According to MR. John Curl, low noise, high transconductance J-fets like 2SK 170 are bad choice for CCS.salas said:
According to what I know from practice, on this circuit of mine, already in service to enough people, cross checking and cross reporting, the 2SK170 there, does the job fine, and does it better than the resistor. They call it the ''R6 mod''. That is the type we have handy because of leftovers from the NJFET simplistic phono matching process and it has very low self noise at least. It does the job in Pass B1 too. There is no reason for not using some other JFET there. Something easy to find like 2N5459 will work too, although I see a bit of a wiggle on my simulator for its current VS time comparing to 2SK170. But I don't think it will practically matter. Nobody claims that this circuit is da best, but its part count is low, its robust, noiseless, works in almost any layout and crazy part substitutions without oscillations, and sounds tonally right.
jameshillj said:
jh, I would avoid using a series reg in front of it. Also, my impression of the salas regulator is that it would not do very well with a load above 150mA. I've simulated that in the past and it did not look too good. Perhaps salas can comment on this.
Are you referring to this one?
http://www.diyaudio.com/forums/showthread.php?s=&postid=1776712#post1776712
If so, I did not get very good results simulating it. I don't trust simulation 100%, but it's usually a good indication on how things can go.
I'd gladly help you, but I think you need a new circuit that can deal with such a load. I could be wrong though.
Hi,
determine the continuous quiescent load current.
Determine the maximum increase in current for worst case output conditions.
The CCS must pass more than that total.
The shunt needs to pass more than the increase.
If the minimum current falls to near zero then the shunt must be able to pass the whole CCS current. The two big questions become:
What is the average dissipation in the CCS pass transistor?
What is the average dissipation in the Shunt transistor.
eg.
A single ended ClassA output stage draws 50mA of quiescent current. The drive circuits draw a further 10mA.
The total current in the quiescent state is 60mA.
The Output stage varies it's current draw by +-20mA.
The minimum current is 40mA and the maximum is 80mA.
The Shunt and CCS use 15mA to ground.
The CCS must pass > 80+15>95mA. Set to 100mA.
The minimum current is 40+15-(CCS current) ~=50mA.
The shunt must be able to pass >100-50> 50mA.
Set the CCS to 100mA, set the amp to work and in the quiescent state the Shunt will pass 100-40-10~=50mA.
Check dissipations in the CCS and Shunt and from there calculate the heatsinking requirements.
To design any Shunt regulator you MUST know the maximum and minimum current draw of the load.
determine the continuous quiescent load current.
Determine the maximum increase in current for worst case output conditions.
The CCS must pass more than that total.
The shunt needs to pass more than the increase.
If the minimum current falls to near zero then the shunt must be able to pass the whole CCS current. The two big questions become:
What is the average dissipation in the CCS pass transistor?
What is the average dissipation in the Shunt transistor.
eg.
A single ended ClassA output stage draws 50mA of quiescent current. The drive circuits draw a further 10mA.
The total current in the quiescent state is 60mA.
The Output stage varies it's current draw by +-20mA.
The minimum current is 40mA and the maximum is 80mA.
The Shunt and CCS use 15mA to ground.
The CCS must pass > 80+15>95mA. Set to 100mA.
The minimum current is 40+15-(CCS current) ~=50mA.
The shunt must be able to pass >100-50> 50mA.
Set the CCS to 100mA, set the amp to work and in the quiescent state the Shunt will pass 100-40-10~=50mA.
Check dissipations in the CCS and Shunt and from there calculate the heatsinking requirements.
To design any Shunt regulator you MUST know the maximum and minimum current draw of the load.
I routinely run its CCS at 250mA to only consume 40mA plus swing on the phono, plus what the shunt circuitry consumes. Why I run it so hot? Because it delivers better firmer tone to my ears that way. I had chosen the IRFP9240s for current waste, and low oscillation possibility for easy variations. Its also things I hear between smaller or heavier Mosfets, I also don't like BJTs as CCS or shunt elements no matter how better they can regulate due to their superb current gain. But don't mind me, this is deeply subjective. Also I find the issue between some more dB or less performance here and there not a real issue because I can always passively pre filter the shunt's input for ripple and HF or RF garbage. This does good to all regs. To the contrary, I don't like the tone when wrapping a Mosfet shunt in an IC pre reg.
