John Swenson's New BDT preamp - Page 2

Sheldon · 21st May 2008, 02:13 AM

Thanks for chiming in Gary, and helping a couple of neophytes. My understanding is that the actual capacitance is not so much the issue as the fact that the slope of capacitance vs Vds is steep at low Vds. That's why I suggested a beefier fet for the bias side. Would this help enough? Sounds like your experience suggests that it would not. In principle, adequate current to the gate should overcome the problem (mostly, at least). If DN2540 were subbed for the LND150, the current could be raised even higher than the 10-20mA I proposed.

Sheldon

oldeurope · 21st May 2008, 07:33 AM

Hello,
the next step is to replace the 6AR7 by (three) MOS FETs.

Kind regards,
Darius

jkeny · 21st May 2008, 11:11 AM

Thanks Gary & Sheldon,
It seems like the approach with Mosfets is not feasible - (in the datasheets for low ultra-low CRSS, it's impossible to make out the Crss at low voltages due to the scaling of the graphs)

How about bi-polars instead - what issues would this give rise to?

As you can tell I'm just a worker bee in this hive - thanks to both of you for your expertise & to Gary for this wonderfully versatile CCS.

It's amazing the number of different configurations this circuit has been transformed into
- Solid State Pentode amplifier
- I/V
- Transconductance amplifier
- Voltage Regulator
I'm probably missing a few?

Oh, BTW, has anybody done a run of Gary's boards yet? I'd be interested in buying some as I wired up the 250 V VR & the CCS on perfboard & it was a pain!

jkeny · 21st May 2008, 06:18 PM

Actually, Sheldon,
I didn't fully pick up on your suggestion about substituting LND150 with DN2540 as a way out of the Mosfet over-heating issues! This would be worth investigating!

Could somebody on that side of the pond email me 10 or 20 of these that I would order from Mouser & I would order whatever number extra for that person to cover postage?

jkeny · 22nd May 2008, 08:59 AM

Please help a dummy in understanding this!

By increasing the current output of the LND150 voltage reference cascode (or substituted DN2540) drives more voltage into the gate of Q1 and this allows for a higher current output from the Mosfet - OK.

But how does this alleviate the problem of Drain to Source needing to be a voltage differential of 20V and at 8A causing Mosfet (circuit) overheating?

Still trying to learn the basics!

Sheldon · 22nd May 2008, 03:17 PM

Quote:

Originally posted by jkeny
By increasing the current output of the LND150 voltage reference cascode (or substituted DN2540) drives more voltage into the gate of Q1 and this allows for a higher current output from the Mosfet - OK.

No. Current and voltage can be set independently. Current in the bias string is set by R1. The voltage across R2, sets the voltage difference between the gates (and outputs) of Q2 and Q1. This drop across R2 is, of course, set by the current delivered by the Q3/Q4 cascode. So if you decrease R1 to increase the current, you have to decrease R2 by the same proportion, to maintain the same voltage relationships as before.

The actual output voltage is set by the combined resistance of R3 and R8 (with Q5 to amplify the error detected at the output and feed back on the input) As for R2, this has to be calculated based on the current set by R1.

Voltage delivered to the gates of Q1 or Q2, has nothing to do with the current they can source. That depends entirely on the load and physical limitations of those FETs. Simplistically, if you need more current capacity, you use a bigger FET or parallel up a bunch of smaller ones (which is what bigger ones are anyway). The problem with that, is that you increase the capacitance seen at the gate. Further, the problem with that capacitance is not so much its absolute amount, it's that it increases rapidly as the voltage across the FET decreases toward zero (check out the curves for Crss - above about 20Vds, the capacitance becomes nearly constant). This changing capacitance means that the frequency response of the output will vary with voltage - not good for sound, as it's not something we would encounter naturally.

My suggestion of increasing the current through the bias string for high current applications was an attempt to provide enough current to push the frequency response issues high enough up in frequency to get any effects out of the audible range. Don't know if it is actually accomplished - that could be simulated. Gary suggests that it wasn't, but maybe he was using the LND150. That leaves the possibility that added bias current would help - maybe Gary will give help us here.

