Babowana said:
I will try to remeasure the harmonics using the notch filter as you say. At present, however, I need to save available time and energy . . . So, it might take certain time . . .
Hi Babo,
I thought that I read that the Tektronics scopes could filter user specified
frequencies through the math functions. I don't know if I understood what
I read correctly, or if this is a new feature, or if it is available on all Tek
scopes... But if your scope has this feature, it would probably be a lot
easier and convenient than building/using a notch filter.
Also, it gives me some comfort to see that your results are in the same
ballpark as mine (scope traces and overall opinion of the sonic results).
Good luck brother Babo, and thanks again for all your help!
Robert
ClassA puts a much lower demand on PSU output impedance, particularly at HF.
There's non of that half wave rectified output appearing on the supply lines and thus none of the HF artefacts on the rails requiring special measures to keep these artefacts out of the speaker signal.
ClassA requires high current continuously.
Regulators do not help by offering low output impedance when the circuit can gain little benefit.
Regulators dissipate all that extra heat when ClassA draws their bias currents continuously, requiring even bigger heatsinks and more space/weight and more energy input.
There's non of that half wave rectified output appearing on the supply lines and thus none of the HF artefacts on the rails requiring special measures to keep these artefacts out of the speaker signal.
ClassA requires high current continuously.
Regulators do not help by offering low output impedance when the circuit can gain little benefit.
Regulators dissipate all that extra heat when ClassA draws their bias currents continuously, requiring even bigger heatsinks and more space/weight and more energy input.
Re: Trafo
I think this is going to depend on a couple of things. Of course, the biggest one is
going to be what parts are available to you. It is hard for me to address that issue,
so I will skip it. You need to consider the dropout voltage of your regulator, and the
dropout from your pass transistor (or other circuit influences). There are also things
like what voltage your transformer will deliver under load (vs unloaded). There are
also some heat dissipation issues that will probably have to be handled.
I started with a 22V (loaded) transformer secondary. After rectification and a CLC
filter, I had about 28.5V to my regulator circuit. I used LM317/LM337 parts, with a
pass transistor (Zen 5 power supply with modifications for regulation - you can see
a circuit diagram in the F5 thread). With a IRF240/9240 MOSFET pass transistor,
there was about a 4V dropout (per rail). With the TIP33/34 bipolar pass transistor,
there was about a 1V dropout. The pass transistor then fed another filtering cap.
The data sheets for the 317/337 parts also show a circuit. The 317/337 parts are
adjustable, so you can set the regulated voltage output, with a minimum dropout being
about 1.6V for my circuit.
So I had 28.5 - 1 (for the bipolar pass transistor) = 27.5V
27.5V - 1.6 (for the regulator) = 25.9.
25.9 - 1.9 (for regulator "headroom" for the varying voltage in my area) = 24V.
Each pass transistor had to dissipate 1V * 1.3A = 1.3W of power. In this case,
it would be 4 times higher if you used the MOSFET.
The voltage regulator passed about 30mA of current to the pass transistor base,
and had a total voltage drop of about 3.5V, so it had to dissipate 3.5V * 0.03A =
105mW. It should be noted that these dissipation figures are for steady state.
Startup is quite a bit harder on them.
I use very small heat sinks on the regulators. I mount the pass transistors on the
same heat sinks as the output transistors (an effort to keep temperatures of the
transistors synced).
I have been using this circuit for about 4 months now (I think). I realize that
sound quality is subjective, but I am firmly convinced that using the regulated
supply made a positive difference. You can read about Babowana's experience
in this thread, and mine in the F5 thread. I think our opinions about this topic are
similar. I wish I could tell you exactly why the regulated power supply makes a
difference, but right now, that is beyond my understanding of this subject. I don't
want to mislead anyone. I have my theories, but I have not tried to prove any of
them yet.
I hope that helps. Please let us know how your project goes!
Good luck,
Robert
kimarin said:
I think this is going to depend on a couple of things. Of course, the biggest one is
going to be what parts are available to you. It is hard for me to address that issue,
so I will skip it. You need to consider the dropout voltage of your regulator, and the
dropout from your pass transistor (or other circuit influences). There are also things
like what voltage your transformer will deliver under load (vs unloaded). There are
also some heat dissipation issues that will probably have to be handled.
I started with a 22V (loaded) transformer secondary. After rectification and a CLC
filter, I had about 28.5V to my regulator circuit. I used LM317/LM337 parts, with a
pass transistor (Zen 5 power supply with modifications for regulation - you can see
a circuit diagram in the F5 thread). With a IRF240/9240 MOSFET pass transistor,
there was about a 4V dropout (per rail). With the TIP33/34 bipolar pass transistor,
there was about a 1V dropout. The pass transistor then fed another filtering cap.
The data sheets for the 317/337 parts also show a circuit. The 317/337 parts are
adjustable, so you can set the regulated voltage output, with a minimum dropout being
about 1.6V for my circuit.
So I had 28.5 - 1 (for the bipolar pass transistor) = 27.5V
27.5V - 1.6 (for the regulator) = 25.9.
