Kindharman, yes you are right things are getting a little lost in the thread and I honestly hope that our highly knowledgeable friends do not lose interest because of the trend of late.
If I can summarize so far is that the capacitance is 3000uF/Amp still to be confirmed the transformer VA rating is > 150% of the rms power rating (I take it into any load specified).
We are waiting for the final simulations to confirm what capacitor value would become irrelevant. It would appear that the capacitor value is as dependent on both the low frequency signal and the line frequency.
We have had members arguing that one could design for a higher than necessary rail voltage since more energy can be stored at the elevated voltage and that rail sagging would not affect the signal. The argument makes sense but has not been confirmed.
We have concluded that there must be distibuted capacitance on the PCBs as close to the power devices as practical, also the cable runs to and from the PCBs need be twisted tightly together and kept as short as practical.
It was also mentioned that multilayer PCBs where there is a positive, negative and earth plane could be very beneficial because of reducing the inductance.
I am a believer in regulating the front end of the amp, but no-one has actually confirmed whether this is an alternative for large reservoir capacitors - probably not.
Have I missed anything?
If I can summarize so far is that the capacitance is 3000uF/Amp still to be confirmed the transformer VA rating is > 150% of the rms power rating (I take it into any load specified).
We are waiting for the final simulations to confirm what capacitor value would become irrelevant. It would appear that the capacitor value is as dependent on both the low frequency signal and the line frequency.
We have had members arguing that one could design for a higher than necessary rail voltage since more energy can be stored at the elevated voltage and that rail sagging would not affect the signal. The argument makes sense but has not been confirmed.
We have concluded that there must be distibuted capacitance on the PCBs as close to the power devices as practical, also the cable runs to and from the PCBs need be twisted tightly together and kept as short as practical.
It was also mentioned that multilayer PCBs where there is a positive, negative and earth plane could be very beneficial because of reducing the inductance.
I am a believer in regulating the front end of the amp, but no-one has actually confirmed whether this is an alternative for large reservoir capacitors - probably not.
Have I missed anything?
Last edited:
I didnt assume, I already mention to look for the Y parameter to see that the transfer from colector to emiter is very low
DF,
Not looking for optimum C, yet. Just simply bounding C, and really only the lower bound.
You say it might not be true (that there must be a "smallest viable C") but then you proceed to the same place I was!
But calculating the C value is NOT straightforward, unless you assume ripple is small and the load is a constant current! THAT is the whole problem.
And non-approximate (non-"small ripple") mathematical solutions are extremely complex. We COULD go that way. I have found the papers with the math done, but not with a sinusoidal load current, although Basso has a nice solution for an SMPS load. I would probably enjoy the mathematical approach more but it seemed quicker and easier, for the purposes of this thread, to use simulations. I am almost done with the first full round of sims, during which I am automating as much as possible (and making it simple so others can use it), doing a single transformer voltage with different VA characteristics. but other transformers will then be much quicker and easier. I will post the semi-automated simulation setup soon. My only worry is that my one available measured transformer model is not great for more than 120 VA, although I have paralleled those to get 240 and 360 cases as well.
We could just make the transformer voltage higher, too. But we want to know the limits for real transformers of different sizes and voltages, don't we?
Cheers,
Tom
Yes its important to stabilize the front end. Its even better to use a separate PSU for the front end.
150% of the amp rms value for the transformer is far not enough.
The efficiency of an Class AB amp is around 35%. That means that you need 3 times more power then the rms value. For a 2x 100Watt amp you need at least a 600Watt transformer.
This is completely different from any previous values I have seen that I can rely on for reasonable accuracy.
The efficiency of a ClassAB amplifier is around 65% when delivering full power without clipping of the sinewave output.
liching,
That is what this thread is about.
If you know how to design the power supply, please post the mathematical steps that arrive at the component values, starting from the variables for the specifications, and allowing for variable transformer and rectifier diode characteristics.
Tom
liching,
Give me a break, man. You insult us.
Many here are electrical engineers, with vast experience and education.
You still totally miss the point of the discussion.
The "basic electronics" books do not even attempt to show the real solution to the "simple" linear PSU.
Tom
Already done that as posted earlier, it appears not "low" enough for all.
Hope this helps
-Antonio
Tom, you're not wrong, it is surprisingly complex. Barton has a detailed analysis in his book on rectifiers (IIRC), but ISTR he assumed a stiff supply, which is far from the case here.
theres nothing wrong with some sort of nonlinear curve-fit of the sim results
I think you'll find the transformer stuff is actually pretty easy, if you normalise the various transformer parameters to its nameplate ratings (Volts & VA).
all we really care about is:
- magnetising inductance (this is much more useful than mag current)
- primary & secondary Resistances
- primary & secondary Leakage Inductances
we could probably reflect the leakage inductance & winding resistance over to the secondary, and ignore the magnetising inductance entirely (as long as the transformer isnt driven too far into saturation - which gives very "peaky" magnetising current (which is just the unloaded primary current)
choose P_base = xfmr VA rating, V_base = secondary voltage rating, w_base = 2pi*Fac
you can then define Z_base = V_base^2/P_base, L_base = Z_base/w_base.
then normalise the xfmr parameters to these base values (aka Per-Unitising):
VA_pu = VA/P_base = 1 (by definition)
Vsec_pu = Vsec/V_base = 1 (by definition)
w_pu = 2*pi*Fac/w_base = 1 (by definition)
Ls_leak_pu = (effective) secondary leakage inductance/L_base < 1 (typ < 10%)
Rs_pu = (effective) secondary resistance/Z_base << 1 (typ < 5%)
[assuming you reflect leakage & resistance to the secondary. If not you need to normalise a few more parameters]
then you can "create" any arbitrary transformer by de-normalising with the desired VA & Voltage ratings.
