Bob Cordell Interview: BJT vs. MOSFET

GK

Disabled Account
Joined 2006
estuart said:



Hi Glen,

let's go back to the beginning. I just gave an answer to forr's question, see above. The subject deals with alternate compensating schemes and definitely not with trivial things like improving the power supply lines. Besides, it makes a big difference in terms of cost, space and elegance, whether to use a big 1mF cap (as Self suggests) or a small 100pF cap. For the rest, I don't disagree with you on the effectiveness of a RC filter. That’s all!


We don't need to go back to the beginning - this is completely irrelevant. In the discussion, D. Self was mentioned. I simply elaborated on a position of his with respect to PSRR and R-C filtering the rails of the low power circuits. Is that OK with you?


Bob Cordell said:



I agree, Glen. But having both approaches in place is certainly best.

Bob


I agree also, but R-C filtering (or another form of conditioning, such as regulation in a fancy design) the low power rails gets precedence in my opinion, because it impacts not only upon the VAS, but relaxes the necessary PSR performance of the rest of the low power/level circuitry as well. In a high performance design, it is much better to deal with a problem such as supply rail noise at the source of the problem first.


Fanuc said:
I am not really sure about that. I ask this in another post with respect to op amps.


A packaged opamp isn't a valid comparison. R-C filtering the supply rails of an opamp of course will not change the devices PSRR, it will only limit the supply ripple/noise on the supply pins.
A discrete power amplifier is a different kettle of fish. Here we can seperate the power supply rails for the low power curcuitry such as the VAS, the input differential amplifier, etc. from the power supply rails for the output and driver stages and filter them.
This isolates the low power circuitry, to a degree, from the rubbish on the main supply rails such as power supply ripple and Class A/B switching spikes, etc.


Cheers,
Glen
 
AX tech editor
Joined 2002
Paid Member
Re: Ayre Acoustics MX-R fuse blown

Bob Cordell said:
[snip]Could there be something going on with the use of the ThermalTrack transistors for bias stability control? I'd guess Charlie has at least four pairs of output devices on each side of the bridge. Might there be an issue concerning how you choose which output transistor's tracking diode (or combiantion of several) to use for bias setting? Could there have been some kind of a thermal runaway situation? I wonder how big Charlie's emitter ballast resistors are.[snip]Cheers,
Bob


Bob,

You may be on to something here. Those ThermalTrack transistors are not well thought out, in that, although the diode is on the same chip as the transistor, the tempco of the diodes and the transistors is not matched. IMHO, ThermalTrack transistors give a completely unfounded safe feeling against thermal runaway.

In contrast, if you look at the Sanken SAP16N and P transistors, you will see that there they were quite aware of that issue. They even recommend specific current through the diodes and a specific bias current for the transistors, so that tempco matching occurs.

I am running the SAP16's at slightly elevated bias currents (60mA versus the reconmmended 40mA for reasons of xover distortion) and noticed that at this higher bias they also show signs of thermal runaway, although they appear perfectly thermally stable at 40mA bias.

Jan Didden
 
EUVL said:
I have a question regarding values of the gate stopper resistor for MOSFETs.
Does anyone use Sampei's formular (US Patent 4168471) to calculate the gate stoppers, or simply trial and error ?

See also :

http://www.diyaudio.com/forums/showthread.php?postid=1166050#post1166050


Patrick

Hi,

Just had a look at that excel formula. One thing I noticed is the values of capacitances are wrong in source follower configuration due to the boot strapping effect. I asked Bob about this up north.

edit:: Just had another look at the capacitances and if they are for the sum of the devices it may be OK.

Have you seen his error correction paper where he talks about input capacitance and slew rate etc. ?

The P Chl having more bandwidth than the N chl seems wrong. (unless I am reading the spreadsheet wrong)

Actually one thing Bob does do that is of interest to me, is use RC networks over the gates of the Fets - allowing lower values of R and hence more bandwidth.

I would like to see if RC networks could be used with Lateral's also, as I have many spares. Hitachi 2SK1056 & 2SJ162. The lowest gate stoppers I have seen is about 100r but not in paralel.

Somebody mentioned a nifty idea of cutting part of the gate lead away on T-O3P or TO-264 types and soldering in the gate resistor so it can just slot in the PCB!!.

I think ahle back somebody linked to a IRF paper (i think) that showed that zener's where inadaquate gate protection for Fets. I cannot rememeber the link unfortunately.

Do you have any links to that patent/paper itself ?

Thanks

Kevin
 
G.Kleinschmidt said:

A packaged opamp isn't a valid comparison. R-C filtering the supply rails of an opamp of course will not change the devices PSRR, it will only limit the supply ripple/noise on the supply pins.
A discrete power amplifier is a different kettle of fish. Here we can seperate the power supply rails for the low power curcuitry such as the VAS, the input differential amplifier, etc. from the power supply rails for the output and driver stages and filter them.
This isolates the low power circuitry, to a degree, from the rubbish on the main supply rails such as power supply ripple and Class A/B switching spikes, etc.

Cheers,
Glen

Would you also say that electronic regulation of the input and driver stages gives the amplifier better PSRR. ?
 