25.9 - 1.9 (for regulator "headroom" for the varying voltage in my area) = 24V.
Each pass transistor had to dissipate 1V * 1.3A = 1.3W of power. In this case,
it would be 4 times higher if you used the MOSFET.
The voltage regulator passed about 30mA of current to the pass transistor base,
and had a total voltage drop of about 3.5V, so it had to dissipate 3.5V * 0.03A =
105mW. It should be noted that these dissipation figures are for steady state.
Startup is quite a bit harder on them.
I use very small heat sinks on the regulators. I mount the pass transistors on the
same heat sinks as the output transistors (an effort to keep temperatures of the
transistors synced).
I have been using this circuit for about 4 months now (I think). I realize that
sound quality is subjective, but I am firmly convinced that using the regulated
supply made a positive difference. You can read about Babowana's experience
in this thread, and mine in the F5 thread. I think our opinions about this topic are
similar. I wish I could tell you exactly why the regulated power supply makes a
difference, but right now, that is beyond my understanding of this subject. I don't
want to mislead anyone. I have my theories, but I have not tried to prove any of
them yet.
I hope that helps. Please let us know how your project goes!
Good luck,
Robert
AndrewT said:
HI Andrew,
Please do not take my comments as arguments. I am trying to gain understanding.
I am unqualified to make any comments about PSU output impedance. I am in
agreement about Class A amplifiers, in general, needing relatively high current and
on a continuous basis.
Speaking specifically about the F5 now:
I think the regulator circuit that I designed can deliver all the current that the F5
requires when driving a "reasonable" load. Ie: this regulator will get strained at
about the same current levels that other components in the F5 circuit will get
strained.
It is my opinion that the amount of power that my regulated circuit dissipates is
insignificant (< 5%) compared to the amount of power that the circuit already
has to dissipate. Have I misunderstood or miscalculated something?
My (possibly incorrect) understanding of a common source architecture like
what the F5 uses, is that it will amplify any noise on the power rails. Hence,
Nelson's suggestion of a very quiet power supply. So it seems logical to me
that the quieter one can get the power supply, the lower the output noise of
the amplifier. I'm sure there are limits to this, but in general, is this not so?
I think this architecture will even put signal onto the power rails. As I use a
single, center tapped transformer for the + and - rails, the 317/337 regulators
significantly help reduce noise going from one rail to the other. At least I
think they are supposed to provide some isolation (about -60db if I recall
correctly).
I'm going off of memory, so I may be wrong, but I think the lowest ripple that I
was able to achieve on my power rails, not using the 317/337 circuit was about
4 to 10mV of ripple. I tried several voltage reference methods, and I think this
was using the shunt regulator. I should have taken better notes... More to
the point is that whatever it was, I could measure the ripple with my scope. With
the 317/337 circuit, I cannot measure any ripple with my scope. The trace is
ruler flat, and posted in the 400A thread.
Also, the line voltage where I am seems to vary a good bit. The 317/337
based regulator circuit has been the most effective circuit I have used so far
to handle this issue. It has been far more effective than the zener diode or
shunt regulator circuits that I have tried. The DC offset on my amps are now
stable to plus or minus 2mV (once they reach steady state). I haven't even
noticed a change when the Air Conditioner turns on (I live in Texas, where it is
95F or so outside right now). I think the shunt regulator got things down to about
+/- 10mV and the zener references were about +/- 20mV. And these last 2 were
in the winter (ie: without the Air Conditioner load changing things).
Would any of those arguments change your mind or have merit? Again, I'm
not trying to argue (in a hurtful way), I'm trying to learn. So if there are areas
or topics that I have misunderstood, and if you have a desire to teach, please
feel free to show me the way.
thanks,
Robert
Re: Re: Re: Trafo
Whoops! You're right about the power dissipation.
The advantage of the bipolar over the MOSFET is that the _minimum_ dropout of
the bipolar is about 1V, while the minimum dropout of the MOSFET is about 4V.
So that would be 4.5V * 1.3A = 5.85W, which hurts my argument about power
losses a little.
AndrewT said:
Whoops! You're right about the power dissipation.
The advantage of the bipolar over the MOSFET is that the _minimum_ dropout of
the bipolar is about 1V, while the minimum dropout of the MOSFET is about 4V.
So that would be 4.5V * 1.3A = 5.85W, which hurts my argument about power
losses a little.
To me, its another engineering decision. Your point that Class A circuits are less demanding on the power supply is well taken.
Here are a couple of papers on Regulated PS that Nelson has on PASSDIY:
Power Supplies
and
Zen V3
I realize the F5 probably has a much greater PSSR than Zen V3, but both are class A
Doug
Hi Jack,
this value seems negligeable. Do you think the value would vary much with voltage and VA ratings? I guess even following what you ahve done would be a good start even if my transformer specs are different.
BTW Andrew, Levinson regulate their rails on class A amps. I hear you, but Im sure it makes a difference if implemented correctly.
I have built very few ClassA amplifiers.
I am no expert.
Go back and read what I have said:
Yes, there are probably some benefits, but your decision is whether the extra resources are worth the extra performance.
