While also considering risks with output stage clipping and loop stability.
In an earlier post you discussed a low current version of the 140 clone where the Vas stage operating current was reduced from 10 m.a. to 3 m.a.
You warned of the consequence of low current drive to the output stage and a need to review the compensation.
There was a party on this thread interested in taking on some of your mods several posts ago - possibly this one.
You are now back on a higher power focus and I think your current proposal is a better and safer one to follow.
With the earlier version your warnings about current limitations would be easily understood. I think the issue of sorting the compensation carries a lot of weight as reducing the Vas current from 10 m.a. to 3 m.a. will lessen stage gain in direct proportion.
Naim run their Vas at higher currents and their LTP transistors at lower currents than other designs. Thus the Vas has higher gm and the LTP lower gm - so in a sense the Vas makes up for the LTP deficit.
A Vas current of 6 m.a. as you are currently proposing is fairly common. It might be more usual with that value to see LTP currents of 2 m.a. in each transistor rather than the 0.5 m.a. value used by Naim.
I have used the transistors you mention. We are lucky that BC546/556 are so readily available.
I fit 4 protection diodes around the output stage.
One across each MF supply rail decoupling capacitor. This effectively ties the Supply rail to Power ground. It prevents an excessive reverse voltage being applied to the amplifier and to it's polar capacitors. It also protects the amplifier from silly mistakes in wiring up the PSU.
One across the main output device to the output rail.
These two prevent the load back emf exceeding the supply rail voltage by more than Vf.
These protect the output devices from excessive voltage.
These 4 diodes can be 1A power diodes. I have never blown a 1n4004 yet.
Some may prefer to use the 3A 1n5404
1A diodes cost 2p each and give a lot of extra protection for very little cost and very little board space.
One across each MF supply rail decoupling capacitor. This effectively ties the Supply rail to Power ground. It prevents an excessive reverse voltage being applied to the amplifier and to it's polar capacitors. It also protects the amplifier from silly mistakes in wiring up the PSU.
One across the main output device to the output rail.
These two prevent the load back emf exceeding the supply rail voltage by more than Vf.
These protect the output devices from excessive voltage.
These 4 diodes can be 1A power diodes. I have never blown a 1n4004 yet.
Some may prefer to use the 3A 1n5404
1A diodes cost 2p each and give a lot of extra protection for very little cost and very little board space.
Thanks Andrew, good idea. I had been asking myself this very question. Even if a couple of big zener diodes could be used back to back. The surge current of a 1N4004 is 30A. That should be enough. IN5404 200A.
If I were trying to protect TIP2955/3055 at 76V perhaps a couple of 30V zeners back to back might work ? A resistor between to limit current. These would be between speaker +/-. This " might " allow 20 Vrms or 50 watts rms @ 8R.
On my experiment 2 x 1N4007 didn't catch the problem. It was on switch off it let go. My friend who built Helmholtz coils found the same. With speakers the coil resistance ( DCR ) is enough to damp the surge I guess. My coil has a DCR of 0R3.
I have a few TIP2955/3055 to risk and won't be happy without an answer. The old Quad 303 where the regulator failed didn't have a problem when > 85 VDC and 2N3055. Real speakers are not as bad as my transformer. Seeing as the Quad was OK it sould be possible work near Vce max. Zener diodes are good because they will go through the origine and have to with this idea. In the non standard connection they are low grade 0.7V diodes ( added to circa 30 V peak or 20 V rms ) . The base emitter diode is often used in current mirrors as a better diode choice. Often the base collector diode is a zener. In the old days some were thought rather good as voltage references.
If I were trying to protect TIP2955/3055 at 76V perhaps a couple of 30V zeners back to back might work ? A resistor between to limit current. These would be between speaker +/-. This " might " allow 20 Vrms or 50 watts rms @ 8R.
On my experiment 2 x 1N4007 didn't catch the problem. It was on switch off it let go. My friend who built Helmholtz coils found the same. With speakers the coil resistance ( DCR ) is enough to damp the surge I guess. My coil has a DCR of 0R3.