The hotter it runs the better, 2 times what is consumed is good, but I prefer more. To get to what is consumed, Andrew gave a very correct way to calculate. My simulator shows it OK in heavy currents, and It is running at 3A, 13V for a 10W T class amp for about a month now. Sounds like a million dollars.
The hotter it runs the better, 2 times what is consumed is good, but I prefer more. To get to what is consumed, Andrew gave a very correct way to calculate. My simulator shows it OK in heavy currents, and It is running at 3A, 13V for a 10W T class amp for about a month now. Sounds like a million dollars.
Thanks Ikoflexor, Andrew and Salas,
Much food for thouht here.
First, Iko,
Yyes that's the Toole shunt Reg all right - I cranked it up asfar as it would go until just below dropping out of Reg and the CCS was passing about 280mA with the shunt sinking only about 30mmA - so not the best or perhaps even realistic, way of operating the reg.
So my dissapointment was perhaps to be expected considering the over enthusiastic expectations! In "ordinary" use" it may well be very good indeed - will try out a final design on my Lightspeed O/P buffers (JCurl, k170 + j74, about 5 - 8mA).
And vwill try the Salas without the Series reg in front , t
Much food for thouht here.
First, Iko,
Yyes that's the Toole shunt Reg all right - I cranked it up asfar as it would go until just below dropping out of Reg and the CCS was passing about 280mA with the shunt sinking only about 30mmA - so not the best or perhaps even realistic, way of operating the reg.
So my dissapointment was perhaps to be expected considering the over enthusiastic expectations! In "ordinary" use" it may well be very good indeed - will try out a final design on my Lightspeed O/P buffers (JCurl, k170 + j74, about 5 - 8mA).
And vwill try the Salas without the Series reg in front , t
.... sorry, damn Gyration keyboard and caress fingers!!)
Thanks for the suggestion about "no pre series reg" - will try.
Evening, Salas - you're up late!
It looks like adding the bigger IRFP after all - no problem and running a good bit of extra power thru the whole thing for 250mA O/P, needs to sink the same amount, or more (thanks Andrew) thru the Shunt (250mA+) for a total thru the reg of 500mA+ - and with a reasonable heatsink (total 500+mA load, both IRPFs), keep the temp below 50*C. No problems, will do.
It appears that this design is quite happy with the heavy working conditions (3A, 13V) - This is good to know, particularly as this simpler cct, or variations, will become widely accepted.
So it looks like the specs for the + and - reg for my "oh so nice" DOA headamp looks like total current of 500+mA with 250+ thru shunt at +/- 15volts.
I'm not sure that simply converting the components to -ve will actually operate the same (j74, etc, IPRF9240, etc) so if it's possible for you to run this thru the SIM, it will save quite a bit of "stuffing about" before putting the "perfboard plumber" back to work!
Many thanks for all the input everybody, too.
Incidently, kamis, could you add the post no. on that mamoth thread so can have a look at the context - I vaguely remember something about this too, but I think it was to do with low current situations but it may actually be quite relevant to try some others. I like the idea of using the out of range k170s for something too, but other deviced perhaps maybe could be considered - thanks for the comments.
I'll add one of my own here - I would seriously suggest anyone that hasn't looked at really good diodes in the supply bridges should do so, as the improvement is not to be sneezed at (sorry, 'tis wot i'm doink now!) and the addition of at least the simple "Haggerman" snubbers is essential - check his website.
... Bye.
Thanks for the suggestion about "no pre series reg" - will try.
Evening, Salas - you're up late!
It looks like adding the bigger IRFP after all - no problem and running a good bit of extra power thru the whole thing for 250mA O/P, needs to sink the same amount, or more (thanks Andrew) thru the Shunt (250mA+) for a total thru the reg of 500mA+ - and with a reasonable heatsink (total 500+mA load, both IRPFs), keep the temp below 50*C. No problems, will do.
It appears that this design is quite happy with the heavy working conditions (3A, 13V) - This is good to know, particularly as this simpler cct, or variations, will become widely accepted.
So it looks like the specs for the + and - reg for my "oh so nice" DOA headamp looks like total current of 500+mA with 250+ thru shunt at +/- 15volts.
I'm not sure that simply converting the components to -ve will actually operate the same (j74, etc, IPRF9240, etc) so if it's possible for you to run this thru the SIM, it will save quite a bit of "stuffing about" before putting the "perfboard plumber" back to work!
Many thanks for all the input everybody, too.
