I fancy if the driver transistor ( JLH BC447/48 ) was beefier and pumped more current something might improve. It might go unstable is the risk. This output stage looks very unlikely to resist oscillation, however JLH say better than the Darlington in place of FET version.
JLH tuning the amp to work on a reactive load and then saying it coincided with what his ears liked makes me think the reactive load test might be the one we have all overlooked as to it's corellation with real world distortion. It should favour the JLH / JV 0R22 output rather than the choke. A choke makes an amplifier look unhappy when very likely it isn't ( stored energy ripples, one can damp the choke with 10R 3 W, not sure my ears favours that ). It seems very reasonable to think not having to overload the VAS to keep everything stable might be better.
Would anyone like to discusss if they see the VAS being easier to drive the JLH way? This amplifier has very little input stage current to drive VAS base + capacitance ( make VAS current gain high if you can ). Many said the VAS input impedance can be as low as you like because this is a I to V converter. If you put an oscilloscope from ground or rail to VAS base a highly distorted triangular wave is seen. However a nice sine wave is seen from the VAS colllector. That's all very well up to a point. Two things. The triangle wave is not completely lost and also the charging of Cdom causes the transient effects to vanish. The rest is history and is called Slewing. Thus the experts have a paradox. It's a I to V converter so will not matter. However we need enough spare current to suck the charge from Cdom. The standard fix is to make the input stage current higher and reduce the input stage gm by restoring the internal re of the transistor with an external Re. As far as I can see input pair degeneration is the best way as we keep DC offset low ( I never care about it, others do , 50mV is fine ). We also keep Cdom as it was because we haven't changed the loop gain much. The bonus being we discharge Cdom faster. I really wonder if we do?
I think why MOS FET's work is they keep that bit of the amplifier very fast and do not present a very non linear load to the VAS. Thus this buffer is a buffer ( more so JLH ). This must make life much easier when thinking slewing. Thus the negative feedback is mostly getting the output impedance down and making the input plus VAS more linear ( and jumping the crossover wobble, FET's wobble, bipolars step ). The FET gate should be a capacitive load more or less with vitually zero current required where music is loudest, even at 10 kHz it won't be much as we seldom need 1 watt at that level. JLH makes things worse than the Hitachi design in that the Cgs is no longer bootstrapped to the speaker. However if the BC447/448 can cope with < 1 nF it's fine ( unlike the Hitachi one BC sees one FET, the capacitive load is halved ). However the Cgd to the load is lower than Cgs which is no bad thing.
I haven't been able to find the Cob of BC447/8. They look like BD139/140 would be similar. I will assume 30 pF for now. Stop press, Towers has it at 5pF for both, books beat the internet. Not bad, MPSA42/92 look to be OK if sellected for gain. That's a VAS I can like.
JLH tuning the amp to work on a reactive load and then saying it coincided with what his ears liked makes me think the reactive load test might be the one we have all overlooked as to it's corellation with real world distortion. It should favour the JLH / JV 0R22 output rather than the choke. A choke makes an amplifier look unhappy when very likely it isn't ( stored energy ripples, one can damp the choke with 10R 3 W, not sure my ears favours that ). It seems very reasonable to think not having to overload the VAS to keep everything stable might be better.
Would anyone like to discusss if they see the VAS being easier to drive the JLH way? This amplifier has very little input stage current to drive VAS base + capacitance ( make VAS current gain high if you can ). Many said the VAS input impedance can be as low as you like because this is a I to V converter. If you put an oscilloscope from ground or rail to VAS base a highly distorted triangular wave is seen. However a nice sine wave is seen from the VAS colllector. That's all very well up to a point. Two things. The triangle wave is not completely lost and also the charging of Cdom causes the transient effects to vanish. The rest is history and is called Slewing. Thus the experts have a paradox. It's a I to V converter so will not matter. However we need enough spare current to suck the charge from Cdom. The standard fix is to make the input stage current higher and reduce the input stage gm by restoring the internal re of the transistor with an external Re. As far as I can see input pair degeneration is the best way as we keep DC offset low ( I never care about it, others do , 50mV is fine ). We also keep Cdom as it was because we haven't changed the loop gain much. The bonus being we discharge Cdom faster. I really wonder if we do?
