Hi folks! I acquired an old Luxkit A3600 push-pull amplifier a couple of years ago and I've enjoyed getting to know how it works, modifying it to take KT120 output tubes (from KT88s when I got it) and 6GU7s (in place of 6240G tubes that are long out of production). I've modelled the power supply in Duncan amp's PSUD2 application and wondered if anyone had advice about the goldilocks level of damping.
The original power supply consists of:
This is what I came up with:
I wasn't sure what to make of this so I did a bit of searching on the internet about optimising PSU components (shout out to DHTRob and many others) and created a new PSU model to make it faster and remove the overshoot and oscillation (see below).
Note that I increased the first resistor (R1) from 1Kohm to 2.2Kohm because I have recently moved house and I'm now experiencing a higher wall voltage. I can see from my PSUD experimentation that spreading out the capacitance across the amplification stages slows the power supply down, while front loading capacitance reduces ripple experienced by the rest of the amplification stages. The total capacitance in the stock configuration is 332uf and 451uf in the second model. The residual ripple for the output stage (I1) is less than half the stock config (1.3V vs 560mV diff min to max) and about the same for the phase inverter (I2) and input (R3) stages.
Do you think it is safe to try this configuration or should I reduce the capacitance of C2 and increase the capacitance of C3 and C4 to increase the damping and reduce ripple at those stages? Are there any stability concerns and what difference might I hear (if any)?
The original power supply consists of:
- a low DCR secondary winding labelled as providing 370V with max current of 400mA into
- a full wave bridge diode rectifier
- two multi-section caps (3x 47uf) split into:
- 2x47uf then a 350mH (11ohm) choke coil
- 2x47uf (provides B+ for KT120 output tubes) then a 2.2Kohm resistor feeds LTP phase inverter stage (twin triode 6GU7 tubes)
- 1x 47uf then a 100kohm resistor feeds input signal preamp stage (twin triode 6AQ8 tube)
- 1x 47uf then a 240Kohm bleeding resistor
This is what I came up with:
I wasn't sure what to make of this so I did a bit of searching on the internet about optimising PSU components (shout out to DHTRob and many others) and created a new PSU model to make it faster and remove the overshoot and oscillation (see below).
Note that I increased the first resistor (R1) from 1Kohm to 2.2Kohm because I have recently moved house and I'm now experiencing a higher wall voltage. I can see from my PSUD experimentation that spreading out the capacitance across the amplification stages slows the power supply down, while front loading capacitance reduces ripple experienced by the rest of the amplification stages. The total capacitance in the stock configuration is 332uf and 451uf in the second model. The residual ripple for the output stage (I1) is less than half the stock config (1.3V vs 560mV diff min to max) and about the same for the phase inverter (I2) and input (R3) stages.
Do you think it is safe to try this configuration or should I reduce the capacitance of C2 and increase the capacitance of C3 and C4 to increase the damping and reduce ripple at those stages? Are there any stability concerns and what difference might I hear (if any)?
What is the 330mA to 430mA I1 step timing? Perhaps make that step occur at 1.1 sec, and change the reporting to say 400ms after a 1 sec delay. It is the response to a step load change of the output stage that is the issue imho, as the start up response (t=0) has no influence on amp performance, and is an unreal scenario given you have current taps.
I would go with the second design with C2 having 367uF and providing better filtering for the KT88 tubes.
Thanks, for the responses! Here is what happens with the 100mA step change in current at I1 (330mA to 430mA) at 3.1sec in the experimental version:
I've done something similar for the original configuration but I had to stretch the delay out to 30 seconds because of the high capacitance (58uf) of C4, which stretches out the voltage rise from startup. This value also means the 260V rail doesn't change much in response to the current step at 30.1 secs in the 500ms I've allowed.
Here, I've showed the response of the 260V rail between 30-35 seconds and it still isn't finished.