I have a few TIP2955/3055 to risk and won't be happy without an answer. The old Quad 303 where the regulator failed didn't have a problem when > 85 VDC and 2N3055. Real speakers are not as bad as my transformer. Seeing as the Quad was OK it sould be possible work near Vce max. Zener diodes are good because they will go through the origine and have to with this idea. In the non standard connection they are low grade 0.7V diodes ( added to circa 30 V peak or 20 V rms ) . The base emitter diode is often used in current mirrors as a better diode choice. Often the base collector diode is a zener. In the old days some were thought rather good as voltage references.
Mjona. Naim VAS being higher and TR1/2 lower is a good point. My TR1/2 run at about 2 mA. That is depending on VAS TR4 Naim can be 500 to 700 mA compared with 1 mA on mine ( 1+ 1 ). My VAS gain is about 67% of Naim. All in all not much difference. 40 V at 6 mA = 240 mW which should suit the 2N5551G. If I have a big collector PCB copper area I could go to 9 mA. A small postage stamp size I suspect. Sit the pad on the edge to make soldering OK. 400 mW should be OK.
This is an amp to repalce one I did in 1993. My friend said he lost 30 sales in the last couple of years. We tried a big TD7293. It didn't last long. It is intended that he can build it without my help.
I will fit a Vbe bias ( TR5 if Naim ). Not really needed for motor drive. 3 x 1N4007 would almost certainly do. I am not using a feedforward resistor as in off the peg Darlingtons. Instead I use circa 220R between the TR7/8 emitters. This marginally switches off the outputs quicker. It will need slightly more bias voltage than a Naim or standard Darlingtons. This makes 3x 1N4007 OK for simple bias. In theory about 0.5V underbiased. If a 68R resistor was added I dare say it might be OK for music. The output stage is like the Naim upper section with all PNP bottom section.
This is an amp to repalce one I did in 1993. My friend said he lost 30 sales in the last couple of years. We tried a big TD7293. It didn't last long. It is intended that he can build it without my help.
I will fit a Vbe bias ( TR5 if Naim ). Not really needed for motor drive. 3 x 1N4007 would almost certainly do. I am not using a feedforward resistor as in off the peg Darlingtons. Instead I use circa 220R between the TR7/8 emitters. This marginally switches off the outputs quicker. It will need slightly more bias voltage than a Naim or standard Darlingtons. This makes 3x 1N4007 OK for simple bias. In theory about 0.5V underbiased. If a 68R resistor was added I dare say it might be OK for music. The output stage is like the Naim upper section with all PNP bottom section.
http://docs-europe.electrocomponents.com/webdocs/12cf/0900766b812cfe18.pdf
I just wanted to put up the 2N5551 spec sheet. I was surprised how good it is. Not too far behind Japanese specials like 2SD756 ( B716 ).
160V
Realistic 400 mW working at 75 C. ( 40V x 9mA = 360 mW ).
Typical gain of >150 at 10 mA in audio band!
Output capacitance 6 pF!!!
> 100 MHz!
2N5401 seems similar for a Naim clone TR4. If I am right G version is better than 100 gain. Insulate base and emitter and add a small piece of copper/ tin can to the collector if wanting to be doubly sure of heat tollerence. Folklaw has MPSA92 as better. I doubt it.
These just about beat them. 500 mW mostly at 75 C to my mind. Centre collector as MJE340/350 BC639/640 BD139/140 etc.
http://www.onsemi.com/pub_link/Collateral/2SA1208-D.PDF
I just wanted to put up the 2N5551 spec sheet. I was surprised how good it is. Not too far behind Japanese specials like 2SD756 ( B716 ).
160V
Realistic 400 mW working at 75 C. ( 40V x 9mA = 360 mW ).
Typical gain of >150 at 10 mA in audio band!
Output capacitance 6 pF!!!
> 100 MHz!
2N5401 seems similar for a Naim clone TR4. If I am right G version is better than 100 gain. Insulate base and emitter and add a small piece of copper/ tin can to the collector if wanting to be doubly sure of heat tollerence. Folklaw has MPSA92 as better. I doubt it.
These just about beat them. 500 mW mostly at 75 C to my mind. Centre collector as MJE340/350 BC639/640 BD139/140 etc.
http://www.onsemi.com/pub_link/Collateral/2SA1208-D.PDF
Last edited:
They are pretty good and considering their price extremely good value.