G.Kleinschmidt said:
We don't need to go back to the beginning - this is completely irrelevant. In the discussion, D. Self was mentioned. I simply elaborated on a position of his with respect to PSRR and R-C filtering the rails of the low power circuits. Is that OK with you?
.......
A packaged opamp isn't a valid comparison. R-C filtering the supply rails of an opamp of course will not change the devices PSRR, it will only limit the supply ripple/noise on the supply pins.
A discrete power amplifier is a different kettle of fish. Here we can seperate the power supply rails for the low power curcuitry ...
Glen

Hi Glen,

Yes, that is OK with me.
I must admit, while we were actually squabbling over the definition of PSRR (not the merits of RC-filtering), that I had wrongly IC related PSRR in mind, in which case, contrary to discrete circuits, device PSRR will not change. What troubles me is that there is apparently more than one definition of PSRR.

Cheers,
 
andy_c said:


Charlie has a number of non-audio-related concerns to worry about. If you don't get my drift, see his blog at ayre.com. The whole turn of this thread frankly makes me uncomfortable reading it. It's insensitive in my view.


Andy,

I understand completely, and am well aware. There was certainly no intention to be insensitive at a time when Charlie continues to need our thoughts and prayers. And indeed, in spite of our interest in what went wrong with the amplifier under some very difficult laboratory conditions, he has built one fine amplifier.

Bob
 
EUVL said:
I have a question regarding values of the gate stopper resistor for MOSFETs.
Does anyone use Sampei's formular (US Patent 4168471) to calculate the gate stoppers, or simply trial and error ?

See also :

http://www.diyaudio.com/forums/showthread.php?postid=1166050#post1166050


Patrick


Wow! I was unaware that someone patented a formula for a gate stopper resistor! I'm unfamiliar with the patent.

I did discuss this issue in my paper on a MOSFET power amplifier with error correction. You can find it on my web site at www.cordellaudio.com. Use of a gate stopper resistor alone is not the optimum solution. One needs to understand the various origins of the parasitic oscillations and then take steps to properly damp the resonant circuits involved. With a proper approach, I was able to get down to gate series resistors as low as 47 ohms, thus preserving more of the inherent bandwidth of the MOSFET devices. But still, I must admit that there was a lot of trial and error involved (but coupled with a good understanding of oscillator topologies that can be formed when all of the device parasitics are considered).

Although power MOSFETs are fairly robust devices, they will pop very quickly if their maximum gate-source voltage is exceeded. When parasitic oscillations occur (sometimes at over 100 MHz), the internal voltage swings at the gate can exceed what is at the external terminal of the device, and this is why external gate zeners may not protect a device from failure due to parasitic ocillations.

Bob
 
I have a copy of the patent in pdf which is 511k, so too large to post here. But if you google the patent number, you should have no problem getting at least the text. I think you can even get pages in pdf at the European Patent Office's website.

No he has no claims of the formular, just the fact that there are gate resistors.

But I have looked through his maths and his model five or six times, and I cannot find anything wrong with it. Which then makes me puzzled, as also from experience, I know I can probably get away with much less, about a factor of 3.

No the capacitances were not wrong. There were values from the datasheet for one single device at 20V Vds, as is in the case of Nelson's F4 design. Also I was myself very surprised, but if you believe IRF's datasheets (I made sure I downloaded the latest copy directly from IRF), then the P devices has less capacitances than the N devices.

The only mistake I can find up till now in the excel sheet is that I Coss and Ciss should be swapped in Column A. The rest is still fine, as far as I can tell.

I shall see whether I can post only the relevant pages of the patent and get them down to 100k or so.


Patrick
 

GK

Disabled Account
Joined 2006
Fanuc said:


Would you also say that electronic regulation of the input and driver stages gives the amplifier better PSRR. ?


With respect to the supply rails powering the output stage and driver circuits, yes. ;)



estuart said:


Hi Glen,

Yes, that is OK with me.
I must admit, while we were actually squabbling over the definition of PSRR (not the merits of RC-filtering), that I had wrongly IC related PSRR in mind, in which case, contrary to discrete circuits, device PSRR will not change. What troubles me is that there is apparently more than one definition of PSRR.

Cheers,


No worries. I don’t think that there is more than one definition of PSRR. For packaged op-amps or discrete power amplifiers, it means exactly the same thing – a measure of how well an amplifier rejects noise and ripple on it's supply rails. With discrete power amplifier circuits there may be multiple sets of supply rails, with regards to which the PSR performance of the amplifier as a whole needs to be defined for each individually.

Cheers,
Glen
 
Re: Re: Ayre Acoustics MX-R fuse blown

janneman said:



Bob,

You may be on to something here. Those ThermalTrack transistors are not well thought out, in that, although the diode is on the same chip as the transistor, the tempco of the diodes and the transistors is not matched. IMHO, ThermalTrack transistors give a completely unfounded safe feeling against thermal runaway.

In contrast, if you look at the Sanken SAP16N and P transistors, you will see that there they were quite aware of that issue. They even recommend specific current through the diodes and a specific bias current for the transistors, so that tempco matching occurs.

I am running the SAP16's at slightly elevated bias currents (60mA versus the reconmmended 40mA for reasons of xover distortion) and noticed that at this higher bias they also show signs of thermal runaway, although they appear perfectly thermally stable at 40mA bias.

Jan Didden


Hi Jan,

I would have thought that one could accomodate any differences in the tempco of the sensing diode in the details of design of the bias circuit (e.g., vbe multiplier, or whatever).

Nevertheless, there may be some real subtleties associated with optimum use of ThermalTrak transistors under what might be hot, high current, stressful conditions. One also may need to wonder about the now rather short thermal delay time constant in regard to the musical waveform (sometimes we have to be careful what we wish for; the painfully long delay of the conventional heat sink mounted diode sensor may cover up other unanticipated behaviors).


I didn't realize that Sanken had devices with an on-chip sensing diodes as well.

Cheers,
Bob