MagneQuest/Peerless Forum

Welcome! Need support, you got it. Or share your ideas and experiences.

Return to MagneQuest/Peerless Forum


Message Sort: Post Order or Asylum Reverse Threaded

A Compact CCS Ver 1.

68.231.117.172

Posted on May 22, 2004 at 14:02:14
VoltSecond
Audiophile

Posts: 2512
Location: Arizona
Joined: October 14, 2000

Here's a CCS which should be suitable for the Compact Amp's output stage.

R3, D2, R7 and C2 make the CCS much more fool proof. They don't cost much so just use them.

Play safe and play longer! Don't be an "OUCH!" casualty.
Unplug it, discharge it and measure it (twice) before you touch it.

. . .Oh!. . .Remember: Modifying things voids their warrantee.

 

Hide full thread outline!
    ...
Hi Paul,, posted on May 23, 2004 at 14:46:13
Bas Horneman
Audiophile

Posts: 3951
Joined: March 28, 2001
In what way is this CCS optimized for use a cathode circuit as opposed to say a DN2540 mosfet?

Cheers,
Bas

 

I'm not ruling out the Supertex DNxxxx FETs, posted on May 23, 2004 at 22:42:58
VoltSecond
Audiophile

Posts: 2512
Location: Arizona
Joined: October 14, 2000
I'm not ruling out the Supertex DN FETs. You don't need me to design a circuit to use them. However, I'll politely point out some of their short comings.

If you are running less than 20 mA, less than 400V B+ and don't mind hand tuning each CCS, the DN2540 may be your beast of choice. From the data sheets I'd actually go for the DN35xx instead if it was actually offered in a TO-220 package.

With the DN series, like all FETs, I'd recommend a resistor in series with the drain to act both as a drain stop resistor and a current limiting resistor under fast rise time fault conditions. Drop as much voltage across this resistor as you can afford and still have the CCS operate at maximum output power at 104V ac low line.

Here we go:

A comparison of a text book FET CCS vs. DN2540 CCS assuming the similar protection measures on both.

1. The DN2540 is sole sourced. There are widely available drop in replacements from different manufacture's for the FETs version. Strangely enough, this is the biggest reason why drew up the FET version.

The IXYS CCS parts aren't drop in replacements.

The curves on the IXYS depletion mode IXTP02N50D and IXTP01N100D aren't given on the IXYS site. So I can't tell if they'll drop in with or without tweaks.

Future-Active has the IXTP01N100D in stock, but I don't think they sell to hobbyists and I think the minimum order is 50 pcs.
The IXTP01N100D isn't in stock at Digikey and NuHorizons, but shows up in the product search.
The IXTP01N100D isn't carried by All American.

2. Let's look at the allowed temperature rise.

A K10 thermal insulators runs 1C/W Rcs.
A DN2545N3 TO-92 runs 170C/W Rja
A DN2540N5 TO-220 runs 8.3 C/W Rjc, 9.3 C/W Rjs with insulator
A TO-220 FETs run 1.0 to 3.5 C/W Rjc, 2.0 to 4.5 C/W Rjs with insulator.
A IRFBC30 FET is 1.7 C/W Rjc, 2.7 C/W Rjs with insulator

With a 30C ambient, 30C sink rise and a 100C junction
A DN2545N3 TO-92 is good for 0.41W..(9 mA at 45V)
A DN2540N5 TO-220 is good for 4.3W..(95 mA at 45V)
A FET is good for 8.9 to 20WW..............(198 mA to 444 mA at 45V)
A IRFBC30 FET is good for 14.8W.........(325 mA at 45V)

2.1 With the FET version, I can easily divert power loss to a cheap resistor. Strangely enough, heatsinks often cost more than the FETs on them. I don't know if I can pull this off with the DN FETs, it may actually work.

3. At currents over 20 mA, the FET current source still swings below 2V drop, the DN2540 can not.

You can see this from the saturation characteristics on the datasheet. For a MOSFET to be a CCS, it needs to be in the "Flat Horizontal Line" area of the curve. The DN2540's -1V curve cuts out at 20 mA with a 0.75V drop. Which is the same voltage drop where the FET CCS cuts out if you eliminate the 0.7V drop from the R2 protection resistor.

At 100 mA, you'll need >5V across the DN2540 for it to operate as a CCS, at 100 mA the DN2530 looks good down to about 2V.

4. With B+ over 400V, the FET current source will have a good chance of surviving a tube arc. The DN2540 doesn't have much of a chance. Below 400V the DN has a chance. We can buy FETs to 1000 to 1200V now.

5. The FET CCS uses an external part (transistor) to set the bias current. This transistor will cause the CCS to have a consistent -0.3%/C temperature coefficient independent at the current set point. This transistor will be thermally coupled to the ambient temperature, not the heatsink temperature.

With 10V drop across the part, the DN2540's temp co will vary from zero to -0.3%/C depending on the current set point. However, the DN2540 gets hot and will see a bigger swing in temperature so it will be easier for the DN2540's to be mismatched at higher current levels. However, the DN2540 will perform better at lower current levels.