Normal 2n5551 have been >170 hFE and 2n5401 >160 hFE
One batch of 2n5401 I bought came in with hFE from 75 to 119.
I keep them for switching duty and similar, i.e. saturated with high Ib.
They may have been fakes, or rejects, so I don't use them for audio.
Last edited:
In the more modest of stakes - no fanfare about excellent linearity of Hfe, I tested a dozen NPN BC639's - unused years old stock Motorola I have in my stash.
Five of these were in the 170 range, 3 in the 160 range,3 in the 150 range and 1 tailing the field at 137. I did likewise with five PNP BC640's these results being, 133,135,136,141,and 152.
I think these are worth a mention as these have been in production with various makers for a long time and should be more common in remote or lower population areas of the world.
Fairchild have a datasheet for these under the grouping https://www.fairchildsemi.com/datasheets/BC/BC635.pdf.
Typical Characteristics Figure 1 on page 2 shows at low base current levels the collector current is a straight line along the horizontal axis. This is a good indication of transistor linearity.
There is some increase in slope with increasing collector current but markedly less than seen with 2SA1208D albeit this is a higher gain device.
Unfortunately Fairchild replaced the datasheets for these devices only recently and I had not downloaded that for BC640 before the change.
However the NXP datasheet shows hFE to be a horizontal straight line in the working range up to 10 m.a.BC636 pdf, BC636 description, BC636 datasheets, BC636 view ::: ALLDATASHEET :::.
These BC devices featured as drivers in a 25 watt power amplifier I built nearly 40 years ago, and more ambitious projects required something beefier BD139 and BD140 supposedly the same devices in a larger package - I am not sure of that anyway I put my BC639/40's aside unloved and unwanted for a long time - until now.
For a Vas stage the BC's should be usable in a Vas stage for 30 volt supply rails - possibly for a 40 watts into 8R amplifier. One could build a reduced power version of the 140 clone along these lines. A bridged pair would give 80 watts so if more power was needed another stereo kit could be purchased.
Five of these were in the 170 range, 3 in the 160 range,3 in the 150 range and 1 tailing the field at 137. I did likewise with five PNP BC640's these results being, 133,135,136,141,and 152.
I think these are worth a mention as these have been in production with various makers for a long time and should be more common in remote or lower population areas of the world.
Fairchild have a datasheet for these under the grouping https://www.fairchildsemi.com/datasheets/BC/BC635.pdf.
Typical Characteristics Figure 1 on page 2 shows at low base current levels the collector current is a straight line along the horizontal axis. This is a good indication of transistor linearity.
There is some increase in slope with increasing collector current but markedly less than seen with 2SA1208D albeit this is a higher gain device.
Unfortunately Fairchild replaced the datasheets for these devices only recently and I had not downloaded that for BC640 before the change.
However the NXP datasheet shows hFE to be a horizontal straight line in the working range up to 10 m.a.BC636 pdf, BC636 description, BC636 datasheets, BC636 view ::: ALLDATASHEET :::.
These BC devices featured as drivers in a 25 watt power amplifier I built nearly 40 years ago, and more ambitious projects required something beefier BD139 and BD140 supposedly the same devices in a larger package - I am not sure of that anyway I put my BC639/40's aside unloved and unwanted for a long time - until now.
For a Vas stage the BC's should be usable in a Vas stage for 30 volt supply rails - possibly for a 40 watts into 8R amplifier. One could build a reduced power version of the 140 clone along these lines. A bridged pair would give 80 watts so if more power was needed another stereo kit could be purchased.
I am thinking of making a 10 mA gain test rig. Anyone done this? My test meter gives about 170 for a 2N5551. I suspect it would be about 70 at 10 mA ? 100 would be ideal.
My friend I am designing this for needs it to be very simple and future proof.
