Emitter resistor in HexFet OPS

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Now, what "stabilize" actually means in this case?
I guess not thermal stability, because we are not sharing current (no parallel devices).
So what is it? Prevents fets from oscillating?

Just trying to understand what exactly these resistors do in this case.

If the mosfet heats up its resistance goes down and it starts to draw more current. If you aren't careful with your design you can get thermal runaway.
If you have source resistors if the current goes up the voltage across the resistor goes up taking power away from the mosfet and creating negative feedback (degeneration.)
 
Rds-on is not a parameter that is relevant for linear output stage. Typically, the very low Rds-on mosfets are the Trench type or similar architecture and suffer hot spot issues in linear applications. These are more suited as switches.

It is used in lateral mosfet output stages to limit max current.
They use a gate-source zener to limit input voltage and that in conjunction with RDSon limits max current to the speaker.
 
Indeed. F7 Manual: http://www.firstwatt.com/pdf/prod_f7_man.pdf
Nelson Pass
It had most of what was on my wish list:
Very wide bandwidth
Low distortion and noise
Large Class A operating region
Less feedback
No degeneration in the output stage
Very low thermal distortion and drift
No capacitors or transformers (apart from the power supply)

attachment.php

But what MOSFET's is he using? Are they lateral's? 6moons audioreviews: FirstWatt F7
Or are there vertical/DMOS F7 versions around?
I have been a bit slack keeping up:eek:. Sorry Nelson. But I did enjoy your 2019 BAF talk:cool:
 

Attachments

  • Baisc-F5-but-with-no-source-resistors-is-F7.png
    Baisc-F5-but-with-no-source-resistors-is-F7.png
    3.9 KB · Views: 523
Since the mid-1990's, the lowest on-resistance vertical power MOSFETs usually suffer from Spirito instability, which is basically a kind of thermal second breakdown. Those are totally unusable for class-A, -B or -AB amplifiers, with or without source resistors, bias loops or whatever. They break down even when the total drain current is kept under control, because at highish drain-source voltages, the hottest spots of the MOSFET die start drawing more current, get even hotter and so on. ....
Hi MarcelvdG,

Thanks for that info. Onsemi AN-4161 https://www.onsemi.com/pub/Collateral/AN-4161.pdf.pdf and Wikipedia Power MOSFET - Wikipedia has a link an interesting paper: "The Trench Power MOSFET: Part I—History,
Technology, and Prospects" Richard K. Williams et al
(free).

The IRF6898 plot you posted (left plot) is interesting that it shows hot-spotting at even 0.2V across the MOSFET at 10A (where the slope increases). It is a 25V device (for switching) but a look at the Id vs vds shows weak avalanche from 15V at any current (plot attached). So these super-trench VFETs are no use for "class-A, -B or -AB amplifiers, with or without source resistors, bias loops or whatever" like you said...

But what about using them in Rush pairs? My Echo amp that I simulated has DMG4800 Trench FETs in the emitters of a 3281+1302 output stage where the FET's dominate the control the BJT current to give a fairly squarelaw crossover region and with no (literal) emitter/source resistors. An earlier Echo circuit using LU1014D's is shown here.

The FETs are both n-channel and wired as diodes but are still in control of the BJTs because the BJT's act as followers and pass their base voltage to the gates which then determine the emitter current. That's the case when the gm of the FETs are less than the gm of the BJT's -- which is arranged by choosing a suitable FET including choosing a FET that never getting into a hot-spot breakdown situation.

BTW Wikipedia says,
"As of 2018, over 50 billion power MOSFETs are shipped annually. These include the trench power MOSFET, which sold over 100 billion units up until February 2017".
I guess most of the power MOSFETs now are trench FETs. Does this mean the early generation vertical DMOS that we need for audio amplifiers will get the EOL flags? Hopefully not, because trench power MOSFETs are unuseable in linear applications which include hot swap current limiters and other applications apart from audio amplifiers. So they have to keep producing the older generation power MOSFET's that we need for audio amplifiers - lucky for us! :grouphug:.

BTW2 the Williams Trench History paper (link above) at the end mentions integrating real time temperature sensing into the Trench cell structure. This allows a fast trip for over-temperature (or to limit voltage drive dynamically) but it isn't a cure for hot-spotting. One manufacturer patent mentions grading the gain of cell regions that hot spot first, but even that doesn't completely get rid of the problem. It would take an integrated circuit within the trench cells to dynamically control gate voltages of cell blocks to overcome hot-spotting and allow safe linear operation. It's probably being looked at.

Cheers,
 

Attachments

  • IRF6898-Id-Vds-25V-214A-78W-1mR-Trench.png
    IRF6898-Id-Vds-25V-214A-78W-1mR-Trench.png
    9.3 KB · Views: 124
There is one practical reason to use emitter resistors beside thermal stability. It is much easier to measure voltage drop on these resistors using DMM than to measure current through one of the rails. If PCB does not have fuse holders or special test points measuring bias current will be difficult withou these resistors.
 
