445 design from my link in 397. I do know that JV asked Bob many questions before starting to think out a Naim design. Bob said to me " I suggested Julian sorted out a Sinclair and give it a big power supply . I didn't exspect him to sell that very same thing" ! Of Bob's P445 design the bootstrap VAS currnet source is the main difference ( I really like them ). Note the Baxendale diode. The phase shift network I suspect were Alan. I strongly feel the RCA is no more than a thing of it's time and resembles the other designs of it's time. See Cavendish Amp example here. The RCA is some ways is doing things I have never seen then or now. Not worth discussing. Just very much it's own thing. The Cavendish is not exactly like the RCA I remember, it is like the Bob Stewart design.

A Paul Kemble web page - compact PA designs.

BDY56 transistors were very fast at that time. I think the arguement was very strong to use Quasi back them.

Someone said Naim amplifiers are 90% power supply. I think that might be the fact that some never get. The power supply can never be too good. The fastest BDY56 were used in the NAP 250 PSU's. As JV said the PSU has to be the fastest part of the amp, if not don't bother. One of JV's other statements was that the Naim amps were often cloned even then. He was surprised to say none sounded like his amps. His feeling was the topology of the amp was not really what a Naim amplifer was, or not all it was.
 
Status
Not open for further replies.