I think why MOS FET's work is they keep that bit of the amplifier very fast and do not present a very non linear load to the VAS. Thus this buffer is a buffer ( more so JLH ). This must make life much easier when thinking slewing. Thus the negative feedback is mostly getting the output impedance down and making the input plus VAS more linear ( and jumping the crossover wobble, FET's wobble, bipolars step ). The FET gate should be a capacitive load more or less with vitually zero current required where music is loudest, even at 10 kHz it won't be much as we seldom need 1 watt at that level. JLH makes things worse than the Hitachi design in that the Cgs is no longer bootstrapped to the speaker. However if the BC447/448 can cope with < 1 nF it's fine ( unlike the Hitachi one BC sees one FET, the capacitive load is halved ). However the Cgd to the load is lower than Cgs which is no bad thing.
I haven't been able to find the Cob of BC447/8. They look like BD139/140 would be similar. I will assume 30 pF for now. Stop press, Towers has it at 5pF for both, books beat the internet. Not bad, MPSA42/92 look to be OK if sellected for gain. That's a VAS I can like.
I remember clearly during my Uni days in winter 1969 when I heard the first JLH 10W constructed appallingly by an Engineer in a shoebox. It used 2N3055, which was SOTA then. The album was Paul Desmond 'Summertime'. My, how things has changed since!
The appearance was terrifying; but the sound was better than most valve/tube amps of the day - except an EL84 Mullard which was very good too, but a little twee.......
The JLH set me on a lifelong pursuit.
HD
The appearance was terrifying; but the sound was better than most valve/tube amps of the day - except an EL84 Mullard which was very good too, but a little twee.......
The JLH set me on a lifelong pursuit.
HD
I think the best amplifer I ever heard considering price was and is the Leak Stereo 20. My BBC friend has a Krell which misbehaved, we used the Leak and found it less musically distorted even when the Krell fault was cured ( wires ). Having found valve designs not to like negative feedback one has to admire the Leak as it has plenty. EL84 seems to be almost a transistor and possibly designed with transistors fast becoming a reality for power amps to see off. The Leak is much tighter sounding than most valve designs.
Hope you won't mind me asking on this thread. Has anyone ever tried this? I need to drive an inductive load at no more than 200 Hz ( an induction motor ). Off the shelf designs have failed or look likely to be gone tomorrow as designs. A LM1875 with 1N4007 x 2 bias and some 2N3055/2955 was considered as nice and simple. However LM1875 seems to be on death row. TDA20/30/40/50 have already gone. I don't like the modern multi pin LM's.
One idea I tried and found to be excellent was a complimentary feedback pair that's just under ideal bias ( 2 x 1N4007 and whatever resistors ). With no loop feedback, just the usual op amp's feedback and gain setting. The amplifier measures much like when a full loop type except more stable. The output impedance isn't prime.
The idea above is a simple class B buffer that uses the imperfect switch off of an FET to advantage. I could add a feedforward resistor or cap. Hopes are for 0.1% THD with < 0.07% above 2 nd harmonic. The amp would have an inverting other side to get 16 Vrms or whatever.
The stability is something I can sort out. My hopes were being zero bias these cheap FET's might be very happy thermally ( could be FET multiples as they are very cheap ). I dare say I can add some bias to nudge them into conduction without going into class AB. This is a project for a friend who want's it very simple. 50 watts would be peak and typical would be 30 watts rms ( 30VA ). Gain would be in a previous stage. FET's are what's to hand. I fancy TL071 will be happier than most.
the opamp has to drive the voltage step in the "dead band" where both devices are off.
That voltage step requires an instantaneous peak current that is way beyond what any opamp or even power amplifier can provide.
It's the feedback that tries to force the opamp into infinite current delivery and it just blows up, or current limits.
That voltage step requires an instantaneous peak current that is way beyond what any opamp or even power amplifier can provide.