Last edited:
But what about using them in Rush pairs? My Echo amp that I simulated has DMG4800 Trench FETs in the emitters of a 3281+1302 output stage where the FET's dominate the control the BJT current to give a fairly squarelaw crossover region and with no (literal) emitter/source resistors. An earlier Echo circuit using LU1014D's is shown here.

The FETs are both n-channel and wired as diodes but are still in control of the BJTs because the BJT's act as followers and pass their base voltage to the gates which then determine the emitter current. That's the case when the gm of the FETs are less than the gm of the BJT's -- which is arranged by choosing a suitable FET including choosing a FET that never getting into a hot-spot breakdown situation.

I should have known you would find a way to use them anyway... :) Another option is a cascode stage, of course.
 
Last edited:
I should have known you would find a way to use them anyway... :) Another option is a cascode stage, of course.
Hi MarcelvdG,

Yes. The down side with Rush pairing or cascodes is the extra parts needed and heatsinking the trench devices with low cost tend to come in DPAK for SMD and are not easy to mount on heatsink with no mount hole and only two wire stubs.

There has to be enough of an advantage over standard output stages to make it all worth the effort.

That may explain why no one else has done it (at least that I know of).
 
Why do source resistors affect sound quality?

Back to source resistors in vertical MOSFET power amps.

Nelson Pass in his BAF 2019 presentation mentions adding source resistors affected the sound quality (start from 9 min and get slides here). Nelson used his own form of blind testing (13 mins). So Nelson was convinced enough to not use source degeneration resistors in his XA25 MOSFET output stage.

So why do source resistors affect the sound quality so much in vertical MOSFET output stages?

The simple answer is source degeneration wrecks the squarelaw of vertical MOSFETs. If you don't use source resistors you can get closer to squarelaws in the Class-A region and closer to squarelaw means closer to no distortion (since perfect squarelaws gives no distortion). And no source resistors widens the class-A region. And further, no source resistors gives more total gain (gm) in the output stage, so the output stage Av as a follower gives less distortion from the higher gm.

I showed in an article in Linear Audio (Vol.1) that adding 0.22 ohm source resistors to IRFP240/9240 has 8 times more distortion than without source resistors (see article Table 1B rows 6 and 7). When the harmonic structure is compared with and without source resistors there is a bigger difference than 8 to our ears. With source resistors there are more high order harmonics produced due to the narrower Class-A region.

What does this mean for power amps. If you use the usual source resistors in the output stage then it produces far more distortion than with no source resistors which means you need to use at least 20dB more feedback (maybe 30dB more) to get distortion below audible levels.

Put the other way, if you don't use source resistors then you can get away with a lot less feedback and still have no noticeable distortion. With no source resistors and >=90dB/W/m speakers the Class-A band (assuming Class-AB biasing) can be wide enough to cover the majority of the signals and then you can get away with just local feedback of a follower stage (and assuming you can use a high level preamp to drive it). Then global feedback becomes unnecessary.

Things change a lot when source resistors are not used in vertical MOSFET output stages.

@ ivanlukic. For monitoring idle current a 0.1 ohm can be added in one of the power rails and removed (or shorted) once the current has been set.
 
I built classic Hitachi lateral MOSfet circuit with one pair of output transistors without source resistors (as on the Hitachi schematic). I liked the sound better than similar APEX FH9 circuit with HexFets. Based on your comment it could be that source resistors in APEX changed things for worse subjectively. I shall remove source resistors from the APEX to see what happens.

I built Hitachi circuit with Exicon outputs without protective Zener inside, and with Exicons with protective Zener inside. I liked the sound of these lateral FETs without Zeners much, much better. Could it be that Zeners also affect sound quality in negative way? Similar as source resistors?
 
and with Exicons with protective Zener inside.
I am not sure if they have internal zener diodes in them. I will ask them since it is not documented like it is for the Hitachi devices. If they do use them they should document that they do.
Even then, I think one should still use external zeners and not rely on the internal ones. The internal ones I believe are meant for static discharge gate protection purposes and not meant to be used as signal level clamps.
Could it be that Zeners also affect sound quality in negative way?
Hum, so how does one determine that? I ? why you think that and why the internal ones would not do the same?
I asked Bob Cordell about using source resistors for laterals, he said he never uses them, only for verticals and bjts.
 
rsavas,

Obviously, I was not precise enough. Exicon produced two versions of lateral Mosfets in TO247 package: one with internal Zener and the other without zeners. Their laterals without internal zeners were considerably cheaper and sounded better. Pic attached is Exicons without internal zener They have suffix Z (08N16Z and 08P16Z). Profusion does not offer them anymore for unknown reason. I built Hitachi circuit with laterals that do not have internal zeners and without external zeners. No problem whatsoever. They just sounded better.
 

Attachments

  • Exicon.JPG
    Exicon.JPG
    111 KB · Views: 279
Last edited:
Anthony Holton HPA (The Saint) said Exicon laterals were changed at his request to not have zener diodes removed to "improve the sound". He also said they used the chip area where the zeners were to increase the gate resistor to improve the stability when two chips were paralleled in one case (double die versions).

The latter change may account for improved sound quality by preventing RF oscillations or ringing between the two chips in parallel.

Now they have discontinued the zener free version I have no idea whether they have reverted to lower value internal gate resistors. Maybe PM Anthony on this.
 