Because of the FET version is coupled to the ambient temperature instead of the heatsink, a pair of CCS's using the FET version are likely to track better than a pair of DN CCSs.

6. I normally see about +/- 0.5V change in the gate voltage thresholds over different lots of parts. The data sheets usually allow +/- 1.0V change. This means at some current levels, you may not be able to adjust in a DN2540. The FET version will always work and will work the same.

Look at the DN curves and go up 0.5V or down 0.5V. If you can adjust it in to your operating point with this variation and keep the junction temperature less than 100C in a 30C ambient, I'd consider using the DN part.

7. After the FET version is tuned up once, you can step and repeat to make multiples of it for years. The DN2540 probably will require adjustments between production lots of parts.

What I've seen with FETs is parts made within 6 months of each other tend to work the same. When they are separated by years or come from different factories, they start to see more variation. I expect the DN's to behave the same as standard FETs in this regard.

Play safe and play longer! Don't be an "OUCH!" casualty.
Unplug it, discharge it and measure it (twice) before you touch it.

. . .Oh!. . .Remember: Modifying things voids their warrantee.

 

CCS issue, posted on May 24, 2004 at 22:39:14
direct
Audiophile

Posts: 10
Location: nj
Joined: May 16, 2004
Hi,
I am not an expert on circuit design. But I am very happy with my experience of CCS chip "IXCP10M45S" .
If I have to build CCS myself, I would use 431 and fixed bias reference with FETs ...

Just my two cents about CCS.

 

Caution: 431's can be fussy parts., posted on May 24, 2004 at 23:11:19
VoltSecond
Audiophile

Posts: 2512
Location: Arizona
Joined: October 14, 2000
431's can be fussy parts. When I've used them, I've had to roll them off at low frequencies to prevent oscillations.

Where'd you pick up the IXYS part?

Play safe and play longer! Don't be an "OUCH!" casualty.
Unplug it, discharge it and measure it (twice) before you touch it.

. . .Oh!. . .Remember: Modifying things voids their warrantee.

 

Can you comment on the IRF710?, posted on May 26, 2004 at 09:23:25
Russ57
Audiophile

Posts: 3754
Location: South Florida
Joined: November 16, 2001
Thinking of using it for a CCS load for a pair of KT88/KT90's. B+ of around 450VDC. A -12 bias supply will be used as well to trim each tube to same idle current.

Russ

 

Re: Can you comment on the IRF710?, posted on May 26, 2004 at 21:11:54
VoltSecond
Audiophile

Posts: 2512
Location: Arizona
Joined: October 14, 2000
I need to know more about what your are doing to give a good comment. Is this a plate load or cathode load? Plate loads need higher performance than cathode loads.

Remember silicon has no sense of humor when it comes to exceeding the rated voltage. The IRF710 is 400V. At 450V it will probably become smoke. Tubes spec the bias voltage as their maximum rated voltage. Transistors spec the 1 picosecond peak voltage.

Have about 10% margin under worst case conditions with silicon. I mean do mean worst case. High line, no or light load.

I'm starting to like the IRFIBC line. The plastic coating on these is fairly rugged and good to many many volts. The down side to the IRFIBC line is that the Rtheta jc for the part runs 4 C/W (5 C/W with a Berg K10 insulator.)

Buy your FET based on Crss at the voltage you are using it. Lower is better. Pay more for a 2:1 difference. Use the less expensive one for a 40% difference. Crss varies with voltage so look at the curves, not the tables. Beware that some manufactures spec maximum values only, other nominal values only. Make sure you compare max to max or nominal to nominal.


Play safe and play longer! Don't be an "OUCH!" casualty.
Unplug it, discharge it and measure it (twice) before you touch it.

. . .Oh!. . .Remember: Modifying things voids their warrantee.

 

Cathode load, posted on May 27, 2004 at 12:17:41
Russ57
Audiophile

Posts: 3754
Location: South Florida
Joined: November 16, 2001
First off let me confess my total ignorance with regards to solid state CCS design and theroy.

Now that we have that out of the way I'll try to explain the circuit but I might get the whole gate/drain/source backwards:)

We have two KT90's, triode connected, in push pull with a B+ of 450VDC applied to the output transformer center tap. We will also apply a -12 bias source to the control grids of the KT90's. Each cathode will have a ten ohm resistor with these resistors terminating at the IRF710's drain(?). The source(?) will be connected to ground through a 39R 2 watt resistor. The gate(?) will have a 1K resistor leading to the positive terminal of a nine volt battery and the negative terminal of this battery will be connected to ground.

Hopefully I have explained this well enough for you to get a concept. Any info/opinions you care on this design would be most welcome. Also an educated guess as to actual tube bias (idle) current would be appreciated. Also, if I happen to feel the need to adjust the idle current would it be correct to assume I would increase the value of the 39R resistor slightly?