My view of linearity is it can only be a manipulation of the parameters. As the more important linearity is how much negative feedback we can use. It is not really a question of the linearity of the devices. The VAS is where the question is possibly more important. I would say if the loop gain was degenerated here would be wisest ( I don't call it local feedback as that's not what I want to say ). It will make the amplifer more stable and the VAS easier to drive. For example if we add 3R to the VAS emitter we might raise Z VAS in from 300R to 600 R. If we halve ( 4.5 mA rather than 9 mA ) the VAS current we do the same except we have less to drive the TR7/8/9/10. If we double the TR1+2 current we should arrive back at the same loop gain with a greatly better VAS compromise ( R send = 500 R and input about the same ). We should find IM distortion is nicely lower. Also I suspect a real resistor is better than 26/ic mA called Re. We will get more DC offset, if below 75 mV it wouldn't bother me. Rotel RA931 is that.
The linearity arguement is worse when valves and is more reasonably so. I was told a ECC81 a total waste of time. The amplifier needed negative feedabck. The ECC81 gave splendid results with feedback. The only reason it looks bad is it has the ability to be very high gain, high current and moderate distortion. ECC82 is low gain, high current and low distortion. ECC82 at gain of 10 is typical. ECC 81 > 100. When ECC81 is at a gain of 26 it uses no cathode bootstrap. Thus like a transistor circut typically seen. It will kill the ECC82 in this situtaion. Linearity is mostly a function of gain and voltage that might be possible.
When I design valve amps the lineartiy of a transistor is very much a question. Only Morgan Jones mentions this. Most valve people won't touch transistors. MJE350 is the usual choice. As a current amp it is fine and 6 MHz is OK in valve designs. I usually use a LED to bias them to get ideal linearity. The CCS in the Naim is not ideal regarless of transistor used. It needs at least 0.4 V more. When valves a 30V bias would be fine as we can have 450 V of which we might use 300 V. 270 V usually would be OK. If we were to use a cascode CCS to the valve anode the linearity might be <0.04 % distortion and gain of 100 ( ECC 83 ) giving 35 Vrms. The anode would seems to be seeing many 1000's of volts whilst being safe. If you look at op amps at gain 100 the valve is doing well.
My friend I am designing this for needs it to be very simple and future proof.
My view of linearity is it can only be a manipulation of the parameters. As the more important linearity is how much negative feedback we can use. It is not really a question of the linearity of the devices. The VAS is where the question is possibly more important. I would say if the loop gain was degenerated here would be wisest ( I don't call it local feedback as that's not what I want to say ). It will make the amplifer more stable and the VAS easier to drive. For example if we add 3R to the VAS emitter we might raise Z VAS in from 300R to 600 R. If we halve ( 4.5 mA rather than 9 mA ) the VAS current we do the same except we have less to drive the TR7/8/9/10. If we double the TR1+2 current we should arrive back at the same loop gain with a greatly better VAS compromise ( R send = 500 R and input about the same ). We should find IM distortion is nicely lower. Also I suspect a real resistor is better than 26/ic mA called Re. We will get more DC offset, if below 75 mV it wouldn't bother me. Rotel RA931 is that.
The linearity arguement is worse when valves and is more reasonably so. I was told a ECC81 a total waste of time. The amplifier needed negative feedabck. The ECC81 gave splendid results with feedback. The only reason it looks bad is it has the ability to be very high gain, high current and moderate distortion. ECC82 is low gain, high current and low distortion. ECC82 at gain of 10 is typical. ECC 81 > 100. When ECC81 is at a gain of 26 it uses no cathode bootstrap. Thus like a transistor circut typically seen. It will kill the ECC82 in this situtaion. Linearity is mostly a function of gain and voltage that might be possible.
When I design valve amps the lineartiy of a transistor is very much a question. Only Morgan Jones mentions this. Most valve people won't touch transistors. MJE350 is the usual choice. As a current amp it is fine and 6 MHz is OK in valve designs. I usually use a LED to bias them to get ideal linearity. The CCS in the Naim is not ideal regarless of transistor used. It needs at least 0.4 V more. When valves a 30V bias would be fine as we can have 450 V of which we might use 300 V. 270 V usually would be OK. If we were to use a cascode CCS to the valve anode the linearity might be <0.04 % distortion and gain of 100 ( ECC 83 ) giving 35 Vrms. The anode would seems to be seeing many 1000's of volts whilst being safe. If you look at op amps at gain 100 the valve is doing well.
