I knew it would be sacrilegious to ask about a solid state current source in a tube forum. 
A tube ccs might be better, but I don't think my budget will stretch that far.
Well, maybe, if it could operate on the same B+ supply. It would definitely look more impressive with two tubes instead of one. As it is, the 6922 will be looking a little lonely.
A tube ccs might be better, but I don't think my budget will stretch that far.
Well, maybe, if it could operate on the same B+ supply. It would definitely look more impressive with two tubes instead of one. As it is, the 6922 will be looking a little lonely.
fdegrove said:Add to that that I have an almost instinctive distrust for semiconductors floating at highish voltages....
Hell, they don't even glow...
Let's face it, if anything is going to break down in a valved product it's 9/10 a zener, a reg or anything else solid state that per definition has it's reference removed by the manufacturer or at best is merely obsolete...which means yesterday nowadays.
Is there anything you can do with transistors you can't do with tubes?
Not that much, I'd say...
Well, what about in the tail of a diff amp? Doesn't need a very large supply for good compliance even with big honkin transmitter tubes, let alone a preamp. Hardly a high voltage risk.
Or as the anode load in a preamp or driver? Hell, in mine, if the power supply magically vanished and supplied rectified AC directly onto them, they'd still be well within limits....
And if you look at Gary's well designed CCS's, you'd see there are two types: the BBMCCS for lowish voltages/powers (even though the FETs are 500V) and the version using a pentode, to amongst other things take most of the voltage and find a way to dissipate power above chassis.
Le bollocks
EC8010 said:
In the application the original poster wanted one for, I fail to see the reason to add a -200V rail, with all the attendant filtering issues when a simple -15V reg supply and an IRF820, two resistors and a cap (and maybe some Li's for bias) will do an even better job according to Gary Pimm's measurements. I don't have a sufficiently resolving test setup to do it myself, so I'll take his data on spec.
Well.... IF you must!
'
The following is an extract from email sent to me by Sparhawk
OK. Please keep in mind that the DIY Tube Hybrid IGC was intentionally developed along the KISS principle. For those not in the know, that stands for Keep It Simple Stupid (definitely not Knights In Satan's Service). And no deference to the intelligence of those inclined to built it. Sorry... a little black humour, I know.
So....
Along with the same principle, here is the simplest way to as a Constant Current source to our Tube Buffer:
As always there are options, but having established KISS then I assume that the circuit is as constructed as above. We can discuss alternatives as we go on. So we assume we have +/- 35V rails as shown.
Changes: The 10K 1W resistor has been halved, hence *4K7. Then *56R is adjusted to give us 3.5 to 4mA. This implies about 20V drop across 4K7, leaving circa 15V for the CS Fet to operate across.
The fet can be a number of different TO-92 types, such as 2N5459 or similar - these are quite easily available. The IDSS needs to be 4mA or higher. Then the *56R value is adjusted for correct current. This is only a starting value.
These smaller TO-92 fets are not suitable for 35V plus and indeed should be kept at or below 20V. The simplest way to to that is the *4K7 - KISS remember. The alternative is add a 7915 regulator and eliminate 4K7.
The 1K resistor is there for safety - isolating the Grid avoids any unwanted parasitic oscilations.
Joe R.
PS: Incidentally, the JLTi Hybrid Amplifer, on which the DIY version is based, is not KISS. The CS is tube and also the Cathode Follower is 'referenced' to the signal and not the power supply, as is usual with CF. The full JLTi and its SuperLinear Cathode Follower (sorry no values given) actually looks like this:
'
'
The following is an extract from email sent to me by Sparhawk
OK. Please keep in mind that the DIY Tube Hybrid IGC was intentionally developed along the KISS principle. For those not in the know, that stands for Keep It Simple Stupid (definitely not Knights In Satan's Service). And no deference to the intelligence of those inclined to built it. Sorry... a little black humour, I know.
So....
Along with the same principle, here is the simplest way to as a Constant Current source to our Tube Buffer:
An externally hosted image should be here but it was not working when we last tested it.
As always there are options, but having established KISS then I assume that the circuit is as constructed as above. We can discuss alternatives as we go on. So we assume we have +/- 35V rails as shown.
Changes: The 10K 1W resistor has been halved, hence *4K7. Then *56R is adjusted to give us 3.5 to 4mA. This implies about 20V drop across 4K7, leaving circa 15V for the CS Fet to operate across.
The fet can be a number of different TO-92 types, such as 2N5459 or similar - these are quite easily available. The IDSS needs to be 4mA or higher. Then the *56R value is adjusted for correct current. This is only a starting value.
These smaller TO-92 fets are not suitable for 35V plus and indeed should be kept at or below 20V. The simplest way to to that is the *4K7 - KISS remember. The alternative is add a 7915 regulator and eliminate 4K7.
The 1K resistor is there for safety - isolating the Grid avoids any unwanted parasitic oscilations.
Joe R.
