I remain curious as global nfb is on the cards to the reason of the 100R preset (or whatever) in the joint 1st cathodes. I use a similar front end circuit with joined common cathodes and dispensed with any resistor as nfb coupled with the near perfect AC CCs impedance transparency is a wonderful corrector of errors.
Only my views, but I also notice the enormous gate capacitor values on the power supply multiplier regulators. Is that "over kill" really necessary ? More important is the correct operating zener current slope. The power (enhancement mode) mosfet will already provide single mV ripple levels with much lower C values.
If global nfb is decided, any PSU ripple will drop proportionally on the output. There is also the issue that the gate-source protection diode has a low current handling, it can be destroyed with the combination of a fast input voltage decay and a large gate cap having to discharge though it consequently destroying the FET. The mosfet inbuilt reverse diode can take this. I´ve come across many circuits in amps akin, that with unforeseen operating circumstances the pass mosfet can be easily zapped to death.
Bench Baron
Only my views, but I also notice the enormous gate capacitor values on the power supply multiplier regulators. Is that "over kill" really necessary ? More important is the correct operating zener current slope. The power (enhancement mode) mosfet will already provide single mV ripple levels with much lower C values.
If global nfb is decided, any PSU ripple will drop proportionally on the output. There is also the issue that the gate-source protection diode has a low current handling, it can be destroyed with the combination of a fast input voltage decay and a large gate cap having to discharge though it consequently destroying the FET. The mosfet inbuilt reverse diode can take this. I´ve come across many circuits in amps akin, that with unforeseen operating circumstances the pass mosfet can be easily zapped to death.
Bench Baron
Gnfb will be switchable, i intend not to use more than 6dB. 100R pott will set perfect balance for AD1 driver tubes, no matter if the EC8010 are perfectly balanced.
I tested even 180uF there, a big value ensures output voltage to rise gently to nominal. 150u gives aprox 15s to rise half voltage. This will be welcomed considering that i am using direct coupling. My test show that will be good if this delay will be at least 20s, 25s is better. If i am not satisfied about the result i will use an delay relay set at 40s.
I tested even 180uF there, a big value ensures output voltage to rise gently to nominal. 150u gives aprox 15s to rise half voltage. This will be welcomed considering that i am using direct coupling. My test show that will be good if this delay will be at least 20s, 25s is better. If i am not satisfied about the result i will use an delay relay set at 40s.
I may be wrong but it seems that the drive voltages of the 845's are anti-phase which is normal for pp operation but not for SE. So make sure that the output signals from the interstage transformer are in phase to prevent a short circuit for ac at the 845 anodes (and possible destruction of the tubes).
No, it is not necessary for a follower, but 150V zeners are fairly noisy.
Increased Rx and non-leaky DC-Link caps give slow rise time, and low value Cs can be used
More importantly, the high values of output capacitor on the FET sources are not necessary, and will degrade the impedance - vs frequency flatness.
Small-value DC-Link capacitors are all that is required, and improve the sound greatly.
AC heat on AD1s produces a long series of IMD spurs, regardless of PP, PSE or any other configuration. KT88, GU-50 etc will sound better, in this case.
I do not see why it may not be better to use both, high value electrolytic and small-value DC-Link?
I am prepared to use DC heat on AD1. I will see.
I do not see why it may not be better to use both, high value electrolytic and small-value DC-Link?
A follower may be a simple circuit, but it is an active circuit. Combining a MOSFET with other randomly chosen parts may well appear to power up and work, but the key to a good audio circuit is even handed behaviour across the full audio band.
THe small-signal output impedance of the follower is a small resistance, whose value is set by the FET channel and the standing current.
If a capacitor is connected across the output, a frequency dependant impedance meets the resistive output impedance. But carelessly chosen electrolytics are big enough to form an impedance crossover in the audio band; this gives needless phase rotation and a lump in the response. Remember, the follower carries the full signal current swing!
Keep in mind that if the supply input gets shorted, the large energy of this big cap has to get burned up in something, on the way out, another unnecessary flaw.
OTOH....
A follower combined with a small DC Link (Power MKP), of a few uF, designed to cross over above the audio cut-off is ideal. It offers a small resistance right across the audio band, and little or no phase shift. The parasitic L and R are also vanishingly small.
Perfect. This is audible, of course.
Simple circuit? Yes, but it requires design work to work optimally.
A follower may be a simple circuit, but it is an active circuit. Combining a MOSFET with other randomly chosen parts may well appear to power up and work, but the key to a good audio circuit is even handed behaviour across the full audio band.
THe small-signal output impedance of the follower is a small resistance, whose value is set by the FET channel and the standing current.
If a capacitor is connected across the output, a frequency dependant impedance meets the resistive output impedance. But carelessly chosen electrolytics are big enough to form an impedance crossover in the audio band; this gives needless phase rotation and a lump in the response. Remember, the follower carries the full signal current swing!
Keep in mind that if the supply input gets shorted, the large energy of this big cap has to get burned up in something, on the way out, another unnecessary flaw.
OTOH....