Virtually all manufacturer datasheets already specify the 2N5551 Hfe @ 10 mA Ic. This is about the limit of their small-signal "sweet spot" for maximum gain and according the flatness of the curves, you should get similar Hfe results right down to <1mA.
https://www.fairchildsemi.com/datasheets/MM/MMBT5551.pdf.
http://www.onsemi.com/pub_link/Collateral/2N5550-D.PDF
https://www.fairchildsemi.com/datasheets/MM/MMBT5551.pdf.
http://www.onsemi.com/pub_link/Collateral/2N5550-D.PDF
This transistor is in my 1983 Motorola "Small-Signal Data" handbook. The graph for DC current gain shows at TJ =25 degrees C you could expect hFE of 150 for currents between 1 and 15 milliamps.
The VCE for the measurement was 5.0 V.
It does not matter what the hFE is in terms of the Voltage gain stage since this =gm (40 times the current in mA) multiplied by the load in k.Ohms.
That being the case the 2N5551 will be good enough for the job requirement for TR4 Vas running 10 mA. Reducing this value will reduce stage and open loop gain of the whole amplifier.
While that would increase the input resistance of the stage, like gm this varies according to the level of current under dynamic conditions.
The snag here is TR4 varying input resistance is in parallel with the 1k collector load resistor of TR1 and that load combination acts to reduce TR1 voltage gain.
TR2 partners TR1 and communicates a corrective signal across the LTP emitter connections.
The aim of running TR4 at 10 mA seems to be to maximise open loop gain rather than interposing a buffer transistor between TR1 and TR4 adding another series device into the overall forward transfer function.
Inserting a resistor in TR4 emitter would provide some local feedback at the expense of a little reduction in open loop gain. The system closed loop is 28 times where 23 times might be adequate - perhaps some scope to claw back the loss caused by the presence of the emitter resistor.[wiki=]%[/wiki]
Last edited:
I use a CCS as the current regulator for the test jig.
The CCS has user selectable resistors to determine the sink current.
A small one using a To126+To92 can run at upto 30mA
So I have the highest resistor permanently in circuit and use a 0.1" pin shorting plug to add in more resistors in parallel.
0.5mA to 20mA using 5 resistors. and lots of steps in between by selecting one or two or three or four of those resistors in paralllel to the permanent resistor.
If you want a high current CCS, then use a 2sa1943, or 2sc5200, with a 10C/W sink and a To92 to get upto 500mA of collector current.
Add in both a base to collector fixed resistor + a series connected variable resistor. You can trim the VR to set a minimum Vce during the test.
I generally attach the CCS to the emitter.
The CCS current is affected by the Tj of the main pass transistor.
It heats up when current is passing and cools down again when you change DUT. You can either keep it warm all the time by having a REF and switching between REF and DUT, or allow time for the Medium power device to heat back up to a stable temperature.
Last edited:
You can also build up an LTP style jig and use the CCS set to the LTP operating tail current to allow "matching" of the LTP transistors.
My 2N5551 have turned up. I lost the instructions of the transistor tester section of my meter years ago. It is anyones guess at what current.
These samples say 204 typical , MPSA 42 about 100 and BD139 140. What might be nice is an AC test at perhaps 4 kHz and 10 mA. Anyone have an idea about that ? 3 x PP3 battery seems where to start and 2K7 load. The input could be AC coupled and negative feedback biasing used to determine ideal conditions. The lower the bias needed the higher gain if we are so lucky. The test could be adapted to any supposed use. It won't give results that others can relate to. It would give relative results showing real differences. I can see many problems with this. Output impedance of the signal needs to be realistic and variable.
These samples say 204 typical , MPSA 42 about 100 and BD139 140. What might be nice is an AC test at perhaps 4 kHz and 10 mA. Anyone have an idea about that ? 3 x PP3 battery seems where to start and 2K7 load. The input could be AC coupled and negative feedback biasing used to determine ideal conditions. The lower the bias needed the higher gain if we are so lucky. The test could be adapted to any supposed use. It won't give results that others can relate to. It would give relative results showing real differences. I can see many problems with this. Output impedance of the signal needs to be realistic and variable.
