I agree.
The LTP gain is reduced by about factor 3 by adding the 100R emitter degen, but this acts equally on Tr1 and Tr2, thereby maintaining the relative distortion profile. In terms of the actual amount of distortion the front end is more linear but the loop gain is reduced by the same X3 factor so the overall distortion level is probably similar.
I suspect Nigel is correct that the reduced loop gain might expose more crossover distortion, but then the emitter degen will help with rejection of input EMI. Swings and roundabouts. Maybe Les should have reduced the compensation cap to match, but he has removed the pole-zero network across the feedback resistor in the latest NCC.
FWIW I use 47R emitter degen and 10R VAS degen in my "005" amp. I think there is an advantage to be had by using a modern complementary OPS (with class A driver) instead of the sluggish quasi. But then it's not really much of a Naim clone, it's just a standard amp topology with the ac characteristics of the LTP tweaked to "undo" the 2nd harmonic cancellation in the IPS.
There is one topic where I don't agree with Les - I select the LTP transistors for the same gain (they seem to be well matched on Vbe without selection) and invert the feedback cap (+ive "down"), but this is splitting hairs.
Last edited:
Looks very similar to Greg B mods. He recommended 47r for emitor degen and 33r for Vas degen
Sent from my Nexus 5 using Tapatalk
Sent from my Nexus 5 using Tapatalk
If you look at Re+ re the current hasn't changed as the VAS base ( TR4 ) is a constant voltage clamp. Nothing in the longtail pair will change that. I estimated it at 650 uV. Therefore what you do in the input pair will not change that until the transistor switches off.
Transistor Re Model
To be honest when tail current is so low I don't see the point of input pair degeneration. The prefect proof of that is the input pair needs a 650/350 mismatch to show significant second harmonic distortion ( note the voltage across the 22K if doubting this ). What people forget is the loop feedback from speaker to TR2 will cause most input pair non linearity ( TR1+2 ) to vanish.
The reason to use input pair degeneration is to have higher current to drive the VAS in an attempt to have higher slew rates. The only reason we add degeneration is to restore the linearity we would have at lower current. We add noise and loose loop gain by doing this. As this stage is run in pure class A and is mostly supplying current there no reason to pretend it is a voltage amp. In my oppinion it is the worst place to seek linearity/gain trade off. This isn't a valve amp.
Transistor Re Model
To be honest when tail current is so low I don't see the point of input pair degeneration. The prefect proof of that is the input pair needs a 650/350 mismatch to show significant second harmonic distortion ( note the voltage across the 22K if doubting this ). What people forget is the loop feedback from speaker to TR2 will cause most input pair non linearity ( TR1+2 ) to vanish.
The reason to use input pair degeneration is to have higher current to drive the VAS in an attempt to have higher slew rates. The only reason we add degeneration is to restore the linearity we would have at lower current. We add noise and loose loop gain by doing this. As this stage is run in pure class A and is mostly supplying current there no reason to pretend it is a voltage amp. In my oppinion it is the worst place to seek linearity/gain trade off. This isn't a valve amp.
An old approach to sub-sonic filtering is to make the turnover frequency of the preceding amplifier and power amplifier stages the same. The thinking at the time I became aware of this nearly 30 years ago was that a -3 dB point of 7 Hz was adequate for DIY designs of the time.
Naim Audio made a point in their early days that no filtering of any kind was needed for their systems, not necessarily with uniformity of turn over points, but I believe still met in their design process including their NAC pre-amp range.
If stepping outside the Naim product circle in my view it is better to look at the whole amplifier chain with an alternative pre-amp than to regard the power amplifier as a stand alone item.
There is information on this topic at http://www.kennethkuhn.com/students/ee351/frequency_response_notes.pdf
Re dc feedback blocking capacitors, Linsley-Hood commented 22 years ago that an electrolytic type used here would have very little effect on THD at 1 kHz.
