The 68uF capacitor, as you can see, is an AC path for the voltage divider in the negative feedback path from the output back to TR2 base. This divider determines the AC gain of the amplifier. In this case, the gain ratio is set by the 1k series resistor and 27k in the feedback path. The capacitor also limits the lower bandwidth which should be chosen to be lower than that imposed by the input cap. Strange effects can occur with different capacitor types if you confuse these priorities.
Increasing the cap. to 100uF would not affect much more than cost and size but could allow you to experiment with other types as you already have.
As long as the cap's polarity matches any small DC measured across the cap, you should be able to use almost anything, remembering that tantalum caps, for example, won't tolerate more than about 1.5V reverse polarity.
Technically, the correct type of electrolytic to use for audio is a non-polarized type (not "bipolar"). This isn't Hi-fi though and probably not intended to be in the original design but many people like to listen with effects like using a wrong component type and say they sound better or worse etc. according to what they like to hear. Fine, have fun with this.
If you want to read about amplifier design and what causes distortion, Douglas Self's and Bob Cordell's Design handbooks are excellent for learning why some choices are made. If you don't want to buy books, there are a few summaries here: Distortion In Power Amplifiers
Increasing the cap. to 100uF would not affect much more than cost and size but could allow you to experiment with other types as you already have.
As long as the cap's polarity matches any small DC measured across the cap, you should be able to use almost anything, remembering that tantalum caps, for example, won't tolerate more than about 1.5V reverse polarity.
Technically, the correct type of electrolytic to use for audio is a non-polarized type (not "bipolar"). This isn't Hi-fi though and probably not intended to be in the original design but many people like to listen with effects like using a wrong component type and say they sound better or worse etc. according to what they like to hear. Fine, have fun with this.
If you want to read about amplifier design and what causes distortion, Douglas Self's and Bob Cordell's Design handbooks are excellent for learning why some choices are made. If you don't want to buy books, there are a few summaries here: Distortion In Power Amplifiers
Last edited:
The Naim is a quite a bit different from most "non musical" amplifiers.
A normal amplifier must ensure that any signal passing remains unaltered and is simply scaled to the new output voltage.
This REQUIRES that the amplifier properly amplifies any signals sent through.
This REQUIRES that the NFB DC blocking cap does not have any significant AC voltage applied to it.
You do this by ensuring that the NFB DC blocking capacitor has a Low Frequency bandwidth that is lower frequency than any signal passed at the input filter.
This is competant amplifier design.
Fit a capacitor that is too small and you will get "processing" of the signal at this capacitor.
For a 1k lower leg resistor, I suggest you use a 150uF or 220uF. This gives an RC of 150 to 220milliseconds.
Then you should find that any make of electrolytic has no "sound effect" on the signal being passed to the output.
But at my intro, I said the Naim is different.
A normal amplifier must ensure that any signal passing remains unaltered and is simply scaled to the new output voltage.
This REQUIRES that the amplifier properly amplifies any signals sent through.
This REQUIRES that the NFB DC blocking cap does not have any significant AC voltage applied to it.
You do this by ensuring that the NFB DC blocking capacitor has a Low Frequency bandwidth that is lower frequency than any signal passed at the input filter.
This is competant amplifier design.
Fit a capacitor that is too small and you will get "processing" of the signal at this capacitor.
For a 1k lower leg resistor, I suggest you use a 150uF or 220uF. This gives an RC of 150 to 220milliseconds.
Then you should find that any make of electrolytic has no "sound effect" on the signal being passed to the output.
But at my intro, I said the Naim is different.
Thanks Ian for your answer.
@Andrew: i'm not sure if i understand to wich 1k "lower leg" are you talking about.
When you say 150-220uf elcap are you talking about NFB cap ?
@Andrew: i'm not sure if i understand to wich 1k "lower leg" are you talking about.
When you say 150-220uf elcap are you talking about NFB cap ?