This is not an argument, it is a discussion, where I and others are offering opinions. Weigh up the pro and cons and spend where appropriate.
I am no expert.
Go back and read what I have said:
eg Levinson regulate their rails on class A amps
Yes, there are probably some benefits, but your decision is whether the extra resources are worth the extra performance.
This is not an argument, it is a discussion, where I and others are offering opinions. Weigh up the pro and cons and spend where appropriate.
The F5 will drive an 8ohm speaker to +-5Apk and stay in ClassAaudiorob said:
Peak currents into that same speaker could be around 10Apk.
The regulator and any subsequent capacitors must be able to supply that transient peak. Let's assume half the transient comes from the capacitors and the other half comes directly from the regulator. That gives us the maximum regulator current before protection triggers of ~5Apk. For a 4ohm speaker that regulator maximum rises to ~15Apk (still 5A coming from the capacitors).
Now look at voltage, the mains here is allowed to vary roughly +-6%.
The regulator must be designed to not drop out at full load and lowest supply voltage.
At nominal mains supply voltage, the voltage drop across the regulator will have increased cf the design voltage.
Let's suppose we use a 115:22Vac transformer on a 120Vac supply. At full load and minimum supply voltage the secondary peak will be ~29.8Vpk.
The maximum DC after the rectifier will be ~29.1V and the minimum will be approximately 27.5V.
The regulator sees the average of 28.3V and is passing 2.6A (ClassA bias)
The maximum output voltage will be ~27.5 - dropout voltage ~= 26V
At nominal supply voltage the regulator sees an average of ~30.4V and passes 2.6A.
The power dissipated in the +ve regulator is ~2.6 * [30.4-26]~=11.4W
The amp dissipates ~ 2.6 * 26V ~ 67.6W
The regulator dissipates an extra 17% of energy when supply voltage is nominal.
Some of your listening will be done when the mains runs in excess of nominal. During these potentially quite extended periods the regulator dissipation will probably be over 20% and occasionally over 25% of the amplifier dissipation. That is a lot of extra heatsink, and energy and semiconductor resource.
One may find that the 22Vac transformer allows the regulator to drop out in worst case conditions. One may have to reduce the output voltage below the 26Vdc I assumed. This will further increase the regulator dissipation.
AndrewT said:
Generally, I think I agree with the concepts of the rest, but I have
some confusion about these first two statements.
Let's say we are using an F5 amp, as Nelson designed it (ie: +/-
24V rails, 1.3A bias/rail), driving an 8R simple resistive load (not
a speaker, and not into any error conditions).
Isn't the maximum amount of current that is delivered into a load
determined by the output voltage the F5 can produce, and the
resistance of the load the F5 is trying to drive? If we have
+/- 24V rails, the output MOSFETS will loose about 4V, and the
source resistors will consume about 0.5V. There is a little give here.
My intention is not to argue the number significant digits so I'm just
using ballpark figures. So the output MOSFETS can deliver about
19.5V. Let's just round that to 20V.
Using the I=V/R form of Ohms law we have 20V/8R=2.5A.
So the maximum current the F5 wil deliver into 8R is about 2.5A.
Is there something wrong about this analysis?
If the analysis is correct, then the only way the F5 will deliver 5A of
current into an nominally rated 8 ohm speaker, is if that speaker's
impedance drops to 4 ohms. While this _could_ happen, it is not
guaranteed to happen. This will depend on the specific speaker and
the rest of the speaker system (the speaker, any enclosure, etc.).
Or am I mistaken? I don't understand where you get the 5A, much
less the 10A figures.
thanks,
Robert
It looks like I was mistaken and assumed the bias was 2.6A.
If the F5 is biased to 1.3A then the maximum ClassA current is ~2.5Apk not ~5Apk.
The maximum transient current is very much music and speaker dependent. It can approach [Vpk / Speaker nominal impedance / factor].
A moderate speaker load can have a factor of 0.4 to 0.3
A severe speaker load can have a factor 0.3 to 0.2
This indicates that the worst case (music) transient peak currents can be up to 5times higher than what the nominal impedance would predict.
25Vpk / 8ohm / 0.32 predicts ~ <=10Apk for short term transients.
The F5 will deliver this by going into ClassAB mode.
With a 4ohm or 2ohm speaker connected, many of the high SPL transients will make the F5 go into the AB mode. The 8ohm ohm load will do this much less often.
If the F5 is biased to 1.3A then the maximum ClassA current is ~2.5Apk not ~5Apk.
The maximum transient current is very much music and speaker dependent. It can approach [Vpk / Speaker nominal impedance / factor].
A moderate speaker load can have a factor of 0.4 to 0.3
A severe speaker load can have a factor 0.3 to 0.2
This indicates that the worst case (music) transient peak currents can be up to 5times higher than what the nominal impedance would predict.
25Vpk / 8ohm / 0.32 predicts ~ <=10Apk for short term transients.
The F5 will deliver this by going into ClassAB mode.
With a 4ohm or 2ohm speaker connected, many of the high SPL transients will make the F5 go into the AB mode. The 8ohm ohm load will do this much less often.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.