If more current still isn't enough, one could simply add enough capacitance across the gate, to swamp out the FET capacitance, and make the capacitance seen by the bias drivers essentially constant. Then, at least any roll off and it's consequent phase shift would be much more benign to the ear.

Sheldon

jkeny · 22nd May 2008, 03:40 PM

Ok, thanks for explaining Sheldon. I'm with you up to the second last paragraph

Quote:

provide enough current to push the frequency response issues high enough up in frequency to get any effects out of the audible range

I don't want to pester you but how does extra current do this? Are we not still operating in the CRSS area of high slope? Is the only way of operating in a more linear part of the graph through inreased voltage accross the FET?

Sorry if all this is EE 101 but I'm still learning the basics!

The last suggestion of extra capacitance to swamp out the FET's seems too easy - there must be a downside to this?

Sheldon · 22nd May 2008, 10:03 PM

Quote:

Originally posted by jkeny
how does extra current do this? Are we not still operating in the CRSS area of high slope? Is the only way of operating in a more linear part of the graph through inreased voltage accross the FET?
The last suggestion of extra capacitance to swamp out the FET's seems too easy - there must be a downside to this?

The issue is the impedance feeding the gate of Q1. I think in this circuit the impedance would be roughly equal to R2. And let's say we are using STW34NB20, a 200V, 34A rated FET. At 2Vds, Crss is about 1nF. Therefore the roll off corner (1/2piRC) would be at about 6kHz, and phase changes would start about a decade lower than that, at around 600Hz - clearly in the audio frequency range. If we want to decrease the impedance looking back from the gate by a factor of 20, we have to reduce R2 by that factor (1.35k). Well, Ohms law tells us, that in order to keep the voltage relationships the same, we need to pass 20 times the original current through it. That would get us to an f3 of 120kHz, and push up the start of phase changes to 12kHZ. Ok, now we are getting there.

If we increase the minimum voltage across Q1 to 5V then the Crss drops to about 300pF, push up the frequency changes by about 3 fold further.

Or, lets add 4nf from gate to source. Now, the corner drops 5 fold, from 120kHz to 24kHz, but the capacitance/Vds slope would be 1/5, or much shallower than before. The benefit is a flatter impedance/frequency response. The downside is the frequency response of supply is lowered overall.

How all this would affect the sound, I don't know - would depend to some degree on the amplifier PSRR. But I suspect that there would be some benefit to having your power supply frequency/phase constant at least up to the limit of the audio range.

Brian Beck describes the issue here: http://www.diyaudio.com/forums/showt...010#post999010

Sheldon

Sheldon · 23rd May 2008, 05:24 PM

One thing to note is that Brian, and I think Gary, are referring primarily to using the CCS in something like a plate load. In that case, the Vds changes with the signal. In the case of the regulated supply, the output voltage is constant, and the input to the control FET is also nearly constant, as it is isolated the the FET above it. Therefore, the slope of capacitance with Vds would not have as big an impact.

Maybe the main issue would be to make sure your power supply does not roll off lower than your amp. This could potentially cause problems of instability.

I think it's worth a try.

Sheldon

jkeny · 23rd May 2008, 10:24 PM

Thanks Sheldon,
So the absolute value of gate capacitance is the more important factor rather than the slope of change?

What I wanted to use this higher current/ low voltage reference for - a Panasonic SAXR57 Digital amp that uses voltage control to adjust the volume. It's range is from 10V to 40V at a max of 6A maybe. So this complicates matters I think & might make it not feasible given the gate capacitance rise at the lowest voltage

As the digital amp is a no feedback design, the quality of the PS is paramount (PSRR of zero).

I will probably give this a try after I've built up the BDT preamp & have learnt some more - thanks for your help in all this.