If you read the vast info in the links I gave ( 1615 ) the capacitor sound does relate to distortion at the limit of a 16 bit system. Sony found the same with vibration ( circa 1979 ). This shows how little we really know. The good news is the trends are reasonably predictable. Non polar caps were the big surprise. Very cheap and very good. 63V polyester not so good, 100V noticably better. I bought a vast number of 33 uF NP 63V from Rapid Electronics at 3 cents each. I re-caped a Rotel 931 with them and a few other NP types. The difference is noticable. One thing to note. Keep the peak voltage below 0.4V with any electrolytic ( see 1615 ). This might explain to me why high gain power amps sound good. There is a critical point where feedback is enough and the cap is not sent above 0.4V at up to 5 watts/8R. If this happens below 6 Vrms output to speakers it should be fine. A gain of 56 like the Quad 405 is a good choice. It means a simple volume pot and CD player is enough to work. The Naim I have seen here is about gain of 28. You could change the 68uF to 100uF 100V NP as I did on the Rotel. The 1K to 680 R. This will give a gain of about 41. DC off set will will rise a little. If below 100 mV it is not very important as far as speakers are concerned, my guess 30 mV typical. I suspect gain changes are as important as capacitors ( more ???). Don't lower the gain as that is risking things. If you increase the gain you might find to your ears it sounds more open. If you go too far hiss and fizzy treble shows through. By then it will be very loud at near zero volume. Real life Naim's were like that with a NACA4 preamps. It claimed 75mV for full power via line level. As a CD player claims 2V that is a vast mismatch. 250mV seems about right and the 500 mV of the Quad 405 is fine. Some claim the Quad is made to suit CD. Not true, the 1967 Quad 303 was 500mV also. The Quad 33 radio input 100mV. It is run from a lower voltage than Naim preamps so is not happy with CD.
100 %.
I reworked my examples to give it original low frequency performance. You might try even less low frequency by not changing the 68 uF. If the PSU is modest it can help the amplifer, less boomy. If you add 2K2 to the 1K associated with the 1K + 68uF you will have a gain of 40 near enough. 1K + 2K2 in paralell is 688R. You can try 1 K also. The capacitor should be left unchanged at first as it might be prefered anyway. I fit 100V types if I can as these most often survive the output devices failing. If non polar that helps even more. Usually the speakers do not do so well. High voltage capaciitors sound better because they are larger. What makes a capacitor survive high voltage also gives it better tan theta ( as a rule ). Not bad for 10 minutes work to know so much more.
Non musical and musical. Yes, but Naim is not coloured except in capacitor choices I feel. When film Kodachrome had enhanced red. Naim has preserved red, it's not Kodachrome. A&R A60 has slightly less red than the signal. I suspect the A60 makes up for it in other sections ( VAS?). The Hitachi MOS FET amp is balanced to 2% without a current mirror on input stage ( almost no second harmonic ). The MOS FET's have second harmonic bias to make it sound right.
I think it is only the figure of speech "lower leg" that is confusing. I know it means the relative position of the two resistors on the relevant circuit diagram but it is introduced here without qualification. Even a translation program will likely botch that.
Ian offered this Avondale version of the Naim. The lower leg ( or arm ) is red. I give the gain as 1+{ 27/1 } = 28 for the shown circuit. That would be at lower frequencies to a reasonable approximation ( the 47pF + 2K2 come in at high frequencies ). If a scope is available one can reduce the 47pF to TR4 to restore loop gain at high frequencies. This should restore the vital HF correction ( > 8 kHz ) . If the gain were doubled one would exspect 27 pF to be OK. 33pf at TR4 should be alright for gain I show. However leave it at 47pF for a simple test. The associated capacitor on the lower leg ( 47 uF here ) can be changed to suit taste. Some systems would prefer it smaller! An underdamped ported speaker for example. That is most ported speakers if truthful. Self argues make this capacitor as large as possible. I tried 1000uF 16V NP with good results. I then used a small film cap at the input. 1uF if so would be 6.6 Hz ( ref 24 K input ). 220uF 63V NP and 4u7 100V polyester input cap would seem OK.
I must question the Avondale 2 x 100 R on TR1+2 . I have tried to attempt a better version ( and failed ). The problem with adding 2 x 100R is this destroyes the loop gain of the amplifer if the TR3 620 R is unchanged from Naim. Loop gain is the only thing we might attempt to repair crossover distortion with. We see this in action when DC offset, one can see it is far from perfect. As said we can restore the loop gain at HF by reducing the TR4 47pF. My experiments suggest this is the better route to take. By the way. Slewing distortion so very badly explained is the crossover distortion showing higher up the waves at high frequencies. This is the loop gain running out. it sounds dreadful.
One can take a different route and this is for the very very brave. Make TR4 47pF larger until the amplifer becomes unity gain stable ( add 220 pF at a guess ). Give the amplier a gain of 3.5 ( 11 K lower leg ). This would need that the preamp give 7 Vrms which is perfectly possible. I suspect some great measurements could be had when optimum ( TR4 cap ). I suspect the sound would be mildly worse than usual. For a sub wooffer it might have something. DC off set might be 2 mV ?
