Copyright © 1998
Cadman Enterprises Ltd
Welcome to the Graham Slee Audio Products Owners Forum Open to all owners plus those contemplating the purchase of a Graham Slee HiFi System Components audio product and wishing to use this forum's loaner program: join here (Rules on posting can be found here) This website along with trade marks Graham Slee and HiFi System Components are owned by Cadman Enterprises Ltd |
1970s Design Indulgence |
Post Reply | Page <1 106107108109110 345> |
Author | ||
Graham Slee
Admin Group Retired Joined: 11 Jan 2008 Location: South Yorkshire Status: Offline Points: 16298 |
Post Options
Thanks(0)
|
|
Another way of looking at the upper and lower output stages is how we always looked at them before the advent of affordable computer modeling. Each half was designed as precisely that - one half - and from it was forecast margins for phase and gain.
Perhaps if we rely on separating the push and the pull, some otherwise unknown characteristics might fall out? An immediate result is seen as the phase margin being far slimmer than if the entire circuit is modeled. Another result is to show that even though the transistors are complementary, the gain margins differ. This should not come as too much of a surprise - PNP's and NPN's are not identical. Perhaps instead of one controlling the other's differences, they could be producing sum and difference signals? This might make sense considering the number of tricks used to counter parasitic oscillations, that doesn't seem to have delivered what was wanted. Perhaps both halves must be made to match, as near as possible, until the gain has dropped at least 10dB below unity bandwidth gain? It might seem weird that one half turning off, while the other half is turning on, could sum to anything other than what the dominant side is doing? In which case, why is it the simulator says so? Old fashioned "slide-rule" or hand calculator methods do not consider everything as SPICE does. Making each side ringing-free, which means a 90-degree phase margin and as close as possible to 20dB gain margin, then we have done things the old way with greater precision. It is then a case of making both halves dB match to a fraction of a dB, because 1dB is 0.25 watts more than 1 watt, and I think sufficient to generate audible sum and difference signals. Just another thing to try. |
||
That none should be able to buy or sell without a smartphone and the knowledge in how to use apps
|
||
BAK
Senior Member Joined: 14 Mar 2010 Location: Kentucky, USA Status: Offline Points: 1744 |
Post Options
Thanks(0)
|
|
"It is then a case of making both halves dB match to a fraction of a dB, because 1dB is 0.25 watts more than 1 watt, and I think sufficient to generate audible sum and difference signals."
And any distortion heard close to the "1st watt" of power is the most noticeable.
|
||
Bruce
AT-14SA, Pickering XV-15, Hana EL, Technics SL-1600MK2, Lautus, Majestic DAC, Technics SH-8055 spectrum analyzer, Eminence Beta8A custom cabs; Proprius & Reflex M or C, Enjoy Life your way! |
||
Graham Slee
Admin Group Retired Joined: 11 Jan 2008 Location: South Yorkshire Status: Offline Points: 16298 |
Post Options
Thanks(0)
|
|
It might be challenging for the average hi-fi Joe to accept that we're listening at a quarter or a third of a watt.
With tone applied (usually 1kHz sinewave), the amplifier meters indicate a reading of 40 at 1 watt, roughly 35 for half a watt, and around 25 for a quarter of a watt. With music playing, the meters are around 20 to 25 most of the time, sometimes going to 30 when particularly loud. The meter ballistics are slow, which suggests peaks are not displayed - just the steady-state average maximum. Digital audio can boast 20dB transient peaks. Vinyl can only boast 14dB. 20dB is ten times, 14dB is five times, but this is referring to voltage. For power, it's different - the square law applies: 14dB is 25 times, and 20dB is 100 times. If the above is true, then we only need 8.33 watts for vinyl and 33 watts for the best transient range that digital can offer. My room size is 9 ft 6 ins wide x 12 ft long x 7 ft 6 ins ceiling height, so there may be louder reflections than in a larger room, but the seating distance for ideal stereo reproduction is roughly the same for all. The power loss might be 2dB more in a larger room, equating to 1.6 times more power. Still, we only need about 13 to 52 watts depending on how good the music medium is, and if some compression has been used in the digital master, we may not need anywhere near 52 watts. The largest transients, being fast and fleeting, might not register accurately with us. Perhaps we don't need super-low distortion on transients, but our sub-1-watt consumption needs to be less cluttered with distortion. Edited by Graham Slee - 27 Jul 2020 at 3:29am |
||
That none should be able to buy or sell without a smartphone and the knowledge in how to use apps
|
||
BAK
Senior Member Joined: 14 Mar 2010 Location: Kentucky, USA Status: Offline Points: 1744 |
Post Options
Thanks(0)
|
|
I should say close to or under the "1st watt"...
|
||
Bruce
AT-14SA, Pickering XV-15, Hana EL, Technics SL-1600MK2, Lautus, Majestic DAC, Technics SH-8055 spectrum analyzer, Eminence Beta8A custom cabs; Proprius & Reflex M or C, Enjoy Life your way! |
||
BAK
Senior Member Joined: 14 Mar 2010 Location: Kentucky, USA Status: Offline Points: 1744 |
Post Options
Thanks(0)
|
|
with my Eminence Beta 8A speakers giving 95dB @ 1watt, they still produce 85dB @ 1/10th watt... and about 80dB @ 1/30th watt...
very revealing
|
||
Bruce
AT-14SA, Pickering XV-15, Hana EL, Technics SL-1600MK2, Lautus, Majestic DAC, Technics SH-8055 spectrum analyzer, Eminence Beta8A custom cabs; Proprius & Reflex M or C, Enjoy Life your way! |
||
Fatmangolf
Moderator Group Joined: 23 Dec 2009 Location: Middlesbrough Status: Offline Points: 8989 |
Post Options
Thanks(0)
|
|
Thank you for sharing all your research into this Graham.
