Hi,
Thanks for the clarification.
Maybe the measurements are partially veiled by the digital processing, the windowing function, the quality of the sound card etc. But as far as you compared apples to apples they give good idea.
I was thinking about loads simulating a real life 2-way speaker similar to those given in D Self book. I don't really care for the load to be by any standard because it will be only for home measurements. One day I'll buy parts and will put them together on a piece of wood. Resistive load gives the option for people anywhere in the world to compare their results. Unfortunately, to my knowledge, not many speakers are even close to resistive load. I've heard about Magnepan only to be like that.
The circuit in post 820 shows Q7 base voltage is less negative than the emitter which may seem counter to the expectation of seeing a positive voltage on the base of an NPN transistor - in those terms Q7 base voltage is more positive than the emitter.
The 22k resistor in the collector circuit will push the base voltage down a bit and some of the charge on Q7 base will be due to Miller effect between collector and base.
Use of tantalum capacitors has been decried widely and using such in the decoupling arm of a nfb network would seem unwise.
This part of the circuit has a critical role since it outside the influence of the nfb loop and in itself is a separate signal source since it passes a.c. signals to zero volts. Quality parts should be used for the decoupling arm.
Is Mr Verekers choice of tantalum and the influence of Miller effect on Q7 base charge one that has a more benign outcome than expected.
The 22k resistor in the collector circuit will push the base voltage down a bit and some of the charge on Q7 base will be due to Miller effect between collector and base.
Use of tantalum capacitors has been decried widely and using such in the decoupling arm of a nfb network would seem unwise.
This part of the circuit has a critical role since it outside the influence of the nfb loop and in itself is a separate signal source since it passes a.c. signals to zero volts. Quality parts should be used for the decoupling arm.
Is Mr Verekers choice of tantalum and the influence of Miller effect on Q7 base charge one that has a more benign outcome than expected.
Last edited:
Member
Joined 2009
Paid Member
I once used an old blue tantalum in my TGM3 amplifier (thread around here somewhere) and I found it way nicer than a small film cap on the input. The film cap was a bit harsh and I now prefer bipolar electrolytics as the best choice. But the blue tantalum did give something subtle to the sound. I have no doubt that they are important for my NAP 'clone' otherwise I would have avoided them as unwise too.
Member
Joined 2009
Paid Member
A few more parts soldered in today.
Question is, to output inductor or not to output inductor, hmmmmm.
I prefer to install one as I have no interest in worrying about speaker cable choices - I won't be using Naim cables.
Question is, to output inductor or not to output inductor, hmmmmm.
I prefer to install one as I have no interest in worrying about speaker cable choices - I won't be using Naim cables.
Same advise from Greg B. only he reduced the resistor to reduce the ringing if there is any.
To those who asked me if I built the Greg variation no not yet. I built 2 pair Naim from Ebay after I used NCC clone to me sound better ( instruments are more precise and much more air in the mid section and deeper more controlled bass)
I am on the way to test the Greg variation sometimes latter.
Member
Joined 2009
Paid Member
Yes, I think the output inductor is the wise choice here, hopefully any audibility is limited to above my current hearing range. It's the mids that I anticipate being key in this project.
I ended up as you suggested but my fdbk cap is 100u not 68u.
I ended up as you suggested but my fdbk cap is 100u not 68u.
Last edited:
A coil isn't authentic but its a long way better than no coil or at the prices asked for recommended lengths of genuine A5 cable. If you don't plan on an all-Naim system, just fit a coil and relax. There will be situations in testing and use where cables will be too short or inconvenient for their inductance to work as intended anyway.
For fence-sitters, I found this amusing reviewer's story too 😉: Listening #80 | Stereophile.com
In the days before A4 or A5 cable Naim used RS 2.5 mm 50 strand twisted 2.5 turns per metre. I suspect the one in the link is the exact same one. As the cable is nothing special shop for price. A5 sounds much better. Mostly because there are fewer strands. If not into the last drop of grunt 0.6 mm single core is in some ways better. A4 I strongly dislike, sibilant and dirty. The RS is the same so don't rush to buy it. QED79 is bad in other ways ( generic is fine for the price, split and twist it ). 1 mm solid for lighting stripped and 2.5 turns per metre is not bad. Be gentle when doing it, it work hardens or whatever which to my ears sounds worse. The big sound thing is a balance of inductance and capacitance ( not too tight ). Minimum 3.5 metres per side and if following Naim's advice equal lengths. The latter can cause problems so do the 3.5 metre minimum as prime. There is no harm going to 0R33 if using less cable. In fact TRY 1R, you might be in for a surprise. It might if well selected be a bit like good valve amps. Theory is the resistance helps the amplifier see the speaker back EMF as a nicer thing. One way to keep a bit of RF out of the amplifer is to decouple the VAS in some way. Simple RC filtering would do something.