It's the feedback that tries to force the opamp into infinite current delivery and it just blows up, or current limits.
I have done it with bipolars and had very OK results. It needs Darlingtons to stand a chance of doing some current. FET's at this low frequency are almost zero current switching ( assuming a bit of bias , 2 x red LED perhaps ). I think I will give it a go.
Here is the classic idea using NE5532 and BD139/140. Note the correction used at 800 mV. Output is not unlike the version without dumpers. The 4 Vrms graphs would be like my use so it looks possible. I am testing the simpler version with purple dot.
Here is the classic idea using NE5532 and BD139/140. Note the correction used at 800 mV. Output is not unlike the version without dumpers. The 4 Vrms graphs would be like my use so it looks possible. I am testing the simpler version with purple dot.
Your sim seems to have omitted the capacitance that has to be driven with this infinitely fast voltage step.
It's the charging of the capacitance by that fast step that requires the driver to output infinite current, or go into current limiting.
I can only see voltage plots for the vertical axes.
Can you sim the currents from the driver output?
It's the charging of the capacitance by that fast step that requires the driver to output infinite current, or go into current limiting.
I can only see voltage plots for the vertical axes.
Can you sim the currents from the driver output?
Last edited:
I think you are right. Getting a critical class B ( not AB ) bias is required. That says Darlingtons to me as I will build one amplifier and he must do the rest. Giving the op amp some gain might be a good idea. One amp I built for this use used a red LED for bias. That was a design where 2.1 VDC was the AB point when T03 darlingtons. It has run for years. 2 x 1N4148 and two Schottky could work. The op amp to the centre.
Thanks Andrew. The TL071 from memory has a 200R resistor in the silicon as protection and the gates might total 3 nF ( less if looking to Cgd as Cgs is bootstrapped to the load ). Fo = 40 to 200 Hz. The losses and low Vrms out look to be other reasons not to bother.
Here is an idea as simple as I need except it needs set up.
https://www.eeweb.com/blog/extreme_circuits/30-watt-audio-power-amplifier
Here is an idea as simple as I need except it needs set up.
https://www.eeweb.com/blog/extreme_circuits/30-watt-audio-power-amplifier
I have done it with bipolars and had very OK results. It needs Darlingtons to stand a chance of doing some current. FET's at this low frequency are almost zero current switching ( assuming a bit of bias , 2 x red LED perhaps ). I think I will give it a go.
Here is the classic idea using NE5532 and BD139/140. Note the correction used at 800 mV. Output is not unlike the version without dumpers. The 4 Vrms graphs would be like my use so it looks possible. I am testing the simpler version with purple dot.
I am not sure whether you are back on subject or discussing the motor subject.
If the latter why are there no catch diodes to protect the output transistors.
Also what is the dc offset voltage and can you get away with capacitor coupling - the motor duty cycle is a bit higher than mains frequency.
Another option instead of booster transistors would be a two diode biased CFP output stage for better rejection of unwanted/delayed load induced signals into the feedback line than a Darlington.
You might need some additional output stability network or whether a simple discrete circuit with less loop gain than an IC op.amp would entail less instability risk.
It was just a question to see if anyone had a future proof idea. The boosted op amps were part of a collection of stepping stones I did for headphone amps. None shown here were thought to be worthwhile, however an opinion should be based on real results. They were better than I guessed and the feedback loop graphs show how nasty the task was for the op amp. Most of the power transistors I could use have protection diodes built in.
I have built similar circuits for motors and they are trouble. The feedforward idea seems to help a bit. As you see distortion can reach target levels.
I had a class D amp running for a while, then it lost the will to live. Here today and gone tomorrow if repairs needed even if I could sort the reliability. Ironically the rather natty protection circuit I came up with might have caused the problem. A boring resistor and relay might have been better.