Is it better for the 260V power supply for the input signal tube to be so damped? And perhaps it doesn't matter much that the 500V B+ is a bit wobbly and has over 1V p2p ripple if the PP output transformer sorts it out?
I've done something similar for the original configuration but I had to stretch the delay out to 30 seconds because of the high capacitance (58uf) of C4, which stretches out the voltage rise from startup. This value also means the 260V rail doesn't change much in response to the current step at 30.1 secs in the 500ms I've allowed.
Here, I've showed the response of the 260V rail between 30-35 seconds and it still isn't finished.
Is it better for the 260V power supply for the input signal tube to be so damped? And perhaps it doesn't matter much that the 500V B+ is a bit wobbly and has over 1V p2p ripple if the PP output transformer sorts it out?
I would also go for better diodes in the bridge, Parts Connexion has some high speed Vishay diodes that are 3 amp 1.5kv that will be more durable and have faster recovery.
TAG, I would like to change to the second version to improve ripple at the output stage! I have the capacitors with these values already...
TR, I would like to understand why the step change from zero to full voltage has no relevance to the behaviour of the power supply - is it only a coincidence there is ringing shown in both the current step change and voltage start up?
BTW- ditch the noisy 1N4007 and use 8x 1200 volt Schottky diodes (two in series each leg) that make no noise.
Start up in a simulation is prone to many unreal conditions - like all loads instantly drawing their 'constant current' even though they have no (or low) voltage across them. The old adage for simulations is 'sh*t in, sh*t out', so it can be a learning curve to separate what may be a reasonable result, to when they need to be deemed not credible.
Even an instantaneous step load change of 100mA from the output stage is somewhat unreal, but has validity in that it can be used to excite any resonant LC circuitry. In your case, I don't see any resonant response from such a change.
For the aim of looking at resonances, there is no benefit in simulating the long distribution circuit to R3, as downstream of the output stage the distribution does not include L, and only RC droppers. Ie. just replace I2 and beyond with either a simple current tap, or an equivalent resistance - then there is no need to extend simulation time out to tens of seconds. Certainly if you wanted a sim to include downstream voltages etc, then include those parts, but the point is to appreciate when and where to use simulation.
How did you derive the 19.2 ohm secondary resistance of the PT ?
A sim can allow you to estimate the peak and average diode currents for your final setup, and that can allow you to choose an appropriate diode. If you aren't copying another amp, then you can go for over-kill on the diode current rating, or try and throw some design effort at it (eg. as per this link https://www.dalmura.com.au/static/Power supply issues for tube amps.pdf).
Ciao, Tim
Even an instantaneous step load change of 100mA from the output stage is somewhat unreal, but has validity in that it can be used to excite any resonant LC circuitry. In your case, I don't see any resonant response from such a change.
For the aim of looking at resonances, there is no benefit in simulating the long distribution circuit to R3, as downstream of the output stage the distribution does not include L, and only RC droppers. Ie. just replace I2 and beyond with either a simple current tap, or an equivalent resistance - then there is no need to extend simulation time out to tens of seconds. Certainly if you wanted a sim to include downstream voltages etc, then include those parts, but the point is to appreciate when and where to use simulation.
How did you derive the 19.2 ohm secondary resistance of the PT ?
A sim can allow you to estimate the peak and average diode currents for your final setup, and that can allow you to choose an appropriate diode. If you aren't copying another amp, then you can go for over-kill on the diode current rating, or try and throw some design effort at it (eg. as per this link https://www.dalmura.com.au/static/Power supply issues for tube amps.pdf).
Ciao, Tim
Thanks, TAG - my amp actually has RU-3C diodes installed and I have been contemplating the value of faster and less noisy Schottky diodes...
Thanks, TIm - your assessment of the value of the startup and current step change behaviour makes sense.
I've checked the current peaks at T1, BR1, and C1 and they appear to stay within the limit of the RU-3C diodes my amp has installed, according to the datasheet.