| RS Pro Red Tri-rated Cable, PVC, 600 → 1000 V, 2.5 mm CSA 100m |
Naim really didn't like output inductors to say the least. If you remember I mentioned Horns of Oxford. Geoff Horn reviewed the NAP 160( ? ) with Quad ESL57's in Gramophone mag. He really liked what he heard. Mr Horn was a subcontractor to Quad building the AM tuner so had a bit of a conflict of interest. Geoff put a 0R22 resistor in the output of a Quad 405 and felt the difference was then minimal. Geoff concluded that the resistor could be better. Self points to this a bit and says the ringing seen on square waves is the choke and not the amplifier in most cases. He then adds various damping resistors. Myself I am not sure if that's ideal. Only my ears to tell me on that one. I don't dislike chokes. The do change the sound ( slightly more boring for want of a description, magnolia sound ). I usually use 16 turn of 1mm copper on a 8 mm drill as mandrel. It seems a good universal strarting point. I ty-rap the coils before soldering.
I notice some 1N4000(7) type diodes in some CCT's. I think it is to get slightly nicer clipping. If so that's a touch of genius. However 3 would be better. It might do other things when back EMF is about.
| RS Pro Red Tri-rated Cable, PVC, 600 → 1000 V, 2.5 mm CSA 100m |
Naim really didn't like output inductors to say the least. If you remember I mentioned Horns of Oxford. Geoff Horn reviewed the NAP 160( ? ) with Quad ESL57's in Gramophone mag. He really liked what he heard. Mr Horn was a subcontractor to Quad building the AM tuner so had a bit of a conflict of interest. Geoff put a 0R22 resistor in the output of a Quad 405 and felt the difference was then minimal. Geoff concluded that the resistor could be better. Self points to this a bit and says the ringing seen on square waves is the choke and not the amplifier in most cases. He then adds various damping resistors. Myself I am not sure if that's ideal. Only my ears to tell me on that one. I don't dislike chokes. The do change the sound ( slightly more boring for want of a description, magnolia sound ). I usually use 16 turn of 1mm copper on a 8 mm drill as mandrel. It seems a good universal strarting point. I ty-rap the coils before soldering.
I notice some 1N4000(7) type diodes in some CCT's. I think it is to get slightly nicer clipping. If so that's a touch of genius. However 3 would be better. It might do other things when back EMF is about.
A warning from history.
Monitor Audio had what was called at the time a Litz type cable that looked like a woven snake. It looked fantastic and set the Naim amps on fire. Naim very quickly gave very accurate reasons. They specualted that the " bad sound " of commersial amps could be the output choke. The MA cable would detune the choke to give back some vitallity. You can choose for youself what you want to believe. My feeling is Naim were not wrong. Naims amps were super relaible so I feel they deserve to be believed. As they said a dealer seldom could remember more than one failed amplifier in many years. Many other makes never got near except Quad who were very slightly worse. Quads were used in more industrial set ups so in truth might have been the more reliable. Both were excellent and showed Sony how to do it. TA 5650 was Sony's best when they worked. Better than? I think so.
Monitor Audio had what was called at the time a Litz type cable that looked like a woven snake. It looked fantastic and set the Naim amps on fire. Naim very quickly gave very accurate reasons. They specualted that the " bad sound " of commersial amps could be the output choke. The MA cable would detune the choke to give back some vitallity. You can choose for youself what you want to believe. My feeling is Naim were not wrong. Naims amps were super relaible so I feel they deserve to be believed. As they said a dealer seldom could remember more than one failed amplifier in many years. Many other makes never got near except Quad who were very slightly worse. Quads were used in more industrial set ups so in truth might have been the more reliable. Both were excellent and showed Sony how to do it. TA 5650 was Sony's best when they worked. Better than? I think so.