However if one measured the inter-modulation distortion (IMD) effects produced when the amplifier is driven by several simultaneous inputs using a spectrum analyser to track down the spurious signals, then the differences between one type and another are more easily seen.
He said changing the type does affect sound quality which is what IMD measurements will point out especially if one of the test signals is of a square-wave type.
While such tests are not accessible to most DIY hobbyists and Linsley-Hood was an advocate of using plastic capacitors, I think UK manufacturers interested in subjective results would investigate electrolytic capacitors to find the best available parts.
If so, all one could do in hoping to emulate the sound of a Naim or an Avondale amplifier would be to look for clues as to brands and makers part descriptions, and, for increased values, selecting from within the same product range.
I have been trying to keep the number of capacitors in the signal path down. If I am playing vinyl my chain is Paradise -> DC-B1 -> HackerNAP -> FrugalHorn XL (Alpair 10.3M drivers), the only caps being the two in the HackerNAP, and I have tried to keep them good quality (I understand that PETP is good for audio). I believe Les at Avondale's personal choice is an axial wet tant for the 47uF feedback cap.
This is another way of doing balance. These values can be soldered on top of the ones already fitted ( under PCB, for quick removal ) . The only small problem is getting a diode that has the same forward voltage as the TR4 Vbe. 1K047 is just to say slightly larger than 1 K. Adding 2K2 to TR3 tail restores the TR4 base drive current ( near enough ).
Nigel, that would work if you replace the ZTX753 with e.g. an MJE350. BUT... the ZTX has a VBE of ~0.5V at ~10mA Ic.
Looks like nobody wants to believe it, but it's true!
Looks like nobody wants to believe it, but it's true!
I was making constant current sources by coincidence yesterday, exactly the same problem. I will explain a little. My idea was to replace a 5V regulator with a constant current source. The possibility was to use the Ron ( CMOS internal resistance ) as the only voltage limiter in the circuit. It worked and yet it didn't. The very same reason as the MJE 350 above. The Vbe of a transistor is anyones guess. That is in terms of the 0.7 V often quoted. With my CMOS circuit I had hoped for 4VDC limited by a 25 mA current limit. Theory was 4V and practice was 6.45 V which was way too high. All was understood and sorted. It was a cost neutral solution that might offer better battery life. The factor I didn't like was Vbe is not a fixed quantity as I would want it to be ( as here ). The better solution turned out to be a LM317 set to 4.2V ( Not 3.9V and not 5V ). LM317 is such a good device ( and LD1085 ). The solution I will use is a 5V regulator and additional 100 R current limiter. Sadly that was what I came up within 10 minutes when I designed the circuit. One thing I did find was the test device died, the CMOS part didn't ! The one thing I didn't use was a Rubber diode as it has multiple problems. TL431 and a MJE 340 might have been an option. One thing I could do is use a 5V regulator and lift -ve on a diode. That would be 4.3V with no great complexity. That saves two components on a very small PCB. Alas it wastes power. The best soulution would be a LDO 1.25 V regulator set to 4.2V. The solution I have is 90% as good as it gets. It will work for months on batteries recovered from smoke alarms at 7V. I only did that to test the theory and was very surprised. I had no idea Duracells have so much energy in them when usually of no further use.
Here is the better solution to the Naim/Avondale. It is hinting at a current mirror. Alternatively one can calculate the current for a given diode and adjust the resistor to suit. 1f it were 1N4148 at perhaps 0.595V and needing 650uA ( 0.65 mA ) it would be 915R4. 22K + 955R. I would prefer this to a current mirror as that vastly increases the amplifier loop gain and would need careful VAS compenstion ( 100 pF? ). This exactly duplicates the original circuit current whilst offereing full balance. Like Swiss cheese it's then up to you which you like best. These mods can be fitted and removed is minutes, no need to remove existing parts. Updated tail current to 1.3 mA ( not 1 mA ). Please ignore 35mV at 100R, it should be about 65mV in each 100R if fitted.