PS: Incidentally, the JLTi Hybrid Amplifer, on which the DIY version is based, is not KISS. The CS is tube and also the Cathode Follower is 'referenced' to the signal and not the power supply, as is usual with CF. The full JLTi and its SuperLinear Cathode Follower (sorry no values given) actually looks like this:
An externally hosted image should be here but it was not working when we last tested it.
'
Joe,
Thanks for the input. Your solution looks very similar to the one that moamps posted at the beginning of this thread. I'm convinced now that a JFET ccs is the easiest way to go, with the lowest parts count.
Is the 1K gate resistor really necessary? I noticed that moamps didn't include it in his version.
What are your thoughts on the operating point for the 6922? I'm leaning towards a higher b+ supply, and a slightly higher current.
BTW, if you don't mind me asking, what tube do you use for the ccs in the JLTi? And what is the B- voltage required?
Thanks for the input. Your solution looks very similar to the one that moamps posted at the beginning of this thread. I'm convinced now that a JFET ccs is the easiest way to go, with the lowest parts count.
Is the 1K gate resistor really necessary? I noticed that moamps didn't include it in his version.
What are your thoughts on the operating point for the 6922? I'm leaning towards a higher b+ supply, and a slightly higher current.
BTW, if you don't mind me asking, what tube do you use for the ccs in the JLTi? And what is the B- voltage required?
@ all:
Another option (KIEMSS) is usage of a current regulator diodes like 50V J509/510/511 (3mA/3.6mA/4.7mA) or 100V 1N5309 thru 1N5314 (3mA thru 4.7mA) or 100V CR300 thru CD470 (3mA thru 4.7mA), but bandlimiting by the gate stopper resistor is not possible this way. Sparhawk, I don’t know if it is important to use it with tubes, but with FET as signal device, it certainly is.
@ Joe:
Do you consider usage of 40V Vgds diodes in 35V application unsafe or it is the availability the reason behind your suggestion?
Options are:
2SK170, Vgds = -40V
2SK363, Vgds = -40V
2SK369, Vgds = -40V
2SK246, Vgds = -50V (Y designated has Idss lower than needed though)
2SK389, Vgds = -50V
2N5546/5556/5566, Vgds = -40V
J111/112, Vgds = -40V
Pedja
PS: KIEMSS = keep it even more stupid simple.
Another option (KIEMSS) is usage of a current regulator diodes like 50V J509/510/511 (3mA/3.6mA/4.7mA) or 100V 1N5309 thru 1N5314 (3mA thru 4.7mA) or 100V CR300 thru CD470 (3mA thru 4.7mA), but bandlimiting by the gate stopper resistor is not possible this way. Sparhawk, I don’t know if it is important to use it with tubes, but with FET as signal device, it certainly is.
@ Joe:
Do you consider usage of 40V Vgds diodes in 35V application unsafe or it is the availability the reason behind your suggestion?
Options are:
2SK170, Vgds = -40V
2SK363, Vgds = -40V
2SK369, Vgds = -40V
2SK246, Vgds = -50V (Y designated has Idss lower than needed though)
2SK389, Vgds = -50V
2N5546/5556/5566, Vgds = -40V
J111/112, Vgds = -40V
Pedja
PS: KIEMSS = keep it even more stupid simple.
Sparhawk said:
Lowest parts count is exactly what I had in mind. Lots of Jfets will do the job. Just adjust a single value R.
The 1K gate resistor necessary? Probably not. But I don't know the exact layout and fet you guys will be using so it is a bit of insurance on my part. Have I had fet cs misbehave on me? Don't know if I have had single fet cs do that - but a cascoded cs (ultra-high Z) with the fet on the bottom, yeah, I have! Once shy...
Re higher voltages and current in the JLTi, they are +/- 55V rails and about 4mA. The tube is Sovtek 'Reflector' 6922 - I had to use a high quality current production easily available tube and a friend imports these. For manufacturing purposes, no NOS tubes.
More current, a little more wouldn't harm. Do I prefer a little more volts on the rails, I do as evidenced in the JLTi. The DIY +/- 35V rails I would consider minimum. I know there are projects (and actual products) out there using 24V with 6DJ8/6922. That's too low for comfort for me. If you up current and lower voltage then grid current becomes a risk. You will get DC through your volume pot.
I would recommend no more than 90V, more like 40-60V.
There is no need to go 100V plus (many samples of this tube are rated no more than 90V) and it was always my intention to provide a Low Tension Tube Project with non-lethal voltages, so that anyone wanting to move up to tube projects from SS (and judging by the feedback I'm getting, that has happened) could use this to cut their teeths on tubes.
It also explains why the Tube Buffer is so simple, sans CS. The more experienced constructor, I'd thought, might figure out for themselves it (cs) could be done. And thus it turned out.
Joe R.
PS: Don't know if any noticed I said Grid on the fet when I meant Gate. It had been a long day.
'
Pedja said:
Hi Pedja
I think there are cases where KIEMS could be desirable.
Mind you not always.
Shouldn't that be Keep It Even More Simple Stupid?