Here are a few musings about the revised NAP140. Don't like the choke.... but.
The red bit is usual. Forgive if reality is different or I didn't spot some current going other places etcetera.
I don't like the 2 x 47R to the long tail pair. The whole point of the NAP140 is to have the sound of a single transistor with the simple DC servo action of a LTP. This changes the amp to semi " blameless ". It also throws away half the loop gain ( ish ). The amplifier can cope with a little class A non linearity here. As the 33R to the VAS also throws away loop gain we possibly will have too little to correct spiky distortion of the class B biasing.
My speculation goes this way and it many miles from standard theory. If we take the Z in of the standard VAS it might be 200(25/10) = 500R. If for fun we pretended it needed to be 3 x 1K3 what might Re be? 19R5. As we have 2R5 in the device 15 or 16R seems OK. Although this is not a 1947 valve amp saying it is of no importance is proved wrong in my previous post where 47R produced less distortion ( Remember my amp is class A so crossover distortion is unlikely in normal use until the load forces AB ). This is 100% against standard thinking. The graphs don't lie. If you see I give 12 watts where it is AB and still nice.
What you will find is the RE if 0 will give more punch ( usual NAP140 ). When 33R it will be sweeter. 4R7 might be what you want. The 22pF might be right with 33R, At first don't adjust the LTP. Keep to Naim values. Tweak at the last minute the LTP as this is about warmth of sound and is not make or break.
As I say I don't doubt my maths will go west somewhere. That's for you to say. My thoughts are how the gear ratios work.
BTW. My class A amp had no measurable distortion above the oscillator when a Complementary feedback pair (Cfbp ) either with OPA 604 op amp or one transistor input and one transistor VAS. This implies we might be talking better than -80dB ( or even better ) at 50 kHz. Rosen states the input needs to be -20dB better to even begin to speculate. As no signal source is, my - 68dB is OK. Notice there are no trends on my better version at - 66dB 50 kHz. Even if you can't measure below a certain figure there are trends.
If you look back to my graphs 47R makes the loop work. The Re= 0R ( VAS ) distortion is the FET's as the Cfbp had non in the same circuit. Cfbp had no obvious improvement feeding the output stage distortion back into the VAS. Furthermore the OPA604 with Cfbp only showed mild distortion with no loop feedback. Would someone like to speculate as to how the 47R caused the FET distortion to be reduced whereas the Re = 0R didn't? My specualtion is the current of the output stage is the better tool and Re insures it works. VAC Rc is my guess. If I had not put in the FET's I would never have known. I did is as the bias of the Cfbp is a complication the FET's don't have if the heatsink large enough.
If you look back to my graphs 47R makes the loop work. The Re= 0R ( VAS ) distortion is the FET's as the Cfbp had non in the same circuit. Cfbp had no obvious improvement feeding the output stage distortion back into the VAS. Furthermore the OPA604 with Cfbp only showed mild distortion with no loop feedback. Would someone like to speculate as to how the 47R caused the FET distortion to be reduced whereas the Re = 0R didn't? My specualtion is the current of the output stage is the better tool and Re insures it works. VAC Rc is my guess. If I had not put in the FET's I would never have known. I did is as the bias of the Cfbp is a complication the FET's don't have if the heatsink large enough.
Better than that, use an old radio AM ferrite rod to form a coil with 7-10 turns with the output speaker cable. This will annihilate any RF entering into your precious amplifier 🙂
Gareth, grow a pair and ditch that coil. I haven't used one since the 1970s and only then because I copied someone else's circuit. 😉
If you are using Litz (!!!) then solder a coil to the end of your cable. If you are using something like Litz then you deserve what you get! 😀
BTW, did you decide to use a 0.22 series R or use the R of the FETs instead?
If you look at my post 813 I think I know why the distortion has near equal harmonics. The distortion type is mostly exponential in harmonics one would think. The powers of the feedback loop are dying exponentially also. Thus the outcome. As said before the Cfbp has none so it's the FET's that introduces this. The mystery is that reducing the loop gain has allowed the VAS feedback to work ( I speculate collector resistance ). Cute? That 47R should be a disaster and it is not. For a class AB with bipolar outputs I don't assume the same.