Just measured the MJE340/350 I will use for the motor drive version. It should be possible to squeeze in MJE15032/33 for posh use.
RS 761 4467 MJE340 g=140 ( TR7 )
RS 486 4577 MJE350 g=80 ( TR8 )
RS 805 1104 2N5551 g=200 ( TR4 )
Rapid 81-0445 BC556 ( Diotec ) g=360 ( TR1/2 )
Outputs are
RS 903 4077 2STC5200 NPN T0-264
RS 686 8026 2STA1943 PNP T0-264
These have almost T03 footprint. They will need to be. 30 watt constant runing which is like 150 watts music. Heat-sink to start with 0.7 deg per watt ( what I have ). I suspect it will need to be larger ( per side ). I will assume ambient of 30 C at 50 % efficient. That means 30 watts of heat to deal with. That looks to be 51 C. Seldom is it as simple as that. I usually think 60 C spells trouble. That's not the way one is supposed to do it. It isn't far wrong. I seem to remember there is even a safety recomendation that 60 C should not be exceeded. Hence the Japanese often put the heat sinks inside ( not sure that solves anything and makes it harder ). If you will forgive the obvious. As a rule of thumb if you can't hold a heatsink for some seconds it is too hot.
RS 761 4467 MJE340 g=140 ( TR7 )
RS 486 4577 MJE350 g=80 ( TR8 )
RS 805 1104 2N5551 g=200 ( TR4 )
Rapid 81-0445 BC556 ( Diotec ) g=360 ( TR1/2 )
Outputs are
RS 903 4077 2STC5200 NPN T0-264
RS 686 8026 2STA1943 PNP T0-264
These have almost T03 footprint. They will need to be. 30 watt constant runing which is like 150 watts music. Heat-sink to start with 0.7 deg per watt ( what I have ). I suspect it will need to be larger ( per side ). I will assume ambient of 30 C at 50 % efficient. That means 30 watts of heat to deal with. That looks to be 51 C. Seldom is it as simple as that. I usually think 60 C spells trouble. That's not the way one is supposed to do it. It isn't far wrong. I seem to remember there is even a safety recomendation that 60 C should not be exceeded. Hence the Japanese often put the heat sinks inside ( not sure that solves anything and makes it harder ). If you will forgive the obvious. As a rule of thumb if you can't hold a heatsink for some seconds it is too hot.
My thinking is along similar lines to your ideas in this and your preceding post.
I don't know what Nigel's is as he has not commented.
@Nigel,
I asked a while ago if the motor project was for a gramophone. While you are forging ahead with your thinking, you are forgetting to keep people informed. You might also clarify if the transformer that caused the fireworks with your output stage is part of the current project or not. If the answer is yes have you resolved the cause.
I asked a while ago if the motor project was for a gramophone. While you are forging ahead with your thinking, you are forgetting to keep people informed. You might also clarify if the transformer that caused the fireworks with your output stage is part of the current project or not. If the answer is yes have you resolved the cause.
Yes it is. Garrard 301/401/501/TD124. The fireworks were me being silly ( V supply +/- 38V and 80 V Vce open circuit TIP3055/2955 ). I would like to force that version to work as I feel it could. It would be stupid to sell it so just to know the truth. I will attempt a propper audio version as it is close to being viable. A version I built with 120V T03 Darlingtons did work. MJ11015/16. They are not a lot worse than 2N3055. They show no feedforward resistor inside the can. If so a 200 R resistor base to base should help them switch off. If I felt confident they will continue I would use them.
My previous design was using 160 V audio FET's. These worked fine and many units built. If need be I can do that again. MPSA 92/42 and BUZ900/905 will work. HH proved that.
Christmas is upon us so progress will be slow. Also I need to tidy my shed as yet again it is impossible to work in.
To match a long tail pair. I don't know if what I do is any good? Sit transistor for about 1 minute in the hFE tester to settle. Group all that are similar together. In clips measure Vbe ( to avoid finger heat ). Use the best matches. It seems very easy and should be saying the devices are similar. I have never tried the Vcb zener voltage effect. It might say something.