Here is the better solution to the Naim/Avondale. It is hinting at a current mirror. Alternatively one can calculate the current for a given diode and adjust the resistor to suit. 1f it were 1N4148 at perhaps 0.595V and needing 650uA ( 0.65 mA ) it would be 915R4. 22K + 955R. I would prefer this to a current mirror as that vastly increases the amplifier loop gain and would need careful VAS compenstion ( 100 pF? ). This exactly duplicates the original circuit current whilst offereing full balance. Like Swiss cheese it's then up to you which you like best. These mods can be fitted and removed is minutes, no need to remove existing parts. Updated tail current to 1.3 mA ( not 1 mA ). Please ignore 35mV at 100R, it should be about 65mV in each 100R if fitted.
Last edited:
My NCC200 clone suddenly died. While listening produced sone strange sound and powered off.
I'm measuring aprox 4.5V acros the two 0R22 power transistors.
Removed from circuit and measured with DMM drivers and VAS transistors and looks ok.
Any idea where to look further ?
LE. For some reason it seems both MJL4281 has died. Measure 50ohm both directions 🙁
I'm measuring aprox 4.5V acros the two 0R22 power transistors.
Removed from circuit and measured with DMM drivers and VAS transistors and looks ok.
Any idea where to look further ?
LE. For some reason it seems both MJL4281 has died. Measure 50ohm both directions 🙁
Last edited:
Strange things happening. Replaced the two MJl burnt, i can adjust bias and is stable BUT i measure 40v on SPK out, the positive rail. What could cause that ?
Hi Atupi. Commiserations for the damage there, I guess it's a real disappointment.
Assuming that there was no short of the output, it's most likely that there was thermal runaway. i.e. the VBE multiplier lost control of the bias current. This or possibly violent oscillation is the most likely other reason for output failure. If that occurred whilst you were listening, you probably would have heard a high noise level if there was significant oscillation. Overheating and runaway normally shows as a rapid rise in volume level and distortion.
You have to agree that your heatsinks are undersized and this could easily lead to rapid runaway and failure. I also raised the point earlier that the VBE multiplier should not simply be mounted to the heatsink for a Quasi-complementary stage as the temp. coefficient and "feedback" does not track correctly.
It is also likely that TR4 or TR6 are shorted. This will drive a large DC offset at the output. The simple circuit is prone to this, like many older designs. Check all the semis after a major failure like that. Also consider an output coil if you don't have suitable (high inductance) speaker cable and have fuses in place or other protections whilst you are messing around at PCB level when powered.
Assuming that there was no short of the output, it's most likely that there was thermal runaway. i.e. the VBE multiplier lost control of the bias current. This or possibly violent oscillation is the most likely other reason for output failure. If that occurred whilst you were listening, you probably would have heard a high noise level if there was significant oscillation. Overheating and runaway normally shows as a rapid rise in volume level and distortion.
You have to agree that your heatsinks are undersized and this could easily lead to rapid runaway and failure. I also raised the point earlier that the VBE multiplier should not simply be mounted to the heatsink for a Quasi-complementary stage as the temp. coefficient and "feedback" does not track correctly.
It is also likely that TR4 or TR6 are shorted. This will drive a large DC offset at the output. The simple circuit is prone to this, like many older designs. Check all the semis after a major failure like that. Also consider an output coil if you don't have suitable (high inductance) speaker cable and have fuses in place or other protections whilst you are messing around at PCB level when powered.
Last edited:
Thanks for supporting me Ian.
You are quite right about my ht and today moved to a bigger one.
In the end fried two MJl4281 and now the beast is singing again.
As you already said the VBE multiplier is quite strange to me. I tried to install some 2n5551 away from heatsinks and the bias is playing games with me. Now i have some TIP41 installed on heatsink but i feel is a bit overcompensated.
I would like to know what is the right way to install this amp into case. I mean how Vbe is working to compensate, or in this project Vbe transistor is not ment to compensate bias according to ht temperature?