Voila, any transformer you like, having the same design characteristics of the original.
likewise you can fiddle with the PU resistance & leakage parameters and create new transformers
Audio mains transformers might be a bit different, but the PU leakage and resistance ought to have similar ranges (a bunch of different xfmr measurements would certainly help though).
theres nothing wrong with some sort of nonlinear curve-fit of the sim results
I think you'll find the transformer stuff is actually pretty easy, if you normalise the various transformer parameters to its nameplate ratings (Volts & VA).
all we really care about is:
- magnetising inductance (this is much more useful than mag current)
- primary & secondary Resistances
- primary & secondary Leakage Inductances
we could probably reflect the leakage inductance & winding resistance over to the secondary, and ignore the magnetising inductance entirely (as long as the transformer isnt driven too far into saturation - which gives very "peaky" magnetising current (which is just the unloaded primary current)
choose P_base = xfmr VA rating, V_base = secondary voltage rating, w_base = 2pi*Fac
you can then define Z_base = V_base^2/P_base, L_base = Z_base/w_base.
then normalise the xfmr parameters to these base values (aka Per-Unitising):
VA_pu = VA/P_base = 1 (by definition)
Vsec_pu = Vsec/V_base = 1 (by definition)
w_pu = 2*pi*Fac/w_base = 1 (by definition)
Ls_leak_pu = (effective) secondary leakage inductance/L_base < 1 (typ < 10%)
Rs_pu = (effective) secondary resistance/Z_base << 1 (typ < 5%)
[assuming you reflect leakage & resistance to the secondary. If not you need to normalise a few more parameters]
then you can "create" any arbitrary transformer by de-normalising with the desired VA & Voltage ratings.
Voila, any transformer you like, having the same design characteristics of the original.
likewise you can fiddle with the PU resistance & leakage parameters and create new transformers
Audio mains transformers might be a bit different, but the PU leakage and resistance ought to have similar ranges (a bunch of different xfmr measurements would certainly help though).
Last edited:
Terry Given,
we could probably reflect the leakage inductance & winding resistance over to the secondary, and ignore the magnetising inductance entirely (as long as the transformer isnt driven too far into saturation - which gives very "peaky" magnetising current (which is just the unloaded primary current)
Does this hold true for both standard transformers wound on a steel core and for toroidal wound transformers. Is there a difference in the saturation of the two that change any of this information you presented or are they similar enough to leave that out of the saturation statement you have made?
Steven
we could probably reflect the leakage inductance & winding resistance over to the secondary, and ignore the magnetising inductance entirely (as long as the transformer isnt driven too far into saturation - which gives very "peaky" magnetising current (which is just the unloaded primary current)
Does this hold true for both standard transformers wound on a steel core and for toroidal wound transformers. Is there a difference in the saturation of the two that change any of this information you presented or are they similar enough to leave that out of the saturation statement you have made?
Steven
So for a 2 x 225W amp, as your Nak Stasis, you'd need a 1350VA transformer.
Truth is that the nominal power figure of the Nakamichi PA-7 is much higher, more like 300W in 8 ohm (the 225W number is for 0.05% THD)
2 times 300W would translate to a 1800VA transformer.
Reality is that the PA-7 runs on a single 700VA for two channels, totally inadequate by your own rule of dumb.
(Papa mentioned years ago that he merely got paid for the Stasis concept by Nakamichi, actual design of the power amps is by Nak. You could ask him yourself how high he ranks the PA-5 and PA-7 designs)
I know that the transformer of the Nak is inadequad for continous full power use. Thats ok for me. However a toroidal can easily deliver much more power then rated for some time.
Last time you show me a picture of a Swiss build amp which costs 50k with a 1kva transformer. In the picture however is only seen a small transformer not more then 200va or so.
Last edited:
The Orpheus lines read : 1050VA total power supply transformers.
The 1050VA total includes the little toroidal you saw on the pics, plus two 500VA toroidals under the black steel box in the middle of the amp.
For a monaural power amp : 1050VA divided by 350W (1% THD, 8 ohm) is 3 times the nominal power rating.
The 1050VA total includes the little toroidal you saw on the pics, plus two 500VA toroidals under the black steel box in the middle of the amp.
For a monaural power amp : 1050VA divided by 350W (1% THD, 8 ohm) is 3 times the nominal power rating.
Are you sure? That there are 2 transformers under the black metal? There was at least 1 brand that put sand in it to make it heavy. How much capacity does they use? I only see 8 small capacitors. Its clever, very clever to let people pay 50k. Piega is also very clever to sell cheap boxes for big money.
350W at 8ohms needs about plus and minus 80 volt supply voltage. A 100 volt elco with the size as in the picture are approx. 1000uF. There are 8 so it means that there are only 4000uF Cap for each PSU. Not much for a 50k amp.
Yeah, I'm sure, they even like low profile custom jobs.
Orpheus started as a sideshow by Anagram technologies in Geneva.
Those guys develop digital solutions they license to others, e.g audio stuff to Cambridge for CDP's, but also for computers, cell phones.
No need for them to screw customers.
Orpheus started as a sideshow by Anagram technologies in Geneva.
Those guys develop digital solutions they license to others, e.g audio stuff to Cambridge for CDP's, but also for computers, cell phones.
No need for them to screw customers.
What is their claim for the high price? 5k is already too much for what I see.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.