Incidently, kamis, could you add the post no. on that mamoth thread so can have a look at the context - I vaguely remember something about this too, but I think it was to do with low current situations but it may actually be quite relevant to try some others. I like the idea of using the out of range k170s for something too, but other deviced perhaps maybe could be considered - thanks for the comments.
I'll add one of my own here - I would seriously suggest anyone that hasn't looked at really good diodes in the supply bridges should do so, as the improvement is not to be sneezed at (sorry, 'tis wot i'm doink now!) and the addition of at least the simple "Haggerman" snubbers is essential - check his website.
... Bye.
salas said:
Guys,
Don't waste your precious J74's on these CCS's, use K170's but
remember to put them in the right way 🙂
From memory there is a very cheap Philips jfet that has almost
identical specs to K170, I'll dig it up.
cheers
T
Well Ill ..... It's as simple as that, eh! Many of the j74s are pretty scattered re Idss so, even tho they're not as common (!) as the k170s, this is as good a place for them.
Thanks Salas - will get myself organised and give this a go, maybe at the weekend. Strange to see it written down, but a 15VA shunt regulators is still a bit new!
I wonder what it would do to the sound of my F3 clone that already sounds rather good - it uses +42V @ constant 1.6A single rail- it looks quite straight forward after your 3Amp one - once you get around the idea of the worse case shunt dissapation, the possibilities are wide
Now look what you've started!!!
Incidently it's 3.30 in the a'noon here (still that stupid daylight saving hour business) but what time is it over there - about breakfast time?
Thanks Salas - will get myself organised and give this a go, maybe at the weekend. Strange to see it written down, but a 15VA shunt regulators is still a bit new!
I wonder what it would do to the sound of my F3 clone that already sounds rather good - it uses +42V @ constant 1.6A single rail- it looks quite straight forward after your 3Amp one - once you get around the idea of the worse case shunt dissapation, the possibilities are wide
Now look what you've started!!!
Incidently it's 3.30 in the a'noon here (still that stupid daylight saving hour business) but what time is it over there - about breakfast time?
Terry Demol said:
Found it, BF862
http://www.datasheetcatalog.com/datasheets_pdf/B/F/8/6/BF862.shtml
Easily available and cheap.
T
A'noon Terry,
A good substitute would be nice.
The BF245C are quite plentiful but I don't think it's anyway up to the k170 results, but for the j74 ....
A good substitute would be nice.
The BF245C are quite plentiful but I don't think it's anyway up to the k170 results, but for the j74 ....
jameshillj said:
6:55 AM here. Use any JFET simply available that comes near, except for Q5, gave the circuit as a guide.
jameshillj said:
K170 is a (slightly) better choice than J74, it is cheaper, has lower
capacitance and is available.
BF862 is readily available, cheaper again and is identical to K170.
FWIW, the zener ref can be changed to a TL431 and biased up with
a resistor. It will have less voltage drift with temp and will be a
more stable reference. However it will need to be RC decoupled to
attenuate noise.
Otherwise the reg is really very good. I simmed JWB's and the
performance is quite spectacular considering the small parts count.
cheers
Terry
Don't substitute the Zener in my opinion, it has no drift in practice and its the least noisy. The shunt gets to nominal at about 10sec and then never drifts. I use such shunts for 5-6 months now. The RC filtering means time, and time means slow down. I avoid.
In practice all the variations are same for core performance. And the Mosfet CCS does sound better. Ok I said it!😀
In practice all the variations are same for core performance. And the Mosfet CCS does sound better. Ok I said it!😀
Sorry Terry,
Where can you get these BF862 fets - can't seem to find them anywhere local here in Melb, and WES, RS, Farnell don't list them.
Unfortunately, it's SMD but quite easy to add legs to it.
Where can you get these BF862 fets - can't seem to find them anywhere local here in Melb, and WES, RS, Farnell don't list them.
Unfortunately, it's SMD but quite easy to add legs to it.
The BF244 and 245 series are medium pinch off voltage jFETs. They are not equivalents to the k170.jameshillj said:
Sorry Andrew,
The headamp draws 250mA (class A) and yes, those 245 are just ordinary - pity, as I've got a lot of 'em!
I see you're following the Ultimate Dac thread,too. I do like that 1541A sound a lot.
The headamp draws 250mA (class A) and yes, those 245 are just ordinary - pity, as I've got a lot of 'em!
I see you're following the Ultimate Dac thread,too. I do like that 1541A sound a lot.