PS. the main source of problem today was that the two power transistors got loose from heatsink and they become very hot.
PPS. now one channel is running with one njl4281 and one NJL0281 and i run out of transistors 🙁
I still like very much the sound of this amp !
You are quite right about my ht and today moved to a bigger one.
In the end fried two MJl4281 and now the beast is singing again.
As you already said the VBE multiplier is quite strange to me. I tried to install some 2n5551 away from heatsinks and the bias is playing games with me. Now i have some TIP41 installed on heatsink but i feel is a bit overcompensated.
I would like to know what is the right way to install this amp into case. I mean how Vbe is working to compensate, or in this project Vbe transistor is not ment to compensate bias according to ht temperature?
PS. the main source of problem today was that the two power transistors got loose from heatsink and they become very hot.
PPS. now one channel is running with one njl4281 and one NJL0281 and i run out of transistors 🙁
I still like very much the sound of this amp !
Last edited:
Quasi-comp. amplifiers are usually UK products and you can find many old designs with pics on the internet. There are pics of several Naim models here: Modifying Naim Audio power amplifiers where you can see that TR5 is just sitting in the middle of the board.
Note: TR5 senses the ambient temp. inside the amplifier. This proves to give a stable but very slow responding bias control where current is kept small (<35 mA) and in a small case that is also the heatsink. That's also typical of similar "shoebox" amplifiers.
DIY clone builders find it simpler to use heatsinks. Whilst this results in very stable bias, heatsinks mounted externally will be even slower in response ( you could say unreponsive) to changing output levels. Obviously, with no case, there is no control at all and bias will float about with room temperature. Build it in small case, and matters will improve.
I learned many years ago not to play old Naims loud with the cover off. Yep, they can easily burn .
Note: TR5 senses the ambient temp. inside the amplifier. This proves to give a stable but very slow responding bias control where current is kept small (<35 mA) and in a small case that is also the heatsink. That's also typical of similar "shoebox" amplifiers.
DIY clone builders find it simpler to use heatsinks. Whilst this results in very stable bias, heatsinks mounted externally will be even slower in response ( you could say unreponsive) to changing output levels. Obviously, with no case, there is no control at all and bias will float about with room temperature. Build it in small case, and matters will improve.
I learned many years ago not to play old Naims loud with the cover off. Yep, they can easily burn .
In theory if TR9 + 10 are removed the DC offset should restore to usual levels ( < 50 mV ) if TR7 and 8 have survived. Often there is enough power to run headphones ( cheap ones and not directly into the ear at first ). The 100 R resistor in both TR7 and 8 emitters should help current limit. TR4,5,6 are at constant current so should be OK ( should!! TR6 sets that). Input pair TR1 and 2 also are on a constant current source ( TR3 ) so should be OK. MJE15030/31 are small power transistors. They should stand abuse. Having said that the worst case would see too much dissipation. That's a good reason to put them on the heatsink if space permits, the circa 0.5 amps would be OK. I have no idea if that's true here? Usually designers don't bother. Even a clip on heatsinks will not work in the worst case. For the want of 4 x M3 tapped holes this problem can be avoided. Check the capacitor shown as 47 uF on TR2 ( to 1K ), it might be damaged. This is why I prefer 63V non polar types as the voltage can be tollerated. This is nicer than protection didodes, although measurements say diodes are OK.
After replacing power transistors everything looks fine, i had i nice 3h listening session after the "disaster". To be in the safe side replaced 2n5551 Vbe transistors with TIP41c mounted om the heatsink. The bias remained much stable.
I dont know if is only my imagination but when Vbe transistor is installed on heatsink wich is keeping the bias more tight the sound looses some dinamics. I remember someone told the same some long ago in this thread.
So i conclude my final case should be more or less without too much ventlation to have a better heat transfer from power dissipated to VBE transistor.
As i plan to redraw a new pcb to etch myself i need to be clarified where shoud Vbe transistor to be installed.