If we calculate the TR1+2 re of the Naim set up it is 26/0.7 = 37R ( re = 26/ic[ mA ] ) and on the 22K side 26/0.3 = 87R. Traditon has it that the internal emitter resistance is as good as a real resistor in making a circuit linear ( it isn't really ). We call external resistors Re when we add extra resistance. I would argue the input pair of Naim is within reason linear and it biased towards a fake single device curve. Naim has slightly less second and fourth harmonic distortion than a real sigle device. It can be argued a single transistor is more like our ears in how it orders the sound. That being true any " unharmonic " series will be heard as a new distortion. A analogy with how the eye sees black would not be too wrong.
I reset the input pair balance as otherwise there is no point in doing anything to the input pair ( TR1 + 2). What on Earth are you doing Les?
To restore a Naim sound one could add a small emitter resistance to TR4 ( adds second harmonic and looses loop gain, make the TR4 easier to drive into the bargain. It adds the distorion by allowing the VAS distortion curve in ). That also would offer a chance to reduce the TR4 47 pF to restore upper frequency loop gain. Douglas Self is very against this idea. He seems to think reducing distortion to - 100 dB at 500 Hz is a valid reason. OK reducing distortion broad spectrum to -80 dB ( 0.01% ) is throwing away a potential - 20 dB of book specification. One might say 0.001% @ 1kHz ( might is a word rather than reality ). Howerver the low frequencies being reduced from - 80dB to - 100 dB is like offering a sportsman a Zimmer frame walking aid. No need. Where as nearly all amplifiers start to need that is at > 10 kHz.
The reason a carefully changed gain amplifer sounds better is the critical lower feedback leg electrolytic capacitor ( next to red resistor 47 uF ) is being worked inside it's very low distortion range ( < 0.3 V rms which is 8 watts when Naim as 0.3 x 27 = 8V rms ). Raising the gain as I show raises it to 18 watts before the sound thickens. As DC offset raises it is less in practice. Oscon on measurements I seem to remember were worse than non polars!
It is possible to change DC offset without using a ultra horrible op amp DC servo. It usually increases hum a fraction. No bad thing if it does as curing that is a general cure for concelled hum. That is hum that is OK at zero sound output, but grows under the music.
Must go get some food so forgive any errors of English or technical.
https://books.google.co.uk/books?id...as self . simple DC offset correction&f=false
Hopefully this link works. If you like it buy the book .
Hopefully this link works. If you like it buy the book .
Ian offered this Avondale version of the Naim. The lower leg ( or arm ) is red. I give the gain as 1+{ 27/1 } = 28 for the shown circuit. That would be at lower frequencies to a reasonable approximation ( the 47pF + 2K2 come in at high frequencies ). If a scope is available one can reduce the 47pF to TR4 to restore loop gain at high frequencies. This should restore the vital HF correction ( > 8 kHz ) . If the gain were doubled one would exspect 27 pF to be OK. 33pf at TR4 should be alright for gain I show. However leave it at 47pF for a simple test. The associated capacitor on the lower leg ( 47 uF here ) can be changed to suit taste. Some systems would prefer it smaller! An underdamped ported speaker for example. That is most ported speakers if truthful. Self argues make this capacitor as large as possible. I tried 1000uF 16V NP with good results. I then used a small film cap at the input. 1uF if so would be 6.6 Hz ( ref 24 K input ). 220uF 63V NP and 4u7 100V polyester input cap would seem OK.
I must question the Avondale 2 x 100 R on TR1+2 . I have tried to attempt a better version ( and failed ). The problem with adding 2 x 100R is this destroyes the loop gain of the amplifer if the TR3 620 R is unchanged from Naim. Loop gain is the only thing we might attempt to repair crossover distortion with. We see this in action when DC offset, one can see it is far from perfect. As said we can restore the loop gain at HF by reducing the TR4 47pF. My experiments suggest this is the better route to take. By the way. Slewing distortion so very badly explained is the crossover distortion showing higher up the waves at high frequencies. This is the loop gain running out. it sounds dreadful.
One can take a different route and this is for the very very brave. Make TR4 47pF larger until the amplifer becomes unity gain stable ( add 220 pF at a guess ). Give the amplier a gain of 3.5 ( 11 K lower leg ). This would need that the preamp give 7 Vrms which is perfectly possible. I suspect some great measurements could be had when optimum ( TR4 cap ). I suspect the sound would be mildly worse than usual. For a sub wooffer it might have something. DC off set might be 2 mV ?
