Thanks. In that case the wire resistance definitely dominates. So we need not push up the current in order to lower the impedance. Lower current, as long as not lower than the maximum current draw from the circuit, would be preferrable to me because there would be less heat dissipation, possibly better stability and longer life.
My heatsinks are running a bit hot on 135mA at the moment. If I use 100mA only, would there be a problem?
My heatsinks are running a bit hot on 135mA at the moment. If I use 100mA only, would there be a problem?
with a shunt regulator there is one over-riding rule:
The Shunt current must be greater than the variation in the client current.
If the client circuit has a quiescent of 20mA and varies by +10mA and -9mA then the output variation is 19mA.
The shunt current must be greater than 19mA when the client is drawing minimum current.
Let's look at an example.
Single ended ClassA amp with quiescent of 20mA and +10mA, -9mA variation with peak signals.
Set the CCS to shunt current + quiescent +1mA = 10+20+1 = 31mA.
At quiescent the CCS passes 31mA, the amp draws 20mA and the shunt passes 11mA.
At maximum peak draw the CCS passes 31mA the amp draws 30mA and the shunt passes 1mA
At minimum peak draw the CCS passes 31mA the amp draws 11mA and the shunt passes 20mA.
Note the variation in the shunt current equals the RANGE of total current that the client circuit draws, i.e. 19mA.
Another way to look at it:
The CCS current must be greater than the highest peak current of the client circuit.
Keep in mind that very few amplifier circuit topologies pass constant current.
All ClassB and ClassAB pass variable current, most ClassA pass variable current irrespective of whether they are single ended or push pull.
A pair of B1s each passing 10mA has a total quiescent current of 20mA. The maximum peak current will be ~15mA/channel i.e. 30mA total.
The CCS must be greater than 30mA. All the recommendations in these threads have been for CCS must more than this and therefore all the recommendations meet that one important rule.
The Shunt current must be greater than the variation in the client current.
If the client circuit has a quiescent of 20mA and varies by +10mA and -9mA then the output variation is 19mA.
The shunt current must be greater than 19mA when the client is drawing minimum current.
Let's look at an example.
Single ended ClassA amp with quiescent of 20mA and +10mA, -9mA variation with peak signals.
Set the CCS to shunt current + quiescent +1mA = 10+20+1 = 31mA.
At quiescent the CCS passes 31mA, the amp draws 20mA and the shunt passes 11mA.
At maximum peak draw the CCS passes 31mA the amp draws 30mA and the shunt passes 1mA
At minimum peak draw the CCS passes 31mA the amp draws 11mA and the shunt passes 20mA.
Note the variation in the shunt current equals the RANGE of total current that the client circuit draws, i.e. 19mA.
Another way to look at it:
The CCS current must be greater than the highest peak current of the client circuit.
Keep in mind that very few amplifier circuit topologies pass constant current.
All ClassB and ClassAB pass variable current, most ClassA pass variable current irrespective of whether they are single ended or push pull.
A pair of B1s each passing 10mA has a total quiescent current of 20mA. The maximum peak current will be ~15mA/channel i.e. 30mA total.
The CCS must be greater than 30mA. All the recommendations in these threads have been for CCS must more than this and therefore all the recommendations meet that one important rule.
Last edited:
What I mean by wire resistance dominates is this:
The picture has not included the impedance for CCS set to 100mA, which would have higher impedance comparing to all other graphs. Even so, we can base on other graphs and assume that the lowest point in the "flat" line for 100mA is about 0.00016R. This is about the same resistance of a 5mm 20aug wire or 12.5mm 16aug wire.
I have thought of mounting the MOSFETs on the chassis on the bottom panel beneath the PCB. In that case the wire length from the MOSFET legs to the load can be minimized to as low as 15-20mm, i.e. the load (opamp) is considered almost directly on the legs of the MOSFETs. But then with this approach, the heat transmitted to the opamp via the wires as well as heat waves from the MOSFETs will increase the temporature of the opamp, something I would definitely want to avoid. For B1, it would probably be OK though. So I still prefer using some thin, long and vertically mounted heatsinks for the MOSFETs. The ones I am using allow CCS current of 150mA running a bit hot but not too hot, and I am sure 100mA would only run warm. I prefer 100mA for long term reliability and peace of mind. With this unique type of heatsink, I can manage to have 30-50mm wires from the MOSFET legs to the opamp power supply pins, which I consider to be quite short, still, some length of thick wire to help with heat dissipation is desirable.
In that situation, the regulator output impedance within the audio bandwidth of 0.00016R is only 1/4 of that of the 16aug wire lengths of 50mm. So even with such pretty extreme implementation, the wire resistance still dominates.
Suppose the opamp draws a current of 30mA (quite extreme as well), at DC with 0.00016R extra resistance an error voltage of 30mA x 0.00016R = 4.8uV is developed. This is still too small to be of any concern.
With 10cm 20aug wire, at 20kHz combined with the inductive reactance the impedance is about 14mR. The error voltage would be as high as 0.42mV. The distortion would probably be quite audible.
No wonder each time I shortened the wire lengths from the MOSFETs to the load I found improvement on the sound!
An extremely low impedance output of a regulator, like this one and Salas v1.2, is necessary to achieve good sound.
Only if you use extreme measures in implementation with wires shorter than about 50mm that you may find increasing the CCS current to lower the impedance beneficial.