I was pondering Andrews extra reverse biased diodes to the decoupling capacitors. The emitter is very slightly disconected from 0V even under ideal conditions. In some circumstances I could see the protection of the transistor being far less effective than thought. If one thing comes from this I would say follow his advice as soon as you can and fit some 1N4004(or7). Even if they have no effect at all they cost almost nothing. 1N5404 probably are worth having. I doubt if high speed is important as forward biased most are. Having said that they also are inexpensive. Always look at the surge current ratings. 30A and 200A for the two quoted. In my application I could imagine 200A as a minimum. I will carefully dismantle my broken one and see if the 1N4007 failed.
I read someone today say " Some of the worlds best amplifiers use 2N5551". I don't know if he was truely qualified to say it. His tone seemed to say he was aware of what he was saying. I thought to myself " I don't doubt it ". I will give you a similar story. The UK climate is very favourable to making dry white wines that are at a similar price to the ones from France. These are not only good but are very wonderful in ways hard to describe. They last a long time in the mouth. Some are made from grapes not usually thought to be usable for high quality wines. It can be the same when amplifiers. As David Mate said to me " Linearity is of no importance in a feedback amplifer. All that matters is the ability to apply feedback. The moment to moment jumps in gain matter greatly and will yield great linearity if no jump is outside of the feedback loop ability". I say it as best I remember. I think David was using a FET input Darlingtons. The linearity much worse than any single device. None the less the feedback could be whatever you want. The sound was supurb.
Correction. MJ110015/16 are 8K and 40 R on feedforward. Gain is said to be >200. I suspect >3000 if wanting 3 amps output peak. They show internal diodes CE. Still fit the Andrew diodes. I have never listened to them as audio devices. I have a hunch they would be OK. I suspect they would be better than standard 2N3055 used with MJE340/350.
My previous design was using 160 V audio FET's. These worked fine and many units built. If need be I can do that again. MPSA 92/42 and BUZ900/905 will work. HH proved that.
Christmas is upon us so progress will be slow. Also I need to tidy my shed as yet again it is impossible to work in.
To match a long tail pair. I don't know if what I do is any good? Sit transistor for about 1 minute in the hFE tester to settle. Group all that are similar together. In clips measure Vbe ( to avoid finger heat ). Use the best matches. It seems very easy and should be saying the devices are similar. I have never tried the Vcb zener voltage effect. It might say something.
I was pondering Andrews extra reverse biased diodes to the decoupling capacitors. The emitter is very slightly disconected from 0V even under ideal conditions. In some circumstances I could see the protection of the transistor being far less effective than thought. If one thing comes from this I would say follow his advice as soon as you can and fit some 1N4004(or7). Even if they have no effect at all they cost almost nothing. 1N5404 probably are worth having. I doubt if high speed is important as forward biased most are. Having said that they also are inexpensive. Always look at the surge current ratings. 30A and 200A for the two quoted. In my application I could imagine 200A as a minimum. I will carefully dismantle my broken one and see if the 1N4007 failed.
I read someone today say " Some of the worlds best amplifiers use 2N5551". I don't know if he was truely qualified to say it. His tone seemed to say he was aware of what he was saying. I thought to myself " I don't doubt it ". I will give you a similar story. The UK climate is very favourable to making dry white wines that are at a similar price to the ones from France. These are not only good but are very wonderful in ways hard to describe. They last a long time in the mouth. Some are made from grapes not usually thought to be usable for high quality wines. It can be the same when amplifiers. As David Mate said to me " Linearity is of no importance in a feedback amplifer. All that matters is the ability to apply feedback. The moment to moment jumps in gain matter greatly and will yield great linearity if no jump is outside of the feedback loop ability". I say it as best I remember. I think David was using a FET input Darlingtons. The linearity much worse than any single device. None the less the feedback could be whatever you want. The sound was supurb.
Correction. MJ110015/16 are 8K and 40 R on feedforward. Gain is said to be >200. I suspect >3000 if wanting 3 amps output peak. They show internal diodes CE. Still fit the Andrew diodes. I have never listened to them as audio devices. I have a hunch they would be OK. I suspect they would be better than standard 2N3055 used with MJE340/350.