If we calculate the TR1+2 re of the Naim set up it is 26/0.7 = 37R ( re = 26/ic[ mA ] ) and on the 22K side 26/0.3 = 87R. Traditon has it that the internal emitter resistance is as good as a real resistor in making a circuit linear ( it isn't really ). We call external resistors Re when we add extra resistance. I would argue the input pair of Naim is within reason linear and it biased towards a fake single device curve. Naim has slightly less second and fourth harmonic distortion than a real sigle device. It can be argued a single transistor is more like our ears in how it orders the sound. That being true any " unharmonic " series will be heard as a new distortion. A analogy with how the eye sees black would not be too wrong.
I reset the input pair balance as otherwise there is no point in doing anything to the input pair ( TR1 + 2). What on Earth are you doing Les?
To restore a Naim sound one could add a small emitter resistance to TR4 ( adds second harmonic and looses loop gain, make the TR4 easier to drive into the bargain. It adds the distorion by allowing the VAS distortion curve in ). That also would offer a chance to reduce the TR4 47 pF to restore upper frequency loop gain. Douglas Self is very against this idea. He seems to think reducing distortion to - 100 dB at 500 Hz is a valid reason. OK reducing distortion broad spectrum to -80 dB ( 0.01% ) is throwing away a potential - 20 dB of book specification. One might say 0.001% @ 1kHz ( might is a word rather than reality ). Howerver the low frequencies being reduced from - 80dB to - 100 dB is like offering a sportsman a Zimmer frame walking aid. No need. Where as nearly all amplifiers start to need that is at > 10 kHz.
The reason a carefully changed gain amplifer sounds better is the critical lower feedback leg electrolytic capacitor ( next to red resistor 47 uF ) is being worked inside it's very low distortion range ( < 0.3 V rms which is 8 watts when Naim as 0.3 x 27 = 8V rms ). Raising the gain as I show raises it to 18 watts before the sound thickens. As DC offset raises it is less in practice. Oscon on measurements I seem to remember were worse than non polars!
It is possible to change DC offset without using a ultra horrible op amp DC servo. It usually increases hum a fraction. No bad thing if it does as curing that is a general cure for concelled hum. That is hum that is OK at zero sound output, but grows under the music.
Must go get some food so forgive any errors of English or technical.
Ported speakers are just wrong. Trying to correct them with the amp characteristics is like trying to mend a broken bone with sticking plasters.
I don't understand your comments about voltage across the feedback cap... it's a short circuit at audio frequencies therefore there is (for a perfect cap) no voltage across it irrespective of the ratio of the feedback transistors.
The NCC200 has decent LTP balance (as per my measurements earlier). What do you want Les to explain?
Simple. It's invalid with the 1K 22K set up to attempt degeneration ( I feel ) . Long tail balance in this context is like removing the holes in Swiss cheese. The twist in that story is it almost happened. The Swiss farmers were told is wouldn't be allowed in future if new rules were followed. After Julian died so did his exact vission. I wouldn't want to say who is right or wrong. Just to say Naim as was seen by Julian had this quirk. The balance factors on twin motorcycles is a similar question. It all depends on what speed you cruise at what is best. The Naim isn't quite that dramatic in it's effect. I still say what was Les thinking of and I show how it might be better ( 0R option replacing 22K ).
To explain a little my main frustration with the mods of Les. The Naim amp has established stability set by TR4 cap ( 47 pF ? ). If this capcaitor is reduced a time will come when the stability changes and oscillation takes place. This is when the phase change reaches a positive feedback situation. Many things determine this. Making the 47pF the dominant pole we simplify the design problems.
Les adding 100 R to the either 37 or 87R internal resistance reduces the loop gain of the amplifer and the 2 nd harmonic balance. 100 R to 87 R isn't exactly swamping the TR2 re. The effects on distortion will be almost zero at 1 kHz. I would exspect the Les mod to increase > 10 kHz distortion and to allow the more nasty upper harmonics to rise up a bit. In my oppinion when tail current is 1 mA the effects of TR1+2 re is to make the stage nicely liinear. If current is doubled 2 x 27 R will restore the linearity.
Effectively Les has reduced loop negative feedback which is mostly a bad thing.
Les adding 100 R to the either 37 or 87R internal resistance reduces the loop gain of the amplifer and the 2 nd harmonic balance. 100 R to 87 R isn't exactly swamping the TR2 re. The effects on distortion will be almost zero at 1 kHz. I would exspect the Les mod to increase > 10 kHz distortion and to allow the more nasty upper harmonics to rise up a bit. In my oppinion when tail current is 1 mA the effects of TR1+2 re is to make the stage nicely liinear. If current is doubled 2 x 27 R will restore the linearity.
Effectively Les has reduced loop negative feedback which is mostly a bad thing.