I came up with the 19.6ohm secondary resistance of the PT by a little jiggery pokery I would be happy to receive advice on! I double-clicked on the transformer symbol and clicked on the ... button next to the ohms box to get this:
The primary voltage is 100V (It is a Japanese 100V unit connected to a 240V step down transformer) and the winding resistance of each side is as measured on my DMM (IIRC). These inputs yield the results as shown and the voltage rails were within a volt of what I was measuring. Is that OK?
I've checked the current peaks at T1, BR1, and C1 and they appear to stay within the limit of the RU-3C diodes my amp has installed, according to the datasheet.
I came up with the 19.6ohm secondary resistance of the PT by a little jiggery pokery I would be happy to receive advice on! I double-clicked on the transformer symbol and clicked on the ... button next to the ohms box to get this:
The primary voltage is 100V (It is a Japanese 100V unit connected to a 240V step down transformer) and the winding resistance of each side is as measured on my DMM (IIRC). These inputs yield the results as shown and the voltage rails were within a volt of what I was measuring. Is that OK?
That's the appropriate method to calculate effective winding resistance.
The original CLC values do show some damped ringing at about 50Hz, so if that is your only choke option, then it would be appropriate to modify the C values, such as your other CLC sim values.
The diode rms current at 330mA dc load is quite high at circa 520mA. Perhaps ok if each diode has a low thermal path to ambient, and ambient temp is not too onerous. It's very easy for diode junction temp to get high unless you have a good awareness of thermal design and actual ambient temps during operation.
The original CLC values do show some damped ringing at about 50Hz, so if that is your only choke option, then it would be appropriate to modify the C values, such as your other CLC sim values.
The diode rms current at 330mA dc load is quite high at circa 520mA. Perhaps ok if each diode has a low thermal path to ambient, and ambient temp is not too onerous. It's very easy for diode junction temp to get high unless you have a good awareness of thermal design and actual ambient temps during operation.
Yes, I agree that simplifying the model to I2 makes it run nice and fast! However, I have been interested in how the capacitance and resistance at each stage affects the 'speed' of response. Large differences in the capacitance at the phase inversion and input amplification stages makes a small but noticeable difference to the response at the B+ rail. Front loading the capacitance for ripple rejection in the first two caps reduces the need for larger caps in the third and fourth spots and the overall speed of response to current shocks is faster. However, I wonder what if any difference this makes to the way music is reproduced? If there is nothing inherently wrong with this approach then I will try it and report back (I already have all the capacitors but the 1uf one, which is on its way)!
BTW- C1 ripple current is high. Many capacitors will not handle the current so make sure your capacitor can handle the ripple current of 1.27 amps.
Actually, I guess it is better to show steady state currents...
It looks like C1 needs to handle 729mA (RMS)?
It looks like C1 needs to handle 729mA (RMS)?
TAG - The caps I found on Mouser and Digikey (75-82uf 550-600V) have ripple current limits that start at 1A at 120Hz and rise to over 2A at 10kHz. Is that good enough? How did you arrive at 1.25amps ripple current?
The first filter cap experiences the highest ripple current requirement. The 729mArms value in the sim data column is the estimate to work with, depending on your amp's final setup and operating conditions.
Digikey has Rubycon MXH range which has an 82uF 550V snap in part. That part has a 0.80A ripple current rating at 105C for 3,000hr - it is the 100-120Hz multiplier of 1x that is the design requirement. Preferably that cap doesn't sit in a high temperature ambient. Are you saying you have a KT120 quad? What idle bias power were you aiming for?
Digikey has Rubycon MXH range which has an 82uF 550V snap in part. That part has a 0.80A ripple current rating at 105C for 3,000hr - it is the 100-120Hz multiplier of 1x that is the design requirement. Preferably that cap doesn't sit in a high temperature ambient. Are you saying you have a KT120 quad? What idle bias power were you aiming for?