Regards,
Bill
Good information.
I draw my opamp circuits in LTSpice and it tells me how much current the opamps draw at what frequencies. I make the AC amplitude in the Small Signal AC Analysis to be 3.54V, which is the peak voltage of my input signal, and check the total sum of all opamp maximum currents. Even with 7 opamps they draw far less than 100mA. I reckon even 20mA may do. Of course, this depends on how we design the opamp circuits.
I can't think of any common audio opamp (filter) circuits that would require over 100mA current, unless you want to drive each opamp with 600R load, etc.
For your intrinsic output impedance in the reg, what is left for the shunt element decides it. So if you run a 20mA drawing opamp with 100mA in reg's CCS you have better reserve than someone that runs a full 150mA preamp on 200mA CCS. On your ohmic wire calculations you are right, but remote sensing is there for addressing that. Its the nH that still lurk...
Sometime ago I posted a question for whether remote sensing can cancel the errors of wire resistance and inductance.
Syn08 was the only person who answered, but without any explanation, and he said that it cancels both.
I can easily understand that it can cancel resistance. But I have not understood how it could cancel inductance with my very limited EE knowledge.
Syn08 was the only person who answered, but without any explanation, and he said that it cancels both.
I can easily understand that it can cancel resistance. But I have not understood how it could cancel inductance with my very limited EE knowledge.
Yes I will be using 100mA CCS for loads between 5mA to 20mA.
In terms of CCS voltage, I understand Iko recommends 15V, and you recommend about 10V for V1.2. Since I have enough CCS current reserve I am thinking about 8V CCS voltage because I DON'T LIKE HEAT!
Do you think that is a problem with enough CCS current reserve?
It can cancel inductance, because an inductor is really just a resistor who's resistance increases with frequency (I don't think cancel is the right word, BTW. Negative feedback can only divide error by a certain amount, reducing its magnitude, but it can't eliminate it completely). Problem comes when the inductance resonates with something at such a high frequency that the amplifier can't compensate for it. IE, the amplifier has to be faster than the sense wire. So what Syn08 said is true in the mathematical case, but the added variables in real life implicate things.
- keantoken
- keantoken
V1.2 does not need much drop. Its hard limit is Vgs+vbe. 10V is safe for mains fluctuations, but had it work on 5 even. Iko will answer 5K.
Kelvin sensing, AKA remote sensing can ignore the drop on the force current carrying lines, but I have not seen anything discussed about inductance.
Because syn08 is a most thorough man and me not even by remote sensing an engineer,
he must have said something positive. Looked it up hastily again, found a link in Wikipedia even, no more news about nH and how it handles it. And I meant real interactions, not leading.Four-terminal sensing - Wikipedia, the free encyclopedia
The ccs in 5k needs more voltage drop because one end of the CCS is at Vin and the other end is at Vout. 1.2 has one end at Vin and the other end at GND, so implicitly it has a high enough voltage drop across it, if you wanted 3.3V output from 1.2 and you supply 7V in, think what will happen.
About the Kelvin connection. I don't see it as canceling the inductance per se. Frequency, inductance, and high current on the same wire will produce a voltage drop. Where the sense wires are connected will affect what they're "reading." The load end of the force wire will have a different "reading" and will need different action from the regulator, than the other end which is at the regulator output.
About the Kelvin connection. I don't see it as canceling the inductance per se. Frequency, inductance, and high current on the same wire will produce a voltage drop. Where the sense wires are connected will affect what they're "reading." The load end of the force wire will have a different "reading" and will need different action from the regulator, than the other end which is at the regulator output.
In Sydney the official main voltage is 240VAC, but I am not sure if lately specified to be 230VAC. In practice, I never measured it to be lower than 240VAC, but often close to or over 250VAC.
10V for mains fluctuations? After the transformer for a 25VAC, it would be 1V fluctuation, wouldn't it?
Do we need such large CCS voltage reserve?
Last edited:
So 15V CCS voltage for 5k, not negotiable? I am running on 13V at the moment and it is a bit hot to my comfort, unless I further reduce the CCS current to be 80mA, or use different types of heatsinks for the CCS.
Last edited:
In Sydney the official main voltage is 240VAC, but I am not sure if lately specified to be 230VAC. In practice, I never measured it to be lower than 240VAC, but often close to or over 250VAC.
10V for mains fluctuations? After the transformer for a 25VAC, it would be 1V fluctuation, wouldn't it?
Do we need such large CCS voltage reserve?
It will work down to 6V for some headroom for ripple, mains line play, if heat is a problem. I prefer 10V if the sink is one piece and chunky. Takes the CCS Mosfet up to a TC relative to the shunt Mosfet. I have seen 223V to 240V steady in different areas of Athens with 230VAC nominal. Or 10V dusk till dawn (Mexican vampire here I come) differences in same areas. Mostly 233VAC +/- 10%. Mine is at 44VAC for instance, more secondary play.
the 2sk170 (or LSK170) that is parallel to the Base/emitter of the BC must operate at <700mV. This one MUST be a high transconductance, low pinch off voltage type.
All the other jFETs have a much higher operating voltage and any cheap low to medium transconductance jFET can be used.
I use bf244b and LSK170b
All the other jFETs have a much higher operating voltage and any cheap low to medium transconductance jFET can be used.
I use bf244b and LSK170b
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.