Thanks in advance,
Russ

 

Re: Cathode load, posted on May 27, 2004 at 21:53:32
VoltSecond
Audiophile

Posts: 2512
Location: Arizona
Joined: October 14, 2000
With the circuit you described, with a new battery, you'll get around 0.13 A tail current and this current will drop over time. I'd be putting a 100 uF 20V capacitor across the 9V battery for noise reasons.

Since your going to have about 40-50V across the CCS, I'd actually recommend a slightly different circuit using two series 6.2V zeners biased from the B+ with a 430K 1W resistor. There'll be about 8.4v across the bias set resistor.

A tail current of 0.17A bias point would be better.


Play safe and play longer! Don't be an "OUCH!" casualty.
Unplug it, discharge it and measure it (twice) before you touch it.

. . .Oh!. . .Remember: Modifying things voids their warrantee.

 

Ain't that the truth!, posted on May 27, 2004 at 22:26:37
Paul Joppa
Industry Professional

Posts: 7006
Location: Seattle, WA
Joined: April 23, 2001
Been spending a lot of time the last week or two sorting out some 431 circuits. Like all the chip regulators (in fact, anything with an opamp inside), thay have stability issues. Worst are those with high impedance collectors at the output, where an external RC sets the dominant pole. It's almost impossible to find out what the gain and poles/zeroes are by reading the data sheets, too - very frustrating! I just today ran across an application note that said (re: a low-dropout series regulator) that the reason the data sheet recommends a tantalum 4.7uF capacitor on the output is that the AVX brand tantalum 4.7uF cap happens to have a 1.5 ohm equivalent series resistance at a few hundred kHz, and that resistance is critical to stability. Apparently they assume ("proof by blatant assumption", we used to call it) that all tantalums have the same resistance, since no mention is made of the resistance in the data sheet. Sheesh!

For anyone else struggling with this, the ON semiconductor 431 data sheet is the only one I've found which has information on design for stability.

 

Re: Cathode load, posted on May 27, 2004 at 22:28:55
VoltSecond
Audiophile

Posts: 2512
Location: Arizona
Joined: October 14, 2000

Simple, unprotected CCS.

Play safe and play longer! Don't be an "OUCH!" casualty.
Unplug it, discharge it and measure it (twice) before you touch it.

. . .Oh!. . .Remember: Modifying things voids their warrantee.

 

Thanks(nt), posted on May 28, 2004 at 04:31:25
Russ57
Audiophile

Posts: 3754
Location: South Florida
Joined: November 16, 2001
.

 

I've considered a dual loop design., posted on May 29, 2004 at 23:21:19
VoltSecond
Audiophile

Posts: 2512
Location: Arizona
Joined: October 14, 2000
Use the TL431 to set the DC bias point and use something else for the high frequencies. If you can afford 5-10V drop in a CCS, local feedback alone works really good for high frequencies.

Play safe and play longer! Don't be an "OUCH!" casualty.
Unplug it, discharge it and measure it (twice) before you touch it.

. . .Oh!. . .Remember: Modifying things voids their warrantee.

 

noise..., posted on May 31, 2004 at 17:35:42
Sector-7G
Audiophile

Posts: 11773
Joined: May 14, 2002
Hey-Hey!!!,
I would suggest a circuit like Gary Pimm uses. Battery positive big resistor, positive terminal( if a 'lytic is used) of a cap, gate stopper, gate. the negative terminal of the battery, negative cap terminal, and other end of the current-set resistor all get strung together. Basically an RC filter for the battery.

Anyway, the rest looks cool. I do find the cascode of two MOSFET's always sounds better than a single device rigged like you just suggest.
regards,
Douglas

 

CCS issue, posted on May 31, 2004 at 19:02:24
direct
Audiophile

Posts: 10
Location: nj
Joined: May 16, 2004
I got them from digikey.

431 works fine with me. just remember that you need caps at the top leg of the voltage divider to avoid oscillations, they can take current but not be able to deal with voltage drop over 37 volts.

I usally put them in the source loop to provide current sensing with good results.

 

CCS issue, posted on May 31, 2004 at 19:19:17
direct
Audiophile

Posts: 10
Location: nj
Joined: May 16, 2004
It is obvious to me that when using shunt regs you need have resistance in serial with the chip, otherwise .....

Anyway, I think 431 is a fairly useful chip with any application need <100 ma total current. CCS just happen to be one of the usages. I have build quite a few shunt regs using 431 and found that once it works, it stays there working forever. The quality control of 431 is not that great, a lot of deviations such that it should be used by experienced builders only.

 

Re: Like so Mike? Parafeed compact 2A3, posted on May 31, 2004 at 19:19:17
direct
Audiophile

Posts: 10
Location: nj
Joined: May 16, 2004
It is obvious to me that when using shunt regs you need have resistance in serial with the chip, otherwise .....

Anyway, I think 431 is a fairly useful chip with any application need <100 ma total current. CCS just happen to be one of the usages. I have build quite a few shunt regs using 431 and found that once it works, it stays there working forever. The quality control of 431 is not that great, a lot of deviations such that it should be used by experienced builders only.

 

Page processed in 0.027 seconds.