OK, let's say we've got 35V rails, add a volt for the approx +1V on the Cathode plus the initial turn-on no load on the rails and you could get awfully close to 40V. Then the current, say 4mA and you have to dissipate 160mW in a TO-92 package?
Yeah, it can be done. I would still like to put a 1K resistor according to my more academically minded friend (hi Terry) who says that will isolate any low level capacitive effects of the fet.
As said, lot of jfets out there can do the job. You can also used an 'enhanced cascoded cs' - Ultra-High Z - but this is neither KISS nor KIEMSS as it requires careful bias componentry (is that a new word I just coined - call me Shakespeare).
Joe R
Hi Joe,
… in fact both KISS and KIEMSS approaches look quite smart to me. Actually, that’s the way (a-ha, a-ha) I like it (there are some things that might be a bit better though).
Attached a 3.6mA (with 100V/475mW ratings) example of KIEMSS.
Pedja
YupJoe Rasmussen said:I think there are cases where KIEMS could be desirable.
Mind you not always.
Shouldn't that be Keep It Even More Simple Stupid?
… in fact both KISS and KIEMSS approaches look quite smart to me. Actually, that’s the way (a-ha, a-ha) I like it (there are some things that might be a bit better though). This sounds interesting (regards to Mr. Demol), will check it as soon as possible…Joe Rasmussen said:
Attached a 3.6mA (with 100V/475mW ratings) example of KIEMSS.
Pedja
Attachments
I stumbled on to this thread and I am offering this schematic to you. I designed this (with a lot of help) back when I was into solid state but the cascaded CCS will work just as well with tubes. You will have to change the values of the resistors of course. The odd values are Caddock mk-132 resistors available from Michael Percy. As you can see it will handle a lot of current. If you only need 5mA then you can change the mosfets as necessary. The diodes are LEDs (super bright blues I think). I don't don't think that I ever wrote out a complete parts list but the LEDs won't be hard to find (Digikey). You are welcome to the design. I hope it helps or at least gives you some ideas. Maybe I'll build the headphone amp one day. I have PCB's designed using Express PCB's software if anyone is interested. As I remember the cost for the full amp was quite reasonable and should sound great.
G
G
Attachments
Hi Guys
OK... This is not exactly KISS and definitely not KIEMSS... so let us go to the max.
This is H-IGC now features an enhanced cascoded constant current source:
It is pretty well self-explanatory. This should handle 60V OK and indeed a little more if the jfets are rated at 40V.
The 15K resistors splits up the bias voltage so that each fet sees approximately the same potential. The Red LED provides the reference for the CS. Notice we can now use a much higher value R (in this case 560R). This higher value plus the cascoding results in Ultra-High Z.
Re 0V this should be "G" in earlier schematic.
The 1K on the Cathode should find approval with Terry (yes, Demol - but I, unlike you guys, know him in the flesh).
Actually should be here Sunday next, and Allen Wright is turning up too.
Joe R.
OK... This is not exactly KISS and definitely not KIEMSS... so let us go to the max.
This is H-IGC now features an enhanced cascoded constant current source:
An externally hosted image should be here but it was not working when we last tested it.
It is pretty well self-explanatory. This should handle 60V OK and indeed a little more if the jfets are rated at 40V.
The 15K resistors splits up the bias voltage so that each fet sees approximately the same potential. The Red LED provides the reference for the CS. Notice we can now use a much higher value R (in this case 560R). This higher value plus the cascoding results in Ultra-High Z.
Re 0V this should be "G" in earlier schematic.
The 1K on the Cathode should find approval with Terry (yes, Demol - but I, unlike you guys, know him in the flesh).
Actually should be here Sunday next, and Allen Wright is turning up too.
Joe R.
Brett said:
Seems like they'll let anyone into the country
Joe, how would you compare your JLTi's to the PP2C? I've built some PP2C (and PP1C) and like them quite a bit.
Hi Brett
Allen, as you may know, spends most of his time in Europe. He is in Australia at the moment, somewhere between here in Sydney and Brisbane, where he was last weekend. He will be here from Sunday to Tuesday, I am then taking him to the airport and he's going to visit familty in New Zealand.
Terry lives about 1/2 hour drive from here, so I see him all the time.
Re JLTi vs PP2C - before I attempt to answer that, what version do you have, the Class AB or the one that is pure Class A differential (hush-hush)?
Off to work.
Joe R.
Joe Rasmussen said:
I know he's in Munich most of the time, and I hope he has a great trip.
I've never met Terry, but I do have him, and you, to thank for the T/S parameters and some measurements on the Goodman's 201's.
Allen made me swear on my life never to reveal the details of it, and it is diiferential. The annoying thing about that though, is I had something very similar on the bench when I received the schematic.....
He also sent a copy to a friend in Melbourne who's madly building a pair now.
You may wish to use a Depletion Mode MOSFET.
Here is some food for thought.
http://home.zonnet.nl/horneman/mosfet.htm
Here is some food for thought.
http://home.zonnet.nl/horneman/mosfet.htm
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.