Huh?
The other way around. Q9 must be absolutely indepent of the ps-dependent Q11-Q19 combo. Make a solid proper psrr cs for q9 whatever. Dont save on r's and q's there fundamentally.
Aah. Now I see. I can considered that before (separate current supplies) when reading about interaction, but using the same reference Q for both seems to be almost universal that I followed along in this amp design.
In the next amp (where I have almost unlimited PCB real estate compared to the small PCB space on this amp bd) I will probably spring for another Q and separate the two current supplies. Very good observation
In the next amp (where I have almost unlimited PCB real estate compared to the small PCB space on this amp bd) I will probably spring for another Q and separate the two current supplies. Very good observation
Would one of them be a NAD3020? Notorious for having no output device emitter resistors.
Schematic available from HiFi Engine.
Me too,. I have never ventured to use less than 0.1 Ohm.
Interestingly the NAD 3030 went in the opposite direction to the 3020 by using 0.47 Ohm resistors. Much too high for good linearity.
NAD 3120- no emitter resistors.
NAD 3130- no emitter resistors.
NAD 310- peculiar output stage half FET, half BJT. no emitter resistors.
NAD 312- no emitter resistors.
I think this explains why NAD amplifiers of that era were prone to blowing up. Just simple thermal runaway, I suspect.
I still have those NAD2240PE's. I plan to recycle the "chassis" and install a PCB of my own design using the case, heatsink, transformer, etc. They were really a rubbish PCB layout - many brown PCB sections from excess heat (especially around the "soft clipping" circuit), crowded ill fitting components. Not well thought out.
I plan to use the higher voltage transformer secondary windings (originally for the upper Class G rails) to supply higher voltage for regulated separate +/- power for the IPS/VAS amp sections.
I plan to use the higher voltage transformer secondary windings (originally for the upper Class G rails) to supply higher voltage for regulated separate +/- power for the IPS/VAS amp sections.
I don't use current mirrors anymore if the signal has to pass one (eg Q5-Q7 and the by Q15 driven Q17-Q13 combo). As a constant current mirror, this self adjusting Q11-Q19 setup is just to simple to serve as a reliable source. I've posted another (Sony) jfet-bjt source with very linear and temperature drift compensation specifications elsewhere on this platform.
Copying is not designing:
I have the impression that many engineers cannot integrate both the DC and AC parameters, to understand that AC is nothing more than 'modulating' the DC setting. Not only the operating values or the flatness of the (closed loop) amplification are important, but the intrinsic (open loop) behaviour under all possible conditions must be considered, verified with simulations and tested in the real world. And with main amplifiers, the signal levels are large and reach towards the rails putting all included devices and the very design itself to the very possible limits.
What does the amplifier do during powering on and off cycles? The correcting feedback loop is not operative or faulty, and many 'normal' things are not normal but doing weird things (for a short time). Hence the relays in the output to protect the load. Sofar, I've only found one design that is intrincic stable on all conditions and needs no protection whatsoever (see my profile). No mirrors, 6 or 8 semi's, resistors and a huge power supply.
Last edited:
Thanks for the ideas MarsBravo. I will definitely add them to the list of ideas to look into.
I've just about wrapped up the optimization of this amplifier design, having worked through optimizing the VBE comp resistor, the bias setting and the TMC compensation component values. (subject of a separate forum post)
The only thing I have left to investigate is how the removal of the phase lead RC in the feedback circuit (R21, C7 on the schematic on #1) affects operation. Peter Baxandall seems to indicate in the Baxandall papers that he thinks it is not needed/appropriate on an amplifier using TMC compensation. Since I have yet to read an explanation that makes clear to me it's function, I can only investigate based on Peter's thoughts. To that end I have built up the left channel on one amplifier with the RC in place, the right channel without. Will be easy enough to test for differences. If only I knew what I was looking for....
I've just about wrapped up the optimization of this amplifier design, having worked through optimizing the VBE comp resistor, the bias setting and the TMC compensation component values. (subject of a separate forum post)
The only thing I have left to investigate is how the removal of the phase lead RC in the feedback circuit (R21, C7 on the schematic on #1) affects operation. Peter Baxandall seems to indicate in the Baxandall papers that he thinks it is not needed/appropriate on an amplifier using TMC compensation. Since I have yet to read an explanation that makes clear to me it's function, I can only investigate based on Peter's thoughts. To that end I have built up the left channel on one amplifier with the RC in place, the right channel without. Will be easy enough to test for differences. If only I knew what I was looking for....
I recently restored a Proton AM-452, and it too had no emitter resistors. Same era and same thinking as NAD maybe. Sounds great though, so I thought it was worth some TLC. Bias was not well regulated, and used fixed resistor values.
I added a trimmer for bias, and placed the sensing transistor on one of the outputs. This stabilized the bias over time/temp, and I was able to increase bias a bit which lowered distortion a little.
Regarding increasing the bias setting. How did you measure the distortion as you adjusted the bias level?
I have seen many advocate increasing the bias level for "improved sound quality" without reference to actually measuring THD levels during the process on the grounds that the amplifier would be more "class A-like". Ouch
I have seen many advocate increasing the bias level for "improved sound quality" without reference to actually measuring THD levels during the process on the grounds that the amplifier would be more "class A-like". Ouch
I was watching FFT in ARTA. Higher order harmonics decreased with the Iq slightly increased, and most importantly, the Iq (and dist) was more stable with time/temp, for example after high power bursts it would settle quickly and nicely. A lot better at cold start too. Also added extra heat sink to be on the safe side. Can't remember the Iq, but it was nothing extreme. I just wrote the set Iq on the heat sink for future reference 
A bit tedious to measure Iq without emitter resistors though. I first ran it off a bench supply to get current limiting if something went wrong, and later removed a supply jumper on the board and inserted an A-meter.