What I think I will do is the simplest LTP with single VAS and bootstrap CCS. 2 x IRF540 2 x IRFP9540 and just enough bias to clear the distortion at 200 Hz. Zobel and SMPS at 48VDC. If careful with the bias I might not need source resistors, often zero bias ( leakage ) is enough up to 1kHz. Cdom can be larger than usual as 200 Hz is the maximum. The SMPS will get me maximum output for least heat loss and the bootstrap CCS will help. I will have solder strips for the FET's as you Yamaha and others. It makes repairs easier. The SMPS can be any generic one if repairs required. On my class D design the SMPS voltage was good from 90 to 253 VAC. That really is a big deal as to get even 50% efficiency is hard with class AB if burning off surplus volts of an unregulated supply.
I had a very good circuit for this using L165. 2 in series bridged. Alas obsolete. It was happy to work as buffer if inverting gain used ( 10K 10 K ).
The problem I have here is to make it simple usually needs set up. I don't really think a boosted op amp is the answer. A boosted LM1875 would be fine except it is on death row.
I have built similar circuits for motors and they are trouble. The feedforward idea seems to help a bit. As you see distortion can reach target levels.
I had a class D amp running for a while, then it lost the will to live. Here today and gone tomorrow if repairs needed even if I could sort the reliability. Ironically the rather natty protection circuit I came up with might have caused the problem. A boring resistor and relay might have been better.
What I think I will do is the simplest LTP with single VAS and bootstrap CCS. 2 x IRF540 2 x IRFP9540 and just enough bias to clear the distortion at 200 Hz. Zobel and SMPS at 48VDC. If careful with the bias I might not need source resistors, often zero bias ( leakage ) is enough up to 1kHz. Cdom can be larger than usual as 200 Hz is the maximum. The SMPS will get me maximum output for least heat loss and the bootstrap CCS will help. I will have solder strips for the FET's as you Yamaha and others. It makes repairs easier. The SMPS can be any generic one if repairs required. On my class D design the SMPS voltage was good from 90 to 253 VAC. That really is a big deal as to get even 50% efficiency is hard with class AB if burning off surplus volts of an unregulated supply.
I had a very good circuit for this using L165. 2 in series bridged. Alas obsolete. It was happy to work as buffer if inverting gain used ( 10K 10 K ).
The problem I have here is to make it simple usually needs set up. I don't really think a boosted op amp is the answer. A boosted LM1875 would be fine except it is on death row.
Have you looked at LM675? At least that is still listed as active and TI are helpful with simulation examples. http://www.ti.com/lit/ds/symlink/lm675.pdf
It looks like a variation on LM1875. I used these in a Gain Clone which disliked my 8R speaker load to the point of the protection circuit dropping out the signal to a point of annoyance. It might not have done that had I followed the recommendations in the application notes and included an output coil.
However the standing current is 100 m.A. so it could be used for a line or headphone amplifier. I have an idea that Ben Duncan might have done that when he wrote his series of articles for Hi-Fi News a couple of decades back.
Mjona,
As my NAP-140 sims have proved frustrating in terms of THD at 20KHz, I came here where I was told there were some simulations.
Already got by page 10, where I found your first Naim sim, set for square waves.
It's 98 pages to go yet, so can you please take me to the posts where you sim for THD and Bode plots?
As my NAP-140 sims have proved frustrating in terms of THD at 20KHz, I came here where I was told there were some simulations.
Already got by page 10, where I found your first Naim sim, set for square waves.
It's 98 pages to go yet, so can you please take me to the posts where you sim for THD and Bode plots?
In theory you can't hear distortion when that high. The benign 2nd harmonic being 40 kHz. All the same I measure these things to know all is well. If all is below 0.5 % and not too many higher harmonics all is OK.
A JLH 10 watt class A was giving 0.08% mostly 2nd and my version 0.03%. The JLH has almost perfect exponential decay of harmonics, a class A trait. My bigger chalenge to measure it. Some are silly enough to say JLH OK for it's time. Seeing as it really needs an Audio Precission test set to say what it really does I think OK will do very well for me.
What sims did you get? What was the 5th harmonic doing?
A JLH 10 watt class A was giving 0.08% mostly 2nd and my version 0.03%. The JLH has almost perfect exponential decay of harmonics, a class A trait. My bigger chalenge to measure it. Some are silly enough to say JLH OK for it's time. Seeing as it really needs an Audio Precission test set to say what it really does I think OK will do very well for me.