Edited by Fatmangolf - 30 Jul 2020 at 5:46pm |
||
Jon
Open mind and ears whilst owning GSP Genera, Accession M, Accession MC, Elevator EXP, Solo ULDE, Proprius amps, Cusat50 cables, Lautus digital cable, Spatia cables and links, and a Majestic DAC. |
||
Graham Slee
Admin Group Retired Joined: 11 Jan 2008 Location: South Yorkshire Status: Offline Points: 16298 |
Post Options
Thanks(0)
|
|
The 10 TH inductor allows us to see a difference in SPICE results between either half of a push-pull, and both working together. SPICE starts to sum in the low MHz region, and it might be because it calculates each half had base-emitter charge "left-over", or it might merely be an unforeseen fault in the transistor models. I doubt if I ever will know.
The old "slide rule" method of calculation (often doomed by false assumptions - we're only human), dealt with one half at a time. However, left-over charge due to base input capacitance (here it is B to E), was known about by RCA, as shown by cross-coupled driver emitters in early schematics. Unfortunately, I cannot see how to increment bode plot results for different power outputs, and neither can I see how each transistor model could incorporate such a thing. Working on each half in isolation, while also realising that in this mode, the other half will not influence VAS loading, I saw that the gain-bandwidth was shallower in the absence of summing. If, in reality, this is the case, then the HF compensation, taking the amplifier as a whole, would be in error. Could I have been tricked into calculating the wrong break frequency? Another error of mine, quite stupid really, was not pulling the Zobel out of the model during simulations. It was a kind of blindness which went unnoticed until my thinking reached the required level. The Zobel does assist in the wanted roll-off rate, and I'm sure many an amplifier design unknowingly relies upon it, but do we want compensation to be on the power output? In effect, we would be relying on the capability of one resistor and one capacitor absorbing power downstream, rather than the compensation being correct upstream. The "upstream" compensation also uses capacitive filtering, but here the AC currents are small. The type of capacitor takes some modeling in SPICE. However, a few manufacturers still publish resonant curves. I think we're safe to imagine that a 47n at the output is going to suffer resonance, possibly within the gain-bandwidth, than a much smaller ceramic in the voltage amplifier. Without the Zobel, the input stage was revealed as having gain, not controlled by the break frequency compensation built between the VAS collector and IPS emitter. Those who have studied non-inverting RIAA stages will know this one - an additional filter required to carry-on the high-frequency roll-off. The input stage (IPS) has gain decided by emitter resistance/impedance, and it is the same for any degenerated stage depending on the degree of degeneration. The only way this can is compensated is by grounding some of its output via its collector load, which, as it is a frequency break-point, has to be a capacitor. Interestingly, some bipolar op-amps feature such a capacitor in their non-inverting differential leg, while not showing emitter degeneration in their simplified schematics. It is a good indication that degeneration emitter resistors exist; otherwise, no such compensation would be required. The op-amp slew-rate will often give it away. Here I am thinking of the NE5532 dual op-amp. Back to the power amp, and adding this capacitor across the IPS collector resistor, has the same effect as connecting it to ground because the top of the local DC filter capacitor is AC-ground. It is quite useful to note that this type of DC-coupled pair, such a capacitor, was often seen, this being what I scolded myself with! Once re-simulated with the right value picked, the roll-off rate gave the preferred phase and gain margins. Further trimming was given to the driver stages in case of parasitics due to negative impedance - in other words, each base was given a stopper resistor. It was found that the negative push-pull half required a 100pF capacitor from the driver base to its collector (ground), for each half to match. The 2.2-ohm base-stoppers of the OP stage transistors, as recommended by Cordell, were kept. Only then was the Zobel reconnected, where it was found too steep, but by upping the resistor from 10 ohms to 22 ohms, the wanted phase and gain margins returned. On test, it was found that a quiescent voltage (Vq) of approx. 13mV gave the best distortion reading at 1kHz and one watt. This happened to correspond to a balance between the 5th and 7th harmonics, below -80dB and the dominant harmonic being the second. The actual reading was approximately 0.02%. Next up: grounding. |
||
That none should be able to buy or sell without a smartphone and the knowledge in how to use apps
|
||
Post Reply | Page <1 106107108109110 345> |
Tweet
|
Forum Jump | Forum Permissions You cannot post new topics in this forum You cannot reply to topics in this forum You cannot delete your posts in this forum You cannot edit your posts in this forum You cannot create polls in this forum You can vote in polls in this forum |