Well it's interesting that no one who has complained about stability of the NCC whereas a man who services/repairs NAPs for a living has experienced a bit of fun trying to tame them:
Is this a better Naim 140 clone? - pink fish media
I must admit I like the sound of my own version of these amps better than the NCC, but I have not yet tried all the variations of emitter degen and ac balance in the LTP. I have however tried the 0.22R versus inductor/10R in the output and the "normal" output inductor is very much preferred in my system.
Is this a better Naim 140 clone? - pink fish media
I must admit I like the sound of my own version of these amps better than the NCC, but I have not yet tried all the variations of emitter degen and ac balance in the LTP. I have however tried the 0.22R versus inductor/10R in the output and the "normal" output inductor is very much preferred in my system.
As I said I don't want to say good or bad of anything. When I see Les again I will ask him what he had in mind. It might be he thought the Naim marginal on stability and thought is wiser? Myself I would not add the two TR1+2 100R. I would try >0<47R to the TR4 emitter ( rail ).
When I tried the resistor and output inductor I was a little dubious and went back to an inductor. The one I use is 16 turns of 1 mm copper on a 8 mm former. It must be a few uH if that.
When I tried the resistor and output inductor I was a little dubious and went back to an inductor. The one I use is 16 turns of 1 mm copper on a 8 mm former. It must be a few uH if that.
Tried both values 100r and 33r. Both seems stable. No other measurements performed.
Sent from my Nexus 5 using Tapatalk
Sent from my Nexus 5 using Tapatalk
It will become more stable as the resistor value increases. The problem is marginally less feedback is available to control HF distortion as the resistor value is made larger. The VAS capacitor ( TR4 ) could be carefully reduced in value to restore the loop gain at HF. This will not help at 1 kHz. The good news is 1kHz is not where there will be a problem. With 100R I would imagine 33 pF to be about right. That is assuming 47 pF was not changed when the 100 R was added. It might have been 68 pF. Hard to say.
As I said previously the feedback can be reduced by changing the 1K feedback resistor to 680R ( red in my dia ) to increase amplifer gain . If so the VAS cap might be reduced further. With the 100 R to both TR1+2 and the increased gain I would imagine 27 pF to be OK. The increased gain being to improve the feedback cap distortion by keeping it away from polarising. An electrolytic will behave much like a diode over 0.4V peak. Oscon whilst a prefered Audiophile capacitor did not measure well in the Wireless World tests ( I gave a link ) . WW seemed to prefer non polar electrolytics. Even the famed Black Gates did not better non polars.
As I said previously the feedback can be reduced by changing the 1K feedback resistor to 680R ( red in my dia ) to increase amplifer gain . If so the VAS cap might be reduced further. With the 100 R to both TR1+2 and the increased gain I would imagine 27 pF to be OK. The increased gain being to improve the feedback cap distortion by keeping it away from polarising. An electrolytic will behave much like a diode over 0.4V peak. Oscon whilst a prefered Audiophile capacitor did not measure well in the Wireless World tests ( I gave a link ) . WW seemed to prefer non polar electrolytics. Even the famed Black Gates did not better non polars.
I am using my HackerNAP (NCC200) with a 15uF PETP feedback cap as I wanted to reduce cone flap on my FrugalHorn XL's when playing vinyl a little, and am also using a 0.47uF PETP input cap, am I causing issues further up the frequency range?
Your re. figures are based on collector currents of .7 m.a for TR1 and 0.3 m.a. for TR2. These differ from those measured by Dave S of 0.499 m.a. and 0.49 m.a. for the Avondale NCC amplifier.
This raises the questions of what version of the Naim amplifier are you looking at and how can these differences be accounted for.
Using Dave S' measurements, re. for the Avondale circuit TR1 and TR2 would be 52R and 53R respectively. If the 100R external emitter resistors are say 1% there is no significant difference in the degeneration amounts between the two halves.
I have a Nait 5i and was a bit peeved when a 5i update appeared a few months later and was regarded by the Hi-Fi press as a subtle improvement.
There were a number of detail improvements however technically the negative feedback had been reduced although the means was not disclosed.
It could have been by emitter degeneration for which there are two possibilities.
You could do it another way. Make the 15 uF much larger and the fillm cap smaller. I would say 220 uF NP and perhaps 0.33 or 0.22 uf input. The advantage of this is the bass filter is closer to 1 st order rather than complex. 63 V type has better HF and survive blown outputs. Also low frequency distortion will be better.
MCNP63V227M16X25 - MULTICOMP - Electrolytic Capacitor, MCNP Series, 220 µF, ± 20%, 63 V, 16 mm, Radial Leaded | Farnell element14