What sims did you get? What was the 5th harmonic doing?
Member
Joined 2009
Paid Member
I thought I'd share a few comments that I provided recently to a builder of this TGM10 amplifier regarding advice when ordering parts against the parts I talked about in the thread:
The information you found in the thread is accurate and would be a great place to start with your parts list. There is a schematic in post #869 which shows the version that I have built. You can cross-check with this diagram.
I haven't finished my journey with this amplifier. There are a two (or more) things that I may change in the future, you might be interested to explore this too.
1) The sound of the Naim amplifier is quite special, it's worth listening. The one thing I didn't like was the treble can sound too forward with some speakers (especially my floor standers with metal dome tweeter that is already got a reputation for being a bit harsh). I might try some tweaks to the amplifier to better voice it to my speakers but I am also considering upgrading the tweeter to a soft dome for better performance.
2) I used a ceramic capacitor (high quality) where the original used a polystyrene type for Cdom (C6 on my as-built schematic) and Traderbam says it's worth trying a polystyrene type of the same value. You could try that too.
3) The speaker disconnect circuit can be improved. This was discussed later in the thread. I posted some updated component values in post #1041 which I haven't tried yet. These changes are to speed up the circuit so that it is more robust. Generally, I don't expect my amplifiers to be abused or to suddenly short out to the power rail so the speaker disconnect circuit is mostly to stop turn on and turn off noises which it works just fine. You can follow my 'as built' schematic and experiment another time.
4) The input capacitor (C1) and feedback capacitor (C4) are tantalum but I thought it would be worthwhile trying Nichicon MUSE bipolar capacitors in one or both positions to learn how much difference it makes. These bipolar aluminium electrolytics are really darn good.
5) My as-built version didn't include an output resistor and inductor coil because they weren't present on the original Naim. I put the option on the pcb as a sensible precaution in case they are needed. I haven't tried to install them, generally I would recommend them as a good idea.
6) Always order a few more parts than you need, some in case you destroy parts during the build, and some for experiments. Often if you buy a few more parts you get a price break too.
Hope this helps.
The information you found in the thread is accurate and would be a great place to start with your parts list. There is a schematic in post #869 which shows the version that I have built. You can cross-check with this diagram.
I haven't finished my journey with this amplifier. There are a two (or more) things that I may change in the future, you might be interested to explore this too.
1) The sound of the Naim amplifier is quite special, it's worth listening. The one thing I didn't like was the treble can sound too forward with some speakers (especially my floor standers with metal dome tweeter that is already got a reputation for being a bit harsh). I might try some tweaks to the amplifier to better voice it to my speakers but I am also considering upgrading the tweeter to a soft dome for better performance.
2) I used a ceramic capacitor (high quality) where the original used a polystyrene type for Cdom (C6 on my as-built schematic) and Traderbam says it's worth trying a polystyrene type of the same value. You could try that too.
3) The speaker disconnect circuit can be improved. This was discussed later in the thread. I posted some updated component values in post #1041 which I haven't tried yet. These changes are to speed up the circuit so that it is more robust. Generally, I don't expect my amplifiers to be abused or to suddenly short out to the power rail so the speaker disconnect circuit is mostly to stop turn on and turn off noises which it works just fine. You can follow my 'as built' schematic and experiment another time.
4) The input capacitor (C1) and feedback capacitor (C4) are tantalum but I thought it would be worthwhile trying Nichicon MUSE bipolar capacitors in one or both positions to learn how much difference it makes. These bipolar aluminium electrolytics are really darn good.
5) My as-built version didn't include an output resistor and inductor coil because they weren't present on the original Naim. I put the option on the pcb as a sensible precaution in case they are needed. I haven't tried to install them, generally I would recommend them as a good idea.
6) Always order a few more parts than you need, some in case you destroy parts during the build, and some for experiments. Often if you buy a few more parts you get a price break too.
Hope this helps.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.