I dont know if is only my imagination but when Vbe transistor is installed on heatsink wich is keeping the bias more tight the sound looses some dinamics. I remember someone told the same some long ago in this thread.
So i conclude my final case should be more or less without too much ventlation to have a better heat transfer from power dissipated to VBE transistor.
As i plan to redraw a new pcb to etch myself i need to be clarified where shoud Vbe transistor to be installed.
If Vbe multiplier has an effect on sound quality then that means the voltage at the output does not match what is at the input.
This may be due to the stage bias becoming non optimal. It could go higher, or it could go lower for the transients and this will be reflected in a change to the crossover distortion. Or the stage could be set higher than optimal and the change could move it closer to optimal, or farther away from optimal. There is a third set up. The bias could be below optimal and the change could move it to even lower bias on transients, or move it towards optimal.
Which of those options/changes is right would be more difficult to determine.
The general consensus is that a Vbe multiplier acts after the fact. They are all too slow to really hold optimal bias. The faster that one can make the multiplier react to Tj changes, the more closely one will hold the bias one chooses to set.
But what was the set point before the changed location?
and what was the set point after the change?
This may be due to the stage bias becoming non optimal. It could go higher, or it could go lower for the transients and this will be reflected in a change to the crossover distortion. Or the stage could be set higher than optimal and the change could move it closer to optimal, or farther away from optimal. There is a third set up. The bias could be below optimal and the change could move it to even lower bias on transients, or move it towards optimal.
Which of those options/changes is right would be more difficult to determine.
The general consensus is that a Vbe multiplier acts after the fact. They are all too slow to really hold optimal bias. The faster that one can make the multiplier react to Tj changes, the more closely one will hold the bias one chooses to set.
But what was the set point before the changed location?
and what was the set point after the change?
Last edited:
I always target around 28mA bias. With vbe off the heatsink there is almost no bias autocorrection as there is no enclosure to "transmit" the heat from ht to transistor.
With transistor mounted offboad to the ht there is a much better bias tracking.
With transistor mounted offboad to the ht there is a much better bias tracking.
Agreed, it is better in that temporary ( I assume) condition, to use the heatsink than nothing at all but the thermal coefficient is incorrect for stability at all powers up to maximum. Keep the level at a comfortable level like this though - I would expect trouble when it was enclosed and warmed up.
Its only my suggestion, but I think the more appropriate place to monitor is TR8, the CFP driver. However, I've only tried this a couple of times and it may prove necessary to use a small thermal coupling to the driver's heatsink rather than directly bolting them together, so that the tempco. is reduced. A smaller transistor like a BD139 or MJE340 is probably easier to attach than a TO220 for VBE multiplier.
The CFP side of the output stage is the one that is over-biased and probably needs a controller that responds to its thermal condition before the EF side. The CFP driver temp. swings with signal demand compared to the EF driver, which is near constant. For that reason, I would always plan to fit separate heatsinks to the drivers rather than lump them together on the main heatsink.
Its only my suggestion, but I think the more appropriate place to monitor is TR8, the CFP driver. However, I've only tried this a couple of times and it may prove necessary to use a small thermal coupling to the driver's heatsink rather than directly bolting them together, so that the tempco. is reduced. A smaller transistor like a BD139 or MJE340 is probably easier to attach than a TO220 for VBE multiplier.
The CFP side of the output stage is the one that is over-biased and probably needs a controller that responds to its thermal condition before the EF side. The CFP driver temp. swings with signal demand compared to the EF driver, which is near constant. For that reason, I would always plan to fit separate heatsinks to the drivers rather than lump them together on the main heatsink.
There is a theory bass frequencies can thermally modulate and the chip power amps can not avoid it. No idea if true. A&R told me about this in 1977 so not new. They had a 1 mm separation transistor to heatsink. The A&R is a variation of how this is done ( not a Vbe multiplier ) , same reasons for the spacing.