Member
Joined 2009
Paid Member
ha! over the past weeks I collected some 'wisdom' from this forum. Here's what I kept of it:
QUOTE=kevinkr ;
Actually RF very frequently couples effectively into the low impedance speaker wiring, your amplifier output may not look like a low impedance (and your speakers don't either) at a couple of MHz and above where even a few inches of wire may have an effect. Often rf picked up on the output wiring will couple through the feedback network particularly if there is a cap across the feedback resistor right back into the input stage which in most cases will rectify it, resulting in unwanted interference with the audio signal at the output.
Jfet based inputs generally don't exhibit this problem to the same degree as bipolar inputs. (no intrinsic diode junctions and higher slew rates, etc.)
This was an all too common problem at the large mid-fi manufacturer I spent most of the last decade at. Extensive testing was done in RF shielded chambers to mitigate this problem.
The inductor is there primarily to isolate capacitance from the output and feedback loop, but it may stop some rf from getting back to the input stage of the amplifier.
QUOTE=Eva;
By the way, some modern amplifiers lack a RL network in series with the output to reduce costs and save space (and you may see people complaining about them oscillating when things like exotic speaker cables are employed). Anyway, adding that RL network is a good design practice because it prevents external stuff to interact with the amplifier above audio frequencies and leaves it only loaded with its internal RC network thus yielding consistent RF behaviour regardless of the load.
Concerning air cored output inductors, they will obviously produce a stray magnetic field proportional to the loudspeaker current and to its frequency, that will obviously cause some degree of feedback over the small signal circuits and signal wires. So these inductors should be placed away from the amplifier circuit itself, in the output binding posts for example. A feet or so (30cm) is already a good distance.I arrived at that number by measuring (or no longer being able to measure) the maximum voltage induced in a 10cm diameter loop of wire at 10Khz (that as you have mentioned, is orientation dependent). You can't realiably measure the crosstalk in the own amplifier because the feedback effects are taking place inside it and there may be positive or negative feedback, or a mixture of them over various stages. I was also able to measure induction with improper wiring, but that's another story.
QUOTE=Bobken;
Otherwise, having carried out numerous *listening tests* on similar areas, myself, in my experience you will *hear* far greater effects from the type of components used than with any minor variations in the square-wave overshoot at say 10kHz, or above.
For good reason (Luminaries like) D.Self, B. Duncan, and JLH all state that the output inductor should be air cored (and subjectively I found this to be true) and two of the above who also comment on the sonic effects, agree that these effects (which are always to give a ripple on the square wave) are of no consequence, sonically.
In my own experiments, I achieved better results with regard to overall stability and 'clean' square waves (initially without any output inductor//damping resistor) through varying the compensation caps. Then, where a *suitable* damped output inductor has been added later, I could hear no difference to the sound, at all.
However, experiments with different components (of the same nominal values) being used for the cap and resistor of the output Zobel, showed marked changes to the overall sound. To show how seriously I take this, I have ended up using $30 worth of components here for this Zobel alone, in a stereo amp.
Such expenditure nearly breaks my heart, but every alternative component I tried out had some different (and more adverse) effect on the outcome, and my present best choice here is to use MIT RTX (multiple-section) polystyrene & tin-foil caps, together with a Caddock resistor.
Varying the values of these (within reason) has little (if any) sonic effect, but changing to an alternative type/make of component alters the sound by an unintuitive degree, but which I cannot ignore.
Running for a short while without any Zobels, sounded even better still, but I am not prepared to take a chance on this on a long-term basis.
I'm using the RCX160N20 FETs which have an Rds of ~110mR. So I'll be seeing enough resistance through these FETs to provide equivalent desired output resistance. And it's quite constant over a wide current range so I'm not anticipating any detrimental impact on sound quality.
"your amplifier output may not look like a low impedance"
To make a serious attempt to stop RF you would need some proper measures like RF chokes and shielding and such. An air-cored inductor with a few, close turns shunted by 10-ohms won't do anything much. But hevan knows inductors are naughty. 😉
I like the FET resistance...very elegant.
To make a serious attempt to stop RF you would need some proper measures like RF chokes and shielding and such. An air-cored inductor with a few, close turns shunted by 10-ohms won't do anything much. But hevan knows inductors are naughty. 😉
I like the FET resistance...very elegant.
- Status
- Not open for further replies.