I too have found over-biasing to be beneficial isn several amps, and this is based on actual measurements like above.
A bit tedious to measure Iq without emitter resistors though. I first ran it off a bench supply to get current limiting if something went wrong, and later removed a supply jumper on the board and inserted an A-meter.
I too have found over-biasing to be beneficial isn several amps, and this is based on actual measurements like above.
Last edited:
Sounds reasonable. I would hesitate to call it over-biasing, perhaps correct biasing.
I have found that the bias level for lowest THD varies with the power level being used for the measurement. You definitely have to measure at a high enough signal level to move the signal swing out from the crossover region which can be quite wide with an EF output stage. I have not found, in my limited experience, that the "optimal" bias setting is the same at all power levels. You have to pick what output level to optimize at.
Also, as you seem to be aware, increased bias level = increased heat dissipation. Manufacturers are sensitive to this given the need to avoid overheating at high ambient temps, poor airflow around the amp. Also, larger heatsinks cost more money, etc.
I have found that the bias level for lowest THD varies with the power level being used for the measurement. You definitely have to measure at a high enough signal level to move the signal swing out from the crossover region which can be quite wide with an EF output stage. I have not found, in my limited experience, that the "optimal" bias setting is the same at all power levels. You have to pick what output level to optimize at.
Also, as you seem to be aware, increased bias level = increased heat dissipation. Manufacturers are sensitive to this given the need to avoid overheating at high ambient temps, poor airflow around the amp. Also, larger heatsinks cost more money, etc.
I had exact to this subject a very old application note (long report) - either from RCA or from Motorola (maybe as part of the introduction of RCA's first generation 2N3055). Unfortunately I don't find it. Maybe one of the members know the exact title resp. headline - thank you very much.
This I have found in the moment:
bjt - VBE Multiplier with Emitter Resistance Cancellation - Electrical Engineering Stack Exchange
Difficulties with Class B Amplifier Biasing - Electrical Engineering Stack Exchange
https://www.diyaudio.com/forums/att...ified-diode-bias-circuit-audio-amplifiers-pdf
Feedforward Class-B Amplifier.
Optimizing the VBE Multiplier
Vbe bias generator insights
THERMAL COMPENSATION OF Vbe MULTIPLIER
Vbe multiplier sensitivity
Check out also this thread:
https://www.diyaudio.com/forums/pas...matic-self-biased-overview-3.html#post6443242
Last edited:
mikebeth: Yes, sometimes it's a compromise. Sometimes I don't see any ill effects from the higher bias at higher output levels either.
Back to topic:
I once experimented with an emitter resistor on the bias transistor to tune an over-compensating bias circuit (high bias class AB amp). I was able to get it really stable by 'tuning' this resistor. It also correlated with spice simulation. Thoughts on this? For example, I did not check for modulation of the bias voltage at high output.
Back to topic:
I once experimented with an emitter resistor on the bias transistor to tune an over-compensating bias circuit (high bias class AB amp). I was able to get it really stable by 'tuning' this resistor. It also correlated with spice simulation. Thoughts on this? For example, I did not check for modulation of the bias voltage at high output.
Last edited:
Yes, that makes sense. At higher signal levels the effects of gm doubling or crossover distortion would be less noticeable as they become a smaller proportion of the overall larger signal level. They don't go away, they just get harder to see as they are now a smaller percentage of the output signal.
An emitter feedback resistor is common on common-emitter amplifier circuits, so I guess it could work if tuned properly. I'd have to simulate it in Spice and see what effect it had as opposed to the common method of applying the resistor to the bias transistor collector.
An emitter feedback resistor is common on common-emitter amplifier circuits, so I guess it could work if tuned properly. I'd have to simulate it in Spice and see what effect it had as opposed to the common method of applying the resistor to the bias transistor collector.
I only 'optimized' it for current vs temperature, I never looked at rail voltage variation like you.
One thing I have found interesting lately is to look at the modulation of the bias voltage when the amp is being driven hard. With caps across the bias circuit you will probably see the cap charging and discharging with high level burst input signals.
You can also plot the Iq to see the effect it has, however effects of the quick temperature transients on the transistor chip on Iq will not be seen.
You can however feed bursts to the amp an observe what happens IRL.
One thing I have found interesting lately is to look at the modulation of the bias voltage when the amp is being driven hard. With caps across the bias circuit you will probably see the cap charging and discharging with high level burst input signals.
You can also plot the Iq to see the effect it has, however effects of the quick temperature transients on the transistor chip on Iq will not be seen.
You can however feed bursts to the amp an observe what happens IRL.
Last edited:
Maybe not related to the thread, but my experiments have proven that BJTs with low Vceo are much faster in temp compensation. For example, the MJE340 with Vceo=300V is very slow compared to e.g. BD135 with Vceo=45V. In a class B amplifier, yet with big supply rails, the Vce of the multiplier is adjusted close to +/-1.75Vdc and never exceeds this level as the signal is in phase at the C and E of it. Simply has an offset of +/-1.75V. I've used BD135 in several amplifiers based on D. Self book (second edition - bought at 2000), with +/-60V, +/-74V, +/-80V supply rails without problem. Excuse me if it is out of topic
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.