A follower combined with a small DC Link (Power MKP), of a few uF, designed to cross over above the audio cut-off is ideal. It offers a small resistance right across the audio band, and little or no phase shift. The parasitic L and R are also vanishingly small.
Perfect. This is audible, of course.
Simple circuit? Yes, but it requires design work to work optimally.
Could you be more specific? you have all the components values at hand.
I am using this circuit for years, and i get frequency response of the stages that it supplies from 0 to 160kHz, with perfect square wave response.
I am using this circuit for years, and i get frequency response of the stages that it supplies from 0 to 160kHz, with perfect square wave response.
When I will have time, I will write an email and ask "Accuphase" engineers to to explain to me what this miserable follower will do to the signal.
For now I am more concerned about what two cascaded transformers will do.
For now I am more concerned about what two cascaded transformers will do.
Got an response email from Accuphase engineers, they said that i need a master degree in electronics to understand what happen in this regulator. But there is a simple way they said, take a simple measurement, scope the output loaded in situ.
So i scoped the output of Reg2 in the amplifier for 10Vrms output into 8ohm at different frequencies. I got no more than 50mVrms at different frequencies.
From here the calculus is simple:
Reg1+Reg2 in series gives 2*50mVrms=0.1Vrms distortion signal(lets call it distortion signal).
10Vrms gets reflected on primary as 10x26,5=265Vrms (26.5 is turns ratio)
So the ratio of the signal/distortion signal is 265/0.1=2650. That`s -68dB or 0.03%. This is far away from what is gossiping here in forum, like 1%.
So i scoped the output of Reg2 in the amplifier for 10Vrms output into 8ohm at different frequencies. I got no more than 50mVrms at different frequencies.
From here the calculus is simple:
Reg1+Reg2 in series gives 2*50mVrms=0.1Vrms distortion signal(lets call it distortion signal).
10Vrms gets reflected on primary as 10x26,5=265Vrms (26.5 is turns ratio)
So the ratio of the signal/distortion signal is 265/0.1=2650. That`s -68dB or 0.03%. This is far away from what is gossiping here in forum, like 1%.
It's very curious that you should ask Accuphase about some DIY power supply you have. Even more curious that they should respond to such a strange request.
Accuphase always concentrated on design of transistor amplifiers with (unsurprisingly) a phase accurate response.
DIY Power supply voltage stabiliser with MOSFET power follower can work very well, but adding too much capacitance to the output will spoil the performance - and most of all the Phase response will be altered.
Just what Accuphase would avoid!
But if you are happy to judge the performance with a single number, that's your choice.
For anyone else, careful choice of the capacitor is advisable.
Accuphase always concentrated on design of transistor amplifiers with (unsurprisingly) a phase accurate response.
DIY Power supply voltage stabiliser with MOSFET power follower can work very well, but adding too much capacitance to the output will spoil the performance - and most of all the Phase response will be altered.
Just what Accuphase would avoid!
But if you are happy to judge the performance with a single number, that's your choice.
For anyone else, careful choice of the capacitor is advisable.
I totally agree with you, the fenomena exist, I just wanted to show you at what scale it shows itself.
You are the first guy that I know that can detect phase distortion by ear, good for you.
You are mean when you say that is just an number, is a real life measurement that show the magnitude of the phenomena.
Interesting enough "the distortion signal" resembles a sine wave, another good news.
Almost forgot: I talked the measurements without GNFB.
You are the first guy that I know that can detect phase distortion by ear, good for you.
You are mean when you say that is just an number, is a real life measurement that show the magnitude of the phenomena.
Interesting enough "the distortion signal" resembles a sine wave, another good news.
Almost forgot: I talked the measurements without GNFB.
Wirewound 10 turn pots are great for power supplies.
I once had an idea to get very repeatable settings of volume controls, and to match gains for stereo, using a 10 turn readout counter.
I tried using Wirewound 10 turn pots for the Volume Controls.
The spike at the beginning of a square wave had about a 100% overshoot.
I will not repeat that mistake.
I once had an idea to get very repeatable settings of volume controls, and to match gains for stereo, using a 10 turn readout counter.
I tried using Wirewound 10 turn pots for the Volume Controls.
The spike at the beginning of a square wave had about a 100% overshoot.
I will not repeat that mistake.
filenet,
Thanks!
I knew you were using the multi turn pots for the power supply, and not as a volume control.
50k Ohms volume control:
I only mentioned that I had a problem with square waves, when I used a 50k wirewound 10 turn pot.
Just trying to prevent anyone else from repeating my mistake.
Thanks!
I knew you were using the multi turn pots for the power supply, and not as a volume control.
50k Ohms volume control:
I only mentioned that I had a problem with square waves, when I used a 50k wirewound 10 turn pot.
Just trying to prevent anyone else from repeating my mistake.
Hey, using the Lundahl LL1623 120mA OPT maybe problem. According to the datasheet, the MAX signal at the primary winding for SE operation is only 270V. With LTspice simulation, the signal voltage at primary is way over...
I believe with parallel 845 SE, it would able to produce over 50 watts rms. Therefore, it needs a BIG output transformer which can handle such high power. Probably a custom made OPT is required.