Last edited:
I take it you are still using a transformer and driving the secondary winding to step the primary up to the required a.c. voltage level. I have a variable frequency circuit to suit a Thorens TD150 record deck, the motor of which is a 115 a.c. type requiring some 8 watts to drive a platter weighing 3.4kg (7.5lb) at 33/45 rpm.
The figure for your subject turntables ranges 11-16 lb. According to the author the TD150 turntable will run on as low a voltage as 25 volts provided it is not allowed to slow down.
While you have stated that your requirement is for 30 watts continuous I wonder if the same characteristic applies to other turntables.
If so, a thought occurs that the signal to your power amplifier could be attenuated by one of those old volume controls incorporating a mains switch.
That might give some boost power control over start up, a cruise running position, and a gentle slowing down before the mains switch turns off the power.
It's a very long time since I experimented with my Thorens motor and drive voltage (1978?)
I think it would not start on 90Vac. It just sat there vibrating.
If one manually started it with a turn between thumb and finger it would run but was very rough. I did not try to run it at a much lower voltage since this roughness (around the 85Vac) would come through as flutter in the audio.
I ended up running at around 110Vac from a reverse connected 240:24Vac transformer fed with ~11Vac from a 50W power amplifier. I suspect it would perform better at slightly higher voltage, maybe 115 to 120Vac.
The Thorens and Linn turntables need about 0.5 watt to maintain rotation. A 5 watt mains voltage PSU is possible which results in about 90 Vrms. LP12 seems to be sounding a bit sterile at about 66 Vrms. The other problem to my mind is the motor itself causes high distortion if measuring the current waveform ( > 10% THD and nasty harmonics at that ). Thus the " real " PSU needs of a TD150/LP12 are voltage and phase. If you own a Linn Valhalla PSU change the 47uF 250V caps to 220uF 250V. If the crysal is changed to one giving 45 RPM 50 Hz mains ripple will be seen at about - 40 dB ( see below ). With 220uF it becomes about -52 dB. The Valhalla is highly asymetrical. If the mains input is floated it improves to about -60 dB. I used a valve transformer usually for EL84 to do that. Having tested a Valhalla to about 280 VAC the 252 VAC I had was OK. The Valhalla is a worthy discussion topic on a Naim clone thread as it is an all NPN output stage. Running at 330V ( +/- 165 ) only NPN will cope. The Garrard is rated at 16 watts, that is 28VA typical. Using the 6 to 1 crest factor that would need a 168 watt design for music. That is about right. The Valhalla is a supurb design that needs a minor upgrade.
In the ideal world a Garrard would be rewound for 18 to 24 V rms.
One option on the Garrard design is to take the Valhalla circuit and run it at 90 Vrms output. If in bridge mode this could be 180 V rms. This reduced to 110 Vrms would suit the 110 V coil option all Garrard 301/401 have. If so engineering options are almost zero. BUL312FP seems to have survived from TV use. Realistic gain is 20 at low current. These would have to be made into an output triple using perhaps MPSA44/94. If you like a Quad 303 at 330VDC. The FP device might not be accepted if chassis mounted. Also a very simple concept is no longer simple. My guess is a minimum of 16 T0220FP ( {3 + driver + MPSA} x 4 and 1R emitter resistors ). Vas might be MPSA44. If we reduced the DC down to +/- 100 V we might have more options. MJE340/350 might be OK. On paper me might just get Exicon 10N/P25 to work.
In the ideal world a Garrard would be rewound for 18 to 24 V rms.
One option on the Garrard design is to take the Valhalla circuit and run it at 90 Vrms output. If in bridge mode this could be 180 V rms. This reduced to 110 Vrms would suit the 110 V coil option all Garrard 301/401 have. If so engineering options are almost zero. BUL312FP seems to have survived from TV use. Realistic gain is 20 at low current. These would have to be made into an output triple using perhaps MPSA44/94. If you like a Quad 303 at 330VDC. The FP device might not be accepted if chassis mounted. Also a very simple concept is no longer simple. My guess is a minimum of 16 T0220FP ( {3 + driver + MPSA} x 4 and 1R emitter resistors ). Vas might be MPSA44. If we reduced the DC down to +/- 100 V we might have more options. MJE340/350 might be OK. On paper me might just get Exicon 10N/P25 to work.