Member
Joined 2010
Paid Member
Back to the OP....
There have been more than a few threads about almost any topology feature in any type of audio circuit possible, at this forum. Googling the output stage type gives you a wide field - adding "resistor" to the search narrows it down.

The emitter or source resistors are not just about current sharing. Here is a detailed, point by point explanation of mosfet transistor operation that I thought I'd throw in, as it treats only a single power transistor for clarity. Cross-over behaviour isn't applicable here but I'm sure you can easily do your own search for that.
mosfet - I don't understand how the resistor between GND and the source in a common-source amplifier stabilizes it - Electrical Engineering Stack Exchange
 
Last edited:
Supercap bypass for vertical MOSFET source resistors

I built classic Hitachi lateral MOSfet circuit with one pair of output transistors without source resistors (as on the Hitachi schematic). I liked the sound better than similar APEX FH9 circuit with HexFets. Based on your comment it could be that source resistors in APEX changed things for worse subjectively. I shall remove source resistors from the APEX to see what happens...
Hi ivanlukic,

Be careful to have the bias setting reduced to minimum before power up. You don't need to remove the resistors physically, just solder a wire bridge under the PCB.

Another option you could try is using a supercap bridging the two sources of the MOSFETs. Leave the source resistors in, then for AC the resistors are no longer degenerating each MOSFET but for DC they provide thermal stability as usual. A suitable super cap is Maxwell Tech 10F 3V BCAP0010 P300 X11 here and is relatively inexpensive (AU$5 ea) and not huge (Dia 10mm 30mm long). I chose this one over say the 505DCN2R7 at 5F, 2.7V, 10x20mm based on longer shelf life (4kh at 60C) when left uncharged. Probably typical room temperatures they will last 20-30 years like standard electrolytics and probably not cause any new concerns with life.

As for the likely benefit I have a simulation attached showing a significant improvement. One simulation shows the change in the wingspread plot with and without the supercap bypass with 0.22 ohm resistors and IRFP240/9240 pair biased at 330mA. A screenshot is below.
attachment.php

The other simulation shows the difference in THD at +/-1A swing (the end of the square-law Class-A region) or about 4W average into 8 ohms. The FFT plot is below
attachment.php

The THD at 4W is reduced from 0.13% to 0.013% or by a factor of 10. The FFT plot shows most of the reduction is from reducing the 3rd harmonic by about 30 times.

The wingspread simulation shows after stopping the input there is no noticeable displacement in the bias point using a 10Hz signal at full output when using with a 10 Farad capacitor. A 5F cap is good also. I tried a standard electrolytic of 66,000uF 6.3V (2x33mF) which is not too costly or large but there is skewing at 20Hz preventing significant distortion reduction at the LF end. So supercaps seem to be the best. I'm not sure about their ESL and resonant frequency as the datasheets don't specify this. Maybe a 4,700uF/6.3V will be needed in parallel to push the resonant frequency above the audio frequency band? A bench test will tell all. Please let us know what you find.

This amount of improvement is after having trimmed the source resistors by 26m ohms to account for differences in the P- and N- MOSFET's. Also using my IRFP240/9240 models shows more improvement with the supercap than using say the LTspice standard library IRFP240/9240 models (to run the inbuilt LTspice models just remove the underscores in my models).
-----
BTW I haven't tried this on a BJT output stage. Will the BJT be happy with this and not go into thermal runaway with their emitter resistors bypassed by this capacitance?:eek: The time constant of 10F with their ESR of 20m ohms gives a pole at about 1Hz. I have no idea if it is safe or not? Don't blame me if you try it and it blows up your BJT amp. It could be simulated. Anyone interested?
 

Attachments

  • DMOS-AB-supercap-bypass.zip
    4.2 KB · Views: 56
  • DMOS-AB-supercap-bypass-FFT.png
    DMOS-AB-supercap-bypass-FFT.png
    9.1 KB · Views: 496
  • DMOS-AB-supercap-bypass-wingspreaad.png
    DMOS-AB-supercap-bypass-wingspreaad.png
    10.8 KB · Views: 484
Last edited:
Supercap bypass for BJT OPS emitter resistors

Here are some sims for a pair of NJL3281/1302 with and without a 10F emitter bypass capacitor. Iq=120mA, Re=0.22 (see attached circuit).
attachment.php

upper Red is with the 10F cap and lower Red without cap. Green shows unity gain.
attachment.php

THD reduces from 0.12% without cap to 0.015% with the cap at 0.5W ave (within the wingspread 'valley').

The PNP has 20mR added to it's emitter resistor and the base resistors are offset to get better symmetry to lower the 2nd harmonic.
 

Attachments

  • Class-AB-supercap-emitter-bypass-wingspreaad.png
    Class-AB-supercap-emitter-bypass-wingspreaad.png
    9.7 KB · Views: 479
  • Class-AB-supercap-emitter-bypass-FFT.png
    Class-AB-supercap-emitter-bypass-FFT.png
    9.2 KB · Views: 479
  • Class-AB-supercap-emitter-bypass.zip
    4.4 KB · Views: 58
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.