21st May 2008, 07:33 AM	#12
oldeurope diyAudio Member Join Date: Mar 2008	Multiplier ... Hello, the next step is to replace the 6AR7 by (three) MOS FETs. Kind regards, Darius

Thread Tools	Search this Thread
Show Printable Version Email this Page	Search this Thread: Advanced Search

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
John Swenson's (In my eyes revolutionary) preamp!	Bas Horneman	Tubes / Valves	71	23rd February 2012 02:58 PM
The JC-2 preamp schematic on the website is wrong...Mr.John C?	X.G.	Solid State	77	26th September 2009 04:14 PM
John Linsley Hood RIAA Preamp?	roger-k	Solid State	5	16th March 2007 07:10 PM
Toshiba 3950 w/ John Swenson Mod vs. 1543 Non-Oversampled Dac	Tuber	Digital Source	6	28th November 2004 01:42 PM

New To Site?	Need Help?
Register to Participate Search Privacy Statement Contact Us	Frequently Asked Questions Did you forget your password? Mark Forums Read


	Please consider donating to help us continue to serve you. Ads on/off / Custom Title / More PMs / More album space / Advanced printing & mass image saving

21st May 2008, 02:13 AM	#11
Sheldon diyAudio Member Join Date: Dec 2003 Location: San Diego	Thanks for chiming in Gary, and helping a couple of neophytes. My understanding is that the actual capacitance is not so much the issue as the fact that the slope of capacitance vs Vds is steep at low Vds. That's why I suggested a beefier fet for the bias side. Would this help enough? Sounds like your experience suggests that it would not. In principle, adequate current to the gate should overcome the problem (mostly, at least). If DN2540 were subbed for the LND150, the current could be raised even higher than the 10-20mA I proposed. Sheldon

21st May 2008, 11:11 AM	#13
jkeny Banned Join Date: Feb 2007 Location: Dublin	Thanks Gary & Sheldon, It seems like the approach with Mosfets is not feasible - (in the datasheets for low ultra-low CRSS, it's impossible to make out the Crss at low voltages due to the scaling of the graphs) How about bi-polars instead - what issues would this give rise to? As you can tell I'm just a worker bee in this hive - thanks to both of you for your expertise & to Gary for this wonderfully versatile CCS. It's amazing the number of different configurations this circuit has been transformed into - Solid State Pentode amplifier - I/V - Transconductance amplifier - Voltage Regulator I'm probably missing a few? Oh, BTW, has anybody done a run of Gary's boards yet? I'd be interested in buying some as I wired up the 250 V VR & the CCS on perfboard & it was a pain!

21st May 2008, 06:18 PM	#14
jkeny Banned Join Date: Feb 2007 Location: Dublin	Actually, Sheldon, I didn't fully pick up on your suggestion about substituting LND150 with DN2540 as a way out of the Mosfet over-heating issues! This would be worth investigating! Could somebody on that side of the pond email me 10 or 20 of these that I would order from Mouser & I would order whatever number extra for that person to cover postage?

22nd May 2008, 08:59 AM	#15
jkeny Banned Join Date: Feb 2007 Location: Dublin	Please help a dummy in understanding this! By increasing the current output of the LND150 voltage reference cascode (or substituted DN2540) drives more voltage into the gate of Q1 and this allows for a higher current output from the Mosfet - OK. But how does this alleviate the problem of Drain to Source needing to be a voltage differential of 20V and at 8A causing Mosfet (circuit) overheating? Still trying to learn the basics!

23rd May 2008, 05:24 PM	#19
Sheldon diyAudio Member Join Date: Dec 2003 Location: San Diego	One thing to note is that Brian, and I think Gary, are referring primarily to using the CCS in something like a plate load. In that case, the Vds changes with the signal. In the case of the regulated supply, the output voltage is constant, and the input to the control FET is also nearly constant, as it is isolated the the FET above it. Therefore, the slope of capacitance with Vds would not have as big an impact. Maybe the main issue would be to make sure your power supply does not roll off lower than your amp. This could potentially cause problems of instability. I think it's worth a try. Sheldon

23rd May 2008, 10:24 PM	#20
jkeny Banned Join Date: Feb 2007 Location: Dublin	Thanks Sheldon, So the absolute value of gate capacitance is the more important factor rather than the slope of change? What I wanted to use this higher current/ low voltage reference for - a Panasonic SAXR57 Digital amp that uses voltage control to adjust the volume. It's range is from 10V to 40V at a max of 6A maybe. So this complicates matters I think & might make it not feasible given the gate capacitance rise at the lowest voltage As the digital amp is a no feedback design, the quality of the PS is paramount (PSRR of zero). I will probably give this a try after I've built up the BDT preamp & have learnt some more - thanks for your help in all this.